xvEPA
             University of
             North Carolina
              Biological Sciences
              Research Center
              Chapel Hill NC27514
             United States
             Environmental Protection
             Agency
              Health Effects Research
              Laboratory
              Research Triangle Park NC 27711
EPA-600/9-77-043
December 1978
Multidisciplinary  Perspectives
in  Event-Related  Brain
Potential Research
                                     JV^^^-^NV     V

-------
                 RESEARCH REPORTING SERIES


Research reports of the Office of Research and Development, U.S. Environmental
Protection Agency, have been grouped into nine series. These nine broad cate-
gories were established to facilitate further development  and application of en-
vironmental technology.  Elimination  of traditional grouping was consciously
planned to foster technology transfer and a maximum interface in related fields.
The nine series are:

    1. Environmental Health Effects Research

    2. Environmental Protection Technology

    3. Ecological Research

    4.  Environmental Monitoring

    5. Socioeconomic Environmental Studies

    6. Scientific and Technical Assessment Reports (STAR)

     7. Interagency Energy-Environment Research and Development

     8. "Special" Reports

     9. Miscellaneous Reports


  This report has been assigned to the MISCELLANEOUS REPORTS series. This
  series is reserved for reports whose content does not fit into one of the other specific
  series. Conference proceedings, annual reports, and bibliographies are examples
  of miscellaneous reports.
                          EPA REVIEW NOTICE
   This report has been reviewed by the U.S. Environmental Protection Agency, and
   approved for publication. Approval does not signify that the contents necessarily
   wflect the views and policy of the Agency, nor does mention of trade names or
   commercial products constitute endorsement or recommendation for use.

-------
                                     EPA-600/9-77-043
                                         December 1978
Multidisciplinary Perspectives
     in Event-Related Brain
         Potential  Research
 Proceedings of the Fourth International Congress
  on Event-Related Slow Potentials of the Brain
                  (EPIC IV)
 Hendersonville, North Carolina, April 4-10, 1976
                    Edited by

                  David A. Otto

           Health Effects Research Laboratory
           Office of Research and Development
        Research Triangle Park, North Carolina 27711

                      and

           Biological Sciences Research Center
             University of North Carolina
           Chapel Hill, North Carolina 27514
                Grant No. R803494-02
             Program Element No. IAA8I7

        EPA Project Officer: M. Thomas Wagner, Jr.
                   Prepared for

     U.S. ENVIRONMENTAL PROTECTION AGENCY
           Office of Research and Development
               Washington, D.C. 20460.

-------
                                          Editor-in-Chief

                              David A. Otto
                              Neurotoxicology Division
                              Health Effects  Research Laboratory
                              U.S. Environmental Protection Agency
                              Research Triangle Park, North Carolina, U.S.A.
                                         Associate Editors
Robert M. Chapman
Center for Visual Science
University of Rochester
Rochester, New York, U.S.A.

Emanuel Donchin
Cognitive Psychophysiology Laboratory
Departments of Psychology and Physiology
University of Illinois
Champaign,  Illinois, U.S.A.

Rathe Karrer
Illinois Institute for Developmental Disabilities
Chicago, Illinois, U.S.A.

John R. Knott
Tufts University School of Medicine
Boston, Massachusetts, U.S.A.

W. Cheyne McCallum
Burden Neurological Institute
Bristol, England

Demetrios Papakostopoulos
Burden Neurological Institute
Bristol, England
Charles S. Rebert
Life Sciences Division
SRI International
Menlo Park, California, U.S.A.

Lawrence W. Reiter
Neurotoxicology Division
Health Effects Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, North Carolina, U.S.A.

Joseph J. Tecce
Laboratory of Neuropsychoiogy
Department of Psychiatry
Tufts University School of Medicine
Boston, Massachusetts, U.S.A.

Patricia Tueting
Maryland Psychiatric Research Center
University of Maryland Medical School
Baltimore, Maryland, U.S.A.

Hal Weinberg
Brain Behavior Laboratory
Department of Psychology
Simon Fraser University
Burnaby, British Columbia, Canada

-------

-------
DEDICATION TO WILLIAM GREY WALTER, M.A., Sc.D.
February 19, 1910-May6, 1977
   The Proceedings of the Fourth International Con-
gress on  livent-Related Slow Potentials  of the Brain
(FPIC  IV)  arc dedicated to the memory of William
Grey Walter, who died May 6,  1977. lie left a world
wiser  in  consequence of his many active  and  imagi-
native  years as an investigator into the physiology of
the mind.

   Grey Walter was one of the  very early pioneers in
the electrical activity of the human brain. Originally
a student of Adrian, he joined Professor Golla at the
Maudsley Hospital  in 1935 to pursue the area of FUG
research opened up by  Berger  and Adrian. His first
and very  major contribution  to FUG was  the dis-
covery of delta activity  and its locali/ing value in
brain tumor suspects.

   Grey  Walter  was  a  pioneer  who  kept  moving
ahead of the frontiers he himself established. In 1940,
he turned his efforts to automatic frequency analysis,
"on line," and sought psychological correlates  of the
changing  voltage-frequency profiles.  Since   single-
channel display was frustratingly limiting, he under-
took to  develop one of  the first truly electro-ana-
tomical displays of the brain  with  his "toposcope"
(1952-53).  Frequency, voltage,  and phase could  be
derived from 22 electrode points on  the  scalp. The
elfects ot sensory modulation of the brain were also
defined in  terms of frequency, voltage, phase, and
area.

   In  1962, Grey  noticed a negative shift that oc-
curred  between  paired  stimuli. His  initial  interest
apparently  was in  possible modification of sensory
evoked  potentials  during Pavlovian  conditioning.
With  this  observation, he opened another frontier,
that of the  event-related slow  potential which  he
named the Contingent Negative Variation. The work
was  presented in England and  the United  States in
1964,  and  published  in  Nature.  While the  initial
instrumentation  that  uncovered  this  phenomenon
was  crude, further  pioneering in computer analysis
and display, up to and beyond his retirement in 1975,
kept Grey abreast (if not ahead) of the times, and
stimulated an  increasingly  large  group of  experi-
mentalists  to explore the neurophysiology  of be-
havior. Here  at  last  was a key that  might fit some
of the locks many had been struggling with  for years.

   These are  some of the things Grey Walter has left
us; other  things are  even more  important:  memories
of  a  probing mind  that  stimulated others  to  think,
memories of a devoted scientist who would  not  re-
quire that others work harder than himself, and mem-
ories of  a  warm and embracing fellowship upon
which all other attributes revolved.

   The continued growth of event-related  potential
research  is a  tribute to  Grey Walter's scientific en-
deavors, his imagination, and his  forceful leadership.
The ongoing series  of International FRP Congresses
is  hut a  small  part  of his  scientific legacy. Grey
Walter  bequeathed  to  future  generations  his  in-
spiration  and  intellect in the form of  extensive and
eloquent  writings.  He was  a giant  in  the field  of
neurophysiology - a  remarkable  scientist, author, and
personality whose memory will always be  cherished
by  those privileged to know him.
                                                                              J. R. Knott

-------
PREFACE
   This volume is the Proceedings of  the  Fourth
International Congress on Event-Related  Slow Poten-
tials of the Brain  (EPIC IV) held in Hendersonville,
North Carolina on April 4-10, 1976. The  volume con-
tains ten  sections  devoted to the following areas of
ERP research:  (1) electrogenesis and neurochemistry,
(2)  motor control, (3)  information processing and
cognition, (4)  language,  (5) development and aging,
(6)  psychopathology, (7) environmental neurotoxi-
cology, (8) scalp distribution, (9) alternatives to  sig-
nal  averaging, and (10) theoretical models.  Sections
are based on plenary sessions  at the Congress and in-
clude, in varying  form, correspondence summaries,
data  and  review   papers,  and  condensations   of
discussion.

In memorium.

   This series of meetings began with Grey Walter's
discovery of the CNV. In failing health, Grey did not
attend EPIC  IV.  As  these proceedings were  being
assembled, we learned with great sadness of his pass-
ing.  In tribute to  the man whose vision and leader-
ship nurtured  event-related  potential research, this
volume is humbly dedicated  to William Grey Walter.

Collective planning.

   A collective approach to  conference planning,
pioneered at Bristol  in  1973, was used extensively
in preparation for  EPIC IV. Correspondence panels
in the areas listed  above were  established a year
before the meeting to (1) define critical issues,  (2)
review and synthesize available  evidence,  and  (3)
formulate experiments or strategies to resolve issues.
Correspondence summaries  were  precirculated  to
conference  participants  and  plenary  sessions  were
organized  around  correspondence  panels.  Corres-
pondence  chairmen   served  as  discussion  leaders
at EPIC  IV  and as section editors of the proceed-
ings. This volume is the final product of this lengthy
but  fruitful  experiment in collective planning and
communication.

Environmental theme.

   Two plenary sessions were devoted to "neurobe-
havioral indices of environmental insult." The objec-
tives of these sessions were to evaluate the utility of
ERP techniques in environmental toxicology and to
encourage neurobehavioral research in problems of
environmental  concern.  Neurobehavioral  evidence
has  played a  relatively insignificant  role in deter-
mining current U.S. environmental standards. Reports
in  this   volume   from  investigators   in  Austria,
Finland, Germany, Italy, and the USSR, on the other
hand,  indicate  that behavioral  and evoked potential
data play an important role in  setting threshold limit
values (TLVs)  in Eastern and Western Europe. Data
presented  in  the  toxicology section and elsewhere
in this volume  document the sensitivity of ERPs and
other  neurobehavioral  measures  to  the effects  of
psychoactive  drugs,  pesticides,  industrial  solvents,
and  other physical insults such as noise  and elec-
tromagnetic radiations. The evidence argues strongly
for increased application of neurobehavioral methods
in studying  the  adverse  health effects of environ-
mental toxicants.

Interdisciplinary bridge-building.

   Progress in  the  ERP field depends  on the inte-
gration of evidence  from many  disciplines within
the neurosciences. ERP investigators include psycho-
logists, psychiatrists, neurologists, and neurosurgeons,
although  the disciplines of neurophysiology, neuro-
anatomy, and  neurochemistry  have not been repre-
sented until recently. A major objective of  EPIC IV
was to recruit investigators from the latter disciplines
to begin to bridge fundamental gaps in  ERP know-
ledge.  The infusion of  new blood and new perspec-
tives  produced an exciting  concatenation  of ideas
reflected  in the series  of  tutorial  papers and theo-
retical models which appear in this volume.

Review procedures.

   The editing of this  volume  took far  longer than
anticipated. Two  factors  deserve mention.  In order
to achieve some  degree of quality control, all data
papers and  many  review  papers were submitted  to
peer  review.  Although few papers  were  rejected,
many authors were required to make  extensive  and
repeated revisions. In order to achieve greater clarity
and  consistency, the editorial   staff further revised
most  manuscripts. The benefits of  these efforts,
however,  must be  weighed  against  the  resulting
publication lag. I am grateful to the section editors
and  individual contributors for  their  cooperation
and patience in this frustrating and time-consuming
enterprise.

-------
                                                  Vlll
 Supplements.

    Dr.  Jon  Peters  prepared  an  extensive "Biblio-
 graphy of CNV and Other  Slow Potentials of  the
 Brain:  Part  III"  which covers the  period between
 the Bristol Congress (August 1973) and EPIC IV. To
 facilitate  distribution and  avoid redundancy with
 other  reference sections in this volume, Dr. Peters'
 excellent  bibliography  was published separately  in
 June 1978  as  an EPA Research Report (EPA-600/
 1-78-042).  This  document  is available   from the
 National Technical Information Service, Springfield,
 Virginia 22161.

    A set of manuscripts submitted by several investi-
 gators  from  the USSR who were unable  to attend
 EPIC IV  will also be published as  a separate EPA
 Research  Report and should be available from NTIS
 in  mid-1979. Further information  on these docu-
 ments may be obtained from D. Otto.

Acknowledgments.

   EPIC IV was sponsored jointly by the Biological
Sciences  Research  Center,  School of  Medicine,
University of North Carolina at Chapel Hill and  by
the  Health Effects Research  Laboratory, U.S. Envi-
ronmental  Protection Agency,  Research  Triangle
Park, North Carolina. We are deeply indebted to both
organizations  and their  respective  directors,  Dr.
Morris   Lipton  and  Dr.  John  Knelson, for  enthu-
siastic  support  and administrative assistance. Fund-
ing was provided by  EPA Grant R803494-02 to the
 University of North Carolina. Dr. Lipton was Princi-
 pal  Investigator and  Dr.  Thomas  Wagner  was EPA
 Project Officer.

   Many  others  generously  contributed  time  and
 talent  in organizing  EPIC FV and  editing this  vol-
 ume.   I would especially like to acknowledge the
 administrative assistance of Bill Heriford and Bobby
 Wagoner (UNC Extension Division), Jeanne Hernan-
 dez  (Institute   of   Environmental  Studies),  and
 Elizabeth  Clark  (BSRC);  the  technical  editorial
 and  graphics support  of Bob Kolbinsky,  Earnie
 Caldwell, Webb  White, Miriam  Harper,  Cathy  Jo
 Poole,  Pam Barnwell, Mary  Woodard, Paul Holder,
 and  Charlie Keadle  (EPA General Services Division);
 clerical  support from Barbara Queen and the HERL
 Word Processing Center, Jo Nichols and the Clinical
 Studies  Division  (EPA), Marty  Byrd  (BSRC), and
 Pat Reefe; and general  assistance from Alex Adams,
 Gayla and  Vernon Benignus, Dick Calvert, Marchell
 Franklin, Mary Hicks, Debbie Markley, Kim Nguyen,
Jim Prah, and Kathy Seiple.

   Finally, I would like to express my deep apprecia-
 tion to  W. Cheyne McCallum,  Congress  President,
 and  John R. Knott, Past  President, for their invalu-
 able advice, encouragement, and  assistance  in  all
 aspects  of  the  organization  of the Congress and
editing of the Proceedings. The Program Committee
also deserves special thanks for strenuous and inspired
efforts  in organizing correspondence,  leading discus-
sion  at  the Congress, and editing  the resultant sec-
tions of this volume.

-------
CONTENTS
                                                                                               Page
Dedication to W. Grey Walter	        v
Preface	       vi
Contents	       ix
Participants	      xvi
Abbreviations 	     xviii
Keynote address: The significance of  neurobehavioral data in  setting air quality standards.
   R. R. Beard	      xix

Section

   I. ELECTROGENESIS AND NEUROCHEMISTRY	        1

     Electrogenesis of slow potential changes in the central nervous system: A summary of issues.
        CS. Rebert  	        3

     Slow postsynaptic responses of sympathetic ganglion cells as models for slow potential changes
        in the brain. B. Libet	       12

     Contribution of neuroglia to extracellular sustained potential shifts. G. G. Somjen	       19

     Neurochemical mechanisms in the genesis of slow potentials: A review and some clinical
        implications. T. J. Marczynski	       25

      Reward contingent positive variation (RCPV) and patterns of neuronal activity in the visual
        cortex of the cat. T. J. Marczynski and G. Karmos	       36

      Acquisition of sustained potentials dissociated from massed action potentials in temporal
        conditioning. V. Rowland  	       39

      Events contingent upon cortical potentials can lead to rapid learning. /. S. Stamm,
        O. A. Gillespie, andB. B. Sandrew	       43

      Sustained activation of cortical neurons in stimulus-recognition tasks. /. M. Faster	       48

      Preliminary study of pharmacology of contingent negative variation in man. /. W. Thompson,
        P. Newton, P. V. Pocock, R. Cooper, H. Crow, W. C. McCallum, and D. Papakostopoulos  ...       51

      Effects of amphetamine and pentobarbital on event-related slow potentials in rats.
        /. H. Pirch	       56

      Cholinergic mechanism of sleep onset positive variation and slow potentials associated with
        K - complexes in cats. T. J. Marczynski	       61

      Electrogenesis references	        64

   II. MOTOR CONTROL	        75

      The present state of brain macropotentials in motor control research: A summary of issues.           77
        D. Papakostopoulos	

       Spinal cord stimulation and event-related potentials. P. Abraham, T. Doherty, S. Spencer,
        A. L. Cook, A. Oygar, L. Illis,  andM. Sedgewick	        82

       Functional significance of cerebral potentials preceding voluntary movement. L. Deecke	       87

       Experimental manipulation of motor positivity: A pilot study. J. Delaunoy, A. Gerono,
        andJ. C. Rousseau  	       92

-------
Section                                                                                            Page

       Telereceptive, proprioceptive, and cutaneous information in motor control. B. Dubrovsky  ...       93

       Methodological criteria for the validation of movement-related potents. L. Gerbrandt	       97

       Effects of movement on sensory input.  P. Hazemann	      105

       Influence of force, speed and duration of isometric contraction upon slow cortical potentials
         in man. P. Hazemann, S. Metral, and F. Lille	      107

       Neuroanatomical organization of the primate motor system:  Afferent and efferent connections
         of the ventral thalamic nuclei.  K. Kalil	      112

       Relationships between Bereitschaftspotential and contingent negative variation.
         W. C. McCallum  	      124

       Slow positive shifts during sustained motor activity in humans. D. Otto and V. Benignus	      131

       Electrical activity of the brain associated with skilled performance. D. Papakostopoulos	      134

       The electromyogram, H-reflex, autonomic function and cortical potential changes during
         the Jendrassik maneuver.  D. Papakostopoulos and R. Cooper	      138

       Oculomotor components of event-related electrocortical potentials in monkeys.  S. Rosen,
         J. Robinson, and D. Loiselle	      143

       Motor control references	      149


  HI.   INFORMATION PROCESSING AND COGNITION  	      157

       Event-related potentials, cognitive events, and information processing: A summary of
         issues and discussion.  P. Tueting	      159

       How many late positive waves are there? W. T. Roth  	      170

       Latency of event-related potentials and reaction time.  W. Ritter	      173

       Equivocation and P300 amplitude. D. S. Ruchkin andS. Sutton	      175

      The late positive component and orienting behavior. D. Friedman	      178

      Does P300 reflect template match/mismatch? /. M. Ford  	      181

      Evoked potentials and feedback. 5, Sutton, P. Tueting, andM. Hammer	      184

      Potentials associated with the detection of infrequent events in a  visual display. R. Cooper,
         P. V. Pocock, W. C. McCallum, andD. Papakostopoulos	      189

      Analysis of nonsignal evoked cortical potentials in two kinds of vigilance tasks. D. Friedman,
         W. Ritter, andR.  Slmson  	      194

      Variations In the latency of P300 as a function of variations in semantic categorizations.
         M, Kutas and E. Donchln	      198

      Topographical study of the emitted potential obtained after the omission of an expected
         visual stimulus, B. Renault and N. Lesevre	      202

-------
                                                   XI
Section                                                                                           Page

       Auditory evoked potentials, skin conductance response, eye movement, and reaction time
          in an orienting response paradigm.  W. T. Roth, J. M. Ford, P. L. Krainz, and B. S. Kopell  . .      209

       Sequential dependencies of the waveform of the event-related potential: A preliminary
          report. K. C. Squires, C. D. Wickens, N. K. Squires, and E. Donchin	      215

       Functional equivalence of signal-present, signal-absent, and threshold-detect P3s. N. K.
          Squires, K. C. Squires, and S. A. Hillyard	      218

       A CNV rebound function: Preliminary report. /. /. Tecce	      222

       Average evoked potentials and time perception. P.  Tueting  	      226

       Attention at the cocktail party:  Brainstem evoked  responses reveal no peripheral gating.
          D. L. Woods and S. A. Hillyard  	      230

       Information processing references	      234

  IV.  LANGUAGE	      243

       Language and evoked potentials:  A summary of issues. R. M.  Chapman	      245

       Issues in Neurolinguistics: Evoked-potential analysis of cognition and language.
          R. W. Thatcher	      250

       Distinguishing linguistic and stimulus effects. D. L. Molfese	      255

       Lateral asymmetry of evoked potentials and linguistic processing. D. Friedman	      258

       Individual differences and similarities  in language effects on evoked potentials.
          W. S. Brown	      261

       Contributions of linguistics and other data bases. T. J. Teyler	      263

       Methods of evoked-potential analysis in linguistic research. R. M. Chapman  	      265

       Event-related potentials associated with linguistic stimuli: Semantic vs low-order effects.
          A. L. Megela and T. J. Teyler	      267

       Visual  evoked potentials to language stimuli in children with reading disabilities.
          5. A. Shelburne, Jr.  	      271

       Choice of active electrode site and recording montage as variables affecting CNV amplitude
          preceding speech. 5. H. Curry, J. F. Peters, and H. Weinberg	      275

       Language references  	      280
  V.   DEVELOPMENT AND AGING	      289

       Development and developmental disorders: A summary of issues. R. Karrer .  . . .'	      291

       Maturation of pattern evoked potentials and visual preference in 6- to 45-day old infants:
          Effects of check size. M. R. Harter, F. K. Deaton, and J.  V. Odom	      297

-------
                                                   xn
Section

        Contingent negative variation, evoked potential, and psychophysical measures of selective
          attention in children with learning disabilities. M. F. Musso andM. R. Harter	      300

        Hemispheric effects of stimulus sequence and side of stimulation on slow potentials in
          children with reading problems, fl. Fenelon	      303

        Task-related cortical potentials in children in two kinds of vigilance tasks. D. Friedman,
           H. Vaughan, Jr., and L. Erlenmeyer-Kimling	      309

        Evoked and slow potentials during sensory conditioning in autistic, mentally retarded
           and normal children. F. Laffont, N. Bruneau, Ph. Jusseaume, and G. Lelord  	      314

         CNV and EEC patterns in children with cerebral palsy and known brain lesions.
           M. Papini, R. Zappoli, A. Pasquinelli, M. G.  Mar tine tti, and S. Guerri  	      319

         Slow potentials of the brain preceding cued and noncued movements:  Effects of
           development and mental retardation. R. Karrer, C. Warren, and R. Ruth	      322

         Age-dependence of the Bereitschaftspotential. L. Deecke, H. -G. Englitz, and G. Schmitt ....      330

         Aging effects on the human evoked potential. G. Marsh  	      333

         Discussion summary 	      337

          Development and aging references	      339
     VI.  PSYCHOPATHOLOGY  ................................................      345

          Event-related potentials and psychopathology :  A summary of issues and discussion.
             J. R. Knoll and]. J. Tecce ............................................      347

          Experimental production of postimperative negative variation in normal subjects.
             J. Delaunoy, A. Gerono, andJ. C. Rousseau  .................................      355

           Effects of visual distraction on contingent negative variation and type A and B CNV
             shapes. /. J. Tecce, J. Savignano-Bowman, and J. R. Kahle .......................      358

           Personality traits and electrophysiological factors during sensory conditioning in normal
             and psychiatric populations. N. Bruneau, P. Dubost, Ph. Jusseaume, F. Laffont, and
             G. Lelord [[[      364

           Contingent negative variation in patients affected by specific phobias. H. Barbas, L. Solyom,
             and B. Dubrovsky ..................................................

           Reliability of contingent negative variation in psychopathology. M. Timsit-Berthier,
             JDeZnoy, B. Xhenseval, and M. Timsit  ..................................       373
                           after osvchosurgery affecting thaiamocortical pathways to prefrontal
                                   *<><< F Denoth> A' A*"l*ldlr' L *OSS/' * G~ MaraaM'
              and S. Guerri  [[[

-------
                                                   Xlll
Section                                                                                             Page

        Some methodological and theoretical issues of ERP in psychiatric populations.
          B. Dubrovsky and M. Dongier	      387

        Morphological analyses of the CNV in psychiatry:  Comparison of resolution mode and
          cumulative curve methods. M. Timsit-Berthier, J. Delaunoy, and A. Gerono  	      389

        Somatosensory evoked potential as a measure of tolerance to ethanol. D. M. Scales,
          P. Naitoh, L. C. Johnson, and M. A. Schuckit	      392

        Stimulant and depressant effects of cigarette smoking, nicotine, and other drugs on the
          CNV in man. H. Ashton, J. E. Miilman, M. D. Rawlins, R. Telford, and
          J. W. Thompson	      397

        Psychopathology references	      401


   VII.  ENVIRONMENTAL NEUROTOXICOLOGY  	      407

        Neurobehavioral assessment of environmental insult:  A summary of issues.
          D. Otto and L. Reiter	      409

        Carbon monoxide and superior colliculus evoked potentials. R. S. Dyer and Z. Annau  	      417

        Effects of ozone on human central nervous system function. N. W. Grandstaff and
          R. R. Beard	      420

        Neurobehavioral assessment of effects of environmental insults early in development.
          /-. D. Grant 	      421

        Neuro- and  psychophysiological effects of moderate carbon  monoxide exposure.
          E. Groll-Knapp, M. Haider, H. Hoeller, H. Jenkner, and H. G. Stidl	      424

        Application of ERP techniques in noise research. N, E. Loveless	      431

        Low-frequency noise, selective attention, and event-related potentials. D. A. Otto and
          V. A. Benignus	      437

        Paradoxical effects of carbon monoxide on vigilance performance and event-related
          potentials.  D. A. Otto,  V. Benignus, J. Prah, and B. Converse	      440

        The use of evoked potential and behavioral measures in the assessment of environmental
          insult.  M. I, Rudnev, A. I. Bokina, N. D. Eksler, andM. A. Navakatikyan  	      444

        Diagnostic utility of neuroelectric measures in environmental and occupational medicine.
          A. M. Seppalainen	      448

        Effect of carbon monoxide hypoxia and hypoxic hypoxia on cerebral circulation.
          R. J, Traystman	  .      453

        Contingent  negative variation as an index of environmental distraction. H. Weinberg,
          S. H. Curry, and J. F. Peters  	      458

        Carbon monoxide, trichloroethylene, and alcohol:  Reliability and validity of neuro-
          behavioral effects. G. Winneke, G. G. Fodor, and W. Schlipkoter	      453

        Dissociation between time course of acetylcholinesterase inhibition and neurophysio-
          logical effects of parathion in rat and monkey. D. E. Woolley and L. R. Reiter	      470

-------
                                                    XIV
Section
         CNV and SEP in shoe industry workers with neuropathy resulting from toxic effect of
           adhesive solvents. R. Zappoli, G, Giuliano, L. Rossi, M. Papini, O. Ronchi,
           A. Ragazzoni, and A. Amantini	      476

         Electroencephalographic analysis of subacute effects of methylparathion in the mouse.
           G. A. Zapponi, C. J. Undsey, and A. Loizzo	      481

         Environmental toxicology references	      485


    VIII. SCALP DISTRIBUTION	      499

         Use of scalp distribution as a dependent variable in event-related potential studies:
            Excerpts of preconference correspondence.  E. Donchin	      501

         Problems in using volume conduction theory to localize evoked potential generators.
            R. Cooper	      511

          Calculated and empirical evoked potential distributions in human recordings.
             T. Allison  	       513

          Methodology and meaning of human evoked potential scalp distribution studies.
             T.  W. Picton,D. L. Woods, D. T. Stuss, and K. B. Campbell	      515

          Intracranial sources of event-related potentials.  W. Ritter  	      523

          Scalp topography in the localization of intracranial evoked potential sources. W. R. Goff,
             T. Allison, P. D. Williamson, andJ. C, VanGilder	      526

           Spatial frequency analysis of an EEC event in the olfactory bulb. W. J. Freeman	      533

           Scalp distribution references	      543


     IX. ALTERNATIVES TO SIGNAL AVERAGING	      547

           Excerpts of preconference correspondence 	      549

           Multivariate analysis of event-related potential data: A tutorial review. E. Donchin and
             E. Heffley  	       555

           Before averaging: Preprocessing slow potential data with a Wiener filter. P. Naitoh and
              S. Sunderman	      573

           Simple digital filters for examining CNV and P300 on a single-trial basis. D. Ruchkin and
              E. Glaser	      579

            Implicit spatial averaging of surface macropotentials. L. Gerbrandt	       582

            Neurometrics:  Quantitative electrophysiological analysis for diagnosis of learning
              disabilities and other brain dysfunctions.  E. R. John, L. Prichep, H. Ahn, D. Brown,
              P. Easlon, B. Karmel, R. Thatcher, and A. Toro  	       585

            Comments on methods of signal analysis and signal detection. H. Weinberg	      593

             Alternatives references	      601

-------
                                                  XV

Section                                                                                           Page

  X.   THEORETICAL APPROACHES AND MODELS  	      607

       A theoretical framework for evoked potential and slow potential research. W. C. McCallum , .  .      609

       The behavioral approach. N. E. Loveless	      612

       A neurophysiological model for regulation of sensory input to cerebral cortex.
          /. E. Skinner  	      616

       A parsimonious model of mammalian brain and event-related slow potentials.
          T. J. Marczynski	      626

       Integrative models:  Macropotentials as a source for brain models. D. Papakostopoulos	      635

       The role of the Bereitschaftspotential and potentials accompanying the execution of
          movement. L. Deecke	      640

       The place of consciousness in brair^ research. W. Ritter	      642

       Significance of slow potential shifts in anticipation of and during task performance.
          R. Nadtdnen	      646

       Some general considerations in formulating electrophysiological brain models.
          R. Cooper	      650

       Bimodal slow potential theory of cerebral processing. R. Cooper, W. C. McCallum, and
          D. Papakostopoulos	      651

       General discussion of theoretical issues related to evoked and slow potential changes	      657

       Concluding remarks. W.  C. McCallum   	      660

       Theoretical models references	      662

-------
 PARTICIPANTS
 Peter Abraham
 Truett Allison
 Rodney Beard
 Warren Brown
 L. Burns
 Enoch Callaway
 Robert Chapman
 Jerome Cohen
 Raymond Cooper
 S. Hutch Curry
 Luder Deecke
 Jacques Delaunoy
 Emanuel Donchin
 Maurice Dongier
 Bernardo Dubrovsky
 Robert Dyer
 Bernard Fenelon
 Judith Ford
 Walter Freeman
 David Friedman
 Joachim Fuster
 Lauren Gerbrandt
 William Goff
 Netta Grandstaff
 Lester Grant
 Elisabeth Groll-Knapp
 M. Russell Harter
 Paule Hazemann
 Steven Hillyard
 Timo Jarvilehto
 E. Roy John
 {Catherine Kalil
 Rathe Karrer
 John Knott
 Gilbert Lelord
 Nicole Lesevre
 Benjamin Libet
Morris Lipton
Alberto Loizzo
Norman Loveless
W. Cheyne McCallum
Thaddeus Marczynski
Gail Marsh
Dennis Molfese
Risto Naatanen
Paul  Naitoh
Helen Neville
David Otto
Demetrios Papakostopoulus
 Southampton, England
 West Haven, Connecticut, U.S.A.
 Stanford, California, U.S.A.
 Los Angeles, California, U.S.A.
 Chicago, Illinois, U.S.A.
 San Francisco, California, U.S.A.
 Rochester, New York, U.S.A.
 Chicago, Illinois, U.S.A.
 Bristol, England
 Bristol, England
 Ulm, German Federal Republic
 Liege, Belgium
 Champaign, Illinois, U.S.A.
 Montreal, Quebec, Canada
 Montreal, Quebec, Canada
 Research Triangle Park, North Carolina, U.S.A.
 Short land, New South Wales, Australia
 Palo Alto, California, U.S.A.
 Berkeley, California, U.S.A.
 New York, New York, U.S.A.
 Los Angeles, California, U.S.A.
 Northridge, California, U.S.A.
 West Haven, Connecticut, U.S.A.
 Stanford, California, U.S.A.
 Chapel Hill, North Carolina, U.S.A.
 Vienna, Austria
 Greensboro, North Carolina, U.S.A.
 Dakar, Senegal
 La Jolla, California, U.S.A.
 Helsinki, Finland
 New York, New York, U.S.A.
 Madison, Wisconsin, U.S.A.
 Chicago, Illinois, U.S.A.
 Boston, Massachusetts, U.S.A.
 Tours, France
 Paris, France
 San Francisco, California, U.S.A.
 Chapel Hill, North Carolina, U.S.A.
 Rome, Italy
 Dundee, Scotland
 Briston, England
 Chicago, Illinois, U.S.A.
 Durham, North Carolina, U.S.A.
 Carbondale, Illinois, U.S.A.
 Helsinki, Finland
 San Diego, California, U.S.A.
 La Jolla, California, U.S.A.
Chapel Hill, North Carolina, U.S.A.
Bristol, England

-------
                                                XV11
Jon Peters
Terence Picton
James Pirch
Leslie Prichep
Charles Rebert
Lawrence Reiter
Walter Ritter
John Rohrbaugh
Steven Rosen
Walton Roth
Vernon Rowland
Daniel Ruchkin
Mikhail Rudnev
Russell Ruth
June Savignano-Bowman
David Scales
Anna Maria Seppalainen
Samuel Shelburne
James Skinner
George Somjen
Kenneth Squires
Nancy Squires
John Stamm
Samuel Sutton
Karl Syndulko
Joseph Tecce
Timothy Teyler
Robert Thatcher
John Thompson
Martine Timsit-Berthier
Meyer Timsit
Richard Traystman
Patricia Tueting
Charles Warren
Linda Warren
Hal Weinburg
Gerhard Winneke
Dorothy Woolley
Roberto Zappoli
Omaha, Nebraska, U.S.A.
Ottawa, Ontario, Canada
Lubbock, Texas, U.S.A.
New York, New York, U.S.A.
Menlo Park, California, U.S.A.
Research Triangle Park, North Carolina, U.S.A,
Riverdale, New York, U.S.A.
Los Angeles, California, U.S.A.
New York, New York, U.S.A.
Stanford, California, U.S.A.
Cleveland, Ohio, U.S.A.
Baltimore, Maryland, U.S.A.
Kiev, U.S.S.R.
Chicago, Illinois, U.S.A.
Boston, Massachusetts, U.S.A.
San Diego, California, U.S.A.
Helsinki, Finland
Washington, D.C., U.S.A.
Houston, Texas, U.S.A.
Durham, North Carolina, U.S.A.
Orange, California, U.S.A.
Orange, California, U.S. A.
Stony Brook, New York, U.S.A.
New York, New York, U.S.A.
Los Angeles, California, U.S.A.
Boston, Massachusetts, U.S.A.
Rootstown, Ohio, U.S.A.
New York, New York, U.S.A.
Newcastle Upon Tyne, England
Liege, Belgium
Liege, Belgium
Baltimore, Maryland, U.S.A.
Baltimore, Maryland, U.S.A.
Chicago, Illinois, U.S.A.
Birmingham, Alabama, U.S.A.
Burnaby, British Columbia, Canada
Dusseldorf, German Federal Republic
Davis, California, U.S.A.
Florence, Italy

-------
 ABBREVIATIONS
   The following standard abbreviations are used in this volume. Other abbreviations are defined in the text when
 first used. Note that numbers in EP component, stimulus, and international 10-20 electrode designations have
 not been subscripted in this volume.
 AEP      auditory evoked potential                   msec
 ARAS     ascending   reticular  activating  system     MRF
 BP        Bereitschaftspotential (synonymous with     Nl.N2.etc
           RP)                                       PI n, etc.
 BPM      beats per minute (heart rate)                 P300
 COHb     carboxyhemoglobin
 CNS      central nervous system
 CNV      contingent negative variation                 PINV
 CS        conditional stimulus                        PMP
 d.c.       direct current or, when  used in relation     PNS
           to amplifiers, 'directly coupled'               po
 ECoG     electrocorticogram                         PSP
 EDR      electrodermal (skin  potential) response     RCPV
 EKG      electrocardiogram  (also ECG)
 EMG      electromyogram                            rms
 EOG      electro-oculogram                           RP
 EP        evoked potential
 EPSP      excitatory postsynaptic potential             RT
 ERP      event-related potential                       SI ,S2, etc.
 FFT      fast Fourier transform                       sc
 GSR      galvanic skin response                       SCV
 HR        heart rate                                  SEP
 Hz        Hertz or cycles per second                   SP
 IPSP      inhibitory postsynaptic potential             SPL
 ip         intraperitoneal administration                TC
 ISI        interstimulus interval                        TLV
 ITI        intertrial interval                            US
MCV      motor nerve conduction velocity             VEP
MP        motor potential
 millisecond
 midbrain reticular formation
.first negative peak, second, etc.
 first positive peak, second, etc.
 positive  wave  with  peak  latency  ap-
 proximately  300  msec  after  stimulus
 onset (also P3)
 post imperative negative variation
 premotion positivity
 peripheral nervous system
 perioral (by mouth) administration
 postsynaptic potential
 reinforcement  contingent positive varia-
 tion
 root mean square
 readiness  potential  (synonymous  with
 BP)
 reaction time
 stimulus one, stimulus two, etc.
 subcutaneous administration
 sensory nerve conduction velocity
 somatosensory evoked potential
 slow potential
 sound pressure level
 time constant
 threshold limit value
 unconditioned stimulus
 visual evoked potential

-------
KEYNOTE ADDRESS

THE  SIGNIFICANCE OF  NEUROBEHAVIORAL  DATA
IN SETTING AIR  QUALITY STANDARDS
 R. BEARD

 Department of Family, Community, and Preventive Medicine, Stanford University,
 Medical Center, Stanford, CA, U.S.A.
   It is an honor to be invited to take part in this
distinguished Congress, and a special honor to be per-
mitted to make some remarks at the opening of the
sessions.  It was suggested that you might be inter-
ested  in  a very practical topic: the significance of
neurobehavioral data in setting air quality standards.
Having been involved in some way with such activities
for about fifteen years,  I  shall  offer  you some
observations.

   First it must  be recognized, that in the United
States, neurobehavioral data were significantly used
in setting  only two community air quality stand-
ards—those for carbon monoxide and for lead, and
the latter  only in  the state of California. The carbon
monoxide standard was set in California by the Air
Resources Board (1970) because the legislature had
ordered that air quality standards must be established
by a specific  date. The only evidence then available
on the effects  of carbon monoxide in concentrations
close  to  those observed in urban  centers were the
studies of McFarland's laboratory on the effects on
vision (McFarland et al.  1944; Halperin et al. 1947,
1950), Schulte's  work  (1963) on  performance of
psychological tests,  and  work from  our laboratory
(Beard and Wertheim 1967) on the discrimination of
short intervals  of time. These all showed effects when
5% or less of the body's hemoglobin had been oc-
cupied by carbon  monoxide. All the other data then
available  referred to industrial exposures at much
higher  levels, with no systematic study of behavioral
responses.

   It was  widely accepted that if the product of the
exposure  time in hours  multiplied  by the ambient
carbon monoxide concentration in parts per million
did not exceed the number 800 there was little prob-
ability of any perceptible effect. Applying this form-
ula as a standard  would  lead to an ambient level of
forty-four parts per million, assuming  that a stable
equilibrium could be reached in eighteen hours. The
California standard was established with 15 parts per
million averaged over 12  hours, and subsequently
changed to  10 parts per  million averaged  over 8
hours. Later a Federal Primary Air Quality Standard
for carbon  monoxide  (Environmental Protection
Agency 1971) was set at  10 mg/m3  or 9 parts per
million, based on very much the same information.

   The California lead standard was set about three
years ago, based on  evidence about the interference
of lead with hemoglobin  synthesis.  Recently, that
standard was reconsidered.  Reports of impairment in
intellectural and behavioral performance of children
were at this time taken into account. The stringent
standard of 1  1/2 jug/m3  was reaffirmed (California
Air Resources Board  1975,1976).
   In order to understand the outlook for the use of
 behavioral data in air quality standard setting, it will
 be  profitable to turn to the standards which have
 been established in industry  for the protection of
 workers' health. Such standards have been  in effect
 for  many decades.  There  are profound  differences
 between the United States and some other nations in
 this regard.  "In [our country] no serious threat to
 health is considered to exist as long as the level of
 exposure does not induce a. . .demonstrable disturb-
 ance of a kind and degree that is accepted  as an in-
 dication  of potential  sickness;  [in  the USSR]  a
 potential for ill-health is said to exist as soon as the
 organism undergoes the first detectable change of any
 kind from its normal state" (Hatch 1972). Eastern
 European  nations  generally follow  the  Russian
 pattern, while other nations generally follow the pat-
 tern used in the United  States,  with a great many
 differences in  the  interpretation of  data. Conse-
 quently,  international agreements on occupational air
 quality standards are  very few  in  number and the

-------
                                                     XX
 limits recognized among the nations may differ by a
 factor of ten.  The leading agency  for  setting such
 standards for the United States does not recognize
 that effects  demonstrated by evoked potentials are
 sufficient evidence on which  to base standards. On
 the other hand, Eastern European countries have used
 such information for many years.

    U.S.  industrial management,  government, and for
 the most part, labor are comfortable with the kind of
 standard-setting  for occupational health which  has
 become familiar. Efforts to bring new methods which
 might lead to more stringent controls will be resisted,
 especially if costly changes in materials, processes, or
 equipment do not promise immediate tangible  bene-
 fits for  workers.  It is not likely to be seen as sound
 policy to invest heavily in protection from a hypoth-
 etical injury which  might, if  prolonged  or  oft-
 repeated, produce only subtle  changes in the mental
 processes of a minority of workers.

    Air  quality standards for industry are generally
 based on two major assumptions: (1) that the  dura-
 tion and frequency of exposure can be controlled and
 (2) that individuals who are unusually susceptible can
 be  detected and transferred to other work. The guide-
 lines for community air quality standards are not so
 clear arid not so generally accepted. Obviously, they
 must be designed  for continuous exposure over many
 generations. But whom must they protect? It is easy
 to proclaim that no citizen should be made to suffer
 the ill-effects of man-made pollution. But does that
 mean protection  for every aged pulmonary cripple,
 gasping away a life  misspent in  the consumption of
 several tons  of  cigarettes, or the protection of the
 prematurely  born  infant  not  yet  ready for inde-
 pendent  life?

    Some years  ago in California, the decision was
 made to protect the most susceptible group of people
 which could be identified. This has led to considera-
 tion  of  the  effects of air pollution on infants, the
 aged,  those  with  chronic pulmonary disease,  and
 those with cardiovascular disease. It is recognized that
 it is necessary for people with extreme susceptibility
 to be protected  by the provision of filtered, purified
 air  on an individual basis.

   In the face of so much uncertainty about the cri-
 teria for air quality standards, what will be the place
 of neurobehavioral observations? To some degree this
 will depend on the extent to which they can be cor-
 related with the threat of disease, or with the impair-
ment of practical  functions which are recognized as
being important.  If  the alteration of the contingent
negative  variation  can be shown to be a reliable sign
of  a  process which leads to increased risk of auto-
 mobile accidents, it will carry a great deal of weight.
 If it is an isolated instance, it is apt to be neglected.
The most interesting case is  likely to be  the one in
                                             Beard

  which the presumed toxicant  at a small dose causes
  enhanced performance on some neurobehavioral task.
  This  phenomenon  is well-known to pharmacologists
  and psychophysiologists, and  has  been observed in
  our laboratory with a small dose of ozone in a test of
  visual   perception  (Grandstaff  and  Beard,  this
  volume).

    The  Environmental Protection Agency  as it  now
  stands is not likely to give much attention to neuro-
  behavioral  data. In setting a new standard, essential
  steps  are first, publication of the document which
  reports  the criteria on which the new standards will
  be based-that is, all the relevant data. The drafting of
  such documents has in the past been done by various
 groups under contract to the agency. The selection of
 contractor  will influence the content of the draft. At
 present,  only one  of the  leading figures in the EPA
 hierarchy  has much feeling  for  neurobehavioral
 science.  The draft, when prepared, will be open  to
 public criticism, but its final form will be decided by
 a sixteen member advisory committee, which as pres-
 ently  constituted  has only one member  who has
 shown an interest  in, and an understanding of this
 kind of work. Finally, the Administrator will act on
 the advice of his staff and still another advisory com-
 mittee to promulgate a regulation.

    As  I  read some of the reports prepared for  this
 meeting,  several areas for comment emerged. One was
 the frequency of seemingly discrepant observations
 and the  analysis of those  discrepancies. In most in-
 stances these could not be resolved  because the con-
 ditions of experimentation were not the same. There
 is a need for better  controls. That, of course, is what
 this meeting is all about,  and  this free and easy ex-
 change of  ideas  will  help to  improve the  com-
 parability of different studies.
   Another thing which emerged is the need for real
 interdisciplinary cooperation. The first rate psycholo-
 gist  cannot really get useful  help  from  a tyro  toxi-
 cologist, nor even from a first rate lexicologist, if he
 keeps him in a purely ancillary  role. I am concerned
 that there is developing an in-group which initiates its
 new  members  by making them learn a  special lan-
 guage. It is, of course, inevitable that these new ideas
 require new modes of expression, and  that there will
 be conventional phrases which will be understood by
 the  initiate  as  symbols  for  large constellations of
 ideas. But we  must beware of  mistaking new com-
 plexities of expression for new ideas. Ultimately, our
 ideas will have value  to our fellow men only to the
 extent that they are understood. The research which
 is  being discussed here this  week should be made
 known  much more widely. It should be publicly re-
 ported in journals which are seen by a wide range of
scientists. It  should be  reported in a language which
can be understood by specialists from other fields.

-------
Keynote Address
                                               xxi
   In the long run, the real contribution of neuro-
behavioral studies will be deepened understanding of
the processes of the nervous system and of the ways
they  can  be affected by external agents. We should
not be  impatient  for practical and applicable results
from this very new field of science. It is progressing
rapidly as shown by the reports prepared  for this
meeting. The insights and cumulative data base that
derive from meetings such  as EPIC IV provide the
foundation essential for future applications of neuro-
behavioral methods to environmental problems in-
cluding the setting of air quality standards.
References

BEARD,  R.R.  and  WERTHEIM, G. Behavioral im-
     pairment associated with small doses of carbon
     monoxide.  Amer. J. Pub.  Health,  1967,
     57:2012-2022.

CALIFORNIA AIR  RESOURCES BOARD. Ambient
     Air Quality Standards,  1970, Sacramento, Cali-
     fornia.

CALIFORNIA AIR  RESOURCES BOARD. Two-day
     hearing conducted on health effects of lead in
     air. CARS Bulletin, 1975, 6(10): 4-5.

CALIFORNIA AIR  RESOURCES BOARD. Standard
     for lead  in air is unchanged. CARB Bulletin,
     1976,7(2):3-4.
 ENVIRONMENTAL PROTECTION AGENCY. Na-
      tional Primary and Secondary Ambient Air
      Quality Standards. Fed. Reg., 30  Apr 1971,
      36(84):8186-8261.

 HALPERIN, M.N., NIVEN, J.I., McFARLAND, R.A.,
      and ROUGHTON, F.J.W. Variation in visual
      thresholds  during carbon monoxide  and hy-
      poxic anoxia. Fed. Proc., 1947,6:120-121.

HALPERIN,  M.H., McFARLAND, R.A., NIVEN, J.I.
     and ROUGHTON, F.J.W. The time course of
     the effects  of carbon monoxide  on visual
     thresholds. J. Physiol., 1959. 146:583-593.

HATCH, Theodore F. The role of permissible limits
     for hazardous airborne  substances in the work-
     ing environment in the prevention  of occupa-
     tional  disease.  Bull. World Health Org., 1972,
     47:151-159.

McFARLAND, R.A., ROUGHTON, F.J.W., HAL-
     PERIN, M.H. and NIVEN, J.I. The effects of
     carbon monoxide and altitude on visual thresh-
      olds. J. Aviat.Med., 1944,15:381-394.

SCHULTE,  H.  Effects of mild carbon monoxide in-
     toxication.  Arch.   Environ. Hlth, 1963,
     7:524-530.

U.S. PUBLIC HEALTH SERVICE.  Air Quality Cri-
     teria  for  Carbon Monoxide.  January 1969,
     Washington, D.C.

-------
I.  ELECTROGENESIS AND NEUROCHEMISTRY
   Section Editor:
   Charles S. Rebert
   Life Sciences Division
   SRI International
   Menlo Park, California, U.S.A.

-------
ELECTROGENESIS  OF  SLOW  POTENTIAL CHANGES IN
THE CENTRAL NERVOUS SYSTEM:  A  SUMMARY OF
ISSUES

C. S. REBERT

Life Sciences Division, SRI International, Menlo Park,CA, U. S. A.
Introduction

    This summary is based on preconference corre-
spondence concerning the electrogenesis and pharma-
cological substrate of slow potential changes in the
central  nervous system. The  following  individuals
contributed to this correspondence:

   1. E. Evarts, National Institutes of Mental Health,
     Bethesda, Maryland.

   2. J. Fuster, University of California, Los Angeles,
     California.

   3. B. Libet, University of California, San Francis-
     co, California.

   4. T. Marczynski, University of Illinois, Chicago,
     Illinois.

   5. V. Rowland, Western Reserve University, Cleve-
     land, Ohio.

   6. J. Skinner, Baylor College of Medicine, Hous-
     ton, Texas.

   7. G.  Somjen,  Duke University, Durham, North
     Carolina.

   8. J. Stamm, State University of New York .Stony
     Brook, New York.

    Discovery of  the contingent  negative variation
(CNV) by Grey Walter and his colleagues (1964) and
Bereitschaftspotential (BP) by Kornhuber and Deeke
(1965) gave significant impetus to the recent study of
SP changes in relation to overt behavior. These
findings were paralleled by discovery of a slow posi-
tive potential  (the P300 wave) and related compo-
nents  that reflect  cognitive  events (Sutton et al.
1965). The basic approach to this work may be char-
acterized  as  "electrophrenerrianology" (EPology)-
that is, the study of the localization of psychological
processes in the brain through analysis of electrical
"bumps" on the human  skull, with the  Skinnerian
assumption  that  the determination of the variables
that  influence  those "overt" events is sufficient to
their explanation. The assumptions inherent in this
approach may  lead to  fallacious conclusions in rela-
tion  to human  perception  and  subjective sensory
experiences unless there  is rigorous validation of
subjective  experience and knowledge  of the phys-
iological bases of the  potentials. Whereas the psy-
chological aspects  of  such  work are usually well
controlled and operationally  defined,  the  neglect
of intracerebral neurophysiology is evidenced by the
lack  of a single investigation of cortical neuronal
firing patterns associated with the CNV, and only one
relating the BP to single  neuron activity ( Gantchev
1974).  Furthermore, very little discussion at three
previous ERP Congresses was devoted to recordings
in animals, and  among those  papers (Borda 1970,
Donchin et al. 1971, Hablitz 1973, McSherry 1971,
Robert 1972), only McSherry directed attention to
causal mechanisms.

    Yet knowledge of the causal mechanisms of the
CNV and other event-related phenomena is critical to
their full understanding  and to their  proper inter-
pretation in relation to behavior and covert psycho-
logical  processes. It is possible, for example, to record
an identical scalp potential under  a variety of intra-
cerebral conditions—that  is,  when the specific neu-
ronal/glia elements involved differ (e.g., generation
of a  surface-negative potential by dendritic depolari-
zation  in  contrast to somatic hyperpolarization).
Whereas the surface potential in the two conditions
could be the  same, the physiological consequence
would  be opposite. Such details must be known to
adequately relate scalp potentials to molar events.

    There has, of course, been a long history of
interest in SP phenomena of the brain aside from the
CNV, BP, and P300-indeed, from the very inception

-------
                                                                                                 Rebert
 of cerebral electrophysiology (Caton 1875).  Because
 Caton  used slow recording  galvanometers that inte-
 grated  oscillatory potentials,  he could  not clearly
 distinguish real  SPs from rhythmic after potentials.
 Modern entry into this field was initiated by Gerard
 and Libet (1940) who were able to clearly distinguish
 SPs from oscillatory  activity.   Their work laid the
 early foundation for  delineating mechanisms of SP
 genesis and it continues  to  this day (Libet, this vol-
 ume).  From  that somewhat independent stream of
 research, a body of evidence has accumulated that is
 relevant to studies of the CNV. The specific neuro-
 physiological and neurochemical mechanisms elabo-
 rated by Libet and his colleagues in the sympathetic
 ganglion provide useful models for the study of CNV
 genesis centrally. Another general stream of research
 that  can  shed light on  the mechanisms of event-
 related SPs  are  studies  of intracerebral neurophysi-
 ology undertaken in behavioral contexts similar to
 those used in  the study of SPs (Fuster 1973). Given
 comparable  situations, interexperimental similarities
 between SPs and neuronal responses can be evaluated.
 There are also several investigations involving direct
 comparisons of SP and unit responses (Fromm and
 Bond   1964,  1967; Rowland 1974; Rebert  1969,
 1973b), but few involving the behaving organism.

    Because it is increasingly obvious that EPology is
 not an  adequate approach to the full understanding
 of event-related   SP  phenomena,  consideration  of
 electrogenesis has been included as part of this fourth
 congress on event-related SPs. The primary objective
 of this  review is  to delineate important  issues  with
 regard to mechanisms of  SP genesis in order to facili-
 tate their early resolution. Investigators from neuro-
 physiology  (Evarts,  Fuster,  Skinner),  neuropharma-
 cology  (Libet, Marczynski,  Somjen), and physiologi-
 cal psychology (Rebert,  Rowland, Stamm) engaged
 in  a  correspondence to  develop a  multidlciplinary
 view of the problem of SP genesis with the following
 objectives:

    1.  Delineate critical issues concerning SP genesis,
       the unknowns at this point in time.

    2. Describe the state of the data to determine
       what needs to be  done experimentally, theo-
       retically,  and  technically  to  resolve critical
       issues.

    3. Determine in a broad sense where  SP electro-
       genesis  research is headed: What  trends are
       occurring, and whether the research is leading
       to any general principles of brain function.

    Three  general categories of  critical  issues were
raised:

    1. Typology  of Sft-Several types of SPs can be
      identified  and defined from different frames
        of reference. From a pheno'menological view,
        SPs have been defined  as "motor readiness"
        events,  "expectancy waves,"  and "postrein-
        forcement  potentials."  In contrast, SPs have
        also been  defined  in terms of their temporal
        characteristics, for example, "phasic"  (CNV-
        like)  versus  "tonic" long-lasting (postrein-
        forcement) SPs.

     2. Experimental /ssHes-These issues include spe-
        cific questions such as the relationship  of SPs
        to neuronal activity, the neurochemical medi-
        ators of SPs, and the role of potassium and
        glia in SP genesis.

     3. Technical Issues—These issues relate to techni-
        cal limitations that preclude or impede  ad-
        vances and interpretations in specific research
        strategies.

 These issues are reviewed below.

 Review of issues

 Typology of SPs

     Several ways  of categorizing or defining SPs are
 discussed here.  This list is probably not exhaustive,
 and the categories are not mutually exclusive. SPs can
 probably be best differentiated  on the basis of multi-
 ple categories.

     Temporal classification: There are relatively con-
 stant potentials between almost any two regions  of
 the brain or between the brain and extracerebral sites.
 The potential of  the  mammalian cortical surface is
 approximately  15 mV positive with respect to corti-
 cal layers V-VI (Aladjalova 1964), and the surface is
 1  to 3  mV  negative  with respect to frontal bone
 (Fromm  and  Bond 1964). These potentials,  while
 relatively constant, may fluctuate  with very long
 periods (Aladjalova 1964) or with the  sleep-wakeful-
 ness  cycle  (Caspers  1963; Wurtz  1965a,b). The
 "resting" potential exists in contrast  to "reactive"
 SPs that are of relatively short duration (1 to 20 sec)
 and  that are  usually evoked by transient sensory  or
 psychological events (e.g.,  CNV). SPs may, then, be
 either oscillatory or nonoscillatory, and the latter can
 be further subdivided  on the basis of their duration
 into two general classes-(l) tonic (resting) potentials
 with durations of minutes or hours and (2)  phasic
 (reactive) potentials with durations of seconds. This
 latter dichotomy might also parallel SP types that do
 and  do  not  correlate with  massed  unit activity
(Sheafor  and  Rowland  1974) and phasic CNV-like
 SPs observed in many parts of the monkey's brain, in
contrast to cumulative postreinforcement potentials
 observed  only in the cortex (Steinmetz  and Rebert
 1973).

-------
 Summary of Electrogenesis Issues
    Functional classification:  The  functional signifi-
cance  of SPs  may  be considered at two  different
levels,  i,e., with respect to neuronal activity or molar
behavioral/psychological events. Negative and positive
postsynaptic potentials (PSPs) measured at the outer
neuronal membrane  surface with microelectrodes are
associated  with excitation and inhibition of the neu-
rons,  respectively, but  the relationship of slow field
potentials  (measured with macroelectrodes) to neu-
ronal  firing  is not clear.  Given clarification of that
relationship, SPs  might be defined in terms of their
excitatory or inhibitory consequences.

    A variety of factors related to sensory input,
behavior,  and covert  psychological  processes  may
elicit  SPs, which  can be differentiated into (1) those
elicited unconditionally by external  events such as
sensory stimulation and reinforcement; and  (2) those
elicited conditionally during situations like  the cued
RT task. Other classes of SPs include  those  related
(1) to "spontaneous"  alterations of arousal, atten-
tion,  and sleep-wakefulness; and (2) endogenous or
induced tissue abnormality such as the SPs preceding
seizures  or accompanying cortical spreading depres-
sion or surgical injury (Irwin et al. 1975).

    Other functional classifications  have been sug-
gested based on phenomenological interpretations of
the most salient psychological process presumed to be
represented by the  SP. For example,  the CNV has
been  related to  cognition, expectancy, motivation,
waiting,  and attention.   Until the functional signi-
ficance of  SPs  is unequivocally  established,  such
presumptive labels should  be avoided.

    Mechanistic classification:  Three general classes
of SPs  are delineated here  based on (1) neuroanatomi-
cal arrangements, (2) membrane and neurochemical
mechanisms in neurons, and (3) extraneuronal mecha-
nisms.

    SPs may be classified with  respect to both par-
ticular and general neuronal arrangements. The orga-
nization of neurons varies from locale to locale in the
brain.  Some  regions like  the cerebral cortex, hippo-
campus,  and cerebellar cortex are characterized by
systematically oriented neurons that develop electric
dipoles directly related to the physical organization.
In contrast, regions like the reticular formation and
caudate nucleus do not have such well-defined neural
arrangements, yet  they  also  exhibit SP responses
(Rebert 1972). The SPs generated in these two types
of tissue might be referred to as "dipole" and "reticu-
lar" types. A difficult  question is  to determine the
mechanisms of SP genesis  in the latter type of tissue.
Since the neurons are not  systematically oriented, the
net field  produced by neuronal membranes might be
zero, the separate fields cancelling one another. But,
as Ubet asked, if a field-recorded SPis generated by a
  type of glial cell, would there  not have to be a suf-
  ficient gradient in SP somewhere along the cell axis to
  develop the external field? What is the evidence for
  such axial gradients in glial cells? In addition, would
  it not be necessary for glial processes to have some
  systematic  arrangement  to preclude  cancelling of
  their  fields? What  alternatives can  be suggested?
  Somjen (this volume) provides  a model of how glia
  might be involved in SP genesis.
      The general intracerebral system involved in the
  production of surface  potentials may  also differen-
  tiate apparently equivalent  surface potentials.  This
  is reflected by data of Skinner and Yingling (1976)
  showing that frontocortical  SPs evoked by  either a
  novel stimulus or a neutral reinforced stimulus appear
  to be the same in that the identical physical stimulus
  evokes them and  their  anatomical distributions are
  the same. However, the subcortical correlates  of those
  two seemingly identical SP events are different. The
  response in n. reticularis thalami to a novel stimulus
  does not require the integrity of the inferior thalamic
  peduncle  (a  bidirectional pathway  interconnecting
  the medial thalamus and frontal cortex), whereas the
  conditioned  response recorded in this structure does
  require intact connections with the frontal cortex.
  Thus,  knowledge of the subcortical correlates of the
  frontal potentials  separates  them into distinct  SP
  events, even  though  the actual local response para-
  meters are the same and are presumably mediated by
  the  same  local  generators. The latter presumption
  requires further  examination because different local
  neural aggregates could be generating morphologically
  similar waveforms.
      Many neurons are "mosiac" in the sense of being
 sensitive  to  a  variety of transmitter substances.  In
 addition, the effect of synaptic events depends upon
 the  influence of nontransmitter chemicals that are
 both endogenous and exogenous to neurons. Chemi-
 cal  "modulators"  are  endogenous substances  that
 function in a homeostatic way to regulate the rate of
' depolarization  or hyperpolarization  initiated  by a
 synaptic  transmitter. Similarly,  blood-borne  sub-
 stances exogenous  to neurons can facilitate (medi-
 ator) or attenuate (moderator) the rate of synaptic
 transmission  or axonal conduction, or act to  alter
 transmitter release or change the  action of a modula-
 tor (Myers 1974).  As indicated  by Libet (this vol-
 ume),  slow SPs are associated with such actions and
 can be classified on the basis of their neurochemical
 substrate.

     A variety   of extraneuronal  factors  are  also
 known to produce  SPs.  They  include  potentials
 generated by  glial membranes, vascular flow, differen-
 tial pH, C02, blood-brain barrier, and perhaps others.

-------
                                                                                                 Rebert
    Tonic  and  phasic  SPs constitute   two  major
 classes that can be  differentiated on a number of
 definitional  factors.  Tonic SPs are contrasted with
 phasic ones, by  definition, in terms  of  temporal
 characteristics.  In addition, phasic SPs appear to be
 closely related to transmitter release, neuronal organi-
 zation (dipoles), and waking psychological/behavioral
 events, while tonic SPs appear to vary with metabolic
 factors, injury, and the blood-brain barrier. Tonic SPs
 may  not  differ from  one intracerebral locale  to
 another, may not relate to neuronal organization, and
 are, perhaps, associated with modulator  and modera-
 tor/mediator actions on neurons.

     We are left with the general question  as to the
 best way(s)  of classifying  SPs. Somjen emphasized
 the  need to  avoid the  kind of confusion that has
 arisen around the  term alpha  rhythm,  customarily
 distinguished from EEC spindles, except by Andersen
 and Andersson (1968) who regard the two  as identi-
 cal because of a suspected common generating mech-
 anism. He suggests that  SP categorization should be
 based  either  on phenomenology or electrogenesis, but
 not  on both, although,  as suggested above, a multi-
 dimensional scheme may be  most appropriate.

 Experimental issues

     Specific experimental issues related to electro-
 genesis may  be summarized in terms of seven major
 considerations:

     1. Spatial distribution   of  SPs  in  the central
       nervous system (CNS).

     2. Neurochemical factors related to SP genesis.

     3. Relations between neural activity /excitability
       and SPs.

    4. The  role of potassium  and  glia  and other
       "nonresponsive" cells in SP genesis.

     5. Intracerebral  relationships inferred from SP
       responses.

    6. The interaction of SP types.

    7. The general question of how knowledge of
       SP  electrogenesis  might improve  under-
       standing of the behavioral/psychological rele-
       vance of SPs.

    Spatial distribution: Given the variety of factors
that  contribute  to SP genesis,  it  is  likely that SPs
can be recorded from any region of the  CNS under
one  condition or another.  The  distinction between
resting and reactive SPs is important in this context,
  as  resting SPs—like those associated with the sleep-i
  wakefulness cycle—may appear with similar polarity
  and time course throughout the brain (Wurtz 1966),
  whereas reactive SPs, e.g., those elicited in the cued
  RT task, exhibit different polarities, waveforms and
  amplitudes in different regions of the brain (Rebert
  1972, McCallum et al. 1973).  The question of distri-
  bution in the brain is, therefore, most pertinent  to
  the latter type of  SP. Event-related phasic SPs have
  been observed in several nuclei of the thalamus, hypo-
  thalamus, basal ganglia, nonspecific reticular nuclei,
  and limbic structures (amygdala, spetum, cingulate
  gyms) as well as over a large extent of the cortical
  surface (Haider et  al.  1968, Irwin and Rebert 1970,
  Rebert and Irwin 1969,McCallum et al. 1973, Rebert
  1972, Hayward et  al.  1966, Vastola 1955, Rowland
  1968; Donchin  et al.  1971, Borda 1970, Hablitz
  1973).  The significance of the scalp distribution  of
  SPs in humans is the subject of a separate section  in
  this volume.

      Neurochemistry: Two related questions can J>e
  asked with respect to the  actions of chemicals in the
  genesis  of SPs. First, which chemicals acting on neu-
  rons or glia cells produce  or modify slow membrane
  potential  changes?  Second, can SP responses in par-
  ticular regions of the brain be  related to internuclear
  pathways defined by their neurochemical properties?
  Some evidence is available with respect to the fust
  question (see Marczynski and Libet, this volume). On
  logical grounds, the second can probably be answered
  in the affirmative, although direct evidence is lacking.
  It is probably true, for example, that positive SPs in
  the caudate nucleus depend to some extent on  the
 dopamine pathway  that originates in the substantia
 nigra.

     There  is  some  question  concerning the  role
 of  transmitter substances  in  SP  genesis since,  as
 Marczynski (this volume) points out, the term implies
 a brisk and quickly reversible effect on neurons. The
 release of acetylcholine (ACh) in the cortex in associ-
 ation with surface  SPs accompanying wakefulness
 suggests a possible transmitter role in SP production.
 On the other hand, ACh has muscarinic and nicotinic
 actions, the former perhaps not representing a typical
 (brisk) transmitter phenomenon, but a modulator ac-
 tion (Libet, this volume). An additional confounding
 observation is  that ACh may also affect glial mem-
 branes,  producing slow, long-lasting  depolarization.
 Whether this is a direct effect on glia or a secondary
 consequence of potassium released from  muscarini-
 cally activated  neurons  is not certain."This consider-
 ation is important since the sensitivity of glial cells to'
 ACh could, theoretically, result in large SPs without
 any obvious correlation  with neuronal activity as ob-
served by Sheafor  and Rowland (1974). Marczynski
marshals evidence from a variety of sources suggesting

-------
Summary of Electrogenesis Issues
that cortical  surface negative  SPs ate mediated by
cholinergic  mechanisms.    Since  the  cholinergic
component  of  the  ascending  reticular  activating
system  (ARAS) exerts a  strong  facilitating  effect
on  thalamic relay neurons, there is reason to conjec-
ture that  thalamic SPs observed by McCallum et al.
(1973)  and Robert  (1972) may  also be choliner-
gically mediated.

    Paradoxically, antimuscarinic drugs such as atro-
pine and  scopolamine, known  to block negative SPs
and the desynchronization induced by ARAS stimu-
lation, in the same dose range also block the occur-
rence  of alpha-type synchronization in men and cats
(post-reinforcement synchronization of 7  to 9 c/sec)
associated with epicortical positive SPs (reward con-
tingent positive variation or RCPV).  Does this mean
there are pharmacologically identical but functionally
different, and even opposed, cholinergic mechanisms?
After systemic administration of scopolamine, the dc
potential  of  the  cat's cortex (posterior  marginal
gyri) remains stable, i.e.,  the SPs show little  fluctu-
ation despite maintenance  of  bar pressing perform-
ance.   Furthermore, subsequent administration of
physostigmine to increase the tone of the cholinergic
systems restores both  the  negative SPs during unre-
warded bar presses and the RCPV during consump-
tion (Marczynski 1971  and this  volume).

     Dopaminergic and cholinergic influences in the
production of slow excitatory and inhibitory post-
synaptic  slow potentials (s-EPSPs and s-IPSPs) are
reviewed  by Libet (this volume). Sympathetic gang-
lion cells are capable of responding  to orthodromic
(preganglionic) volleys with a  slow IPSP and a slow
EPSP,  the  s-IPSP  in  response  to  the  transmitter
dopamine and  the s-EPSP in response  to  acetyl-
choline,  acting muscarinically (Libet  1970). Both
slow PSPs  have  synaptic  delays in the  tens and
hundreds of msecs and durations of many seconds.
Both are  generated without any detectable change in
membrane resistance by electrogenic mechanisms that
may depend  on active ion transport. Generation of
these  prolonged responses without the necessity of
ionic leakage, in contrast  to that of the "classical" or
 "fast"  PSPs,  allows their achievement without the
excessive energy requirements for restoration of ionic
concentrations.  If such monosynaptic s-PSPs provide
possible models for some neuronally generated SPs, it
would  be profitable to study  their incidence in the
 CNS in relation to SP responses.

     An additional  function of dopamine consists of
 a true modulating action on  neuronal reactivity to
 ACh (Libet  and  Tosaka  1969,  1970;  Libet et  al.
 1975). Brief exposure to dopamine enhances  slow
 muscarinic response to ACh (s-EPSP responses) up to
 several hours or longer. This modulatory function has
 important implications for behavioral, motivational,
and memory processes.  It could provide a basis for
long-lasting  alterations  in  SP  responses,  especially
when such alterations are concomitants of behavioral
modifications.

     The neurochemistry of the CNV is an issue of
specific concern.  Thompson et al. (this volume) pro-
vide  preliminary evidence and a brief review of phar-
macological effects on the CNV in man.

     Relation  of SPs  to neural  excitability:  Slow
potentials at the cortical surface may reflect a variety
of events  occurring  in  cortical neurons,  including
somatic and dendritic depolarization or hyperpolar-
ization in any combination. Thus, surface SPs could
occur in the absence of neuronal firing,  reflecting
only a change in the dynamic equilibrium of den-
dritic  depolarization  and somatic hyperpolarization,
perhaps resulting in a more reactive state (McSherry
and  Borda 1973). In contrast, extracellular slow field
potential changes may  accompany changes in  the
average  firing  rate of neurons. Study  of these two
questions requires substantially different experimen-
tal approaches. The first necessity is to determine the
relationship  between SP  responses and  neuronal
firing,  and  in  the  absence  of changes in firing, to
determine if SP responses reflect changes in neuron
thresholds.

     There are two  types  of evidence  available  for
assessing the relationships  of  SP and neuronal activ-
ity:   (1)  indirect,  inferential data  based  on inter-
experimental  comparisons  and (2) direct correlation
of units and SPs. Indirect evidence regarding cortical
SPs  comes  from SP  and unit  behavior observed by
different experimenters using  similar  experimental
paradigms.  Fuster (1973),  Fuster and  Alexander
(1971), Stamm (1964), and Stamm and Rosen (1972)
have studied  units and  SPs in  monkeys performing
delayed response  tasks. The  results  suggest  that
surface-negative SPs are  associated, in some portions
of the delay  period, with increased neuronal dis-
charge. Another comparison can be made between
studies  of  promotion  behavior  of cortical  units
(Evarts  1966,  1973) and  studies of slow  readiness
potentials (Kornhuber and Deeke 1965). The occur-
rence of increased cellular firing in the cortex in asso-
ciation with  surface-negative  SPs implies  that  SPs
are due, at least in part, to dendritic depolarization,
since  a preponderance of  somatic depolarization
would be reflected by a surface positive SP.

     Data from direct comparisons of SPs and single
units in immobile preparations are inconsistent (Li and
Salmoiraghi  1963) and has rarely been obtained in
behavioral contexts. However, Gantchev (1974) con-
currently recorded single unit  activity  and  readiness
potentials in monkeys and found a preponderance of
units increasing firing during surface-negativity, with

-------
  8
                                           Rebert
  a  mixture of no  responding and decreased cortical
  cell  firing in association  with surface-positive SPs.
  Marczynski  (this   volume)  found  surface-negative
  components  of postreinforcement  synchronization
  in phase  with cortical  cell firing.  A  small sample
  study  of  ten neurons suggested  that marked reduc-
  tion of firing accompanied  postreinforcement posi-
  tive  SPs.  Yingling  and Skinner (1975) and  Skinner
  and  Yingling (1976)  observed that single units  in
  n.  reticularis  thalami  increased firing in  association
  with negative SPs  and decreased firing with  positive
  shifts.

     Measurements  of massed  unit (MU) activity  in
  the cortex and subcortical regions give similar results.
  Rebert (1969, 1973b) showed that negative and posi-
  tive SPs evoked by light flash in the lateral geniculate
  accompanied  increased and  decreased  MU activity,
  respectively.  Sheaffor and Rowland (1974) found
  that reactive cortical surface-negative SPs  accom-
  panied increased  cellular discharge.  This  result was
  complicated by additional findings that long-duration
  "expectancy" SPs  were not  associated  with  system-
  atic changes  in MU activity. The technical consider-
  ation of the sensitivity of MU measures to small unit
  discharge  was raised.  It appears that, despite differ-
  ences in neuronal morphology and specific generating
  mechanisms,  the polarity of surface SPs and those  in
  subcortical regions reflect the state of neuron  activity
  in a similar  way,  i.e., negativity is  associated with
  increased  firing. However,  either  relationship is pos-
  sible and  at  this time the  determination  of  SP/unit
  relations  requires substantiation  in each  case.  Ulti-
  mately, repeated demonstration of the same relation-
  ship in a  variety of experimental settings  and nuclei
 might  allow a firm generalization about SP polarity
 and unit firing to be made.

     Role  of  "nonresponsive" cells and potassium in
 SP genesis: Neural  membrane  shifts may occur via
 regulation of the extracellular ionic concentration of
 K+, a mechanism that Ranson and Goldring (1973a,
 b,c) have linked with glial cell transactions. The elec-
 trogenesis  of  extracellular SPs could derive, in part,
 from glial membrane  shifts,  or SP generators could
 be  associated  with other biophysical processes  of neu-
 rons  and  glia  that  have  not as yet  been  explored.
 Rosen thai  and Somjen (1973), for example, showed
 a correlation between cellular metabolic activities and
 local SP shifts. Metabolic end products could, in turn,
 affect cerebrovascular  flow, a phenomenon that has
 also been  associated with  SP shifts (Woody et al.
 1970).  Somjen emphasizes  the  importance of recog-
nizing the  source of potassium when it  accumulates
in extracellular fluid. Sources include terminal axon
arborizations;  cell bodies,  dendrites,  and axons of
spiking  neurons; cell bodies and dentrites exhibiting
nonspike synaptic currents; and small  (short axon)
 cells. Determination of the precise source is difficult
 due to the complicated architecture of tissue.

     The role  of neuroglia as a generator source of
 sustained potential shifts  is extensively reviewed by
 Somjen (this volume). He suggests that the theory of
 glial generation is compatible with the theory ofcho-
 linergjc mediation of event-related SPs (Marczynski,
 this volume). It is conceivable that, in normal cortex
 under  physiologic  conditions,  the  most  abundant
 sources  of  potassium are activated  by  cholinergic
 input.  The idea of direct ACh action on  the glial cell
 membrane is  less  attractive because it  implies  the
 extrasynaptic diffusion of ACh.

     Whereas glia have received considerable attention
 as possible SP sources,  Rowland points out that not
 all "silent" cells  of the cortex are  glia.  He suggests
 that attention be directed to a possible role  of such
 cells in the production  of SPs.  Skinner suggests too
 that calcium has not been adequately studied in terms
 of its role in the production of SPs.

     Intracerebral relations: In the cued RT task, SP
 polarity  appears to  differentiate mesencephalic and
 thalamic nonspecific arousal  mechanisms from basal
 ganglia/limbic structures. The former exhibit negative
 and the latter exhibit positive SPs (Rebert 1972). It
 has been suggested (Rebert 1977) that these SP para-
 meters reflect  a  reciprocal interactive state between
 the  two  arousal systems  described  by Routtenberg
 (1968).

     Observing SP changes in  one region of the brain
 in response to manipulations of other regions appears
 to be a useful method of studying intracerebral rela-
 tionships (O'Leary and Goldring  1964, Arduini 1958,
 Skinner and Yingling 1976). However, caution needs
 to be exercised in generalizing from experimentally
 derived relationships under one set  of conditions to
 relationships in  other circumstances.  For example,
 evidence  suggests  that activity in mesencephalic reti-
 cular regions has a primary role in the production of
 surface-negative SPs (Arduini  1958), but trial-by-trial
 analyses of SPs in nonspecific nuclei and  on the cor-
 tical surface in the cued RT task indicate that the two
 regions were not coupled in terms of SP  covariation
 (Rebert 1977).  In contrast, the  latter analyses sug-
 gested that the caudate  nucleus  and  amygdala were
 coupled during the foreperiod of RT. Therefore, one
 cannot  conclude from the demonstration of anatomi-
 cal connections and  functional interaction between
 nuclei in one situation that the nuclei are functionally
 coupled in other situations.

    The foregoing indicates the necessity for deter-
mining  intracerebral dynamics related to particular
behavioral/psychological   states.   Analysis  of  SP

-------
 Summary of Electrogenesis Issues
fluctuations appears to be one way to achieve this ob-
jective.  Skinner's work  (this volume) combining SP
measures and selective intracerebral blocking has also
been fruitful in defining functional paths related to
specific  behaviors. He suggests an important role of
tnalamocortical  projections, especially  the  inferior
thalamic peduncle, in frontal SP genesis. In contrast,
Zappoli  et  al. (this volume)  argue that CNVs  in
humans occur  despite extensive prefrontal isolation
and disruption of thalamocortical projections.  It has
been alternatively suggested that the ventral motor
thalamus is  involved in  CNV  genesis (Haider  et at.
1968). Also since the caudate nucleus exerts an inhib-
itory  influence  on the  cortex, reflected  in surface
positive SPs (Buchwald et al. 1967), and positive SPs
occur in  the caudate in  association  with  cortical
CNVs (Rebert 1972, McCallum et al. 1973), it is con-
ceivable that the CNV could be due, in part, to re-
lease  from  caudate  inhibition.  In  addition, Fuster
(this volume) has proposed that prefrontal unit activ-
ity, and correlatively SPs, are modulated in part by a
visual transcortical pathway involving inferotemporal-
prefrontal connections.  These  findings represent the
humble beginning of an understanding of how dif-
ferent brain regions interact to generate SP responses.
These  results  also indicate the  usefulness of em-
ploying SP  responses to assess intracerebral relation-
ships.

    Interaction of SP types:  The determination of
how different SP types  interact is an essentially un-
tapped  area of inquiry.  Some specific questions and
possible  ways  of studying interactions include the
following. How might slow glial  membrane depolar-
izations  affect neural membranes?  Could glial cells
regulate  the neural equilibrium potential of K+ and,
therefore, the membrane resting  potential?  Can glia
be selectively blocked (e.g., by application of mor-
phine to the cortical surface-Roitbak 1969) to deter-
mine their contribution to SPs?

    Tonic and phasic SPs and their interaction can
be observed concurrently in some  behavioral con-
texts. Rebert et al. (1976) observed  dc level changes
on the  human scalp preceding CNV trials in  associ-
ation with forearm muscle tension induced by having
subjects lift weights with a hand grip. There was  a
direct relationship between the amount of weight
pulled and  the tonic dc level up to a point (30 Ib),
after which additional  tension produced no further
SP shift. However, CNV amplitude showed no incre-
ment between0 and 15 Ib, but increased S/uV between
 15 and 30 Ib. It remained at the larger value at 45 Ib.
While  the pretrial tonic dc shift reached asymptote
at 30 Ib, an enhanced CNV was superimposed on it,
suggesting that the two potentials were independently
generated. Similarly, in monkeys there was no appar-
ent interaction between CNVs and a tonic postinges-
 tion  frontal negativity  observed by Steinmetz and
Rebert (1973). However, when the cortical  dc level
shifted several mV negative  for an unknown reason
in one instance, CNVs were absent. These findings are
relevant to the hypothesis that CNV reduction under
stress in highly anxious subjects is due to the presence
of a physiological ceiling for SP genesis (Knott and
Irwin 1968, 1973). Although a ceiling appeared to be
produced by muscle tension (which is perhaps analo-
gous to excessive arousal produced by anxiety), CNVs
were still apparent. Thus the data did not support the
type  of interaction  between  tonic  and  phasic  SPs
predicted by the ceiling hypothesis. The monkey data
were partially supportive, but only in the context of
extremely large background shifts, which should be
readily detected on the  human scalp. Pirch (this sec-
tion) presents further evidence concerning the rela-
tionship  of tonic and phasic negative shifts in rats,
suggesting that the ceiling hypothesis is valid.
     Relevance of SP genesis to behavioral interpre-
 tations: All the issues discussed above are pertinent
 to the attempt to relate SP phenomena to behavioral/
 psychological  processes. The location of SP phenom-
 ena in the brain provides information about particular
 nuclei and general systems involved in specific behav-
 iors, while SP changes produced by anatomical and
 neurochemical manipulations can  reveal  functional
 pathways that couple nuclei. SP and unit recordings
 may  reveal excitatory and  inhibitory patterns in
 general intracerebral systems that mediate particular
 behaviors.  The delineation of ionic mechanisms in SP
 genesis would provide additional detailed information
 about brain-behavior relationships.  With  respect to
 SPs, however, efforts to define systems are at a primi-
 tive  level. With few exceptions (Skinner, this volume),
 SP measures are not currently being used as a tool to
 delineate general cerebral  systems related to the cued
 RT  task  or conditioning paradigms or  to  specify
 dynamic intracerebral interactions. In other contexts,
 however,  this  kind  of  systems  approach is making
 significant strides (John et al. 1973).

     It is  worth reiterating that  a given behavioral/
 psychological  state is  a dynamic  affair in terms of
 configurations of brain states, and that the behavioral
 significance of a given electrophysiological event such
 as the CNV,  no doubt,  depends  upon the simulta-
 neous state of SPs (and other events) in many brain
 regions. The dynamic and  multivariate nature of brain
 activity is, perhaps, related to  the difficulty in pre-
 dicting even  simple behaviors from physiological
 events like the CNV. Reaction time, for example,
 correlates  weakly  with a  wide variety of biological
 indicators  (heart rate  deceleration,  CNV  amplitude,
 alpha  abundance,  alpha  phase, alpha asymmetry,
 muscle tension). It would seem useful to apply multi-
 variate  techniques (cf.  Donchin,  this volume) to
 determine  what combination of biological indicators

-------
 10
                                                                                                Rebert
 taken together best accounts for RT variability. In
 this case, SP  responses recorded from  an array of
 intracerebral  electrodes  would  contribute substan-
 tially  to our understanding of  the general psycho-
 physiological  system  that  underlies  reaction time
 behavior.

 Technical issues

     Rowland raised the question  whether MU re-
 cordings  are sensitive enough to reflect small  unit
 changes  associated with SPs. This is  important for
 interpreting  apparent dissociations between SP  and
 unit activity.  He also raised the issue of SPs being
 confounded with impedance shifts. Study of the rela-
 tionship  of this  parameter  to other electrical events
 is badly needed.

     Another question of interest is the  adequacy of
 sytemic administration of pharmacological  agents in
 the study of SPs. For example, if a shift is produced
 or altered following a systemic  dose of atropine to
 block cholinergic synapses, what can  be concluded?
 The nucleus in which recordings are  being  made
 might be reacting  to  a local change in adrenergjc
 synapses consequent  to cholinergic  alteration  of a
 nucleus  which sends efferents to the  recorded area.
 For example,  cholinoceptive cells in pars compacta
 of the substantia nigra  project  dopaminergic fibers
 to the caudate nucleus.  Systemic administration of
 an anticholinergic  agent could  block the nigra and
 indirectly cause  a change in the caudate,  a change
 that would, in all likelihood, be erroneously attri-
 buted to direct cholinergic mediation in  the caudate.
 Thus, the use of  intracerebral perfusion  of  neuro-
 chemicals (Myers  1974) in conjunction with electro-
 physiological recordings is another valuable combina-
 tion too little used in SP research.

     The  growth of glia around electrode tips seems
 to impair the quality of dc  recordings (Skinner,
 personal  communication), and  electrode impedance
 changes  also appear to do odd things to Emde on-line
 calibrators. In  the absence of changes in event-related
 potentials, on-line  calibration  signals have been ob-
 served  to either slowly appear or  disappear  over
 several weeks of recording from a particular electrode
"in a particular monkey, while not changing in other
 monkeys (Rebert, unpublished observations).

     Ldbet expressed the need  for  improved intra-
 cellular recording techniques that would  permit rapid
 impaling  of  cells combined with the capability  to
 hold the cell for long periods without significant leak-
 age across the membrane. Without better techniques,
 the  critical questions  of neuronal  PSPs and excita-
 bility in relation to various SPs and behavioral func-
 tions  will probably remain  unresolved.  A  recent
 development of a "sharpened" glass micropipette by
 Brown  and Flaming (1975) offers some hope along
 this line.  Chronic implantation of subdural and  sub-
 cortical nonpolarizing dc recording electrodes, which
 are not toxic to neural and glial tissues, is also neces-
 sary for the longitudinal study of behavior in relation
 to SPs.  Rowland (1961) and Rebert and Irwin (1973)
 have described the construction of nonpolarizing elec-
 trodes suitable for this purpose.

     Libet expressed the need for geometric models of
 different types of neural and glial configurations from
 which slow potential parameters can be  predicted.
 Volume conduction models have been  proposed for
 auditory (Vaughan and Ritter 1970), visual (Vaughan
 1974),  and somatosensory EP components (Goff et
 al., this volume). Models  relating field  potentials to
 neural architecture have been worked out in greater
 detail for the  olfactory  bulb  (Rail and  Shepherd
 1968; Freeman 1975,1976).

 Conclusions

     This  summary has  only partially addressed the
 goals set  forth  in the  introduction.  The ultimate
 objectives  of this  area of research have  not been
 clearly  defined  nor have  the practical applications
 been adequately examined, although there are  cer-
 tainly pursuits in the applied direction, e.g., studies
 related  to epileptogenic mechanisms  (Prince et al.
 1973).  A pertinent question here is the  extent to
 which obtaining "molecular" information about SPs—
 especially the CNV, which may have diagnostic uses-
 increases  the  power of their  medical application.
 Would the CNV be more  useful if its exact electro-
 genesis and pharmacological genesis were known?  Per-
 haps, then, abnormalities in various CNV parameters
 would indicate specific intracerebral abnormalities. In
 this  same vein, changes in  the dynamic intracerebral
 patterns of SP responses might be sensitive  indicators
 of brain alteration by low levels  of environmental
 toxicants; and,  if so, knowing the mechanisms of SP
 genesis  might help to  pinpoint the  mechanisms of
 action of toxic agents. An exciting possibility for
 future application is presented by Stamm et al. (this
volume),  who suggest  that biofeedback procedures
that alter  frontal SP levels influence the rate at which
learning occurs.  Some critical issues have been delin-
eated in this summary  and several points of direct
experimental attack have been suggested. A summary
of the  general   state of knowledge concerning  SP
genesis has been provided, and an indication of trends
in research has been included implicitly.

    It is  apparent that  the electrophrenerrian  ap-
proach  to the  study of SP and behavioral relation-
 ships is  inadequate. Elaboration of SP electrogenesis
is necessary and demands  interdisciplinary coopera-
 tion as  much as any other specialized area in neuro-
science.  Because  SPs   are sensitive indicators  of

-------
 Summary of Electrogenesis Issues
                                              11
endogenous and exogenous influences on the brain,
the  experimental  psychologist in league with  the
electrophysiologist  possesses  a powerful  tool  for
studying brain-behavior relationships.  Such studies
have ranged from the evaluation of memory in ro-
dents (Rebert et al. 1974) to assessment of cognitive
divisions of the human mind (McAdam and Whitaker
1971).  At  the  molecular level, an understanding of
SP electrogenesis requires  evaluation by the physio-
logist,  neurochemist,  and  neuroanatomist of neural
spiking patterns, synaptic transmitters,  modulators,
mediators  and  moderators,  glial functions,  ionic
mechanisms,  blood  flow, energy metabolism,  and
macro- and microanatomy.  Because  of the variety,
ubiquity and involvement of slow  potentials in so
many  cerebral  processes, the interdisciplinary study
of SPs should greatly enhance  the understanding of
brain function in general. An outstanding question is
whether SPs are mere signs  of  neurophyriological
activity or whether they have a direct role in neural
function per se. Some evidence suggests that  the
latter is true (Schmitt et al. 1976). If so, the study of
SPs may contribute to our understanding of synthetic
and  integrative properties  of the brain not easily
accounted for in terms of digital information trans-
mission systems. It is known that graded potentials at
the neuronal level  act to integrate converging inputs
and  provide  a decision point in information  flow.
Might  similar  functions be attributed to  SP events
occurring at widespread points in the brain? The
detailed  explication of  SP  genesis will  hopefully
provide an answer to this intriguing question.

     Has SP research contributed to  the development
of any general principles of brain function? The final
section of this volume provides an  impressive af-
firmative in the theories proposed by Cooper  et al,
Marczynski,  Papakostopoulos,  and Skinner. These
general  models are  complemented  by  molecular
models at the level of cellular membranes and ionic
mechanisms proposed by Somjen and  Libet (this sec-
tion). These heuristic contributions are not surprising
in view of the integrative nature of SP research. Slow
potential phenomena thus appear to  provide an in-
creasingly important bridge between  the  mind and
matter of the human brain.

-------
SLOW  POSTSYNAPTIC RESPONSES OF SYMPATHETIC
GANGLION  CELLS AS MODELS  FOR SLOW  POTENTIAL
CHANGES IN THE  BRAIN


B. LIBET
University of California School of Medicine, San Francisco, CA, U.S.A.
  When Gerard and I initially explored the existence
 and significance of slow potentials (SPs) in the brain,
 we  utilized the relatively simpler system  available
 in  the  isolated cerebrum of the  frog (Gerard and
 Libet 1940, Ubet and Gerard 1941). The absence of
 vascular circulation eliminated any possible blood-
 brain SPs,  and the thin-walled  outer pallium of  the
 cerebrum made it feasible to record and apply SPs
 across a layer of similarly oriented cerebral neurons.
 These studies led us to  suggest that SPs were gen-
 erated  as  gradients  along  the dendroaxonal axes
 of neurons and that substantial extracellular  fields
 of current could be developed when this occurred in
 masses  of  similarly oriented neurons.  In spite of
 many developments since that time, questions about
 the  precise cellular origins and electrogenic processes
 for SPs  are obviously still  open ones.

  In the late 1950's, my  interest was aroused in the
 potentialities of an even  simpler system as a model
 for  SP mechanisms.   Findings  of slow  potential
 responses to orthodromic  inputs were being reported
 for  isolated sympathetic  ganglia (see  Libet 1970).
 These structures have no  complicating interneuronal
 networks (although the  existence  of  one  type of
 intemeuron has now  been  established);  they are
 quiescent  except  when  activated  by  neuronal  or
 chemical inputs; composition  of  external  medium
 is easily manipulated; and the postsynaptic responses
 of the principal neurons  or  "ganglion cells"  them-
 selves can be readily studied.  We have been able to
 establish that the ganglion cells can exhibit, in addi-
 tion  to  the  well-known excitatory  postsynaptic
 potential (EPSP),  a slow inhibitory  postsynaptic
 potential (ffSP) and an  even  slower  EPSP (Libet
 1970).  The unique features  of  these slow  post-
 synaptic  potentials  (PSPs)  and  of an  additional
 enduring modulatory action (Libet and Tosaka 1970)
 will now be listed, with a  brief reference in each case
 to potential significance for SPs and for other slow
 functions of the brain,
 Neuronal origin

  Slow  ganglionic  potentials can be recorded with
 surface  electrodes in response  to brief  trains of
 orthodromic, preganglionic volleys.  These can be
 shown to be independent of the true afterpotentials
 that follow cell-firing by their appearance in the
 partially curarized ganglion  (Fig.  1). In the latter
 the (fast) EPSPs are  depressed to below firing level;
 superimposed  upon  and  following each  train of
 EPSPs is a  surface-positive (hyperpolarizing)  and a
 later  surface-negative  (depolarizing) slow potential.
 These  slow  potentials  are  also  demonstrable  by
 direct  intracellular recordings in  almost  all cells
 proven  to be  principal neurons (i.e.,  ganglion cells,
 Fig.  2).  The slow hyperpolarizing and depolarizing
 postsynaptic potentials (Fig. 2 II)  are properly to be
 regarded as a s-IPSP and s-EPSP, respectively: they
 are obviously neuronal, not glial, in origin; they can
 be  elicited  independently,  in the  absence of  any
 EPSPs (Fig. 2 II, tracings E.F); they are selectively
 blocked not by curare but by  atropine (Fig. 2 II,
 tracings B-D) and, in the case of the  s-IPSP,  by an
 oadrenerglc antagonist; and they appropriately alter
 neuronal   excitability (Libet  1964i  Dunant  and
 Dolivo 1967,  Brimble and Wallis  1974). It is clear
 that true postsynaptic potentials with the slow timing
 features described below can be generated by  a syn-
 aptic  action at the single-cell level; the corollary is
 that an extracellularly recorded SP of neurons! origin
 need not be an envelope  of a series of faster indiv-
 idual cell responses.

 Repetitive input

  Although a single orthodromic volley can elicit a
relatively small s-IPSP and s-EPSP, the amplitudes
and  durations of these  responses can be built  up
greatly by a repetition of 5 or 10 volleys (Fig. 1 and
2 IQ.  Optimal repetitive frequency is about  10 to 20
per sec for the s-EPSP and about 40 per sec for the

-------
 Slow PSPs as Basis for Recorded SPs
                                               13
Fig. L  Surface recordings of responses of isolated superior cervical ganglion (rabbit) to stimulation ofpregang-
lionic  (cervical sympathetic) nerve.   Upper section, I,  taken  before, and lower section, II at higher gain, after
partial curarization.   Responses are to a single volley (A,E), or to a train at the pulse frequency Indicated and
recorded at the slower sweep speed of I sec per dlv, (For the upper horizontal row in each section, stimulus inten-
sity is below threshold for preganglionic C fibers; for lower rows, stimuli are supramaxtmal for both B and C fiber
inputs.)  Note the following points: (1) The compound action potential, and its surface-positive afterpotential,
are absent in the curarized state; single responses in  the latter consist of the (fast) EPSP (section II, A and E),
and repetition at 40/sec produces a sustained EPSP with almost no spike components (J).  (2) In spite of the ab-
sence of cell firing and of any true afterpotentials, the train responses of the curarized ganglion exhibit a slow
surface-positive (hyperpolarizing) and an  even slower surface-negative (depolarizing) component, superimposed
on  and outlasting the summated EPSPs produced during the stimulus train.  These are the s-IPSP and s-EPSP,
respectively,  (3) Evidence of the slow  depolarizing component is visible as a negative  hump in the form of the
posttetantc afterpotential of the firing, uncurarized ganglion.  This hump can be selectively eliminated by atro-
pine (see  Libel 1964), with  the positive after-potential then showing a long, exponential decay. (From Eccles
and Libet 1961.)
s-IPSP, but increases are substantial even with 2-per.
sec trains.  This large effect  of repetition appears
to be a postsynaptic function, not a presynaptic one
as in classical posttetanic  potention (FTP): (1)  It
occurs without a parallel effect on the EPSP response.
(2) The  effective repetition interval is much longer
than for presynaptic FTP.   (3) Effective train dura-
tions are much shorter than for presynaptic FTP. The
slow  PSPs  thus can exhibit considerable  temporal
facilitation  with physiologically  reasonable inputs,
both  as  to frequency and duration  of the arriving
group of impulses.
 Durations

  The slow PSPs have  durations from a few seconds
after a  single  preganglionic volley to 10 to 30 sec
following a brief train of S to 20 preganglionic volleys
(at 2 to 40/sec). It should be reemphasized that these
prolonged PSPs cannot be assigned to any sustained
presynaptic  delivery of acetylcholine (ACh);   the
fast EPSP, which is also cholinergic, shows a relative-
ly sharp termination after the end of each brief pre-
ganglionic  train.   The  long  period of heightened
excitability  that accompanies  the  s-EPSP  provides
a  form  of  postsynaptic PTP, which can summate
with  a  heterosynaptic  input  (Libet  1964).   This
feature is far more  significant for brain function than
that  of classical presynaptic  PTP,  which is only
effective homosynaptically  (i.e., when  test volleys
are delivered via the same  fiber inputs  that carried
the train of conditioning volleys). There has been a
tendency to regard  as  "modulatory" all  synaptic
changes that are slower than  the  classic  PSPs.   It
would  be better, however, to expand the view of
possible types of excitatory and inhibitory synaptic
actions so as not to exclude responses like the s-IPSP
and s-EPSP simply on the  basis  of  slower  temporal
properties or different electrogenic mechanisms. The
term  "modulatory"  should be  reserved for cases
in which a  transmitter alters the reactivity in a syn-
aptic  pathway by actions that go beyond the changes
in  cell  excitability  associated with  its  own PSP
(see  discussion  under  "Dopamine modulation..."
below), and for the additional  cases  in  which  a

-------
                                                                                                   Libet
 II.  A
                             50 msec
                                                                                                 0.2 sec
Fig. 2.  Intracellular recordings from a principal neuron (ganglion cell) in rabbit SCO.   Upper section, I:  Res-
ponses of normal, uncurarized cell; successive fast EPSPs appear as the stimulus strength  applied to the pre-
ganglionic nerve is raised, until the EPSPs reach firing level for both the earlier arriving B fiber input (D) and the
later arriving C fibers  (E).  Lower section, II: Responses ofcurarized cells. In bottom row, note the single EPSP
in G, with its small s-IPSP component building up  with  repetition in H and (at slower sweep in another cell) in
I.   Upper two  rows at slow  sweep (see 1-sec bar).   Top row responses (A,B,E) are to 10/sec, 1-sec trains and
second row  (C,D,F) to 40/sec, 1-sec trains.  In  A and C,  note the small fast EPSPs (mostly blocked by curare and
having a spike-like appearance  at this slow sweep), with the s-IPSP and s-EPSP superimposed and following
these.  In B and D, both slow PSPs have been  eliminated by addition ofatropine (0.3 Hg/ml), leaving behind the
fast EPSPs alone.  In  E and  F,  another cell, curarization is strong enough to block the fast EPSPs completely,
but the cell still exhibits s-IPSP and s-EPSP responses.  In both portions of figure note  that the slow potentials
are recorded in typical neurons that also exhibit fast EPSPs and firing, but that the slow PSPs are also producible
independently of fast EPSPs and of firing.  (From Libet and Tosaka 1969.)
"neurohormone" is delivered to the affected neuron
via the  blood  stream  rather  than by  presynaptic
terminals in its vicinity.
Long synaptic delays

  The slow PSPs have  extraordinarily long synaptic
delays,  in marked contrast to  all the  known delays
for all fast PSPs, which are on the order of 1 msec
or less  (Eccles  1964).   In mammalian  ganglia, the
delay for the s-IPSP is about 25  msec (80  to 100 msec
in amphibian ganglia), and for the s-EPSP, it is in the
range of 200  to 300 msec (Libet 1967, 1970); an
example  of   the   latter  is   seen  in   Fig.  3.
It should  be  emphasized that  these  are  delays
between the arrival of the impulse in the  preganglion-
ic fiber  terminals and the onset of the  PSPs them-
 selves; they are not delays in cell discharge. Further-
 more, the delays are for PSPs mediated by only one
 (s-EPSP) or at most two (s-IPSP) synaptic steps in the
 ganglion; they are not due to  the addition of many
 short delays in an interneuronal chain. Such delays
 and durations of response could easily accommodate
 the  possibility  that slow event-related   potentials
 (ERPs) in the brain, like the P300 wave or the contin-
 gent negative variation  (CNV), are  developed by a
 single class of neurons activated  monosynaptically.
 It should also be obvious  that the traditional method
 for estimating the number of successive interneurons
 in  a  pathway-dividing   the net  central  delay  by
 about 0.5  to 1 msec for  the assumed delay at each
junction  - could be invalid by a factor of two orders
 of magnitude!

-------
Slow PSPs as basis for recorded SPs

             I

 A
                                             15
                   II atropine
                                                                                              20msec
                                                                                              4-f-
                                                                                               100msec
                                                                                              4H-
                                                            i
                                                                                              200 msec
                                                                                             +H-
                                                                                               1.0 sec
Fig.  3.  Synaptic delays for (fast) EPSP vs. s-EPSP.  Surface-recorded responses of curarized celiac ganglion of
cat, column I before and column II after adding atropine (0,1 Hg/™l), A-D show ganglionic responses to single
preganglionic volleys at progressively slower sweep speeds.  (E, at 50 msec per div, shows a brief train of res-
ponses,  with  only a small population  action potential appearing on some.) Note that the onset of the s-EPSP
is delayed until about 300 msec after the onset of the (fast) EPSP, as indicated by the arrows.The initial 5- to 6-
msec latency before onset of the (fast) EPSP (see A) already  includes conduction time into  the presynaptic
terminals of the preganglionic fibers here.  Therefore, the extra 300-msec latency for  the s-EPSP represents
additional actual synoptic delay.  (The relatively small amplitude of s-EPSP is due to use of a single  volley rather
than a train; but its reality is indicated by its selective elimination with atropine.) (From Libel 1967.)
Electrogenesis with no increases in membrane
conductances

  The slow PSPs do not depend on increases in mem-
brane  conductance for their  electrogenesis,  unlike
all of the well-known (fast) IPSPs and EPSPs (Kobay-
ashi and  Libet 1968, 1974).   In  addition, changes
produced by polarizing the  resting  membrane poten-
tial are not those expected on any ionic permeability
hypothesis.   Both the s-IPSP and s-EPSP responses
are  enlarged  by increasing  the resting membrane
potential in  the  range of from -50 mV to about
-65  or -70 mV and  are rapidly decreased  (but not
reversed) by further hyperpolarization  (Fig.  4);
both are reduced rapidly by  depolarizing  to levels
less  than -50 mV, but this may be partly an indirect
effect  of a reduced  membrane resistance associated
with "delayed rectification."  The actual  electro-
genie mechanisms for the slow PSPs  are yet to be
determined.  They could involve activation of various
ion  pumps  that  may be electrogenic, although the
specific  case of the  ouabain-sensitive Na+-K+pump
appears to be excluded (Kobayashi and Libet 1968,
1970; Libet 1970; Libet et al. 1977). It has been pro-
posed that  the s-EPSP may result from a decrease or
"inactivation" in  the resting membrane conductance
for K+, GK+ (Weight and Votava 1970 for sympa-
thetic ganglia, Krnjevic et al. 1971 for cerebral cor-
tex, also see Krnjevic 1974), as opposed to the increa-
ses in ionic conductances for the fast PSPs. However,
a careful evaluation of all the evidence disproves the
validity of  this proposal, at least for s-EPSPs in sym-
pathetic ganglia (Kobayashi  and Libet 1974, Libet
and Kobayashi 1974).
  The hypothesis that inactivation of GK+ provides
the  electrogenic  mechanism  for the s-EPSP  was
initially based on findings of (1) an increase in mem-
brane resistance (rm) accompanying  the s-EPSP in
frog ganglion cells,  and (2) a kind of reversal of the
polarity  of  the  s-EPSP when  these same cells were
hyperpolarized by  passing  steady currents  across
the membrane (Kobayashi and Ubet 1968,  1970).
But these  characteristics  are relatively  restricted
to the case  of frog ganglion cells to which nicotine

-------
16
                                            Libet
                      20 mV   DP
                                             0.5 sec
 Fig. 4.  Effects of polarizing currents on slow PSPs.
 Intracellular recordings of  ganglion  cell In rabbit
 superior cervical,  partially curartzed; each tracing is
 response to 0.25-sec train of supramaxlmal pregang-
 llonic volleys at 40/sec.  Top tracing, response during
 continuous outward current, with resting membrane
 potential  depolarized  by 20 mV; second  tracing,
 with no polarizing current (resting potential of this
 impaled cell = -4SmV); third tracing,  during passage
 of inward  current  that  hyperpolarized  by 15 mV
 (i.e., present membrane potential kept at -60 m V);
 bottom tracing, during hyperpolarlzatlon of 30 mV
 (membrane potential at  -75  mV).  Note:   (1) The
 summated  EPSPs produced during each  train In-
 crease  progressively  as the  transmembrane poten-
 tial is increased,  as expected.   (2) The  s-IPSP and
 s-EPSP  responses both  Increased  with  moderate
 increases in membrane  potential, but both decreased
 with hyperpolarizing  levels  greater than  the pre-
 sumed  normal or  physiological resting potential
 of -60 to  -70  mV.    (From  Kobayashi and  Libet
 1968.)
has  been applied (for  the  purpose of blocking the
fast  EPSP responses). In curarized frog cells, there is
either  a small or no change in rm during the s-EPSP
(Kobayashi and  Libet  1970); hyperpolarizing these
cells produces no reversal of s-EPSP at all, but rather
an enhancement of its late  phase (Kobayashi and
Libet  1974).   Curarized mammalian ganglion  cells
show no detectable changes in  rm with the s-EPSP,
and  moderate  hyperpolarization  produces an in-
crease,  rather  than the  predicted  decrease, in the
amplitude of the s-EPSP (Kobayashi and Libet 1968;
see Fig. 4).   Even  in the nicotinized frog cells,  an
increase  in rm  with s-EPSP is  found only when
the  cell membrane is  depolarized.  At normal  or
hyperpolarized levels, no change in rm accompanied
 the s-EPSP response (Nishi et al. 1969). Additionally,
 the  apparent  "reversal" of polarity of the s-EPSP
 when nicotinized frog cells are hyperpolarized is not
 a true reversal; only a brief initial portion of the re-
 sponse  reverses polarity,  and the  latency of this
. portion is  distinctly shorter than that of the normal
 s-EPSP (Kobayashi and Libet  1970, 1974).  Finally,
 the s-EPSP of nicotinized frog cells does not behave
 appropriately in relation to EK+ (equilibrium poten-
 tial for  K+); the apparent  "reversal" of initial phase
 of  s-EPSP has been  found to occur at  membrane
 potentials different from EK+ (Kobayashi and Libet
 1974), and the s-EPSP is relatively insensitive to
 the removal of K+ from the external medium (Kobay-
 ashi and Libet 1968, Nishi et al. 1969).

   A  related proposal  for  generation of the s-IPSP
 involves  an  inactivation  of  resting conductance
 for Na+,  GNa+ (Weight and Padjen 1973a).   But
 the  evidence  for  this hypothesis is all based  on
 studies of nicotinized ganglion cells of frog, in which
 ACh is  able to elicit a hyperpolarizing response by
 a  direct   postsynaptic  action  on  ganglion  cells
 (Weight and Padjen  1973b,  Libet  and  Kobayashi
 1974); this form of  "s-IPSP" appears to be an ab-
 normal component of the response, made possible by
 an  interesting pharmacological interaction  between
 side-effects of nicotine and ACh (Libet and  Kobaya-
 shi 1974).  In frog  or  rabbit cells  treated with a
 "cleaner,"   a  competitive  nicotinic  blocker   like
 curare,  the direct transmitter for the s-IPSP is a
 catecholamine  (Libet and Kobayashi   1974); the
 latter must be released by an indirect action of ACh
 on  an interneuron (Libet  1970, Libet and Kobaya-
 shi 1974).  The physiological s-IPSP mediated  by
 a  catecholamine  is  obviously  a different  type of
 response  from the  ACh-hyperpolarization  elicited
 in  nicotinized  frog  ganglia.    The characteristics
 of this response (Kobayashi and Libet 1968, 1970)
 and  of  non-nicotinized  cells  do  not fit with the
 requirements of the  GNB+ inactivation hypothesis
 (Libet and Kobayashi  1974).

     Whether an inactivation  (decrease)  of resting
 ionic conductances underlies those slow postsynaptic
 responses  found in  the  central nervous system  re-
 mains to be seen. Pyramidal cells in neocortex exhi-
 bit  a slow  muscarinic depolarizing response  to  ACh
 applied  iontophoretically  (Krnjevic  and Schwartz
 1967, Krnjevic 1974).   This response strongly resem-
 bles  the s-EPSP of mammalian sympathetic ganglion
 cells (see  Marczynski, this section), except for  a
 reported increase in rm (Krnjevic et al. 1971).  Slow
 hyperpolarizing  responses  to  norepinephrine  have
 been  described  for  spinal motoneurons (Engberg
 and Marshall 1973) and for cerebellar Purkinje  cells
 (Siggins et al. 1971).  These  hyperpolarizing responses
 resemble  in important  ways the s-IPSP of sympa-
 thetic ganglion cells,  except for  reported increases

-------
Slow PSPs as basis for recorded SPs
                                               17
in rm.   It is not clear  at  present whether electro-
genie mechanisms for slow PSPs in the  central ner-
vous  system  are  actually different  from those in
sympathetic  ganglia, or whether the experimental
conditions in central nervous system  studies have
introduced additional characteristics  that may have
a more  pharmacological  than physiological signifi-
cance.  For example, it  is generally recognized that
in tests  with iontophoretic application  of chemical
agents the local concentrations  are  unknown and
may be  higher than those available in physiological
inputs.   Also, the  exogenously  applied agent can
reach and act on sites  other than  those involved
in normal local synaptic transmission; such sites may
be nonsynaptic  receptor sites on the  same cell, with
different  response properties  from the  postsynaptic
receptors, or  sites on other types of adjacent cells
which  indirectly  affect  the  function  of the cell
under study.  One example of the latter is the pro-
duction  of presynaptic  blockade  by  exogenously
applied  catecholamines;  this could depress ongoing
background excitatory inputs into a cell and thus
indirectly result in   a hyperpolarizing  shift  and a
rise  in  r  back to  nonexcited levels (e.g.,  Jordan
1973).
  In any case, it is generally agreed that there are
at least  no  increases in  ionic conductances (i.e.,
no decreases in rm) in all the slow  responses studied,
and  even  this presents some unique  implications
for  slow functions in the brain. It means that long-
lasting active  PSPs  could be generated without the
increased  ionic leakages  and possible  alterations
in cell concentration  gradients  that characterize the
"fast" PSPs:

  1. The cellular energy costs for slow potentials are
    thus greatly reduced.

  2. Possible  effects of otherwise altered  extra-and
    intracellular  changes  in  ionic  concentrations
    are eliminated; for example,  the rise in extracel-
    lular  K+  concentration that  is  an  expected
    feature of sustained fast EPSPs  would  not be
    produced  during slow PSPs.

  3. The  "anomalous"  ways  in  which slow  PSPs
    are altered by  the passage  of polarizing cur-
    rents  could result  in  effects of  the latter  on
    slow potentials of brain that are different from
    the effects predicted  with electrogenic  mech-
    anlsms based on changes in  ionic permeability.

  4. Additionally, the i-EPSP  mechanism it  selec-
    tively  and relatively rapidly  depressed  by de-
    preisanti  of  oxidative metabolism, such   as
    dinltrophenol, sodium  azide, cyanide, or simple
    anoxia (Kobayashi and Ubet 1968).
This  sensitivity of the s-EPSP  to loss of oxidative
energy  supply,  and its synaptic delay of 200 msec
or  more, point  to the probable involvement of a
sequence of chemical  reactions in its electrogenesis.
There is already evidence that the synthesis of cyclic
GMP (guanosine 3 ' 5 -monophosphate) is  an early
step in  this mechanism (McAfee and Greengard 1972,
Kebabian et al. 1975), but the succeeding reactions
are yet  to  be  unravelled (Ubet et al.  1975).  The
special  sensitivity of the s-EPSP to anoxia is remi-
niscent  of a similar one for many cerebral processes,
and is in contrast to the relative insensitivity of the
fast PSPs in both ganglia and brain.

Transmitters mediating the slow PSPs

  The intraganglionic pathways and transmitters for
fast and slow PSPs are schematized in Fig. 5.
                                                                                           CURARE BLOCK
Fig. 5.  Schema for synaptic mediation of fast EPSP,
s-IPSP,  and s-EPSP,   Note;  (1) The two  types of
postsynaptic receptors for ACh, ntcotintc and mus-
cartnic, on  the same ganglion cell.   (2) The  small
monoamtnergtc  tnterneuron (SIF or granule-contain-
ing cell), which "converts" ACh input into a dopa-
mine  (DA)  transmitter here. (3) The two actions of
DA on the ganglion cell; these are represented in two
types of receptors here, a concept supported by more
rectnt evidence. (From Ltbet 1976.)


Muscartnic transmitter action for s-EPSP

    The s-EPSP is elicited by a direct (monosynaptic)
action of preganglionically released ACh  on postsyn-
aptic  muscarinic  receptor  sites (Eccles  and  Libet
1961, Libet 1970). Deeper lying neocortical pyrami-
dal cells have been found to exhibit a slow muscarinic
depolarizing response to ACh, which has important
similarities to the s-EPSP in ganglia.  Indeed, most of

-------
18
                                           Libet
 the receptor sites involved  in responses of cerebral
 neurons to ACh are also of the muscarinic type, and
 atropinic agents are potent modifiers of various brain
 functions. Evidence for participation of such respon-
 ses in certain SPs and brain  functions is summarized
 elsewhere in this symposium (Marczynski).

 Adrenergic transmitter action for s-IPSP

    The  s-IPSP is elicited in  ganglion cells by a direct
 action of dopamine, although exogenous norepine-
 phrine can produce a similar response (Libet 1970,
 Libet and  Tosaka 1970, Libet and Owman 1974).
 Dopamine  is released  by a special class  of  inter-
 neurons, now  recognized as the  small,  intensely
 fluorescent  (SIF) cells  or,  in  electronmicroscope
 studies, as  the  small  granule-containing cells.  The
 SIF  cells are  activated by  a muscarinic action of
 preganglionically released ACh  (Eccles  and Libet
 1961,  Libet 1970,  Libet and  Owman 1974).  In
 the brain, it has been found  that the norepinephrine
 transmitter  delivered by fibers arising in the  locus
 coeruleus can elicit a slow hyperpolarizing response
 in cerebellar Purkinje cells (Hoffer et al. 1972) and
 in hippocampal pyramidal cells; this response  has
 important similarities to the s-IPSP of sympathetic
 ganglia. The cellular nature  of the  responses  to
 various  monamines elsewhere in the brain are  yet
 to be worked out.
Dopamine modulation  of the s-EPSP response
to ACh

    In addition to its role as an inhibitory transmitter
in ganglia,  dopamine was  found to induce another
neuronal change,  which  could  persist for hours.
This change is manifested  in an enhancement of the
response  to ACh, but it  is selective for the  slow
muscarinic  or s-EPSP type  of response (Libet and
 Tosaka  1970).  This novel type of synaptic action,
 in which one  transmitter  alters  the postsynapiic
 response to another, has  now  been shown  to pos-
 sess  features  of a memory  process at the neuronal
 level (Libet et al. 1975). At  least one cerebral exam-
 ple of such a modulatory action has already been re-
 ported (Yamamoto 1973).  Also,  a morphological
 substrate for monoaminergic actions, in the form of
 widespread cerebral distributions of various mono-
 aminergic fibers  that  arise  in  certain brain  stem
 nuclei,  has  more recently  been  provided.   These
 nuclei have  already been  implicated in the  control
 of sleep  and waking states  (Jouvet 1973),  and the
 self-stimulation  or  "reward"   mechanism of  Olds
 (German and Bowden  1974).  The notion has been
 gaining ground  that these  systems may in part  func-
 tion  to  change, in various ways, the  reactive levels
 of cerebral neurons to  other synaptic inputs  (e.g.,
 Jasper 1975, Reader et al.  1976).  In the case of
 those ERP components or  SPs in the brain  that
 may   be  similar  to the cholinergic, muscarinically
 mediated  s-EPSP of ganglion  cells,  the  possibility
 arises that monoamine inputs  might produce  long-
 lasting alterations in the  amplitudes and durations
 of such ERPs produced at  selective sites. Such modu-
 latory changes  by  monoamine inputs could  con-
 ceivably  account for  certain  ERP changes which
 accompany psychological  processes  discussed  else-
 where in this volume.
Acknowledgment

I dedicated this paper to Ralph Waldo Gerard, whose
scientific imagination helped to  pioneer the field of
slow potentials in the central nervous system. The re-
search was supported by Public Health Service Re-
search Grant NB-00884 from the National Institute of
Neurological and Communicative Disorders and Stroke.

-------
CONTRIBUTION OF NEUROGLIA TO EXTRACELLULAR
SUSTAINED  POTENTIAL  SHIFTS1

G. G. SOMJEN

Department of Physiology, Duke University Medical Center, Durham, NC, U.S.A.
Evidence for neuroglia acting as a generator of
extracellular current

    That sustained shifts of extracellular potential
associated with protracted neural activity are, under
certain definable  experimental conditions generated,
in part or entirely, by glial elements is indicated by
two  sets of observations. First, the time course of
membrane  potential  changes- of glia cells observed
during neural activity  is comparable  to sustained
shifts of extracellular potential. Second, extracellular
potential shifts are altered by experimental manipula-
tion  in ways more similar to concomitant alterations
of glial responses than to those of neurons.

    In mammalian brains, membrane  potential re-
sponses related to neural activity, but recorded from
cells originally designated as "idle" or "unresponsive"
and  subsequently identified  as glia, were  first re-
ported from the laboratory of Goldring (Sugaya et al.
1964; Karahashi  and  Goldring 1966; Castellucci and
Goldring 1970; Ransom  and Goldring 1973a, b, c).
Kuffler and collaborators also described the prop-
erties and responses of neuroglia in leech and mud-
puppy nervous systems  (Kuffler and Potter 1964,
Kuffler et al. 1966, Orkand et al. 1966).

    The classical work of Kuffler's group  has  pro-
vided a theoretical outline of the possible mechanism
of glial current generation. These investigators found
that  the membrane potential of glia cells conforms to
the Nernst equation for potassium much more closely
than that of neurons. In other words, the degree to
which  glial membranes favor permeation by potas-
sium over other ions is significantly greater than the
similar but lesser preference of neuronal membranes.
Potential responses were  observed in glia cells of the
amphibian optic nerve, and it was suggested that they
are related to presumed increments of extracellular
potassium consequent to neural impulse activity. An
electrotonic coupling between glia cells was, further-
more, demonstrated. The presence of low-resistance
intercellular junctions was seen to provide the condi-
tion of glial  contribution to electrical activity of
tissue, for glia cells depolarized by external potassium
could,  through such junctions,  draw current from
distant, non-depolarized glia cells. The electric circuit
would  then be completed  by  return current flow
through the extracellular medium. The latter would
generate a voltage drop which could be registered by
external electrodes.

    A highly simplified  model describing the behav-
ior of such a quasi-syncytial electrotonic network has
been developed (Joyner and Somjen 1974; Sornjen
1973, 1974).  The model is reproduced in Fig. 1 and
2. The elements of the  model are: (1) a set of bat-
teries representing the EMF of depolarized glia cells;
(2) three sets of resistors simulating the conductances
of glial cell membranes, the intercellular electrotonic
junctions, and the extracellular medium; and (3) the
bulk resistance of the body  interposed between nerv-
ous system  and reference  (or ground).  Capacitive
components are ignored because the time course of
the potential changes to be mddeled is very long com-
pared to the (probable)  time constant of the circuit.
The model demonstrates that the spatial  profile of
extracellular potential shifts would  accurately map
the spatial distribution of membrane responses (and,
hence, of potassium accumulation) provided that the
"space constant" of the network was small, i.e., mem-
brane resistances were low  compared to other resis-
tive components. The profile of extracellular poten-
tial changes might extend beyond the spatial limit of
 1 Since the adjective "slow" in electrophysiology can mean any event that lasts longer than 1 or 2 msec, it seems
 desirable to distinguish potential shifts that ire maintained for seconds and/or rise and fall with half-times of 0.3
 sec or more by another name. "Sustained potential" seems an accurate descriptive term. Processes that have an
 hourly or daily rhythm should be given still other names, for they probably represent yet different classes of
 events.
    In cases where an electrical event has a known generator, that cellular element may justifiably be named,
 Thus, evoked sustained potentials of the spinal gray matter appear to be the sum of two components, a smaller
 abruptly  rising  neural potential (probably related to synaptic currents) and a larger, more slowly rising glial
 potential.

-------
  20
              PASSIVE CELLS
•ACTIVE CELLS
              Somjen


PASSIVE CELLS-)
       VR?>,
                                      RX|1  , VT
                          -'-S    i   VMSr-i"
                    B.
                                                                                  INTRACELLULAR
                                                                                      PLANE
                                                                                      ROUND
                                                                                  EXTRACELLULAR
                                                                                      PLANE
Fig,  1,  Two  models simulating electrotonic syncytia,  A: Simulating a linear array of cells; B: A  two-dimen-
sional layer of cells.  The cytoplasm of cells is assumed to be isoelectric, represented by the potentials at the
points marked as Vj. Cells are assumed to be coupled by resistances, RX, and connected to their environment by
the membrane resistances RM- The potential immediately outside the cells is given by VE, and the environment
of one cell connected to that of another by the extracellular resistances Rg.  The extracellular fluid of the nervous
system is grounded through  the tissue resistances Rj\ Some cells are assumed to be depolarized; these are desig-
nated "active," and the EMF of the potential change is represented by batteries ^Ey. The resting potential of
cells is not represented, since it does not contribute to current flow. Also neglected is  membrane  capacitance
(see text). (After Joyner and Somjen, reproduced, by permission of the publisher, from Somjen 1973).

-------
Neuroglial Basis of SPs

         A. VARYING MEMBRANE RESISTANCE
                                   B. VARYING COUPLING RESISTANCE
                                                                                                   21
           INTRACEUULAR
 ,.—x     a=Ru=io
,'      \        *
           EXTRACELLULAR
          C.  VARYING EXTRACELLULAR RESISTANCE
                                   D. VARYING TISSUE TO GROUND RESISTANCE
                                                                                      a   RT=0
 Fig. 2.  Computer solutions of the model network of Fig. 1A.  Twenty-one "cells "are represented, the central
 five of which are assumed to undergo depolarization. The voltage of the depolarizing effect is taken to be iden-
 tical for all cases.  The computer-generated profiles show intracellular and extracellular voltages referred to
 "ground," as they would be measured in conventional microelectrode recordings. The change  of transmembrane
 potential would be given by the difference between intracellular and extracellular potentials shown. The absolute
 membrane potential would be  the resting transmembrane potential (neglected in the model),  less the change of
 transmembrane  potential; i.e.,  the  horizontal "zero" lines  of  the diagrams represent the "collapsed" resting
 membrane potential. In real cells the intracellular potential profiles would lie in the negative domain, below the
 zero line and below the extracellular potential profiles. Resistances that are not Indicated in inset legends were
 assumed to  be uniform within one set of computations. (After Joyner and Somjen, by permission of the pub-
 lisher, from Somjen 1973,)
 depolarizing EMF (the distribution of the "batteries"
 in the model) if the space constant were taken to be
 large (Fig.  2A).  With the aid  of the model, it  also
 becomes  apparent  that a  relatively small positive
 extracellular potential shift may or may not be de-
 tectable at the (inactive)  extracellular "source" of
 current. It  furthermore appears that the depolariza-
 tion  registered by  an  intracellular probe may fall
 short of the value predicted by the Nernst equation,
 even if the only charged particles admitted through
 the  membrane were potassium ions. An erroneously
 low reading would  always be expected if,  as is some-
 times done,  intracellular  potential responses  were
 referred to a distant reference (or to ground) so that
 extracellular  potential shifts  were  neglected  (cf.
 Somjen  1975, Lothman and  Somjen  1975).  True
 attenuation of the transmembrane  potential would
 occur if a significant fraction of current flow would
                               spread  well  outside the activated  region (note the
                               cases of high membrane resistance  and low coupling
                               resistance in Fig. 2A and B).

                                   The  analysis  of glisl  function by Kuffler and
                               collaborators  (Kuffler  and  Nicholls 1966) concerned
                               the nervous  system of cold-blooded organisms. For
                               the mammalian central nervous system, experimental
                               evidence  now available supports the glial theory as
                               follows:

                                   1.  The   development   of  potassium-selective
                               microelectrodes enabled the measurement of activity
                               of this ion in central nervous tissue in situ. With these
                               devices, it was shown that potassium activity can
                               readily be induced to rise above resting levels signifi-
                               cantly  and  reproducibly.  Such  responses  can be
                               evoked by electrical stimulation of tissue or afferent

-------
 22
                                                                                                Somjen
 paths leading  to it (Vyskocil et al. 1972; Vyklicky
 et al. 1975; Krnjevic and  Morris  1972, 1974; Kriz
 et al.  1974;  Lux  and  Neher  1973;  Somjen and
 Lothman  1974; Lothman et al.  1975; Lothman and
 Somjen  1975)  and under more physiological condi-
 tions, e.g., in the visual cortex by "adequate" optical
 stimulation (Singer and Lux 1975).  This response also
 occurs spontaneously during "spindle" activity  in
 barbiturate anesthesia (Somjen et al. 1976).

     2. A close  correlation between depolarization of
 glia cells and sustained shifts of potential in the extra-
 cellular environment evoked by repetitive stimulation
 of afferent nerves  was  demonstrated in the spinal
 cord by  Somjen  (1970), Strittmatter  and  Somjen
 (1973), and Lothman and Somjen (1975).  No  such
 correlation exists  for extracellular potential  shifts
 and neuronal membrane potential responses (Somjen
 1969,1970).

     3. Dependence of the membrane  potential  of
 mammalian  glia  cells on extracellular potassium
 activity was indicated by Pape and Katzman (1972)
 and Ransom and  Goldring (1973a), and conformity
 of this  relationship  to the  Nernst  function by
 Lothman  and   Somjen (1975) for the  spinal  cord
 and by Pedley et al. (1976) in cortex.

     4. The close  correlation between transient ele-
 vation of  potassium  activity and sustained shifts  of
 extracellular potential, evoked either by direct  elec-
 trical stimulation of cortex (Lothman et al. 1975), by
 afferent  nerve stimulation in gray (but not  white)
 matter of spinal cord (Somjen and Lothman 1974,
 Lothman and Somjen 1975), or occurring spontane-
 ously  during seizure activity (Lothman  and  Somjen
 1976) is also well documented.

     The required electrical continuity of intracellular
 conduction within glial tissue of  mammalian  gray
 matter has not been demonstrated so far. Electro-
 tonic junctions  between glial cells  have  been shown
 only in invertebrate and cold-blooded vertebrate nerv-
 ous systems (Kuffler and Potter 1964, Kuffler et al.
 1966b, Cohen 1970).Yet, even if mammalian glia cells
 prove  to be electrically  insulated  from  each other,
 they could contribute to extracellular current flow,
 provided that  the  processes extend over sufficient
 distances in sufficient profusion to  provide for cyto-
 plasmic continuity between excited and quiescent
 regions. Miiller cells of the retina and Bergman fibers
 of the cerebellar cortex fulfill this requirement. Only
 three-dimensional reconstructions from  serial electron
micrographs could determine  the  issue for other
regions (Katchalsky et al.  1974, Somjen 1975).

    In the mammalian nervous system, glial responses
have been  studied most intensively in the cerebral
 neocortex and in the spinal cord.  Although quali-
 tatively similar, the two regions appear to differ in
 the relative magnitude of response. In the spinal cord,
 the amplitude of glial depolarization is approximately
 twice  the amplitude of the corresponding extracellu-
 lar sustained potential shift. In the cortex, this ratio
 is  consistently greater.  Furthermore, in the spinal
 cord,  smaller potassium  transients appear to be asso-
 ciated with relatively larger electrical events than in
 cortex (reviewed in Somjen 1975). The close corres-
 pondence  of the  spatial profile of potassium  re-
 sponses and of sustained potential  shifts in spinal
 cord (Lothman and Somjen 1975) indicates that  the
 "space constant" of the glial network (see Fig. 2) is
 small  (i.e.,  membrane resistance is low).  The rela-
 tively  large extracellular potential shifts observed in
 spinal  gray matter  suggest a higher tissue resistivity
 (in relation to membrane resistance  and cytoplasmic
 coupling resistance) than that of cortical tissue.

     Oxidative metabolic activity is another function
 normally tied closely to pertubations in the distribu-
 tion of ions. Oxidative metabolism can be monitored
 by measuring the fluorescence of NADH in neocortex
 (Rosenthal  and Jobsis 1971, Rosenthal and Somjen
 1973,   Lothman  et al.  1975)  or  in hippocampus
 (Lewis and  Semite 1975). The  close coupling  be-
 tween  | potassium| responses,! 8UStained |potential
 shifts,  and  oxidative activity  is broken  during con-
 vulsions, as though seizures placed a load on energy
 metabolism in excess of the demand made on active
 transport  by extracellular potassium (Lothman  et
 al.  1975, Somjen  et al. 1976). The relationship of
 cellular oxidation  and ion transport is illustrated
 schematically in Fig. 3.

 Significance of potassium and of extracellular
 electric current flow for neural function

     Even  if neurons are less sensitive than glia cells
 to small elevations of extracellular potassium, potas-
 sium discharged  by active cells should, theoretically,
 influence the membrane potential of inactive neigh-
 bors and synapses between them. Actual intracellular
 recordings made from neurons in spinal cord suggest
 an  indifference to prevailing extracellular potential
 (Somjen 1970) and potassium activity (Lothman and
 Somjen 1975). Nerve  cells appear  to be controlled
 by  specific  synaptic transmitters. Any superimposed
 influence by changes of potassium  activity appears
 to  be   negligible  or  undetectable  in electrical  re-
 cordings. Transient  responses of extracellular potas-
 sium and  extracellular electric current  flow appear
 also to be  of little  significance in the generation of
 primary afferent  depolarization  and the attendant
 dorsal  root potentials evoked by afferent nerve stimu-
lation (Somjen 1970, 1973, 1974; Somjen and Loth-
man 1974, Lothman and Somjen 1975).

-------
Neuroglial Basis of SPs
Fig.  3. Schematic representation of the relationship
of perturbations  of Ion distribution and metabolic
activity  in  central nervous  tissue. Neurons  lose
potassium  (and  gain  sodium)  when  firing action
potentials and also during certain types of synoptic
activity.  All  lost  potassium  must  eventually  be
retrieved by neurons.  Part of it may be recaptured
immediately by active transport ("pump") through
the  neuronal cell membrane, but part may be tem-
porarily dissipated by diffusion through extracellular
space from  high  to low activity, and by uptake into
glta.  Active  transport is fueled by Na+-K+ dependent
membrane-bound ATPase. The A DP generated in the
process enters mitochondria and activates the oxida-
tion of NADH to NAD. NADH*, being fluorescent,
is accessible to  optical monitoring. Events in the
intramitochondrial  enzyme chain lead to rephospho-
rylation  of  ADP to ATP, at the expense of oxygen
and  substrate utilization. Neuroglial membranes are
also   known  to  contain  ATPase.  Glia  cells,  like
neurons, are exposed to the excess potassium accumu-
lated In  extracellular space but, unlike neurons, do
not gain sodium in the course of activity. Whether or
not  glial ATPase is significantly stimulated by extra-
cellular K+  is not known at this time. Potassium dis-
persed either by way  of the glial syncytium or by
diffusion through extracellular space must eventually
return by a reversal of the process. Note that extra-
cellular space in central gray matter (unlike in dia-
gram) occupies only about 15% of the tissue volume.
(Reproduced  by permission from  Somjen  et al.
1976.)

     During  seizure activity, the  distribution of potas-
sium responses within  the spinal cord shifts  dramat-
                                              23
ically.  Cells and fibers in  ventral grey matter are
exposed to potassium levels greatly in excess of that
seen under normal conditions. These paroxysmal ele-
vations of potassium may well be the agent respon-
sible for  paroxysmal dorsal root  potentials which
behave differently from dorsal root potentials evoked
by   afferent   stimulation   in  normal  spinal cords
(Lothman and Somjen 1976).

     The  possibility  that  extracellular current gen-
erated by neurons may influence other neurons in the
vicinity was much discussed a few decades ago (Libet
and Gerard 1941; Terzuolo and Bullock 1956) and
has not been  resolved.  The interactions  of glia, neu-
rons and IK+Io in physiological  and  pathological
conditions are the subject of two forthcoming publi-
cations (Varon and Somjen, in press; Somjen,in press).

Validity of the glial  model for the contingent
negative  variation  and  other event-related
potentials

     Potassium-induced  depolarization of neuroglia is
a candidate for supplying currents for potential  de-
flections  which  can  be measured only in  seconds
rather  than milliseconds (half rise times  and half de-
cay times of 0.3 sec or more). Even for these it is not
a unique source, for we have recorded small negative
potential  shifts in white matter and in dorsal roots of
the spinal cord that were not associated with elevated
potassium activity  and were, nevertheless, sustained
for  seconds. Such non-potassium related potentials
were small with sudden onset and decay compared to
those related to potassium and glial depolarization.

     Although not  the  only possible source  of sus-
tained extracellular current, neuroglia deserves special
attention for  two reasons:  (1) glial theory has with-
stood  rigorous  experimental  testing,  albeit  under
limited and  physiological  conditions;  (2)  its role
could  be tested again, under more life-like circum-
stances.

     In  principle, the  experimental testing of glial
theory is simple. To validate glial theory, it is neces-
sary to show  that elevation of potassium activity and
shift of potential are closely and reproducibly corre-
lated. Construction of a potassium-sensitive electrode
that can be used under conditions in which CNV can
be induced will be difficult, but certainly possible.
It  will probably prove easier  to record potassium
activity  in  a  CNV-eliciting situation than to record
the intracellular potential of glia cells.

Source of extracellular potassium

     The  source of excess potassium to which the
membrane  potential of  neuroglia responds is neu-
ronal.  The  question is, what kind of neural activity

-------
 24
 is most likely to add potassium ions to extracellular
 fluid (ECF) at a rate that exceeds their clearance.
 If the glial theory of SP generation proves generally
 valid and if we  can exactly define the  conditions
 of potassium  accumulation  in  ECF, we  then  shall
 understand the significance  of  SP shifts. While the
 relative  contribution of possible sources has not yet
 been assayed, clues can be derived from the following
 three  observations  (Lothman  and  Somjen  1975):
 (1) No rise of potassium activity could be  detected in
 dorsal white matter of the spinal cord near the entry
 zone of an excited dorsal root unless stimulation was
 intense  enough  to  activate  C-fibers; (2)  potassium
 activity  did not  detectably  increase in the ventral
 horn of the spinal  cord when the ventral root was
 antidromically stimulated;  (3) by  contrast,  sizable
 potassium responses (and sustained  potential shifts)
 were readily evoked in substantia gelatinosa of the
 spinal cord, a region where spike potentials are rarely
 recorded.  Neurons,  though  abundant,   are  rarely
 entered  (by  intracellular  microelectrodes), while
 "unresponsive"  (presumed  glia) cells are quite  fre-
 quently entered.

     From  these  observations,  one  suspects  that
impulses carried in myelinated axons and the action
potentials  of the somata of large neurons  contribute
relatively  little  potassium  to  extracellular fluid.
 Hence,  one must  seek the source either in synaptic
currents generated by dendritic trees, in the activity
                                          Somjen

 of unmyelinated fibers and  axonal arborizations, or
 else in small neurons, the impulses of which are hard
 to detect and difficult to  penetrate  with micro-
 electrodes (see also discussion in Varon and Somjen,
 in  press). Furthermore,  the  possible contribution of
 non-spiking neurons cannot be discounted (Rowland,
 this section). Electrical  activity  of non-spiking neu-
 rons  would,  theoretically,  be distinguishable from
 that of glia cells, for  synaptic potentials and synaptic
 noise  bear  readily  recognizable  features  (Somjen
 1973, 1975).

    The theory of glial generation is compatible with
 the theory of cholinergic mediation of event-related
 sustained potentials  (Marczinsky,  this  section).  In
 normal cortex under physiologic conditions, it is con-
 ceivable  that the most abundant sources of potassium
 are activated by cholinergic input. The idea of direct
 action of ACh on  the glial cell membrane is less at-
 tractive because it  implies extrasynaptic diffusion of
 ACh in  tissue rich in cholinesterase. Besides, depolari-
 zation by ACh is associated with  a  rise in membrane
 resistance (Krnjevic et al.  1971), and no such change
 was observed during the  depolarizing response of glia
 cells associated with neural activation (Somjen 1970,
 Ransom  and Goldring 1973b). Finally, the conform-
ity of glial depolarization to the prediction of the
Nernst equation leaves  no  requirement of an  in-
fluence   other  than   that  exerted  by potassium
(Lothman and Somjen 1975).

-------
 NEUROCHEMICAL MECHANISMS  IN THE  GENESIS  OF
 SLOW POTENTIALS: A  REVIEW  AND SOME  CLINICAL
 IMPLICATIONS

T, J, MARCZYNSKI
Department of Pharmacology, College of Medicine, University of Illinois, Chicago, 1L, U.S.A.
  Since  there U  evidence that  SPs are  related  to
neuronal activity (Robert 1973b, Rowland  1974),
there is  reason to consider whether particular neuro-
hormonal  substances  are  involved  in their genesis.
Several  crucial questions arise in this connection. Is
the concept of a neurotransmltter, which implies a
brisk and quickly reversible effect on ionic traniport
across a neuronal membrane, compatible with a role
in the genesis of SPs, or should  our attention be di-
rected to substances capable of acting as modulators
rather than transmitters, I.e., to substances with slow
onset  of action  and a sustained effect on neuronal
and glia  cell membranes? Can the influence of the as-
cending  reticular activating system (ARAS)  on epi-
cortical  (Gaspers 1961) and subcortical SPs (Hayward
et al. 1966) be  related to a particular transmitter or
modulator? What is the role in SPs of the cholinergic
component of the ARAS, defined histologically  by
Shuts and Lewis (1967,1969) and pharmacologically
by Bradley and his colleagues (Bradley  1958) and by
Rinaldi  and Himwich (1955)? Finally, is it possible
to relate the  neurohormonal mechanisms which un-
derlie-SPs to a general concept of information flow
in the mammalian brain? (The latter question it dis-
cussed in the final section of this volume.)

Cholinergic  mechanisms  in epicortical nega-
tive SPs

  The   cholinergic  component  of the ARAS ori-
 ginating in the  legmen tal  nuclei projects  to  the
neocortex (Shute and Lewis 1967), and the release of
 acetylcholine (ACh) from the cortex is related to the
tone of the ARAS (Mitchell 1963, Kanai and Szerb
1965, Celesia and  Jasper 1966,  Pepeu and  Mante-
 gazzini  1964).  The  role of ACh  In EEC desyn-
 chronization  has been postulated (Rinaldi and Him-
wich 19SS, Phiilis 1976) and, because surface negative
 SPs accompany ARAS stimulation (Arduini 1958), a
 role in SP genesis Is also possible.

   If ACh  is  involved in SP genesis,  an important
 question concerns  the type  of cholinergic receptors
 (muscarinic or  nicotinic) that  mediate  its effect.
There Is  strong evidence that the  apical  dendrites
of some  neurons in the neocortex are  endowed
with depolarizing muscarinic receptors that  can be
selectively blocked  by atropine (Sigg et al. 196S).
One  can  assume  with  reasonable  certainty  that
these dendrites belong to large pyramidal cells located
between layers IV and V which are  not sensitive to
nicotine,  but are very responsive (in terms of spike
generation) to  iontophoretlc application  of ACh.
These neurons are endowed  with "pure" muscarinic
receptori that can be  blocked by  atropine  or sco-
polamine applied topically or systemicaUy (Krnjevic
and  Phiilis 1963, Krnjevic  1974).   Some cortical
neurons are not excited by ACh, and others are even
inhibited by micro-iontophoretlc application of ACh
(Phiilis 1976).  Hence,  the question arises as  to
whether  the cholinoceptive "minority"  of  large
pyramidal cells is capable of determining the surface
negative SP upon release of ACh from terminals of
ARAS projections.

  Several observations support an affirmative answer
to  this question.   First, tracing the polysynaptic
cholinergic ARAS projections to the neocortex in the
rat  and  monkey,  Shute and  Lewis (1967, 1969)
showed that these  projections  relay  in the septal
region. In addition, Pirch and Norton (I967b) demon-
strated that rats with septal lesions show significantly
smaller negative SPs in response to conditional stimuli
than do normal animals.  Secondly, laminar recording
of SPs in cortex treated with various concentrations
of ACh  showed that the  initial  surface  negativity
rapidly  spread to  deeper  layers  and reached  its
maximum at the level of large pyramidal cells, or
layers IV and V (Ferguson and Jasper 1971).  Finally,
 the  pharmacological study  of surface  negative long-
 latency components of sensory-evoked potentials and
concomitant measurements  of  ACh release clearly
 point to causal relationships between ACh output and
 the  amplitude  and  duration of the negative waves.
 Topical  application of low concentrations  of cho-
 linesterase  inhibitors  prolonged  the  duration and
 increased the  amplitude,  while  hemicholinium-3,
 which is known  to interfere with choline uptake

-------
 26
                                     Marczynski
 and ACh synthesis, blocked the negative waves.  A
 similar blocking effect  can  be produced by appli-
 cation of antimuscarinic drugs (Szerb 1965). These
 data  strongly  indicate  that ACh  is  released  pre-
 synaptically  in the cortex  from cholinergic com-
 ponents of the ARAS originating in tegmental nuclei,
 and that ACh  plays a  significant role in  the genesis
 of  surface-negative SPs.   A state  of hyperarousal
 associated  with  sustained  depolarization  of  apical
 dendrites, however, may reduce event-related negative
 SPs (vide infra).

 Functional  differences between  muscarinic
 and nicotinic receptors

  Cholinoceptive  neurons  excited by ACh in  the
 cortex,  caudate   nucleus,  thalamus,  and  hypo-
 thalamus show responses to ACh markedly different
 from those of Renshaw cells in the spinal cord, which
 are endowed predominantly with nicotinic receptors.
 ACh applied to the the latter neurons acts as a typical
 transmitter:  excitation  of Renshaw cells is almost
 instantaneous and lasts only as long as administration
 of  ACh (the late  and prolonged  response caused by
 activation of muscarinic receptors is  usually negli-
 gible).  On  the other hand, the response of neurons
 with predominantly muscarinic  receptors is slow  in
 onset and outlasts the application of ACh for many
 seconds  and even  minutes (Krnjevic 1974).  More-
 over,  ionic  mechanisms of ACh action  on  mem-
 branes with  muscarinic  receptors are substantially
 different  from  those with nicotinic  receptors.   In
 the former  instance,  ACh  reduces  the potassium
 current during  the late phase of the action potential.
 This reduction results in delayed repolarization after
 each action  potential  and in  increased  membrane
 resistance (Krnjevic 1974).  Thus, cortical neurons
 "primed" with ACh in subthreshold concentrations
 show  a  long-lasting lower threshold  to  incoming
 volleys or to application of excitatory amino acids,
 and a tendency  to  repetitive  firing  that can be
 blocked  by  antimuscarinic drugs.   Therefore,  the
 effect  of ACh  on neurons  with muscarinic receptors
 can be better described  as a modulator  rather than
 transmitter action.

  In view of the physiological role  of cholinergic
 mechanisms  at the cortical  level and their contri-
bution to the  genesis of negative SPs, it should be
 mentioned  that   ARAS-induced  facilitation  of
 neuronal  responses to visual stimuli can be mimicked
with micro-iontophoretic application of ACh. More-
 over, both facilitatory influences can  be  blocked
by  antimuscarinic  drugs (Spenlmann  1971).  This
 observation   and  the   negative  SPs  induced  by
increased  tonus of the  ARAS associated  with  the
 release of ACh suggest that the diminishing gradient
 of negativity toward the depth of the cortex should
 be  associated with an  overall reduced threshold to
 sensory volleys and enhanced firing of most cortical
 neurons.  This conclusion is not in agreement with
 Fromm  and  Bond  (1964), but agrees well with
 Rowland's   data 0974) and with  our own obser-
 vations  that  surface  positivity  as well  as  deep
 positivity,  accompanied  by alpha-like  oscillations,
 are always  correlated  with conspicuous inhibition
 of neuronal firing in freely moving animals during
 operantly  conditioned  behavior (Marczynski  and
 Karmos, this volume).

     Most likely, there are no substantial differences
 in the neurohormonal  mechanisms  of SPs that  are
 topographically  restricted to specific  sensory pro-
 jections, e.g., those  described by Gumnit and Gross-
 man 0^61) and Picton (this volume) over the auditory
 cortex,  compared  with  negative  SPs  distributed
 diffusely over larger cortical regions.  An enhanced
 release  of  ACh,  topographically restricted to  the
 visual  cortex, is observed in response to visual stimuli
 while   the  remaining cortical  regions  show only
 moderate enhancement  of ACh  release (Collier and
 Mitchell 1966, Neal  et at. 1968).  Our studies in cats
 show  that both types of SPs are almost equally sen-
 sitive  to systemic administration of antimuscarinic
 drugs (Marczynski, unpublished).

    Results  of pharmacological  studies of  CNV in
 man are in harmony with cholinergic mechanisms
 discussed above.  Atropine  (0.4 to 0.5 mg, i.m.)
 markedly reduced the  mean  amplitude of CNV,
 while  small doses of  nicotine,  taken in cigarette
 smoke, increased these  SPs in subjects  classified  on
 the  basis  of psychological  tests  as  extroverted
 (Thompson et al., this volume; Ashton et al. 1974).
 Since  cortical  Cholinoceptive neurons in man, like
 those  in  experimental  animals, are  most  likely
 endowed with  "pure"  muscarinic  receptors  and,
 therefore, are  not sensitive  to  nicotine, the locus
 of nicotinic action  is  probably  restricted  to  the
 midbrain part of the ARAS or to the pons as shown
 in  cats  and dogs (Kawamura and  Domino, 1969;
 Knapp and Domino,1962).

Catecholaminergic systems and SPs

     Norepinephrine  (NE)  pathways  originate  in
 tegmental nuclei and project to midbrain and fore-
brain structures, including  n. ventralis anterior (VA)
and n. reticularis (R) of the thalamus (Fuxe 1965).
 VA  and R regulate thalamocortical EEC synchroni-
zation  by generating rhythmic sequences of IPSPs
in neurons  of specific thalamic relay nuclei (Sasaki
et  al.  1976; Frigyesi 1972; Skinner, this volume).
Since  NE projections to most areas, including  the
thalamus and  cortex,  may be  regarded as inhi-
bitory   (Hoffer   and  Bloom 1976),  NE-mediated
modulation  of inhibitory neurons in  R and VA may
be  of  crucial  significance in  normal functioning  of

-------
Neurochemical Mechanisms in SP Genesis
                                              27
 the  sensory  gating  system and selective attention.
 NE  terminals, present in  all areas of neocortex,
 arise from cell bodies located in the tegmental nu-
 clei.   NE  axons run through  the  medial  forebrain
 bundle,  bypass  the  septal   region,   and  project
 diffusely to the cortex (Fuxe et al. 1968).

     Dopaminergic  (DA)  neurons  with cell bodies
 in  the midbrain also  contribute  to innervation of
 large  cortical  areas,  including the  frontal region
 (Fuxe  et  al.  1974,  lindvall et  al.  1974, Berger
 et  al.  1974).   These  DA projections may be as
 important  in  the  regulation  of integrative   pro-
 cesses as  the  DA projections  to the limbic and
 mesolimbic   areas   described    by   Ungerstedt
 (1971).

     At the cortical level, the function of NE and
 DA  terminals  may be  different.  NE  terminals
 establish   contacts   preferentially   with   apical
 dendrites  of pyramidal  cells and, therefore, have
 a distribution characteristic of nonspecific afferents
 (Fuxe  et  al. 1968).  On the other hand, DA ter-
 minals in  frontal  cortex, entorhinal  cortex, and
 pyriform  cortex  show highest  density in deeper
 layers, particularly  V and  VI.  Only  in the cingu-
 late  gyrus do  DA terminals show a distribution
 comparable  to that  of  NE  terminals-i.e., highest
 density  in  layers  1,  II,  and  III  (Lindvall et al.
 1974, Berger etal. 1974).

     Experimental evidence  indicates that, at the thal-
 amocortical level, catecholaminergic projections may
 influence the genesis of SPs by inhibitory action on
 inhibitory   interneurons  that are  responsible  for
 the phasing  of EEC  activity  and  positive SPs.  Al-
 ternatively,  these   projections  may modulate  the
 sensitivity  of excitatory  cholinergic  receptors of
 the muscarinic  type,  present at  the membrane of
 cortical pyramidal  cells  and  their apical  dendrites.
 Both mechanisms, which  are of great clinical import-
 ance, are dicussed below.

 Catecholaminergic   modulation   of  GABA-
 mediated hyperpolarizing inhibition

     Surface positive SPs  are believed  to reflect the
 hyperpolarization  of large pyramidal cells and their
 dendrites  (Creutzfeldt et  al.  1969; Marczynski  and
 Karmos, this volume).

     The following  observations support the conten-
 tion that  recurrent inhibition of cortical pyramidal
 cells (as well as thalamic relay neurons) is mediated
 by GABA-ergic  neurons, which, in turn,  are modu-
 lated by  inhibitory  catecholaminergic projections.
 These relationships are  illustrated in  a diagram of
 information   flow   in    the  mammalian  brain
 (Marczynski, this  volume).   (1) Micro-iontophoretic
application  of GABA  to  pyramidal  cells  lowers
membrane  resistance and  causes  inhibitory post-
synaptic potentials  (IPSPs)  whose   patterns  and
time course are virtually identical with spontaneous
IPSPs  or  those  elicited  by  antidromic stimulation
of the  pyramidal tract that activates the inhibitory
interneurons via axon  collaterals (Kinjevic 1974,
Johnston 1976).  (2) The highest  concentration of
GABA-forming neurons  (i.e.,  those  that contain
glutamine  decarboxylase) is found  in  layers  III and
IV (Albersand Brady 1959), where  inhibitory basket
cells are located (Marin-Padilla 1975).   (3) GABA is
released from the cortex only during EEG synchron-
ization (Jasper and  Koyama  1969) when inhibitory
interneurons show bursts of action potentials  that
coincide   with  inhibition   of   pyramidal   cells
(Steriade and Deschenes 1973).  By contrast, GABA
release from the  cortex  is totally suppressed during
strong arousal or electrical stimulation of the ARAS
associated   with  EEG  desynchronization.   During
that  time, inhibitory interneurons are suppressed
(Steriade and  Deschenes 1973).  (4)  lontophoretic
application of NE to cortical, as well as cerebellar,
neurons  and  electrical  stimulation  of  tegmental
nuclei from which NE  projections  originate (e.g.,
n. coeruleus) elicit  IPSPs with different character-
istics  from  those induced by  GABA  (Hoffer and
 Bloom 1976). (5) GABA releasing inhibitory neurons
(e.g., Purkinje cells of the cerebellum) are under power-
 ful  inhibitory control of  NE  projections  from  n.
coeruleus  from  which  a substantial  contingent  of
 both  neocortical and thalamic NE projections also
 arise (Hoffer and Bloom 1976).

     In conclusion,  despite  the lack   of  direct evi-
 dence, there are strong indications that the function
 of  catecholaminergic terminals in deeper  cortical
 layers is  restricted to  the process of disinhibition
 by suppressant action on the inhibitory interneurons
 in  the recurrent circuits responsible  for the phasing
 of EEG patterns. Since rhythmic phasing of neuronal
 activity in  the  cortex  and  thalamus  is associated
 with  hyperpolarization  of large neuron populations
 (Andersen and  Andersson  1968), and with surface
 positive SPs (Marczynski and Karmos,  this volume),
 catecholaminergic modulation of the GABA releasing
 system may play  an important role  in the genesis
 and topographical distribution of SPs.

 Catecholaminergic modulation  of thalamo-
 cortical recruiting responses

      Recurrent  hyperpolarizing inhibition in specific
 thalamic  relay  nuclei,  as well as the powerful in-
 hibition generated by   R and VA,  is most  likely
 GABA-mediated (Curtis and Johnston 1974).  Thala-
 mocortical  recruiting responses and EEG synchroni-
 zation depend  on the  integrity  of these  pools 6f
 inhibitory neurons (Skinner, this  volume) and there

-------
 28
  Is evidence that  a ciitcclioliimliicrglc component of
  the  ARAS modulates  their  function.   First,  IPSPs
  In   ihaltitnlc  relay   nuclei   arc  abolished  during
  increased (onus  of the  ARAS (Purpura  et al. 1966).
  At  the  same  time,  rhythmic  firing of inhibitory
  intcrneurons Is  suppressed (Steriade  and Deschenes
  1973).    Secondly,  spontaneous  arousal and elec-
  trical stimulation  of the ARAS are associated with
  a  large  positive  SP in R, which reflects  tonic hyper-
  polarizing   inhibition   (if this  nucleus  (Skinner,
  this  volume).   Thirdly, the  catecholaminergic com-
  ponent  of  the  ARAS  projects diffusely to R and
 VA  (Fuxe  1965).   Finally,   physostigmine-induced
 activation of the ARAS and  the resulting blockade
 of thalamocortical  recruiting  responses  can  be pre-
 vented  by  depletion  of  catecholamines (VanMeter
 and Karc/.mar 1971).

     Excitatory  influences  from  frontal  association
 cortex  impinging on inhibitory  pools of neurons in
 R, which, in turn, project to  specific thalamic relay
 nuclei,  determine  the  gating of sensory  input-a
 mechanism  which  may be the  basis for selective
 attention (Skinner, this volume). Sensory gating  re-
 quires very selective, well-limed excitation of neurons
 in  R which  can  only  be effective if there is a mod-
 erate  inhibitory  background  in this nucleus. The
 inhibitory  surround  is  most  likely  provided   by
 catccholamincrgic projections.


Catechi>laniinerf>ir  activation  of subcortical
cholinergic system

    In the cat preparation with  the brainstem tran-
sected at the midpontinc pretrigeminal  level, d-arn-
phi'iamme administered  syslemically increases ACh
output  fioni the  cortex (Nistri  et al.  1972).  This
effect  could be prevented  by septal lesion  or pre-
Ireatmenl of the animal with alpha-methyl-p-tyrosine,
a drug thai blocks synthesis of NE and DA.  Appar-
ently, such  activation of the  ACh system does not
depend on  the integrity  of brainstem ARAS nuclei
since  (1)  pentoharbital  does  not block  the effect
of  amphetamine  on ACh output even in  doses  that
block F.Hi  activation and  ACh  output  in response
to  electrical  stimulation  of  the ARAS and (2)
amphetamine  activates EEC patterns  and increases
ACh  output in  cats transected  at the pretentorial
level,  a preparation which excludes nuclei from which
the cholinergic component of the ARAS originates
(Nistri et al.  1972).   Amphetamine, however,  also
activates  KEG patterns in cats with septal lesions  in
which  no cortical  enhancement of ACh output  is
obseived. This pu//Jing observation can be explained
by  the previously  discussed possibility that amphet-
amine, by activating  the catecholaminergic system,
may  suppress the  GABA-ergic inhibitory pools of
neurons  in the thalamus  and cortex, thus  preventing
the emeigeiue of synchroni/.cd EEG patterns.
                                     MatT/ynskl

 Cholinergic  medinnlMiis  of ruibcorticnl  SPs


 Nenronal mechanisms

      Large  negative SPs have  been observed In the
 thalamus  and  other  suhcortical  areas  during the
 reaction time  forcperiod  (McC'allum  et al.   1973;
 Rebert  1972,1973b)  and  in  icsponsc  to  electrical
 stimulation  of the ARAS (Hayward  et al. 1966).
 These  SPs  seem  to  reflect  ARAS modulation  of
 transmission  in  sensory thalainic  nuclei.  Evidence
 from the  microelectrophoretic studies of the lateral
 geniculate  (Steiner 1968),  the  ventrobasal nuclei
 (McCance et al.  1968), and  the  medial geniculate
 (Tebecis  1970) show  that  the  effect  of  electrical
 stimulation of the ARAS  can be mimicked by ACh
 application, and  that  both  effects can  be  blocked
 by antimuscarinic drugs given systemically or applied
 topically.  These  studies suggest  that the cholinergic
 ARAS component exerts a strong facilitatory effect
 on sensory transmission. Since negative SPs in specific
 thalamic nuclei closely mirror frequency of neuronal
 firing (Rebert 1973b), it appears probable that these
 SPs reflect increased excitability of relay neurons in
 response  to afferent input, an effect that is likely to
 be  mediated by  activation  of  cholinergic  ARAS
 pathways.


 Participation ofglia cells

     Preliminary observations  (Krnjevic  1974)  show
 that iontophoretic  application  of ACh to cells iden-
 tified  as  glia causes  a slow onset and prolonged
 depolarization that outlasts the  application of  ACh
 by many  minutes.  At  the  same  time,  there  is an
 increase   in  membrane resistance  i.e.,  an  effect
 strikingly  similar  to that seen in neurons with mus-
 carinic  receptors.   If  this  is a  general property of
 glia  cell behavior,  it could help  explain  the genesis
 of  SPs in  both  cortical and subcortical areas  and
could  account  for certain  discrepancies between
neuronal  firing  patterns  and  SPs, e.g.,  those  ob-
served by  Rebert (1973b) in the  lateral  geniculate.


     Are glia cells  depolarized by ACh, or is glia cell
 depolarization  secondary to  an  increase in extra-
 cellular potassium  caused by  neuronal firing?   The
 increase  of membrane  resistance in both glia  cells
 and  neurons  upon application  of ACh  (Krnjevic
1974)  suggests  an  independent sensitivity to ACh.
 Rowland  (1974)  has also suggested that the depo-
larization  of glia  cells in  response to  ACh release
may  account  for  negative  SPs without  obvious
correlation with  grossly integrated  multi-unit  acti-
vity  in some  regions  of the brain.   Somjen  (this
section)  discusses  in   detail   the  possible  contri-
bution ofglia cells  to SPs.

-------
Neurochemical Mechanisms in SP Genesis
                                             29
Genesis and pharmacology  of positive  epi-
cortical SPs

Common  features  of surface positive  SPs

    Most  positive SPs in man, such as the  P300,
P450,   skilled   performance    positivity   (SPP)
described  by Papakostopoulos et al. (this volume),
and  detection  positivity  (DP)  of  Cooper  et  al.
(this  volume) increase  in  amplitude  with  greater
a priori uncertainty of  outcome  of performance
and/or  greater  difficulty, in correct  event  recog-
nition.  The  unusually large SPP and DP apparently
reflect  the demanding task involved in  the eliciting
paradigm.   If the a priori  uncertainty  is held con-
stant,  positive  SPs in  most instances increase in
proportion to the a posteriori  reduction of  uncer-
tainty, i.e., when there is  less  doubt that the  per-
formance  was successful (Ruchkin and  Sutton, this
volume).  In  the cat, the level of a priori uncertainty
also plays a major role  in the  emergence of reward
contingent positive  variation (RCPV) recorded over
the pane to-occipital cortex (Marczynski et  al. 1969,
197la). Presentation of  milk triggers an augmented
RCPV when  the  reinforcement  schedule  for  bar
pressing is variable (i.e., unpredictable) in comparison
to a schedule in which  all  bar presses are rewarded
(Marczynski  and  Burns  1976). RCPV  is, however,
more  sensitive to a posteriori reduction  of uncertain-
ty. For  instance, as  shown in Fig.  1, when an animal
finds  that there is a perfect match between the ex-
pected and actual result of a lever press, RCPVs are
of greater amplitude. If, however,  the animal expects
pure milk, but receives water or adulterated milk in-
       hoo/tv

   2 sec

  Fig. 1. The effect of change in quality of reward on the amplitude and duration of the reward contingent positive
  variation (RCPV) in the cat trained to press a lever for water (W) and milk (M) rewards.  A continuous record.
  Epidural electodes were used, and the positivity of '3' and '1' with respect to the reference '2'is downwards  In
  thJsDe^r°fe "SfMSWi the mirrorjeversal Patterns of RCPV are explained by a decreasing potential gradient
  ItSSZS^^SSj^          RCPV ^ aSS°Ciated ^ ™ ClSeC high ^P^^chronization.

-------
30
                                     Marczynski
stead,  the  electrocoiticogram (K'oC.)  synchruni/a-
tion  (Sternum and  Wyrwicka 1967)  and associated
R( TV  are suppressed or blocked (Marc/ynski et al.
    There  is  suhstanlial  evidence  tliat  all longer-
duration surface  positive  SPs  result  from hyper-
polan/ing  inhibition  of  pyramidal  cells  and  the
electrcilonic  spread  uf  IPSPs  to  apical  dendrites
(Creut/k'ldt   e!   al.  196')).     Ilyperpolari/.ation
lesults  tiom  recuiieiit and,  perhaps,  leed-lorward
inhibitory  phasing  o|  nemonal  activity,  as  indi-
cated by the  relationships between  inhibitory  neur-
unal discharge, the  time  course  of  IPSPs (Steriade
and  Deschenes 11'73), and  firing patterns of neurons
dining  emergence of  the RCPV (Marc/ynski and
KarniDs, tins volume). The close correlation between
cortical release of (JAHA  and level of cortical  hl-.G
syiii hmni/ation (Jasper and Koyama  1%9) supports
this  interpretation. Since  inhibitory  neurons in  the
coilex  and  thalamus are themselves subject U>  in-
hibitory modulation  by the ARAS (vide supra),  the
general  ouulusioii can tie  drawn that all aloremcn-
I loiied positive Si's occur during a transient hut  pow-
e i fnl  suppression of the  ARAS  by  influences orig-
inating  in  the  torebrain  (Marc/ynski, this volume).
HecUophysiotogical  evidence  for ARAS suppression
during  emergence of the  R(TV  is convincing  (Mar-
c/ynski and H;n.keft  l''6°, llackett  and Marczynski
ll'71 , Marc/ynski et  al. ll>71a).

C'hulincrKic   mechanisms  in  surface  positive
     I he physiological (onus of the cholinergic ARAS
component  plotting to the  th.ilamus and cortex is
necessary foi  the  emergence  ot  RCPV.   Doses  of
scopolanune or  atiojiine,  whicli  do  not  interfere
with operant behavior, block  the  RCPV, which can
be  restored  promptly by  administration of  physo-
sUi'imne (Tig. 2 top). I he most important aspect ot
this  antimtiscarimc blocking action is that the alpha-
like  bursts  associated with RCPV become irregular
and  'Yhoppy," suggesting that  SP  production  is
directly  related  to  the  rhythmic  hyperpolari/.int!
phasing ot neuioiial activity. Moreover, the duration
ut  choppy  alpha-like activity  is  not  significantly
dittaent  t'lom  control response (Fig.  -'). Hence, it
ian  he  concliuled  that aiitimuscaiimcs do not block
the  "pmiuuy" effect of  toward  (or  goal  achieve-
ment).  but  meiely  interfeie  with the "execution
<>t  ihal.niHKoitical  synclmmi/ation and  emergence
nl RCPV (.Vluiv/ynski I 97 I).

     The  phasing,  theory  ot  neuronal  activity  ex-
pl.iiinuj'  alpha  inechanisuis, which  is  based  on re-
cnnent inhibition (Andersen  and Andersson  1968),
implies that a certain level  of synaptic  drive is neces-
sary  to mutate  the  functioning, of  the recurrent
 inhibitory circuits.  Taking into account the facil-
 itatory  cholinergic  influences  on  sensory  trans-
 missions at thalaniic and cortical levels, one can pro-
 pose  that antunuscarinics, by  reducing  the cholin-
 ergic  ARAS influences, lower synaptic  drive to  a
 level  that  is incompatible  with normal operation  of
 recurrent inhibitory circuits (Marczynski and Burns
 1976).  Blockade  of alpha activity by antimuscari-
 nics in man (Longo 1966)  and blockade of alpha-
 like activity associated with  K-complexes normally
 triggered   by sensory  stimuli  during sleep  onset
 (Marczynski, this volume)  are in agreement with our
 interpretation of  cholinergic  mechanisms.   Further
 supporting  evidence  has  been  presented  elsewhere
 (Marc/ynski  and Burns 1976; Rick and Marczynski
 1976; Marczynski and Karmos, this volume).

    Assuming  that the proposed interpretation  of
 cholinergic  mechanisms is  correct, one can  make  a
 second  assumption  regarding  the  role of synaptic
 drive  in  the  topographical  distribution of  positive
 SPs.  Upon  completion of task  performance, slack-
 ening or suppression of the catecholaminergic ARAS
 component  allows   inhibitory   interneurons   to
 function.   Only  those  thalamocortical projections
 and  corticocortical  pathways  that  received  the
 strongest   synaptic   drive   can  be  expected   to
 activate   recurrent  inhibitory   circuits  promptly.
 It is,  therefore,  not surprising that the RCPV in the
 cat depends  on visual input and occurs over primary
 and  secondary  visual  projections  (Marczynski  et
 al.  1971a,b).  Similarly,  the SPP of Papakostopoulos
 et al. occurs over  the motor and somatosensory cor-
 tex, while the DP of Cooper et al. (this volume),
 which requires intensive visual searching but only  a
 minimal motor response, occurs over the  vertex and
 midline occipital cortex. On the basis of  anatomical
 and electrophysiological evidence (Frigyesi 1972), it
 is probable that  the emergence of SPP depends on ac-
 tivation of motor  cortex, as well as strong synaptic
 drive  from the cerebellum, which reaches the cortex
 via the  bracchium conjunctivum and n. ventralis lat-
 eralis. Pyramidal neurons of the cortex, in turn, send
 axons  back  to n. ventralis lateralis as well as axon
 collaterals  to intracortical  inhibitory circuits and to
 R.  The latter,  in  turn, project inhibitory axons to
 n. ventralis lateralis (Fig. 4).  Immediately  after post-
 performance  slackening  of   the  inhibitory  ARAS
 component (catecholaminergic?), both thalamic and
 intracortical  recurrent  inhibitory  circuits may  be
 activated, resulting in  a positive  SP and  alpha-like
 I-.F.Ci oscillations.

Cholinergic activation   of the catecholamin-
ergic  system

    Antimuscarinic  drugs  block  RCPV  and alpha
activity  in humans.  Physostigmine  restores RCPV
and  associated  alpha-like  bursts,  but also  blocks

-------
Neurochemical Mechanisms in SP Genesis

           CONTROL             SCOPOL.
                                                                                            31
                                                                              PHYSOST.
                                                                                 ,I.M.. 35min.
                                                                                     W
                                   ,,,  /I'          ttLU
                               HMMNQifr
                                    • i   ir                 ' if
                                                                       *•
        M-SCOPOL.
    60 /ng/kg, I.M., 30 min
                                                                                SCOPOL.
                                                                          50 A/g/kg. I.M., 30 min.
PHYSOST. 60 jug/kg. I.M.. 30 min
                                                                         2 sec
Fig. 2.   Top:  The effect of scopolamine HBr and physostigmine salicylate on reward-induced EEC synchron-
ization  and associated RCPV, recorded  over the primary visual cortex (posterior marginal gyrus or PM) with
reference to the subjacent white matter.  The dc  record was filtered out to half-amplitude response at 3 c/sec.
This channel is integrated in the first trace (INT) in which two full-scale upward deflections are equal to one unit
of RCPV.  As marked by dotted lines in the upper right record, only those INT deflections were counted that
occurred in the time interval between  the reinforced (REINF) and  the subsequent rewarded or nonrewarded
(NOR)  lever press.

Bottom:  48 hours later, the same cat was  pretreated with methylscopolamine HBr; physostigmine suppressed
the RCPV, which  subsequently could be partially  restored  with scopolamine  (From:  Marczynski  1971J. For
further explanation, see the text.
these responses if given to animals whose peripheral
muscarinic receptors have  been  "protected" with
methylscopolamine, which does not penetrate through
the blood brain barrier (Marczynski 1971). In some in-
stances, as shown in Fig.2 (bottom right), RCPV and
alpha-like responses can be restored with scopolamine.
Better results have been obtained with chlorpromazine,
a drug known to block dopaminergic and adrenergic
receptors. Hence, it appears that  physostigmine and
perhaps  other acetylcholinesterase inhibitors, by in-
                                                   creasing ACh levels, can  activate catecholaminergic
                                                   systems. Since similar blockade of EEC responses has
                                                   been  observed after systemic administration of nico-
                                                   tine,  an  effect  reversible  with  chlorproma/ine
                                                   (Marcznski,  unpublished), one  can  postulate that
                                                   cholinergic activation of catecholaminergic systems
                                                   results from  action on nicotinic receptors. The latter
                                                   are primarily located in ARAS nuclei at the brainstem
                                                   level  (Kawamura and  Domino  1969,  Knapp  and
                                                   Domino 1962). The mechanism by which activation

-------
 32
                                                                 Marczynski
 of the catecholaminergic system may modulate and
 block  thalamocortical inhibitory  and synchronizing
 circuits has already been discussed.
 Ginical  implications
 chemical mechanisms
    of  impaired  neuro-
 Catecholaminergic
 brain dysfunction
hypothesis  of minimal
     Selective attentional  deficits  are characteristic
 of children with Minimal  Brain Dysfunction (MBD).
 The role of n. reticularis thalami (R) in gating sensory
input and  the putative  role of catecholaminergic
projections in  regulating  the inhibitory  background
of this nucleus have  been reviewed above. It  may
be hypothesized that the inhibitory background in R
of MBD children  is deficient, a condition which is
likely to impair the selective gating of sensory input.
Moskowitz and Wurtman (1975), for instance, obser-
ved that the catecholaminergic system in  MBD child-
ren seems  to be impaired which may lead to uncon-
trolled functioning of  the thalamic gating system.
Moreover, MBD children have an unusually low thres-
hold to photic driving which sometimes leads to myo-
clonus. These children also tend to be hypersensitive
                                               CONTROL
   M«to^^
                      ATROPINESULFATE
                                                    0.8 mg/kg, I.M., 25 min.
     (l(jM4%i^HA^
                  PHYSOSTIGMINE SALICYLATE
                                                            0.1 mg/kg, I.M., 25 min.
     llT^t^UiM^^
                                                                                       1444,
                                                                  5 sec

Fig. 3. Comparison of control reward-induced EEC synchronization (top) with that recorded  25 min  after
administration of atropine sulfatefO.8 mg/kg, i.mjin the cat trained to press a bar for 1 ccofmilk. LICK signals
caused by lapping and licking; NOR  nonrewarded bar press; REINF rewarded bar press; PM posterior marginal
gyrus /with reference to the anterior ectosylvian gyms).  The heavy horizontal bar below each EEG response
marks its duration  from the first to the last EEG wave that, if measured peak-to-peak, exceeded 100 pK. The
numbers after each EEG response tell how many oscillations occurred that exceeded lOOuV,  Note that the dur-
ation  of the "choppy"  EEC responses is not significantly different from the control. Note also that physos-
tigmine promptly restored the responses (bottom). For further explanation, see the text.

-------
NiMiroehemicnl Mechanisms in SP Genesis
to amphotnmine and mothylphenldate,  as tested  by
the blocking  action of these drugs on EEC photic
driving and  myoclonus (Shetty  1971).  Hypersensl-
tivity  to amphetamine Implies that catecholamlnergic
pathways and terminals are  damaged. Furthermore,
amphetamine and  methylphenldate usually  amelio-
rate MBI) symptoms.
 Fig. 4.   Pathways most  likely  responsible for the
 emergence of the skilled performance positivity (SPP)
 of Papakostopoulos et al. The arrows and the small
 bars denote  the excitatory  and inhibitory (hyper-
 polarizing) synapses respectively. nR nucleus reticu-
 laris tlralami: VL n. ventralis lateralis; BC brachium
 conjunctivum, conveying input to the  VL from the
 cerebellum: MCx motor cortex.  Filled circles denote
 inhibitory neurons. Note  that the rhythmic phasing
 of ncuronal  activity and  the emergence of the SPP
 are possible only when  the inhibitory neurons are re-
 leased from  tonic inhibition exerted by  the ARAS
 (ascending reticular  activating system;  probably  its
 catecholaminergic component}. For further explana-
 tions,  see the  text, (Modified from Frigyesi 1972).
     Stamm and his colleagues (personal communica-
 tion) studied the Nl component of auditory evoked
 potentials in normal and MBD children. The mean Nl
 amplitude  uifference between attend and non-attend
 conditions was 44 percent for the normal group, but
 only 14 percent for the MBD group.  Karrer and his
 colleagues  (this volume)  found  that  children with.
 cognitive difficulties show an increased surface posi-
 tive SP over frontal and central cortex prior to and
 during motor response, both observations support the
hypothesis  that  "sculpturing" of specific excitatory
spatio-temporal patterns in R and other pools of thai-
amlc inhibitory neurons Is Impaired, leading to a low
signal-to-noise ratio of input to  the cortex (Skinner,
this volume).


Cholinergic  interpretation   of dopaminergic
hypothesis of schizophrenia:  a new working
hypothesis

    Relationships  among catecholaminergic  systems
and  thalamocortical  inhibitory  pathways  in MBD
children  may be  opposite  to those in adults with
acute  anxiety,   schizophrenia,  or  amphetamine-
induced  schizophreniform  psychosis.   That is, in-
creased tonus of catecholaminergic systems  is likely
to produce tonic inhibition of R and impaired thala-
mocortical  EEC synchronization.  This condition is
incompatible with effective gating of  sensory input
and leads thus to a state of confusion.  Thalamocor-
tical relationships are discussed  further in the "Par-
simonious Model of Mammalian Brain"(Marc/ynski,
this volume).

     During the last seven years, considerableil.ua have
accumulated indicating that a hyper function  of the
DA system  and/or increased sensitivity of  DA post-
synaptic receptors  may  be  responsible  for many
symptoms of schizophrenia  (Meltzer and Stahl ll>76).
Moreover, Libet and his colleagues (this section) have
discovered  that  DA has a unique and powerful poten-
tiating effect on  slow  muscarinic  KPSPs in sympa-
thetic ganglia. As already mentioned, pyramidal cells
of cortical  layers IV and V  are endowed with "pure"
muscarinic receptors.  Stimulation of these receptors
with ACh triggers EPSPs characterized by slow onset
and long duration (Krnjevic 1974).  The specificity
of this  action  is indicated by the fact that these
 EPSPs are  associated with an increase  in membrane
resistance,  in contrast to the brisk and short-last ing
EPSPs mediated by nicotinic receptors and associated
with a sudden drop of membrane resistance.


    Following ACh application,  neurons in the stria-
turn, a nucleus known to be profusely  innervated by
DA  projections, show  typical  muscarinic  responses
comparable to  those seen  in  the cortex. Therefore,
they may provide a clue regarding the  possible inter-
action between DA and ACh systems. Indirect  evi-
dence  for such  a  heterosynaptic interaction at post-
synaptic sites comes from intracellular recording of
striatal  neurons.  Electrical  stimulation  of  the  sub-
stantia  nigra activates  DA  projections to  the stria-
turn and causes release  of ACh in  this  structure (Por-
tig and  Vogt  1969). This  dual effect is probabK
caused  by  the  proximity of the  ARAS cholinergic
component  described by Shute and  Lewis  (l^e>7).
Electron microscopic  study of  DA terminals in  the

-------
  34
                                    Marczynski
 striatum shows that DA  is  likely  to be  released
 through  narrow  pits directly into  relatively large
 extracellular  spaces, allowing  for  diffusion of  the
 transmitter away from  the terminal (Tennyson et
 al. 1974). These morphological characteristics imply
 that a heterosynaptic interaction at postsynaptic sites
 is possible, an assumption supported by an intracellu-
 lar study of  striatal neurons by Hull et al. (1970).
 These investigators  found  that a single volley of
 electrical stimuli applied to subs tan tia nigra, or its
 close vicinity, produced in most neurons a short-last-
 ing hyperpolarization (DA effect ?), followed by an
 exaggerated depolarization lasting 30 to 40  seconds.
 This  phenomenon, which resembles the paroxysmal
 depolarization shift  of neurons in epileptic foci, was
 not observed in other structures.

    Although  the information  concerning the  intri-
 cate relationship between DA and  ACh terminals in
 the cortex is not yet  available, the possibility of inter-
 action  between these  two   neurotransmitters  at
 muscarinic receptor   sites in schizophrenics should
 be  seriously considered  as a   working hypothesis
 because all the necessary elements are present in the
 cortex, including the DA receptor, i.e., DA-sensitive
 adenylate cyclase (Bockaert et  al. 1977).  Consider-
 able  evidence  supporting this hypothesis is already
 available. First, there is little  doubt  that increased
 (onus  of the DA system plays  a crucial role in the
 emergence of most schizophrenic symptoms  (Meltzer
 and  Stahl 1976).   Secondly, several  electrophysio-
 logical observations indicate that exaggerated depol-
 arization of cortical neurons and their apical dend-
 rites may be associated with  schizophrenic symptoms.
 For   instance,   schizophrenics   show   considerable
 impairment  of amplitude   recovery  of auditory,
 somesthetic,  and  visual  evoked  potentials if the
 stimuli are presented at short intervals (Shagass 1976).
 Thirdly,  acute or chronic schizophrenics with florid
 symptoms show significantly lower CNV amplitude
 than normal adults (Dongier  1973).

    Later components of cortical evoked potentials
 are produced  by postsynaptic  potentials and  their
electrotonic  spread  to apical dendrites (Creutzfeldt
et al.  1969).  Tonic depolarization of apical dendrites
 by  cathodal current (Purpura 1967)  or by appli-
 cation of ACh  (Sgg et  al.  1965) reduces  the  slow
 negative  component  of evoked potentials.   It is
 likely  that CNV amplitude is  similarly diminished
 relative  to  background  depolarization of apical
dendrites during hyperarousal.  Hence, modifications
 of the CNV paradigm,  making it more demanding-
e$.,   by  introducing  distracting  stimuli-lead  to
 reduction of CNV amplitude (Tecce and Cole 1976).
 Administration  of  amphetamine  to   experimental
 animals  reduces negative Sft  over frontal cortex
 elicited by  conditional auditory stimuli, an effect
 that  can be reversed by haloperidol or barbiturates
 (Pitch, this  section).  Likewise, the  administration
 of fluphenazine to schizophrenics with florid symp-
 toms  considerably enhances   CNV  amplitude and
 improves the clinical picture (Tecce and Cole 1976).
 On the other hand, when no drug-induced or disease-
 related hyperarousal is present, one would expect a
 reduction of CNV amplitude to parallel the slacken-
 ing  of ARAS tonus  or blockade of postsynaptic
 receptors.  Indeed, Thompson  et al.  (this  section)
 found  that  atropine  or a DA antagonist, metoclo-
 pramide, decreased CNV amplitude in normal volun-
 teers.   In conclusion, the  above physiological and
 pharmacological considerations support  the theory
 that  CNV  amplitude  is monotonically  related to
 selective attention, but nonlinearly related (inverted
 U) to increasing ARAS tonus and arousal level (Tecce
 and Cole 1976).

    Finally,  the prolongation of the spiral aftereffect
 (SAE) described  by Herrington and Qaridge  (1965)
 appears to  represent one of the most characteristic
 phenomena associated  with schizophrenic symptoms
 (Abraham and McCallum 1973).  A recent study of
 SAE in schizophrenics and controls revealed that its
 duration faithfully reflects the actual clinical state as
judged  by Schneider's first-rank  symptoms.   This
 correlation was present even when amplitude reduc-
 tion or prolongation  of the CNV was marginal or
 absent (Abraham and McCallum,  personal communi-
 cation).

    Little  is known  about the neurophysiological
 basis of SAE  This phenomenon may, however, be
 related to transient entrainment of neuronal circuits
involved in motion detection.  A classical example of
such entrainment  is the "waterfall effect."  If one
stares at a waterfall and then  looks away, a part of
the background corresponding  to the size  of the
waterfall appears  to  move  upward,  while the rest
of the landscape remains stable.  This illusion is not
caused  by eye movement,  but by entrainment of
neuronal circuits  most likely involving edge and/or
motion detector cells in the striate cortex.

    It  is quite probable that SAE duration is regu-
lated by a cholinergjc  mechanism of  the muscarinic
type.    Neurons  which  normally discharge  briefly
(0.5 to 1.0 sec)  in  response  to  a sensory volley,
respond to the same input after iontophoretic  appli-
cation  of ACh with a continuous burst of action po-
tentials lasting  10  to 20 sec (Krnjevic 1974).  Thus,
a primary role  of  the cholinergic ARAS component
projecting to the cortex is  to  amplify in time the
transient sensory  input, a  process believed to  be
necessary for the emergence of  conscious experience
(Tibet  1965, Kmjevic 1974). Hence, prolonged SAE
in schizophrenics is likely to reflect  a pathological
potentiation  of the normal  function  of the cholin-
ergic  ARAS component, an  action   that may  be
tentatively  ascribed  to  DA modulation of post-
synaptic muscarinic receptors.

-------
Neurochemical Mechanisms in SP Genesis

 Summary

     The enormous  complexity of dynamic inter-
 actions  among neuronal  ensembles and  chemical
 transmitters  or  modulators  necessitates  a multi-
 disciplinary approach to the study of brain function.
 This review attempts to integrate  many anatomical,
 electrophysiologjcal,  and  pharmacological  aspects
 of the genesis  and  functional significance of  EEC
 patterns, evoked  potentials, and  slow potentials.
 Evidence is presented  that BEG  desynchronization
 and slow  negative  potentials are  mediated by  a
 concerted  action  of  the  cholinergic and  catecho-
 laminergic  ARAS  components.    The cholinergic
 component depolarizes neuronal dendrites,  reduces
 the  tiring threshold  of large  populations of neurons
 in specific thalamic  nuclei and cortex,  and prolongs
 their discharge to incoming volleys.  Simultaneously,
 the  catecholaminergic  ARAS component  appears
 to  block  the  function  of  GABA-ergic  recurrent
 and/or  feedforward  inhibitory circuits responsible
 for  hyperpolarization of large populations  of neu-
 rons.  Conversely, bursts of alpha-like EEC patterns,
 surface  positive  SPs and  most positive SPs in the
 specific   thalamic   nuclei  seem   to  result  from
 slackening in  the tonus  of catecholaminergic  pro-
 jections, thus  allowing the function of GABA-ergic
 hyperpolarizing circuits, the emergence of alpha-like
 EEC patterns, and positive SPs.
     Antimuscarinic  drugs (atropine  or scopolamine)
 block at the thalamocortical level not only the norm-
 al cholinergic  facilitation (amplification in time) of
 sensory input, the negative SPs and EEC desynchron-
 ization, but also abolish opposite phenomena, such as
 the bursts of EEC alpha-like patterns and positive SPs
 in specific thalamic nuclei and cortex. These dual and
                                             35
seemingly contradictory actions of these drugs can be
accounted for by two assumptions  based on experi-
mental results:  (1) the functioning of the recurrent
inhibitory GABA-ergic circuits depends on a certain
level of excitatory synaptic pressure and its physio-
logical amplification by the cholinergic ARAS com-
ponent; and (2) this pressure can either be converted
into a desynchronized EEC pattern  and negative SPs
or into a hypersynchronized EEG alpha-like pattern
and positive SPs, depending upon the functional state
and  "readiness" of the GABA-ergic circuits. If the
latter are tonically inhibited by  the ARAS (catecho-
laminergic projections?) e.g. during arousal, the ex-
citatory  synaptic pressure can  only result  in  EEG
desynchronization and negative  SPs in specific thal-
amic nuclei and cortex. However, if the GABA-ergic
system, e.g.  during relaxed wakefulness,  is released
from tonic  inhibition  (slackening  of the  catecho-
laminergic ARAS component?), the excitatory  syn-
aptic pressure in specific  thalamic nuclei and cortex
can  be readily  converted  into hyperpolarizing inhi-
bition, bursts of high voltage alpha  activity and posi-
tive SPs in specific thalamic nuclei and cortex.

     The  issues discussed  here  have several clinical
implications. For instance, it appears that a moderate
(catecholaminergically  mediated) inhibitory   back-
ground in reticularis thalami (reflected as a positive
 SP in this structure)  is necessary for the gating  of
sensory input by this nucleus.  This function is likely
to be impaired in children with MBD.  On the other
hand, an excessive hyperpolarizing blockade of this
nucleus by catecholaminergic projections could render
it  nonresponsive to cortkofugal   modulation and
cause  an overload  of sensory input and confusion-
characteristic  symptoms  of   schizophrenia.    To
account for  changes in SPs in schizophrenic patients,
heterosynaptic interaction between dopaminergic and
cholinergic systems is postulated.

-------
REWARD CONTINGENT  POSITIVE  VARIATION  (RCPV)
AND PATTERNS OF  NEURONAL  ACTIVITY IN THE
VISUAL CORTEX OF THE CAT


T. J. MARCZYNSKI AND G. KARMOS
Department of Pharmacology, College of Medicine, University of Illinois, Chicago, IL, U.S.A.
    The relationship between the  firing patterns of
cortical and subcortical neurons and slow potentials
(SP) has been studied by several investigators. Fromm
and  Bond  (1964) found that in the  enclphale isole*
cat preparation, cortical neurons in the visual pro-
jections  fire more frequently during surface positive
EEC waves and cease  firing during  negative waves.
These relationships were reversed, however, when the
visual cortex became more  negative  with respect to
the  frontal sinus.  Rebert (1973b)found that flash
stimuli produce in the lateral geniculate body a slow
negative SP associated with  increased neuronal firing.
In the chronic cat preparation, under conditions of
nonoperant training,  Rowland  (1974)  found that
surface negative SPs over the visual cortex were asso-
ciated with increased multiple unit activity.

    Clemente  et al.  (1964) observed postreinforce-
ment synchronization (PRS) of EEC in cats trained
to press a lever for milk reward.  Marczynski et al.
(1969, 1971a) have  shown that PRS  (7 to 9 c/sec)is
always associated with an epicortical SP which has
been designated  reward  contingent positive varia-
tion (RCPV).  This phenomenon, like PRS (Sterman
and Wyrwicka 1967), depends on the quality and de-
sirability  of  reward. The PRS-RCPV responses are
topographically restricted to the primary and second-
ary visual  projections and a part of the association
cortex (the posterior marginal and the suprasylvian
gyri). The RCPV usually  outlasts the PRS burst  by
1  to 2 sec (Marczynski et al. 197la, Rick and Marc-
zynski 1976).

    In the present  study,  the relationship between
patterns of single-unit activity and  the PRS-RCPV
phenomenon was investigated. Since PRS-RCPV  re-
sponses depend on unpatterned light input, even in
cats trained in the dark (Marczynski et al.  1971b,
Rick and  Marczynski 1976), the effect of ambient
light was also tested.

Methods

     Eight  adult cats were trained to press a lever for
 0.8  ml of milk reward. Ag/AgCl electrodes were im-
 planted epidurally over the posterior marginal (PM)
 and anterior ectosylvian gyri in three cats under pen-
 tobarbital anesthesia. Since the latter electrodes were
 relatively "neutral"  during  PRS-RCPV  responses
 (Marczynski et al. 197la), they served as reference.
 In two other cats, the reference electrode was placed
 in the white matter of the PM gyms approximately 6
 mm below the surface and 3 mm lateral from the
 epidural electrode.  The subcortical  reference con-
 sisted  of a 0.5 mm Ag/AgCl pellet, enclosed in a glass
 tube, which was closed with  agar.  In the remaining
 three cats, no dc recording was attempted: PRS and
 unit activity were monitored using standard epidural
 stainless steel electrodes in the cats. In all eight cats,
 single*  or multiple-unit  recordings from  PM  gyms
 were obtained by means of bundles  of "floating"
 platinum-indium or stainless steel wires, 15 to 35/jrn
 in diameter, insulated except at the tips which were
 cut  with scissors.  The bundles of wires,  stiffened
 with sucrose  solution, were  implanted through  a
 guide  cannula, which was tangentially oriented and
 made  out of a polyethylene or stainless steel tubing
 0.5 mm in diameter. The tip of the cannula was cut at
 approximately a 20° angle, and its orifice  was pro-
 vided  with a 1-mm rim. Through a small opening in
 the  dura, the rim was placed above the pia mater, and
 the  cannula was fixed to the skull. Wires and epidural
 electrodes  (including the  white  matter reference)
 were connected to separate miniature sockets. Unit
 activity  was  recorded with a miniature high-input
; impedcnce preamplifier  (Sherry et  al.  1975), with
 wide frequency band responsiveness-(1.0  to 1.5 Hz).
 Notch filters and Grass 7-PS 11 amplifiers permitted
 simultaneous  recording  of unit activity  and slow
 wave  ECoG patterns from the  electrode tip. Of 52
 "floating"  wires, 10 yielded good single or multi-
 ple  unit recording for  2  to  9  days in unrestrained
 animals. Data  were  stored  on magnetic tape. An
 amplitude discriminator was  used for separation of
 multiple unit activity. Frequency histograms of unit
 firing  were  obtained with a Computer  of Average
 Transients Model 1200 and displayed by means of an
 X-Y plotter.

-------
RCPV and Neuronal Activity in Visual Cortex

Results

    All eight cats showed typical PRS responses  in
the presence of ambient  light. In the dark, consum-
matory responses were not associated with any sig-
nificant changes in ECoG or SP. In all cats implanted
with Ag/AgCl  electrodes, PRS was always associated
with a typical  RCPV,  as shown in Fig.  1A(S: surface
lead referred to white matter). Steep negative waves
of 7 to 9 c/sec recorded with the intracortical micro-
electrode (D)  were always in phase with bursts  of
unit activity. During a fully developed PRS response,
these  intracortical  negative  waves  were  abruptly
terminated by  much longer duration (approximately
100 to 130 msec) positive  waves.  Surface negative
waves  were, in most  instances,  approximately 20 to
35  msec out  of  phase with  respect to intracortical
ones.  The  duration  of  surface  positive waves was
comparable to that of intracortical waves.

     In the presence of light,  all 10 successfully stud-
ied  neurons showed  markedly  reduced firing rate
                                                         37
            during consummatory responses associated with the
            PRS-RCPV phenomenon as compared to the time
            period immediately  prior to, during, or immediately
            after the lever press.  As shown in Fig. IB (L: fre-
            quency histograms obtained in  the light), the higher
            amplitude unit H and the small amplitude unit S were
            maximally suppressed during the time period heuvtvn
            2 and  4  sec after  reinforcement (R), i.e., during
            the consummatory responses and the  occurrence of
            PRS-RCPV.  Since cats  performed on a 6-sec fixed
            interval  reinforcement schedule,  the  histograms en-
            compassed approximately  two  consummatory  re-
            sponses  and,  therefore,  were  almost symmetrical on
            both sides of reinforcement.

                Basically, two types of units were encountered:
            one type  showed no significant change in firing rate
            in the dark (D), i.e., when PRSRCPV  responses were
            suppressed,  but  showed  strong  post-reinforcement
            inhibition in the  light (the histograms S-D and SL,
            respectively). The second type of unU, illustrated by
            neuron H, showed increased firing rate approximately
                 REWARDED BAR PRESS
                                                LIGHT
                             +600
R                    R
   TIME, milliseconds
 Fig.  I A: Patterns of unit activity in the visual cortex, simultaneously recorded intracortical ECoC from the tip of
 the same electrode (D) and surface ECoG(S) during lever pressing performance for 0.8 ml of milk reward The ref-
 erence electrode placed in the white matter approximately 3 mm laterally from the intracortical and epicortical
 electrodes. B: frequency histograms for higher amplitude (H) and smaller amplitude (S) units during performance
 of the  cat in the presence of light (L) and in the dark  (D). Numbers above the unit channel show discharges for
 unit  H after reinforcement in  light and dark.  Bin width in histograms:  100 msec.  Each histogram represents 10
 rewarded (R) lever presses.

-------
38
                                                                              Marczynski and Karroos
 1.5 sec prior to, during, and 1.5 sec after reinforce-
 ment, both in the presence and absence of light (his-
 tograms  H-L and H-D, respectively).  Both types of
 units, however,  were strongly inhibited during con-
 summatory  responses  in  the  light i.e., during the
 occurrence of PRS-RCPV.  As shown by the  dashed
 line  in  L histograms,  firing rate during PRS-RCPV
 reached markedly lower levels than the average firing
 rate observed during a relaxed state after satiation.
    During sleep onset after satiation, ECoG and SP
patterns were indistinguishable from PRS-RCPV re-
sponses. In contrast to operant behavior, however,
the characteristic 7- to 9- c/sec  bursts of discharges
and an overall decreased firing rate were observed in
all units both  in the dark and light.

Discussion

    The  PRS phenomenon and associated suppres-
sion of unit activity in the cortex seem to represent a
typical example of phasing  of  neuronal  discharges
based on recurrent  inhibition as  postulated for thala-
mocortical relationships by Andersen and  Andersson
(1968). The phasing theory that  explains recruitment
of larger populations of neurons into alpha-like dis-
charges implies that a certain level of synaptic pres-
sure is necessary to initiate and drive the phasing cir-
cuitry. Hence, neurons that show sustained  on-re-
sponses  to light (Corazza et al. 1971, Horn  1965,
Jung et  al. 1963, MacLean et al. 1968) may be the
main  source  of "electromotive  energy" that  could
initiate the phasing mechanism during the brief (but
apparently strong) reward-induced suppression of the
brainstem  reticular activating  system (Marczynski
1972a, Marczynski and Burns 1976). This  suggestion
is  supported  by the observation that, in the dark,
 single and relatively weak electrical stimuli applied to
 the optic nerve during consummatory responses trig-
 ger in a stimulus-bound manner typical  PRS-RCPV
 responses  whose  patterns and  topographical distri-
 bution are virtually identical with  those of spontan-
 eous responses in the presence of ambient light (Rick
 and Marczynski  1976). The  specificity of the effect
 of this "noisy" electrical stimulation was  demonstra-
 ted by the fact that the same, or even stronger, stimu-
 li applied during other behavioral states, such as non-
 rewarded lever pressing  or relaxed  wakefulness after
 satiation, produced no  effect. In  conjunction with
 previous data on  the role of unpatterned light input
 in the emergence of PRS-RCPV responses (Marczyn-
 ski et al. 1971b), it can be  concluded that "noisy"
 photic input is effectively utilized in reward-induced
 inhibition of neuronal activity in the feline visual cor-
 tex,  therefore,  in the  broader sense, in  the inte-
 gration of input  associated  with consummatory re-
 sponses.

     The occurrence of the epicortical positive shift,
 i.e., RCPV in association with the PRS bursts, can be
 interpreted as resulting  from a phasic tendency to-
 ward hyperpolarizing inhibition of larger populations
 of neurons in the cortex and the electrotonic spread
 of IPSPs to apical dendrites,  which  is reflected as sur-
 face positivity (Creutzfeldt et al. 1969). A more de-
 tailed  analysis of the PRS-RCPV phenomenon, path-
 ways involved, pharmacology, and  physiological  sig-
 nificance has been presented elsewhere (Marczynski
 and Bums 1976).

Acknowledgment

    The help of Dr. C. J. Sherry in the early phase of
this study is appreciated.

-------
ACQUISITION  OF SUSTAINED  POTENTIALS
DISSOCIATED FROM MASSED ACTION POTENTIALS
IN TEMPORAL CONDITIONING

V.  ROWLAND

Case Western Reserve University, School of Medicine, Cleveland, OH, U.S.A.
    The CNV may be seen as a sustained, nonoscilla-
tory field potential of the shortest end of the spec-
trum of sustained elec trophysiologic durations. This
spectrum extends from a half second  up to several
minutes in duration, as can be shown by conditioning
experiments (Rowland and Goldstone 1963, Sheafor
and Rowland  1974).  Also, the potassium-induced
glial depolarization mechanism of genesis of sustained
potential fields may be seen as a possible temporal
integrating  or  timebinding mechanism, although, to
my knowledge, no direct experimental demonstration
of such a role has yet been offered.

    Interest in such a role is enhanced by the failure
to demonstrate  actual circulating or  reverberatory
iterated fast-decaying transients such as action poten-
tials,  short  postsynaptic potentials, or wave  trains
(oscillating field potentials) to account for identifi-
able electrophysiologic states sustained for more than
a half second. The only known alternative is for the
action of transients to be transformed by other cell
parts or systems to  dynamics having much  slower
decay-time  constants,  thereby producing temporal
summation.

    In  order  better  to understand the relation of
nontimebinding  transients  (action  potentials)  to
 the sustained field  potentials reflecting the tem-
 poral integration of generators of slower decay, we
 (Rowland  and Dines  1973) developed small hybrid
 macroelectrodes and  circuitry  for  simultaneously
 registering  an artificial integration of  massed action
 potentials (integrated  mass units, or IMU), the con-
 ventional oscillatory field potential patterns (electro-
 corticogram, or ECoG), and sustained slow potentials
 (SP).

     Many  examples  of the expected association of
 IMU increments with ECoG and  SP activation were
 observed in cortex (Rowland and Dines 1973), as
 described by Robert (1973b) for the lateral geniculate.
 Also  IMU  decrements were  observed  with  ECoG
 synchrony and clear mass unit bursting in phase with
individual high-voltage, low-frequency waves of spin-
dle activity.

    By  serendipity, in relation to studies of the
IMU-SP relation occurring  with  food rewards,  the
appearance of marked anticipatory SPs without asso-
ciated  IMU changes  was  encountered (Rowland
1974).   The same  system  continued to show the
associated IMP-SP relations  formerly seen in reaction
to stimuli or association with an emitted potential,
the lambda  wave.  The reactive, associated  IMU-SP
relation is identified  in this  paper as SP-1,  and the
proactive  (anticipatory) dissociated  relation  (SP
present in the absence of IMU change) as  SP-2.
Fig.  1,  taken  from Sheafor and  Rowland 0974),
shows  the replacement of SP-1 in the  naive animal
by SP-2 after 3 weeks of training.  Essential to the
demonstration   of  SP-2  is  a  state  called  quiet
expectancy.  This was developed in  the cats in this
study  by  their being trained to restraint in a stock
that permitted them to  sit comfortably.   They
received food directly into the mouth every 4 min.
They  had   5  sec  of tone (R)  prior to each  rein-
forcement,  and 8-cm3  feeding.  The same  tone
was  repeated 2 min later at the middle of each 4-fnin
interval but not reinforced (N) in a program  called
single  alternation (SA).  The cat's cortical SP devel-
oped a clear 2-min phase of anticipation of forth-
coming reinforcement while  the  animal  remained
behaviorally  quiet.   If  it  spontaneously moved,
a transient SP-1  would   appear  superimposed  on
the sustained ramp of the SP-2,

     No evidence was observed in the IMU record of
 the  graded expectancy or proactive response prior to
 either N or R tone, whereas the SP was maximally
 positive or least negative prior to the N tone and grad-
 ually  and regularly became maximally negative prior
 to the tone.

     Fig.  1 shows,  in  the  course of acquisition,  the
 development of a proactive SP-2 through  an initial
 phase at 11 days, interpret able as related to the animal's

-------
  40
                                                                                               Rowland
  generalizing  the  N and  R  tones. The  N  tone  pro-
  duces marked change in the  IMU and SP. Subsequent-
  ly (day  14), both of these disappear, suggesting that
  the animal has begun  to discriminate the difference at
  the same time as the  sustained ramp over the second
  half of the 4-min interreinforcement interval emerges.
  Since  N  and R tones are the same, the  animals have
  only  their capability of monitoring the lapse of time
  (independently of direct cueing)  to enable the dis-
  crimination between  forthcoming reinforcement and
  nonreinforcement. This discrimination is not seen just
  prior to  the  tones in the IMU record (Fig. 2).  It is
  seen clearly in the SP-2  pattern. If further tests  sup-
  port  the  interpretation  that a true dissociation  be-
  tween the two variables exists, the conclusion follows
  that  the  SP-2 generating mechanism has  access to
  stored experience independently of the action poten-
  tial generating mechanisms. The proactive response
  can only develop in the context of temporal oreinta-
                                tion arising from the relatively remote memory of
                                prior days  of training and  the Fecentf memory of
                                when the last reinforcement occurred.

                                   Nothing  is   known  from direct  experiments
                               about the  mechanisms of genesis  of the proactive
                               SP-2. It appears to be some other mechanism than
                               the release of potassium generated  by mass action
                               potential activity inducing the presumed glial  depo-
                               larization  postulated for reactive  SP-1  genesis  (Ran-
                               son  and Goldring 1973b).   A more extensive con-
                               sideration of possible  mechanisms  such as synaptic
                               potentials  operating  independently of action  po-
                               tentials,  intracellular electrotonus,  etc., is presented
                               in Sheafor and  Rowland 0974).   A further candi-
                               date mechanism  to be  considered is the diffuse
                               release of  a neurotransmitter by  the  "varicosities"
                               of axons that  originate in cells of median  raphe or
                               other brain stem nuclei.
               62




              DAYS


               2
              11
       SAIN LIGHT
                             R
                                                 ECoG
r^^^^

y*ti*J5^^
                                                                         I.M.

               14
C$^^^

                             IjfV^^iMAAAa
              19
               TONE

                FOOD
                                     2 min
                                                                                I  100/JV
 _n
                              u
Fig. 1.  Representative pain of ECoG-IMU tracings for individual trials taken from days indicated illustrate the
SP-IMU relationships observed during SA training in Cat 62.  Close SP-IMU association (SP-1) was characteristic
during early acquisition (Day 2) and was obtained with R and during poit-R intervals throughout training (right
half of the figure).  As the pre-R (N to R) negative SP was acquired (left half of the figure),  it became increas-
ingly dissociated from the IMU measure (SP-2).  Multiple unit integrator  output is filtered from dc to half-
amplitude frequency response at 1.5Hz in all figures except Fig. 3. Negativity is down for Fig. 1-3.

-------
Dissociation of IMU and SPi During Conditioning
                                                               41
                    62
                    TRIALS
ACQUISITION
    DAY 10
                      10
                      12
                      14

                                                                                       ECoG
                                                                                        100/uV
                                                           I • .
                                                           t*^^
                                                                                        I.M.
                      14

                      TONE

                      FOOD
 Ffr 2.   Representative  ECoG tracings of seven consecutive trials  (28-min continuous recording) illustrate
 reliability of the SP pattern (top) acquired in the SA schedule in Cat 62.  In relation to corresponding IMU
 tracings  (bottom),  reliability  of SP-IMU dissociation  (SP-2)  during the pre-R  (N to R)  interval  and SP-IMU
 association  during  the  post-R (R to N) interval is demonstrated.   IMU tracings also show incrementing in
 CS-UCS interval of the R trials but little or no change to the same tone in the N trials.
     Slope and duration of SP-2 has been shown  to
 depend  on  the  interval between  reinforcements.
 Also, it is found to be relatively independent of the
 N  and  R tones.  The  graded  sustained  potential
 is  also  of interest  as  a possible  alternative to  an
 oscillator system as an internal clock, but it remains
 to be shown  that the SP itself is not dependent  on
 underlying oscillators.
     The above description pertains to SPs of long
 duration as dependent variables. Their roles as an in-
 dependent variable have only been indirectly implied
 in the behavioral studies of Roy Anderson carried out
                 in our laboratory.  Anticipatory  SP shifts recorded
                 without  benefit of associated observation of IMU
                 were seen to appear in cortices of rats on both fixed-
                 interval and fixed-ratio responding.  Again, the slopes
                 of the expectant portion of the interval were propor-
                 tional to the time intervals governing the animal's be-
                 havior. To my knowledge, there has been no specific
                 test of whether the distribution of SP levels reached
                 just prior to a timed lever press correlates with the
                 distribution of times of responding in a specific tim-
                 ing task in animals; however, Anderson  observed  a
                 statistical proportionality between  the  degree of an-
                 ticipatory cortical SP developed  and the proportion
                 of lever  presses produced in three fixed-interval pro-
                 grams (Fig. 3). A constant intersection point for the

-------
  42
                                                             Rowland
                                                  30 sec
                                                                                              LM	RAV
             FI30

             -<*^lf
-------
EVENTS  CONTINGENT UPON  CORTICAL POTENTIALS
CAN LEAD  TO  RAPID LEARNING


 J. S. STAMM, 0. A. GILLESPIE, AND B. B. SANDREW
 Department of Psychology, State University of New York, Stony Brook, NY, U.S.A.
     Research reported in this volume deals primarily
 with HRP  parameters as functions of experimentally
 controlled  conditions—i.e., these  potentials are con-
 sidered as dependent variables. In the present investi-
 gation, a  different experimental  strategy was em-
 ployed:  the occurrence of specified  electrocortical
 events as dependent variables was utilized for training
 monkeys  on a spatial  delayed response (DR) task.
 The start of each DR trial was made contingent upon
 computer-detection of an  endogenous cortical poten-
 tial of predetermined magnitude. The  dependent
 variable-the  rate  of task  acquisition-was  then
 assessed as a  function of the characteristics of the
 pretrial contingency.

     In previous research (Stamm  and  Rosen 1972),
 macropotentials were  recorded from several cortical
 areas during performance  of a DR task. The  averaged
 recordings delineated several distinct surface-negative
 slow potential (SP) shifts  during the course of a trial,
 one  of which  occurred in prefrontal cortex toward
 the end of cue presentation and the start of the intra-
 trial delay. This shift seemed important in relation to
 task performance because its  magnitude was signifi-
 cantly correlated with  the level of correct responses,
 but not with the duration of either the cue presenta-
 tion or the delay. The physiological significance of
 this  SP shift  is  suggested by the  finding that  pre-
 frontal  units  increase  their firing  during corres-
 ponding epochs  of the DR trial (Fuster  1973, Niki
 1974).

     Investigations of the functional role of prefrontal
 cortex in  humans and monkeys  have  provided evi-
 dence for  its implication in the regulation of attentive
 processes  (Fuster  1973,  Picton  and  Hillyard 1974,
 Pribram and  Luria 1973). In accordance with this
 view, the occurrence of prefrontal SP shifts and unit
 activation during DR performance may be considered
 as an expression of a heightened attentive state. At-
 tention is especially  important  during  cue  presenta-
 tion when the monkey must recognize  the  signifi-
cance of the location of the cue in order to program a
subsequent instrumental response. This interpretation
may be relevant to the general observation that mon-
keys normally learn the  DR task very slowly,  requir-
ing many hundreds of trials to attain criterion per-
formance. Slow acquisition may be a consequence of
training with constant intertrial  intervals because  the
spatial cue is then often presented while the monkey
is  inattentive. Conversely, if cues are presented during
periods when monkeys  are in a state  of high atten-
tion, more rapid task acquisition should result.

     This hypothesis was examined by using the mon-
key's endogenous prefrontal surface-negative  SP shift
for  starting  the DR  trial. Verification that  such  a
contingency  would lead  to  rapid  task acquisition
required control of other  factors that could  account
for any observed effect. Since  enhanced acquisition
rates could result from generalized cortical activation,
monkeys  were also trained with pre-cue negative SP
shifts  from  precentral  cortex.   In  addition, animal
training procedures have shown that the requirement
of a preparatory response for trial initiation,  such as
the  pressing of a "readiness"  lever,  leads to more
effective  and consistent task performance (e.g., Niki
1974). Therefore,  monkeys were also  trained under
conditions  of trial initiation contingent upon  the
occurrence of a behavioral response (lateral eye devi-
ations). This ocular movement  was selected because
it  is relatively simple,  clearly  identifiable,  and has
been found concomitant with operantly conditioned
prefrontal SP shifts (Rosen et al. 1974).

Methods

     Methods have been  further detailed elsewhere
 (Sandrew et al.  1977).  A total  of  13 stumptail
monkeys  (Macaca speciosa)   were trained.  Under
sodium  pentabarbital anesthesia,  pairs  of  Ag/AgCl
 nonpolarizable  electrodes  (Rowland  1968)  were
 chronically  implanted   bilaterally  in  prefrontal,
 precentral, and  occipital  cortex with one electrode

-------
  44
                          Stamm et al.
  of each pair on the cortical surface and the other In
  subjacent white matter. Prefrontal surface electrodes
  were situated on cortex In the  depth of the posterior
  third of the principal sulcui. Miniature epoxy resin
  electrodes were cemented to the bony orbit to record
  electrooculograms (EOGs).

      During  testing the monkey was placed In a re-
  straining chair, and its loft arm was attached by wrist
  cuff to the shelf of the chair, The chair was in front
  of a vertical panel that contained two circular display
  windows, situated at the monkey's eye level, with 12
  cm  between centers.  This separation  corresponded
  approximately  to a  36°  angle from the monkey's
  nose. Each window served as both projection surface
  for the stimuli and manlpulandum for  the monkey's
  response.  A  plastic  food  cup  was mounted at the
  center, below the windows. The DR trial started with
  cue presentation, a 1-sec white  illumination of either
  the  left or  right  window, followed by  a  blackout
  period-the intratrial delay. Both windows were then
  illuminated with blue light, and the monkey's press
  on either window started  another blackout period,
  the intertrial interval (ITI). A  correct response (on
  the cued  panel) was  rewarded with delivery  of a
 sucrose  pellet (4'S  mg) to the food cup. Each testing
 session consisted of 98 trials, with a pseudo-random
 sequence for left and right cue presentation.

     During  testing sessions,  the monkey was  con-
 nected by shielded cables to dc preamplifiers of a
 Crass Polygraph.  Transcortical  electrocorticograms
 (ECoGs) and  horizontal  EOGs were  recorded on
 seven-channel magnetic tape. Selected channels were
 monitored on-line by a  PDP-12A computer  pro-
 grammed to  detect SP or EOG events that fell within
 a defined voltage change  over time. Upon detecting
 a criterion event, the program initiated cue presenta-
 tion. ECoG,  EOG, behavioral data, and trigger pulses
 were stored on digital tape.

     Prior to  surgery all monkeys were adapted to the
 testing situation and trained with 8-sec ITIs on 0-sec
 DR until they  made 88  correct responses  in one
 session. Following surgery, each monkey was assigned
 to  one of four training groups (Table 1): (1) Group
 FSP (n-4)-cue presentation contingent upon  com-
 puter detection of a  left prefrontal surface-negative
 SP shift of 50-100pVand 2.5-sec duration; (2) Group
 MSP (n=2)-cue presentation contingent upon detec-
 tion of a SP shift of the same parameters from left
 precentral cortex; (3) Group  LEM (n=2)-cue  pre-
 sentation contingent upon detection of an EOG volt-
 age change that corresponded to  an eye movement to
 the  right of  approximately 40° during  a  1.5-sec
epoch; and (4) Group YC (n=5)-no pre-cue require-
ments. Each YC monkey was yoked for ITIs to one in
another group (three  to FSP monkeys and one  each
  to an MSP and an LEM). This was accomplished by
  recording  ITIs  for  each session  and programming
  these for the corresponding session  with the yoked-
  control animal.

      Each monkey was first trained under the contin-
  gent  (on-line) condition  on 2-sec DR until 90% cor-
  rect response was reached in one session, In order to
  aisess  performance  transfer to  noncontlngent  (off-
  line)  conditions, each  monkey was then tested (one
  or two sessions) with constant ITIs equal to the mean
  ITI during  the  previous on-line session. If perform-
  ance  fell below criterion, on-line  testing was repeated.
  This procedure was repeated with successive delays of
  4, 8, and 12 sec. The final testing phase was designed
  for more direct evaluation of the monkey's attentive
  functions by reducing  the duration of cue  presenta-
  tion.  Monkeys were retrained under on-line condi-
  tions  on 12-sec DR and 1-sec cues. During the subse-
 quent  two  sessions, trials were programmed  in  a
 random sequence for cue durations of 0.5, 0.2, and
 0.1 sec.
 Results

     Averaged ECoGs and  EOGs for an  FSP and an
 LEM  monkey are presented in Fig.  1.  Recordings
 for every  FSP monkey showed that  the large pre-
 cue negative  SP  shift  was  preceded  by a smaller
 positive wave  that started about 4 sec  before cue
 onset  (Fig. 1A). Criterion  SP shifts were localized in
 left prefrontal cortex; no concomitant SP events were
 observed at right  prefrontal, precentral,  or occipital
 electrode locations. ECoG recordings for MSP mon-
 keys indicate a similar sequence of pre-cue SP shifts
 localized to the left precentral area. EOGs for LEM
 monkeys (Fig.  IB) show that the large pre-cue right-
 ward eye deviation was preceded  by  a smaller devi-
 ation to the left and was followed by return toward
 baseline level. For  trials with left cues, these monkeys
 attained maximal leftward  eye deviations at- approxi-
 mately 0.5 sec after cue onset, and their EOGs indi-
 cated a subsequent drift to baseline levels. Compari-
 sons of ECoGs and EOGs  during the pre-cue period
 indicated no  systematic  relationships between SP
 shifts  and  lateral  eye  deviations. The pre-criterion
 leftward eye deviation seen in Fig. 1A was not found
 for  the  other  FSP monkeys. During testing  under
 off-line (constant ITI) conditions, pre-cue  ECoG and
 EOG traces  remained  near  baseline levels  for all
 monkeys, including YCs.

    Behavioral  results (Table 1) show substantially
 faster acquisition by FSP than by other groups. Com-
 parisons between  FSP  and  YC monkeys show no
overlap of  individual scores for  either total errors or
sessions required to reach criterion performance on

-------
Cortical Contingencies of Learning
                                45
                                          R.PREF.ECoG&EOG
                                                                        Rt.t
12°Rt.llOOj-tVneg
                        i—i  i sec
                         t  Response
Fig.  1.  Averaged left and right prefrontal electrocorticograms (solid lines) and concomitant electrooculograms
(dotted lines, superimposed)  during criterion performance of 12-sec  delayed response for: (A) FSP 297 and (B)
LEM 293.  Separate averages  (49 trials each) were obtained for trials when the cues were presented to the right
(Cfc) and left (C^j positions. Horizontal lines denote 1-sec cue presentations. Arrows signify the beginning of the
choice response period. Calibrations for ECoGs and  EOGs are indicated. Upward deflections indicate cortical
negativity or eye-deviations to the right.
    Table 1. Acquisition of 12-sec Delayed Response Task by Groups of Monkeys with Differing
                                   Requirements for Trial  Initiation"
Errors at
each delay
2 sec
4 sec
8 sec
12 sec
Total errors
Total sessions
Experimental

295
7
4
4
4
19
4
FSP
297
7
16
5
8
36
5
b
314
14
2
34
10
60
9

298
24
20
34
20
98
7
group
and monkeys
YCC
302
21
8
89
3
121
10
301
78
41
88
24
231
15
315
88
136
95
31
358
19
292
171
84
74
21
350
21
294
337
42
25
76
480
20
MSPd
313 316
32 51
4 86
42 124
7 65
85 326
8 15
LEMe
293 114
59 306
41 7
22 109
2 30
124 452
10 16
 "Scores of error* and sessions Include criterion performence of 88 correct responses In a 98-trial session.
 bFSP • lurface negative SP ihlft from left prefrontal cortex,
 CYC • Intertrlal intervals yoked to those of monkeys In enothar group,
  MSP - lurface negative SP ihlft from left precentral cortex.
 aLEM • eye deviation to the right

-------
 46
                         Stamm et al.
 12-sec  DR.  Rapid acquisition by  the FSP group is
 also  indicated by  the  finding that one monkey re-
 sponded at 90% correct during every training session
 and the others  required few additional sessions. Al-
 though one MSP monkey (No. 313) acquired the task
 almost  as  fast as the FSP group, the other MSP was
 much slower. Since MSP and LEM groups consisted
 of only two monkeys each and their acquisition scores
 overlapped, they were combined for statistical analy-
 ses. Comparisons  of total errors indicate significant
 differences between  FSP and  YC  (t=3.59;  p<.01),
 and  between FSP  and MSP-LEM  groups  (t=2.19;
 p<.066), but not between YC and  MSP-LEM groups
 (t=0.59).  Performance  during off-line  testing  was
 evaluated by  transfer scores between the last on-line
 and first off-line session at each delay setting. Mean
 transfer scores were above 92% for  every group, indi-
 cating that monkeys indeed learned the DR  task and
 could perform well without pre-cue requirements.

    Results for  12-sec DR testing with brief cue pre-
 sentations  (Fig.  2) also showed better performance
 scores by FSP than by other  groups. Only FSP mon-
 keys continued to respond at the 90% criterion level
 during  testing with brief cues. Performance differ-
 ences between FSP and YC  groups were significant
 for 0.1-sec cues (t=4.91; p<.01), but the differences
 were  not significant  for 0.2- or 0.5-sec cues.  Low
 scores by  the MSP group were the consequence of
 poor  response by  one monkey (316), which seemed
 unable to perform adequately under these conditions.
 Low scores by LEM monkeys for 0.2-sec and 0.1-sec
 cues may be attributed  to an inability to detect cues
 in the left position.
  100
 Sao
 ee
 oc
 o
 o
 t80
ui
CJ
K
 •70
  60
                                v1'    N MSP
        1.0       O.S             0.2    0.1
                    CUE DURATION, sec
Fig.  2.  Performance on 12-sec DR (two sessions, at
96 trials) with randomly presented cues of 0.5-sec,
0.2-sec, and 0.1-sec duration.  The scores with 1-sec
cues were obtained during  the previous session. The
testing  contingencies for the different groups  are
indicated in Table -1.
  Discussion

      The  results show impressive  enhancement in
  learning the DR task  when trial initiation is contin-
  gent upon computer-detection of prefrontal surface-
  negative SP shifts. Criterion  performance to  12-sec
  DR by the FSP group was attained with  about one-
  third  the  sessions and  one-fifth the errors of the
  yoked control group.  These scores indicate substan-
  tially  greater facilitating effects on task  acquisition
  than previously found with either low-voltage stimu-
  lation (Stamm  1964)  or direct  anodal polarization
  (Rosen  and  Stamm  1972)  of prefrontal  cortex.
  Enhancement was  more pronounced when training
  was made contingent  upon prefrontal than when it
  was contingent  upon precentral negativity, although
  further study is needed to establish, clearly, the corti-
  cal specificity of effect. The functional significance of
  the  prefrontal contingency is reflected, furthermore,
  by high transfer scores to noncontingent testing con-
  ditions. This finding indicates that the monkeys had
 indeed learned the  task; i.e., their high performance
  was  no  longer dependent upon  the  contingent
  training procedure.
     The  present  results support  the  view that  the
 surface-negative SP shift is an important physiological
 phenomenon and an  expression of neuronal activa-
 tion (Fuster 1973, Niki 1974). The hypothesis that
 this  electrocortical process  reflects  a heightened
 attentive  state  is supported .by the results of DR per-
 formance  with brief cues. Cues of 0.1-sec duration
 clearly require the monkey to be highly attentive in
 order to recognize the relevant cue. The finding that
 correct performance with a  brief cue presentation
 was obtained only by FSP monkeys provides strong
 support for the attentive hypothesis.
     Facilitated learning may only be possible when
 the crucial demands of the task occur during the trial
 epoch of maximum prefrontal negativity. This is the
 case for DR, which requires the monkey to recognize
 the location of the cue and program  its choice re-
 sponse at  that  time.  For other tasks that require
 optimal attention for selection of the choice response
 during other epochs of  the trial, a requirement  of
 pre-trial negative SP shifts may not lead to enhanced
 acquisition. This prediction was confirmed  by an
 additional experiment  (Gillespie  1977) in which the
 present monkeys were trained on a visually delayed
 matching-to-sample  (DMS) task  with the same pre-
 trial requirements as for  DR. The FSP monkeys did
 not acquire the DMS  task faster than their yoked
controls.

-------
 Cortical Contingencies of Learning

    The  present investigation demonstrates that the
acquisition of a difficult  task can be facilitated by
presenting cues contingent upon the occurrence of a
specific cortical event. The prefrontal negative shift
selected  as the criterion event in monkeys is  analo-
gous to the CNV in humans (cf. Low et al  1966),
Walter (1966, 1967) demonstrated several years ago
that external stimuli  could  be triggered by com-
puterized detection of CNV-like negative shifts from
ihumans.  With  the  development of techniques for
scalp  detection of single electrophysiological events,
jit may be feasible to apply this method to the study
of human learning and performance. With the detec-
tion of  endogenous  electrocortical events for trial
                                              47

filiation, it might be possible to enhance a subject's
.performance on  attentive  and cognitive  tasks.  The
iputative effects with this method may be of special
benefit to subjects with certain cognitive and emo-
tional disorders, such as mental retardation, learning
'disabilities, or schizophrenia.

Acknowledgments

     This research was supported by National Science
Foundation Grant GB 35743X1. We wish to express
our appreciation to Steven Rosen and Richard Reeder
for their technical assistance in all phases of experi-
mentation and data analysis.

-------
  SUSTAINED  ACTIVATION  OF  CORTICAL NEURONS  IN
  STIMULUS-RECOGNITION TASKS1
 J. M. FUSTER

 Department  of Psychiatry and Brain Research  Institute,  School of  Medicine,  University of
 California at Los Angeles, Los Angeles,CA, U.S.A.
     Studies of  the effects of brain  lesions  on be-
 havior have implicated certain cortical areas of the
 monkey  in the integrative processes underlying per-
 formance of  stimulus-recognition  tasks.   Delayed
 response  (DR)   and  delayed   matching-to-sample
 (DMS) are  prime examples of  such  tasks.   Their
 most distinctive  feature is a  period  of delay inter-
 posed  between  a particular  stimulus and a parti-
 cular behavioral  act predicated  on  the perception,
 retention, and recognition of information  pertaining
 to  that stimulus (Fig. 1).  Inasmuch  as correct per-
 formance of  these  tasks  requires  the temporary
 retention  of  sensory  information,   it  is  appro-
 priate to consider them as tests of a certain form of
 memory  variously designated as short-term memory,
 image   memory,  recent   memory,   or   transient
 memory.

     In the adult animal, the functional integrity of
 the  prefrontal cortex is important for short-term
 memory.   Animals with  reversible  dysfunction of
 the  prefrontal cortex show a reversible  deficit in
 performance  of  DR  and  DMS  tasks that require
 temporary   retention  of   spatial,  kinesthetic,  or
 visual  information  (Fuster and Alexander  1970,
 Fuster and  Bauer 1974,  Bauer  and  Fuster  1976).
 The deficit  is closely dependent on  the  length of
 the  delay.   When that delay  is  zero or very brief,
 the  deficit is  not apparent; the  deficit appears at
 longer  delay and increases as   a  function  of its
 duration  (Fig.  2).   Consequently,  it seems  that
 there is a neural process occurring during  the delay
 for  which  the  functional integrity  of  the  pre-
 frontal cortex is important.

    In  order to  investigate that process, the  ac-
 tivity of nerve cells  in  the prefrontal cortex and
 in related structures  during performance  of delay
 tasks has been explored.  This exploration revealed
a large proportion of neurons exhibiting  increased
 firing during  the delay  (Fuster  and Alexander
1971, Fuster 1973).  The  discharge of  these neurons
is higher  on the  average during this period than it
is  spontaneously, namely,  between  trials.   Inter-
 estingly,  many  units  show  sustained  activation
 during  the  delay but no  activation during the pre-
 ceding  cue  or the succeeding response of the animal
 (Fig. 3).  Concomitant recording of EEC and EOG
 indicates  that  the  phenomenon of  sustained  acti-
 vation cannot be simply attributed to arousal or to
 eye movements.  Both EEC arousal and eye move-
 ments  are  maximal  during  cue  presentation and
 during  the  response  period, not during the delay.

Fig. 1.  Diagrammatic representation  of a  monkey
performing  stimulus-recognition  tasks.    Delayed
response  (DR):   the  stimulus (cue)  is white  light
in  one of the  two lower stimulus-response  buttons,
the animal turns: it off by pressing the button; a delay
ensues, at the end of which both buttons are lit; the
monkey is then rewarded with fruit luice for pressing
the button  that has been  lit  before the delay.   De-
layed   matching-to-sample  (DMS):    the  stimulus
(sample) is a color, red or green, presented on the
top button; after the delay, both colors appear in the
lower  buttons; reward is given for pressing the button
with  the sample color.  Position of cue (in DR) and
of colors (in DMS) is changed at random from trial
to  trial.
'This research was supported by NIMH Research Scientist Award KO5 MH-25082 and NSF grant GB-41867.

-------
Cortical activity and short-term memory
                                                                49
                 :
      MATCHING-TO-SAMPLE
 DELAYED
RESPONSE
 *-»NORMAL TEMP
 v~>FRWTAL &•
 '-"PARIETAL W
 Fig. 2.  Effecti of cortical cooling on performance
 of DMS and DR.  Note  the delay-dependent deficit
 produced by cooling of lateral prefrontal cortex
 on the two tasks.  No significant effect Is produced
 by cooling parietal cortex.

    There are some indications that prefrontal unit
 activation during the delay has something to do with
 the process of retention  of information. A relation-
 ship has been seen  between  the level  of activation
 of some units and the level of correct performance
 of the animal.   Furthermore, certain control pro-
 cedures Indicate  that the  phenomenon is determined,
 at  least  in part,  by the  mnemonic attribute of the
 cue (Fuster  1973);  however,  the  role of prefrontal
 units in retention is still obscure.

    The temporal characteristics of,cellular activation
 in  the  prefrontal cortex  are  particularly interesting
 and may bear on the generation of cortical slow poten-
 tials.   Many units show the highest levels of discharge
 at  the end  of the cue period and beginning of the
 delay,  when  the  information has just been presented
 and the waiting  period begins.   Sharp  increases in
 firing  are commonly observed at that time.  Judging
                                  CUE
from  individual unit records, there then appears to
be a peak in probability of cell-firing within the pre-
frontal cortex.  It  ii at that same time that Stamm
and  Rosen  0972), using  a  similar set  of behavioral
operations,   have  found  a  slow negative  surface
potential.   This  potential, which  they  relate to
memory  formation  or  registration, may be  a  mani-
festation of the underlying cellular phenomena men-
tioned above.

    The  search for  the source and function of pre-
frontal unit activity In short-term memory led to
the study of the visual transcortical pathway.   This
pathway  has been anatomically  well demonstrated.
It is composed of a series of interlocking  cortico-
cortical   connections that  originate in the  striate
cortex and  lead  to  the  prefrontal  cortex (Pandya
and  Kuypers 1969, Jones  and Powell 1970).   The
last step of this pathway is constituted by  connec-
tions  between  the  inferotemporal  and prefrontal
cortices.    Histological studies with  silver  impreg-
nation   techniques   have  established that  these
connections are  bidirectional:  the prefrontal cor-
tex is not  only the target of afferent fibers from
the  inferotemporal  cortex  but  the  source  of
efferent  fibers   running  back  to   the  infero-
temporal cortex (Pandya et al. 1971).
                        This anatomical evidence provided  part  of the
                    rationale for  investigating  the  neuronal  activity of
                    the  inferotemporal  cortex  during  visual  short-
                    term memory tasks.  This  research was also encour-
                    aged by reports in the literature, mostly derived from
                    ablation  studies,  indicating that the inferotemporal
                    cortex is  implicated in certain  aspects  of  visual
                    memory (Gross 1972, Mishkin 1972).
                                             II  «:l,111 II   11 111 111 Ml Illl   Ill
                                                                                        fCR
                                           llilllll  illUM.JllJI.Jl.il
                                                                                        tCL
                                     JJL	I   1    I MM II III I    I   Hill
                                                                    •fl.iilJJJ.Illl
                                              ii Hiiuinin    i
                                                                                    .
                                                                             5 sec
                                                      fCR
                                                      II
                                                      fCR
  *'    D'sc*ar8e °J a ce" in! the P™fr°"tal ™rtex during five trials of a direct-method DR task. A t left  spon-
taneous discharge between trials Horizontal bars mark the period of presentation of a positional cue-placement
of food ™der one of two Identical objects, one on the right and the other on the left.  Between trials and du ing
the  mtratrua delay the objects are concealed behind a screen.   Arrows mark the termination of delay
                      R^f^S Dented for choice.  Ad,acent notations refettc" £
              t. R, right, L, left).  Note the increased discharge of the cell during the 32-sec delay.

-------
 50
                                             Fuster
     A DMS task, using color as the sample or memo-
 randum, has been recently utilized as the behavioral
 paradigm for examining the activity of inferotemporal
 neurons in  visual short-term memory.  So far, most
 of the records have been obtained by microelectrode
 penetrations of the anterior inferotemporal region,
 including areas of the middle  temporal gyrus and
 posterior wall of the superior temporal sulcus. Many
 color-coded cells have  been found  in  these areas.
 Such cells react differently to the sample depending
 on its  color.  When using a task with two  colors
 (red and green),  about the same proportion of units
 have  been  found to  be activated preferentially  by
 one color as by the other.
     Again, as in  the  prefrontal cortex,  sustained
 activation  of  discharge has  been observed during
 the delay.   Here,  however, the level of activation
 is generally more related to the specific information
 which is relevant on each  trial.  This information in
 DMS depends exclusively  on  the  wavelength of the
 sample color and is not spatially defined by  posi-
 tion  or configuration.   This  implies that whatever
 differences of firing are induced  on a unit by the

       25 r
     sample  cannot be  attributed  or related  to  differ-
     ences in motor  activity, including eye  movements.

         In many inferotemporal units the level of firing
     during the delay is different depending on the color
     of  the  sample.  In some of the units, differential
     firing is  in  fact  only evident  during the delay and
     not during  the sample-presentation period  (Fig. 4).

         In our view, these observations are presumptive
     evidence  that  at least  some  units in  the infero-
     temporal cortex  are constituent elements of corti-
     cal  neuronal assemblies  that code and  retain visual
     information.   Because of the  mentioned  relation-
     ships between  inferotemporal and prefrontal cortex,
     and because of the latter's  role in delay  tasks, we'
     suspect   that  the  retentive  function  of  infero-
     temporal units is  to some degree dependent on
     tonic  influences  from  the  prefrontal  cortex.   We
     have obtained some  evidence, still  sketchy  and
     preliminary,  that  supports  this hypothesis.   Pre-
     frontal cooling results  in  the  diminution or  dis-
     appearance  of color-dependent  differences  in  the
     discharge of some  inferotemporal units during the
     delay.   This occurs in conjunction with a drop in
     performance, which is indicative of poor retention.
        21)
                       GREEN SAMPLE TRIALS
                       RED SAMPLE TRIALS
  FREQUENCY COMPARISONS

POST-GREEN VERSUS BASELINE
POST-RED VERSUS BASELINE
POST-RED VERSUS POST-GREEN
         P
3.68     0.001
8.21   <0.001
5.66   < 0.001
Fig. 4. Discharge of an inferotemporal cell during performance of DMS with two colors.  S:  sample; M:  match.
Firing is markedly increased during the delay of red-sample trials.

-------
PRELIMINARY  STUDY OF PHARMACOLOGY OF
CONTINGENT  NEGATIVE  VARIATION IN MAN
J. W. THOMPSON, P.  NEWTON, P.V. POCOCK, R.  COOPER, H. CROW, W. C.  McCALLUM,
D. PAPAKOSTOPOULOS

Burden Neurological Institute, Bristol, England, and University of Newcastle Upon Tyne, England
    Although   the  contingent   negative  variation
(CNV)  was fust  described in 1964 (Walter  et al.
1964), the electrogenesis of this phenomenon remains
to be determined.  The possibility  that at least one
putative neurotransmitter is involved in the  gener-
ation of  the  CNV seems  likely for  at least  two
reasons. First, evidence is accumulating from  neuro-
physiological, histological, and histochemical studies
which  suggests strongly that neurotransmitters are
involved in neuronal transmission within the central
nervous system, and there seems no good reason to
exclude  the genesis  of the CNV  from this possi-
bility.  Second, a  large number  of drugs have  been
shown  to modify  the CNV.  These include ethanol
(Kopell et al. 1972), carbon monoxide (Groll-Knapp
et al. 1972), cannabis (Kopell et  al. 1972, Low et al.
1973), flurazepam (Hablitz and Borda 1973), caffeine,
 nitrazepam, diazepam and cigarette smoking (Ashton
et al. 1973,1974,1976), barbiturate and amphetamine
 (Kopell  et  al. 1974), and chlorpromazine  (Tecce
 et al. 1975).  On  the other hand, little attempt has
 been made to use drugs as tools to analyse, system-
 atically, the neuropharmacological mechanisms that
 subserve the CNV.  In  animals,  Pirch  (this volume)
 has  shown that   event-related slow potentials can
 be recorded  (dc)  from anterior cortex of the un-
 anaesthetised rat and that drugs can be used to study
 the  neuropharmacological mechanisms  of the  CNV
 under these conditions.  The present study appears
 to be the first attempt of its kind  in man.

     The basic  principle used  in  this pilot study was
 to examine the effect on  the CNV of administering
 a drug with well-established  selective  blocking pro-
 perties  against a  putative  central  neurotransmitter.
 It  was predicted  that,  if any or all of the neuro-
 transmitters played  a role in the electrogenesis of
 the  CNV, this fact would be reflected by an  alter-
 ation in the record,  most probably by a reduction
 in magnitude.  Of several putative  neurotransmitters
 likely  to  be  involved, three were considered ini-
 tially:    (1) acetylcholine, (2)  noradrenaline,  and
 (3)  dopamine, acting, respectively,  on  receptors
designated as muscarinic cholinergic,  a-adrenergic,
and dopaminergic.   Several pieces of evidence sug-
gest  that  one  or more  of these neurotransmitters
may  be involved in the  electrogenesis of the CNV.
First, there  is  histological and  histochemical  evi-
dence (Shute and Lewis 1967, Dahlstrom and Fuxe
1964,  Ungerstedt  1971) on the  distribution  and
interconnection   of  cholinergic,  adrenergic,   and
dopaminergic systems in  the brain.   Second,  the
case  for the  role of acetylcholine has been cogently
argued by Marczynski (this volume), while evidence
for the role  of dopamine  comes from Libet (this
volume).

    The  blocking   drugs  (phannacological   antag-
onists) used in this  study are listed in Table 1. Each
drug  acts by  competing  with  the  corresponding
neurotransmitter  for  the  appropriate  pharmacol-
ogical  receptors.   A  placebo,  physiological  saline,
was also used.

Methods

    Sixteen  tests  were conducted  on  10 healthy,
paid  volunteers (6  male and  4  female),  aged
18-30 years  and of  widely different occupations.
Two  of the male  subjects and  one of  the  female
subjects  took part in two, four,  and  three experi-
ments,  respectively; the  remaining seven subjects
took part in  only one experiment.  The  project
was   approved  by   the   local   Hospital  Ethical
Committee.

     A single-blind  and  pseudo-random  design  was
used  with one  drug  administered per experiment.
 EEC from  the  nasion + 2 cm, F3, F4,C3(Cz,C4,
 P3,P4   referred to  linked mastoids was  recorded
 with amplifiers modified  to provide  a  5-sec  time
constant.   Eye movements  were monitored  by
 recording the  vertical  EOG.    Movements  of the
 right (button  pressing)  hand were recorded  via the
 EMG.   Electrocardiogram, respiration, and inter-
 mittent blood pressure were also recorded.

-------
 52
                                Thompson et al.
 Table  1.   Blocking  Drugs  (Pharmacological
 Antagonists) Used in This Study
Neuro-
transmitter
acetylcholine
noradrenaline
dopamine
Antagonist8
atropine sulphate
thymoxamlne
HCI
metoclopramida
HCI
Dose
0.4-0.5 mg
0.08-0.1
mg/kg
5 • 7 mg
 a Placebo was  physiological saline solution  0.5 ml.

     After control responses were obtained, a  drug
 was  injected  into the  right  deltoid  muscle,  and
 the  responses  repeated.  The intramuscular (I.M.)
 route  was chosen  as  a compromise between  oral
 and  intravenous  methods  to avoid,  respectively,
 the   vagaries   of gastrointestinal  absorption  and
 the  risk of subjective  effects of rapid  attainment
 of blood levels.

     CNV tests commenced  at times (min):  -28, -8,
 +2, +8, t]6, +22, +36, and +56; and blood pressure
 (BP) at:  -34, -22, -2, + 14, +28, +42, and +62 relative
 to drug injection. For each set of recordings the sub-
 ject received 24 trials. Sixteen  trials uncontaminated
 by eye movements were selected for off-line  compu-
 ter averaging. Subjects lay on  a  bed  with eyes open
 and fixated throughout.

 Results

 Atropine

     Fig.  1  shows a  series of  eight CNVs recorded
 from one subject who received a dose of atropine af-
 ter the second CNV. The figure clearly shows the sub-
 sequent decrease in CNV amplitude, which reached a
 minimum at  16-22 mln with partial but irregular re-
 covery in the CNVs  that followed. The mean result
 obtained from  four subjects  is shown in Fig. 3a
 (t-test of CNVs  at -8 and +22 min, p <.02).

 Metoclopramide

     Fig.  2 shows the result obtained  in one subject.
 The mean result  obtained in four subjects (Fig. 3b)
 shows  a  steady decrease in  CNV amplitude, which
 reached a minimum  at  36 min (t-test at -8 and  t22
 min, p <.05 and at -8 and +36 min, p <.01).

 Thymoxamine and placebo

    Neither thymoxamine (Fig. 3c) nor the placebo
(Fig. 3d) produced a significant change in  the mean
 CNV in four subjects.

 Topographical effects

     Corresponding changes in the CNV were observ-
 ed in recordings made from adjacent electrode sites.
 Preliminary analysis indicated no  topographical lo-
 calization of drug effects.

 Reaction time

     RT  to  the imperative stimulus (S2) showed no
 change with atropine or placebo, but it was lengthen-
 ed by as much as 30% with metoclopramide and shor-
 tened by 15% under the influence of thymoxamine,
 in spite of the fact that the latter drug had no detect-
 able effect  on the  magnitude  of the  CNV. These
 changes  in  RT were statistically significant (t-test
 p<.001).

     Blood pressure and respiration were not affected
 by any of the drugs used in this study.

 Subjective effects

     With one  exception, none of the subjects com-
 plained  of any subjective effects during the experi-
 ments. About  10 min after the injection  of atropine,
 one subject complained of transient blurring of vis-
 ion, making it difficult to see the fixation spot.

 Discussion

     Preliminary results indicate  that the CNV is de-
 pressed  by atropine and  metoclopramide, but is un-
 affected  by thymoxamine and  placebo administra-
 tion. It  was important that  the  doses of drugs used
 did not produce significant subjective effects which
 might have  confounded interpretation of the direct
 drug-Induced effects on  the CNV. Since thymoxa-
 mine has a very short plasma half-life (c. 10 min), it
 is  possible  that no  effect would be seen with intra-
 muscular drug administration. It  is important, there-
 fore, to study the effects of thymoxamine  adminis-
 tered by intravenous infusion to maintain more con-
 stant blood and brain  levels during CNV recording.
    These  results  suggest  that  both  cholinergic
muscarinic   and   dopaminergic   mechanisms  are
involved  in  the generation  of the  CNV, although
the data  do not provide any direct indication  of the
neuroanatomical slte(s) involved.  The results support
hypotheses  that  both  a cholinergic  mechanism
(Marcrynski,   this  volume)  and  a  dopaminergic
mechanism (Libet, this volume) are  involved  in the
genesis of the CNV.

    The  present findings are not incompatible with
Somjen's proposal  (this volume) that the CNV may

-------
Pharmacology of CNV in Man
                                                                                             53
                                     EFFECT OF ATROPINE ON CNV
                                                     E.B.
 -  S2=tone
   N+2
    /2 sec
•28 min.
                                             -8
                                         A
                                                                         +8
    EMG
    EOG
tie
                                V
                             +22
+36
+56
     Cz
     EMG
  FJ^. 1. Effect ofatropine 0.4 mg I.M. injection (at arrow) in one subject.  N+2 refers to an electrode 2 cm above
  the nation.

-------
54
                                                            Thompson et al.
   Si=click
                          EFFECT OF METOCLOPRAMIDE ON CNV  B.H.
 EGG
           1  ,0      2 sec
            -28 mm
 N+2
 EMG
EOG
+16
+22
+36
+56
Cz
EMG
              g. 2. £//ecr of metoclopramtde 7 mg I.M. injection (at arrow) in one subject.

-------
 Pharmacology of CNV in Man
                                                                               55
   24 -


   22 -
V.
   16
<
    1
          *
                   B
          EFFECT OF ATROPINE
               ONCNV
                       FOUR SUBJECTS 04 osmgiM
   ©®  (D@    cs>
                  DRUG
      -30  -20   -10
                    0   10   2O   3O   40

                         TIME.min.


                                                            *
 b. EFFECT OF METOCLOPRAMIDE
           ON CNV
FOUR SUBJECTS5 7ma'M
                                                                        © ® (D ©    ©
                                                                      IG
                                                                             H
                                     JO  -20   -10   0   10   20   30   40   SO

                                                      TIME, min.
                                                                                 d. EFFECT OF PLACEBO
                                                                                       ON CNV
                                                                                   FOUR SUBJECTS
                                                                                05ml physiological saline I M
    c. E :ECT OF THYMOXAMINE
             ON CNV
FOUR SUBJECTSooe-oimafcgiM
      -30  -20   -10
                        10   20   30   40   SO
                         TIME.min.
                                      .IO   -'0   -10   0   10   20   M   40   50   SO

                                                       TIME.min.
Fig. 3.  Mean effect in four subjects of (a) atropine (0.4-0.5 mg I.M.), (b) metodopramide (5-7 mg I.M.), (c)
thymoxamine (0,1 mgjkg I.M.) and (d)  placebo (physiological  saline 1 ml I.M.) on CNV.  Drug injected at t0.
be generated by depolarization of glial cells. Somjen
conjectures  that  glial  cell depolarization is  deter-
mined  by the extracellular potassium concentration
which, in turn, depends upon  the  activity of adja-
cent nerve fibres.  If atropine  or  metoclopramide
block transmission in  a substantial number of nerve
fibres,   the  reduced  neuronal   activity  would  be
accompanied by a decrease in extracellular potassium
concentration and, consequently,  a reduced CNV.

    Whether or not glial cells are involved in the elec-
trogenesis of the  CNV, it is reasonable to postulate
that the generation of the CNV depends upon the in-
tegrity of cholinergic muscarinic pathways in  the as-
cending  reticular  activating system.  Dopamine may
enhance cholinergic transmission in  these pathways
                                  by inducing a synaptic change in cholinergic receptors
                                  similar to those concerned with slow  muscarinic or
                                  s-EPSP responses in ganglia (Libet and Tosaka 1969).
                                  Further experiments in man are planned in which a
                                  range of drug doses will be used to extend the present
                                  pharmacological analysis.
                                  Acknowledgment
                                  The authors wish to thank Dr. B. M. Guyer, William
                                  R.  Warner & Co. Ltd., for the donation of thymox-
                                  amine HC1 and Miss D. Mustart and the Department
                                  of  Photography  and  Teaching  Aids  Laboratory,
                                  University of Newcastle upon Tyne, for preparing the
                                  figures.

-------
EFFECTS OF  AMPHETAMINE AND PENTOBARBITAL
ON  EVENT-RELATED SLOW  POTENTIALS IN RATS1


J. H. PIRCH
Department of Pharmacology  and Therapeutics, Texas Tech University  School of Medicine,
Lubbock, TX, U.S.A.
    Investigations concerning the effects of drugs on
event-related slow potentials (ERSP) in unanesthetiz-
ed, behaving animals are relatively few (Caspers 1963;
Marczynski et al.  1969; Marczynski 1971, 1972b;
Norton  and Jewett  1967; Pirch and Norton 1967a,
1967b;  Pirch and Osterholm 1975), and studies of
the effects of drugs on the human contingent negative
variation (CNV) have yielded varying results (Kopell
et al. 1972, Low et al. 1973,  Braden et al. 1974,
Tecce and Cole 1974, Kopell et al. 1974). Thus, the
effects of different classes of drugs on ERSPs are in-
completely characterized, and, as a  result, interpre-
tation of the meaning of drug  effects is uncertain.
For example, it appears to be widely accepted that
ERSP amplitude is  increased by  "stimulant" drugs
and decreased by "depressant" drugs. Tecce and Cole
(1974)  reported,  however,  that amphetamine may
either increase or  decrease CNV amplitude and Ash-
ton et al. (this volume) found  that nicotine has a
biphasic action on CNV amplitude. Additional know-
ledge of the effects of drugs on slow potentials would
be of value not  only  for interpretation of drug-in-
duced alterations of ERSP but also as a data base for
evaluating changes induced  by  toxic environmental
substances. Accordingly, this paper is a brief report
of some of the studies conducted in our laboratory
to characterize drug effects on slow potentials record-
ed from the cortex of rats trained on various behav-
ioral paradigms.

Methods

    Female albino rats were housed individually dur-
ing the experimental period. Twenty-two gauge silver
wires were  coated with silver-chloride and were im-
planted  under pentobarbital and ether anesthesia af-
ter atropine pretreatment.  The electrodes were in
contact with the active or reference site via agar-saline
pools (1% agar-1% saline). Artifact due to pulsation
of the brain was reduced by making the bridge with
two pieces of polyethylene tubing  filled with the
agar-saline.  One piece (PE 90) was fitted onto the e-
lectrode and was  inserted into  the other  (PE 240)
which rested on the recording site. The dura was left
intact over the active recording site. The bone refer-
ence electrode was inserted into  an agar-saline pool
formed by the polyethylene tubing  placed in a small
depression drilled in  the bone. The active  electode
was placed over the cortex 2-3 mm anterior to the
bregma and to the right of midline. A reference was
placed approximately 2 mm anterior to the parietal-
interparietal suture on the left side  either in bone (in
the discrimination experiment) or on the dura (in the
operant  experiment).  In other experiments  with the
operant  paradigm used  here,  there  was little differ-
ence between  SP responses  when  the reference at
this site was in bone, on dura or on lesioned cortex.
On the  other hand, if  the cortex  under the active
electrode was damaged, SP response amplitude was
markedly  decreased. EEC was recorded using pre-
amplifers set for dc recording. The 3-Hz half-ampli-
tude high  frequency filter on  the driver amplifier of
one channel was used to obtain  a  recording of the
dc potential change without the associated EEC.

Discrimination procedure

    For conditioning and recording, animals (n=8)
were placed individually in a clear Plexiglas  chamber
10 x 10 x 7.5-in. high with a grid floor. Before elec-
trode implantation, the animals were trained  to assoc-
iate the delivery of reinforcement (45-mg food pellet)
with the sound of the delivery device  so that they
would approach the feeding dish on hearing this cue,
thereby  allowing associative conditioning  to other
auditory cues.  Two tone bursts of 0.5-sec  duration
were  employed  for discrimination  conditioning;
Supported in part by DHEW MH 23653, MH 29653,
by  Tarbox Parkinson's Disease  Institute  at  Texas
Tech University School of Medicine, and by the Mul-
tidisciplinary Research Program  in Mental Health at
the University of Texas Medical Branch. A portion of
this work was performed in the Department of Phar-
macology and Toxicology at the University of Texas
Medical Branch, Galveston.

-------
Drug Effects on ERSPs in Rats
                                              57
one (Sd) was followed after 3 sec by the delivery of a
food pellet while the other (SA) was not reinforced.
The  two stimuli were 400-Hz and 4500-Hz tones at
SOdB.  For some animals the 400-Hz  tone served  as
Sd,  whereas the 4500-Hz tone was  reinforced  in
others.  S" and SA  were  given on alternate trials.
Intertrial  intervals varied  between 45 and 90 sec.
Thirty trials were given in a single session (IS trials
with each stimulus),  and only one session was  con'-
ducted  per day. The animals had free access to Pur-
ina Laboratory Chow during one hour at  the end of
each day.

    Voltage changes  during the 3-sec  period follow-
ing the onset of the  conditioned stimulus were meas-
ured at each 0.25-sec interval using a Tektronix 31/53
data acquisition system. The area of the averaged SP
response to each stimulus was  calculated  and these
values were used for statistical  comparisons and ill-
ustrations.                           '

    d-Amphetamine sulfate was dissolved in 0.9% sa-
line  and was  administered intraperitoneally 30 mln
before trials began. Saline was given before each con-
trol session during the drug testing phase.

Operant procedure

    Before  electrode implantation,   eight  animals
were trained to obtain food reinforcement by press-
ing a lever when it  extended into the chamber. The
lever was made of nonconducting acrylic. In the final
behavioral schedule, the  retractable  lever began to
move into the chamber 2  sec following the onset of
a  20-msec  auditory  warning  stimulus (1600  Hz,
90dB). Once  the lever was activated (approximately
3.9 sec after the onset of the  trial), the rat had 2 sec
to obtain  reinforcement. The lever was inactivated
and  retracted either  when pressed or 2 sec after be-
coming activated. Intertrial Intervals  varied between
18 and 70 sec. Sixty to 65 trials were given In a single
session and only one session was conducted per day.
Acquisition of slow  potential data began  after the
first 20 acclimation trials.

     EEC was recorded  as described  above. Output
from the polygraph was fed into a Nicolet MED-80
Data Acquisition and Analysis System. An additional
channel was used to record  the lever-press, and the
signal was also fed  to the computer  for analysis of
response latencies.  The data for each  trial (including
500 msec of prestimulus  baseline data) were  digit-
ized at a rate of 78  samples per second and stored on
a floppy disk. At the end of the session, artifact-free
trials were averaged, lever-press latencies were meas-
ured,  and the maximum  negative amplitude of the
averaged waveform was determined. The mean of 39
samples obtained during the 500 msec prestimulus
period  served  as zero baseline.
    d-Amphetamine sulfate was  administered intra-
peritoneally 30 min before  data acquisition began
and pentotarbital sodium (dissolved in distilled water)
was given subcutaneously  15 min before  acquiring
data. Results are reported for those animals in which
SP responses remained sufficiently artifact-free for a
long enough period of time to allow administration of
all doses of a drug to an individual rat.
Baseline correlations

    In most animals,  the recording electrodes were
sufficiently stable that little or no drift occurred dur-
ing the experimental session. During the initial 20 ac-
climation trials the potential difference between the
two electrodes was balanced and no further adjust-
ment was made. Through the use of the data acquired
with  the 3-Hz half-amplitude high frequency filter,
the prestimulus dc baseline values obtained for each
trial  during the operant  schedule  were stored, the
most positive baseline  for any trial  in the session was
found, and this value was subtracted from all baseline
values (i.e.,  the most positive baseline was adjusted to
zero  and all other baselines were negative relative to
this point). The  maximum negative amplitude was
also determined for each trial. The Pearson product-
moment correlation  coefficient was calculated to
examine the relationship between baseline and amp-
litude. All measurements and calculations  were per-
formed by the computer.
 Results

     When the  conditioned  stimulus  was presented
 without  reinforcement,  Initially  observed negative
 slow potential (SP) responses decreased in amplitude
 after a few  sessions. This habltuation developed to
 each conditioned stimulus used in these experiments.
 When reinforcement was instituted, the amplitude of
 the SP responses increased several fold and reached
 maximum within a few sessions. Polygraph tracings
 of SP responses during four trials in one rat trained in
 the operant procedure are shown in the left panel of
 Fig.  1. The  right panel  of Fig. 1 illustrates the aver-
 aged SP response of 40  trials in the session. The neg-
 ative slow potential response began shortly after the
 stimulus  and retruned  to baseline following the de-
 livery of reinforcement. In some cases a positive shift
 developed after reinforcement, similar to  the postre-
 inforcement positive shift observed in cats by Marc-
 zynskietal.(1969).

-------
 58
                                            Pirch
    4-2 »c-*
F& 7. EEC recordings and averaged SP response from
a rat  performing the  operant  task.  Left panel:  re-
cordings from four single trials; first arrow indicates
onset of warning stimulus; time  between  arrows is
2 sec; calibration is 250 //K;  negative is up.  Right
panel:  average of 40  trials in session from  which
the single trial recordings  were obtained; sweep time
is 2 sec; calibration is 50 n V, negative is up.

Discrimination procedure
    The first 10 training  sessions on the discrimina-
tion schedule demonstrated a phase of stimulus gener-
alization during  which large SP responses developed
to both stimuli. By session 12, the responses to the
reinforced  stimulus (S")  were significantly greater
than  responses to the nonreinforced stimulus (S^)
(t-test for  paired  comparison  of SP response  areas,
p <.05).  This  difference increased  and   persisted
throughout the remaining 25 or more sessions.

    Amphetamine produced a dose-related  depress-
ion  of  SP responses  to  the  reinforced stimulus at
doses  of 0.25 to  2 mg/kg (Fig. 2a). The effect  on
responses  to  the nonreinforced  stimulus was, how-
ever,  biphasic and depended  upon  the dose. The
lower doses (0.25 and 0.5 mg/kg) enhanced the  re-
sponses, the  intermediate  dose (1 mg/kg)  produced
no change, while the  high dose (2 mg/kg) depressed
the  SP responses (Fig.  2b).  Examination of the
effects of  0.5  mg/kg clearly  shows that  this dose
of amphetamine depressed  the responses to Sd at the
same  time that responses to SA were enhanced. These
experiments illustrate  the importance of dose and of
stimulus significance  in  determining the  action of
drugs  on event-related slow potential responses.

Operant procedure

    Fig. 1  shows samples  of single trial SP responses
in  one animal, recorded during performance of the
operant task, and Fig.  3 illustrates  averaged  SP re-
sponses in four different animals, obtained in control
sessions.  Mean   lever-press  latencies  for  control
sessions ranged from 70 to 750 msec after lever acti-
vation, depending upon the animal.

    Both  d-amphetamine  and  pentobarbital  caused a
dose-related  reduction of  the  amplitude of SP res-
ponses  (Fig. 4).  The doses  of  amphetamine were
0.25, 0.5, and  1.0 mg/kg while  pentobarbital was
200 a. RESPONSE TO S

150-
a.
LU
f 100-

111
B
85 50

n.
, b. RESPONSE TO&A
•°,5 <:°1 <.ooi
rl J^T 1







!













JL
1
1














1.














1








D CONTROL
El DRUG



<.06
I
r1!
V,;
[if



<0°5..<.005
1


1


IMS
1 r1!
;
,' P'
1


1,
          0.25  0.5  1.0  2.0    0.26  0.5  1.0  2.0
             DOSE OF d AMPHETAMINE, mg/kg

 Fig. 2.  Effect of d-amphetamine on SP responses to
 reinforced (Sd) and nonreinforced (S&) stimuli. Each
 bar  represents  the mean of SP responses (calculated
 areas) from eight animals.  Vertical lines are standard
 errors.  Open  bars  show  the responses after saline
 treatment while shaded bars show the responses after
 treatment with various doses of d-amphetamine. Re-
 sponses were  negative in polarity. P values are based
 on paired-t comparisons.
Fig. 3.  Averaged SP responses from control sessions
in four different animals performing the operant task.
Average of 40 trials.  The time of each response is the
2-sec period following the onset of the warning stimu-
lus. Negative is up.

given in doses of 5,10, and 15 mg/kg. The number of
lever-presses was  unaffected by the two lower doses
of amphetamine and  there were no consistent changes
in latency. After  1 mg/kg, the four animals shown in
Fig.  4  obtained  reinforcement in 98, 93,  63,  and

-------
Drug Effects on ERSPs in Rats
                                               59
50% of  the  trials and the lever-press  latencies were
not  significantly altered. The  three animals treated
with  pentobarbital  obtained  100%  reinforcement
after 5  rng/kg; 98, 88, and 50% after  10 mg/kg; and
38,  10, and  0% after 15 mg/kg. There were no con-
sistent changes in lever-press  latencies,  with  some
animals  showing  increases  and others decreases  as
compared  with the  previous  control  session. It  ap-
pears that  if lever presses occur after drug treatment,
they are within the range of control latencies. The av-
eraged SP responses for trials in which the animals ob-
tained reinforcement were larger than the responses
for trials in which no lever press was made.

    These  experiments  indicate that both stimulant
and  depressant drugs can suppress event-related slow
potentials  recorded  from the  rat cortex during  an
operant task.  Under the conditions of these experi-
ments, the slow  potential response was a more sen-
sitive indicator of drug effect than  the behavioral
measure.
          correlations
    Slow potential responses to the warning stimulus
were greater in amplitude during those trials in which
the baseline dc  level was more  positive as compared
with the amplitude  of responses in trials with more
negative baselines. Fig. 5 illustrates this relationship
for a session in one animal. Averaged SP responses for
the 20 trials with the more positive (larger response)
and the 20  trials with the more  negative baselines are
shown in the insert. Significant inverse correlations
between baseline negativity and SP response ampli-
tude were observed  in eight  rats  in which baseline
   75-
   50-
o.
Z 25
   0 J
       0     0.25     0.5     l!o  0      5     10   15
       DOSE d-AMPHETAMINE,  DOSE PENTOBARBITAL,
               mg/kg                  mg/kg

Fig.  4.  Effects of d-amphetamine and pentobarbital
on  SP responses associated with the  operant  task.
SP  response amplitude is plotted as percent of the
amplitude  of the  vehicle control sessions with  each
rat  serving as its own control. Hach curve represents
an individual animal.
correlations were  studied. These experiments indi-
cate that  the magnitude of slow potential change in
response to a conditioned stimulus may vary accord-
ing to the dc potential level at the time of stimulus
presentation.
     -240 -i
     -200 -
   UJ
   a

   t -150-j
   a.

   a.
   w -100
   O
   5
   X
   <  -50 ^
                                                               0-
r = 0.701
p <0.01
            0      -50     -100     -150     -200
                  RELATIVE BASELINE, JUV

Fig.  5.   Relationship  between baseline dc potential
and  SP response amplitude. Data from 40 trials of a
single operant session  in one rat.  See text for com-
plete description. Averaged SP responses for the 20
trials with the more positive (larger response} and the
20 trials  with the more negative baselines are shown
in the  insert. Calibration  is 25 \lV.  Negative is up.


Discussion

     The results of these studies indicate that event-
related  slow potential responses recorded from the
cerebral cortex  of rats are  altered  by stimulant and
depressant  drugs  in a dose-dependent manner.  Fur-
thermore, the experiments  utilizing a discrimination
procedure suggest that the  drug effects are  also de-
pendent upon the behavioral significance  of the  elici-
ting stimulus. Thus, with a single drug one may ob-
serve enhancement  or depression  of SP responses.
With appropriate doses and behavior schedules, drugs
which  are members  of different  pharmacological
classes  can produce similar effects on SP response.
The fact that drugs with different modes of action
can  cause similar alteration of event-related slow po-
tentials suggests that caution should be exercised  in
interpreting the  meaning of a drug effect, especially
when based on single-dose  or even two-dose studies.

     It is  clear that chemical agents which affect  arou-
sal  mechanisms  produce   changes  in ERSP ampli-
 tudes. Other factors which  alter arousal such as the

-------
60
                                                                                                 Pirch
experimental setting or behavioral task can also be ex-
pected to influence slow potential responses. A role
for arousal mechanisms in determining amplitude of
the CNV was proposed by Irwin et al. (1966) and Re-
bert et al. (1967).  Tecce (1972) further suggested
that an inverted-U relationship exists between level of
arousal and  amplitude of slow  potentials.   Using
Tecce's model, one might predict  that if the arousal
level were optimal for maximum SP amplitude under
the conditions  of the experiment, either an increase
in arousal induced by a stimulant drug or a  decrease
in arousal produced by a depressant drug could cause
a reduction of the slow potential amplitude.  Such a
model might apply to the effects of amphetamine and
pentobarbital in  the present experiments using the
operant task. In other studies with rats trained on a
simpler paradigm  in which food reinforcement was
automatically  delivered, both d-amphetamine  and
pentobarbital reduced  SP response amplitude  (Pirch
and  Osterholm  1975). When doses  of each agent
which  depressed  SP amplitude were combined, the
SP response  was  restored to control levels.  Those
results would be consistent with the concept of an
inverted-U relationship between arousal and SP amp-
litude. Arousal is  difficult to define, however, and
some other  process such as attention may be more
directly  related to  amplitude of  slow potentials.
Although some consistent  relationships appear  to
exist  between arousal  and  ERSP  amplitude,  the
effects of amphetamine  on the discrimination pro-
cedure suggest that other processes may also play a
role.
    An additional factor which has been proposed to
affect  ERSP  amplitude  is the baseline dc potential.
Knott  and Irwin (1968) suggested a "ceiling effect"
whereby greater negativity of the dc level at the time
of stimulus presentation results in a limitation of the
negative slow potential change which can develop in
response to the stimulus. The results of the dc base-
line  studies reported here are consistent with  the
"ceiling"  hypothesis. Additional  experiments  are
under  way to determine whether the drug-induced
changes in SP responses are related to alterations of
the baseline dc level.

-------
CHOLINERGIC MECHANISM OF SLEEP ONSET
POSITIVE VARIATION AND SLOW POTENTIALS
ASSOCIATED WITH  K-COMPLEXES  IN  CATS
T. J. MARCZYNSKI

Department of  Pharmacology,  University  of Illinois at  the Medical Center,Chicago,IL, U.S.A.
    In the cat, sleep onset is characterized by high-
voltage, rhythmic (7 to 9 c/sec), alpha-like bursts
over the pane to-occipital cortex with highest am-
plitude  over primary  and  secondary visual  pro-
jections.   These electrocorticogram  (ECoG)  pat-
terns  are  always associated with large epicortical
positive slow potentials (SPs) the amplitude of which
ranges from  200  to SOO^iV if recorded epidurally with
reference  to subjacent white matter or a  distant
relatively  indifferent  electrode  located  over  the
anterior ectosylvian  gyrus.  Since this localized SP
always occurs during sleep onset, it was termed "sleep
onset positive variation" (SOPV)  (Marczynski et at.
1971b).

    During  slow wave  sleep (SWS) mild  auditory,
somatosensory, or visual stimuli trigger K-complexes
that are immediately followed by bursts  of alpha-
like activity  and  SPs whose topograpical distributions
are  virtually  identical   with   those   of SOPV
(Marczynski  et  al. 1969).  The assumption that the
SOPV phenomenon  reflects an  active process of
internal inhibition in the Pavlovian sense is supported
by the observation that, in a relaxed cat, 7- to 9-c/sec
flash stimuli produce responses indistinguishable  from
SOPV and  shorten  the  period  necessary for the
emergence  of SWS  (Marczynski  and Sherry 1972).
Hence, the SP associated with K-complex (K-SP) can
be regarded as reflecting a homeostatic mechanism
aimed at maintaining SWS.

    In the present study, the effect of mild auditory,
somatosensory,  visual,  and  vestibular  stimuli  were
investigated  in  the  cat on  the emergence  of  K-SP
prior to and after  administration  of scopolamine
hydrobromide.   In addition, the SOPV phenomenon
was also quantified  prior to and  after drug adminis-
tration.

Methods

    Four adult cats were used.  Under pentobarbital
anesthesia, Ag/AgCl  electrodes  were implanted epi-
durally  over the  posterior  marginal (M), medial
suprasylvian (S)  and anterior ectosylvian  gyri (E).
The  technique   of implantation,  recording,  and
integration  of SPs, using Grass cumulative  inte-
grators,  has been previously  described (Marczynski
et al. 1971b). The output of the integrator was ad-
justed to produce  two  pen deflections in response
to a positive shift of lOO^V lasting 1 sec.  This value
was accepted as one unit of SP. Only the SP between
the S  and  M gyri were quantified and  used for
statistical  evaluation,  using  the Student's  t-test.
Experiments were conducted  in a 1-m sound-atten-
uating  Lehigh  Valley  test chamber  provided with
dim light and a one-way window. Auditory stimuli
of approximately 6 dB above  background noise were
presented  through a loudspeaker attached to the
ceiling.  The Grass visual stimulator unit,  modified
to produce noiseless flashes, was also attached to the
ceiling.  Electric somatic stimulation was  delivered
through a pair of stainless steel electrodes implanted
subcutaneously on the  cat's  back.  The electrodes
were connected to  a Grass S-8 stimulator and stimu-
lus isolation unit; the parameters of stimulation em-
ployed were 3 - 5V and 0.5 • 0.8 msec  duration.
Experiments were conducted between 2  and 5 p.m.
One hour  prior to the session, the cat was provided
with food  and milk.  After  satiation, the cat was
allowed to sleep  in the basket suspended by strings
from  the  ceiling  of the chamber such  that,  when
one  string was pulled, the basket  would sway to
provide vestibular stimulation. K-SPs were quantified
by counting the  number  of integrator  deflections
during a 7-sec time period following stimulus pre-
sentation.    At least 10  K-SPs  were obtained for
each  modality  and  each  dose  of scopolamine
hydrobromide.   SOPV  responses were quantified
by  counting  integrator  deflections  from  the
moment  the cat  assumed  sleep posture to the
emergence  of SWS lasting at least 10 sec. Output
of  the  integrator  was set to zero when the cat
assumed sleep  posture.  If for any  reason the cat
did not develop  SWS,  raised the head,  or left the
basket, the record was disregarded.

-------
 62
                                     Marczynski
  Results

 SOPV responses

     In most instances, 4 to 7 SOPVs, each lasting 3
 to 12 sec, were  observed prior  to  the emergence of
 SWS.  Fig. 1A shows three SOPV responses in one cat.
 Usually, after the first and second SOPV, a "residual"
 positivity was  observed, larger over the M gyrus and
 smaller over  the S gyrus, thus causing a tonic SP shift
 below baseline in the S-M lead and above baseline in
 the S-E  lead. Residual SPs did not  continue to accu-
 mulate with  subsequent SOPVs and  tended to dissi-
 pate with the  onset of SWS (marked with a vertical
 dashed line)  as shown in Fig. 1A (right) and IB (left)
 of the continuous record. The use of the S gyrus as
 the reference caused the  typical mirror  reversal pat-
 terns due to the decreasing potential gradient along
 the M-SOE line.  A detailed description of the topo-
 graphical distribution  of the SOPV has been report-
 ed elsewhere  (Marczynski et al. 1971 b).

     During control sessions, the sum across cats of
 SOPV responses  observed  from  the moment  the
 animal assumed sleep  posture' to emergence of SWS
 ranged from  46 to 105 Units (mean 56.7; S.D. + 3.4;
 N  =  50).  In cats treated with scopolamine hydro-
 bromide  (0.02   mg/kg,  i. m. ) 30 min prior to a
 session,  this  value was reduced  by  27.6%  (S.D. +
 0.6; p < 0.01).    The dose  of 0.04  mg/kg reduced
 responses by an  average of 55.3% (S.D.  +  0.08;
 p < 0.0001).    After the  latter  dose, the alpha-like
 bursts during  SOPV  responses  never  occurred in
 long  rhythmic  trains, but  were  often  interrupted
 by irregular  delta  waves.   Despite  the conspicuous
 reduction  in  the amplitude of  SOPVs, their dur-
 ation (6.7 sec per  single response;   S.D. +_  1.4) was
 not significantly  different than  that of the control
 SOPV   (7.1  sec;  S.D.  +. 1.9;p< 0.05). After higher
 doses of  scopolamine (0.06 mg/kg, i.m.), the back-
 ground  amplitude of  the  ECoG  increased and it
 was difficult  to  decide  when  the  SWS  emerged.
 Choppy  alpha-like  bursts,   if present,  were  asso-
 ciated  with  small SP shifts (Fig.  1 bottom  left).
K-complex SP (K-SP)

    The control K-SP in response to an auditory (A),
somesthetic (S), vetibular (V) or flash (F) stimulus
are shown in  Fig. IB. The patterns and topographical
distributions of these responses were virtually identi-
cal for all  modalities tested  and closely  resembled
those of the SOPV. In cats treated with scopolamine
(0.02 mg/kg i.m.),  K-SPs were reduced by  an average
of 32.2% (S.D. ± 0.7; p< 0.01), and  doses of 0.04
mg/kg reduced these responses by 67.4% (S.D± 0.9;
p< 0.001). As shown in  Fig. 1C (right), a dose of sco-
polamine 0.06 mg/kg almost totally suppressed these
 responses and the EcoG and SP remained  "fro/en"
 within a relatively narrow band of fluctuations.

 Discussion

    Since  scopolamine  blocks  SOPV  and   K-SP
 responses,   it  most  likely  interferes   with  the
 process of internal inhibition.  With moderate doses
 of  scopolamine,  the  duration of "abortive" SOPVs
 did  not change,  although  their amplitude was  re-
 duced.   This indicates that "primary"  hypnogenic
 influences  which  trigger SOPV responses  are  not
 affected, but the "execution" of SOPVs  is impaired,
 most likely at the thalamocortical level. Likewise,
 scopolamine  suppresses alpha-like postreinforccment
 ECoG  synchronization  (PRS)  and  the associated
 reward  contingent  positive  variation  (RCPV)  in
 cats trained to lever press for milk  reward, although
 the  duration  of  the choppy  PRS  and  that of the
 abortive RCPV responses is not significantly changed
 (Marczynski 1971).

    The phasing  theory of neuronal  activity,  based
 on  recurrent  inhibitory  circuits  in  the thalamus,
 which are reputed to control  alpha rhythm, implies
 that a certain level of "synaptic pressure"  is necessary
 to initiate  and drive  the phasing circuits (Andersen
 and   Andersson  1968).    Numerous  microelectro-
 phoretic studies of single neurons in specific thalamic
 relay  nuclei  (Steiner  1968,  McCance et  al.  1968,
 Tebecis 1970, Phillis  1971)  suggest that the  primary
 role of chohnergic projections  described by  Shute
 and  Lewis (1967) is facilitation of  sensory input.
 Hence, blockade of cholinergic influences may result
 in reduction of synaptic pressure  to  a level insuffi-
 cient  to initiate and drive phasing circuits  in thalamus
 and  cortex.   Evidence  that  further supports this
 contention  and  the  hypothetical  interactions  be-
 tween cholinergic and monoaminergic  projections
 that seem  to  control  the phasing mechanisms have
 been  discussed  elsewhere  (Marczynski  and  Burns
 1976 and this volume).

    The occurrence  of  SOPV and K-SP in associ-
 ation with  alpha-like  bursts  can  be  interpreted  as
 resulting  from a phasic tendency  toward hyper-
 polarization  of  large populations of  neurons  in
 the  cortex  and  electrotonic spread of IPSPs  to
 apical  dendrites   which  is  reflected  as  surface
 positivity   (Creutzfeldt  et   al.   1969).    During
 bursts of  alpha activity  in  the  thalamus (Andersen
 and Andersson 1968) and visual cortex (Creutzfeldt
et al. 1969),  many  neurons show  primary  IPSPs
 (i.e.,  not   preceded  by EPSPs) and may remain
 silent  in   a   state  of  hyperpolarization.   These
 neurons are  the  most  likely  source  of  phasic sur-
 face positive SPs  associated  with alpha-like bursts.
The   participation of  glia  cells  may  be secondary
 to phasic  changes in  concentration  of  ions in  the
extraneuronal fluid.

-------
Cholinergic mechanism of SOPV
                                             63
     The  virtually  identical  patterns  and  distri-
 bution  of SOPVs  and K-SPs  triggered by  various
 modalities can  be explained  by  the fact  that,  in
 the  visual cortex, there  are  neurons that respond
 to   auditory,  somesthetic, and  vestibular  stimuli
(cf. Morrell 1972). The equipotential role of various
modalities  in generating K-SPs and their localization
over  primary  and   secondary  visual  projections
indicate that striate  and parastriate  cortex plays  a
dominant   role  in  internal  inhibitory  processes.
                                               CONTROL
                                                                                             100/iV
           INT

                                SCOPOLAMINE HBr 60 \i g/kg, I. M. 30 min
Fig. 1.  Typical SOPV responses and K-SPs during slow wave sleep to auditory (A), somatosensory (S), vestibular
(V) and flash (F) stimuli (top and middle, respectively).  Scopolamine  hydrobromide  blocks these responses
(bottom). For further explanations, see text.

-------
 ELECTROGENESIS  REFERENCES
 Abraham, P. and McCallum, W. C. The CNV and the
    spiral  aftereffect.  Electroenceph.  clin.  Neuro-
    physiol., 1973, Suppl. 33:205-208.

 Aladjalova, N. A.  Slow  electrical processes in the
    brain. Elsevier, Amsterdam, 1964.

 Albers, R. W. and Brady, R. O. The distribution of
    glutamic decarboxylase in the nervous system of
    the rhesus monkey.  J. biol.  Chem., 1959, 234:
    926-928.

 Andersen, P. and Andersson, S. A. Physiological basis
    of  the  alpha rhythm. Appleton-Century-Crofts,
    New York, 1968:235.

 Anderson,  R. Unpublished data cited  in A. Katch-
    alsky, V. Rowland,  and  R.  Blumenthal (Eds.),
    Dynamic patterns  of brain cell assemblies. Neu-
    rosci. Res. Prog. Bull., 1974, 12:111-121.

 Arduini, A. Enduring  potential changes evoked in the
   cerebral cortex by stimulation of brain stem reti-
   cular formation and thalamus. In H. Jasper, L.
   Proctor,  R. Knighton, W. Noshay and R. Costello
   (Eds.),  Reticular formation of  the  brain.  Little,
   Brown, Boston, 1958:333-350.

 Ashton, H., Millman, J.E., Telford,R. and Thompson,
   J. W. Stimulant and depressant effects of cigarette
   smoking on brain activity in man. Br. J. Pharmac.,
    1973,48:715-717.

Ashton, H., Millman, J.E., Telford.R. and Thompson,
   J. W. The effect of caffeine,  nitrazepam and cig-
   erette smoking on the CNV  in man. Electroen-
   ceph. clin. Neurophysiol., 1974, 37:59-71.

Ashton, H., Millman,  J. E., Telford, R. and Thomp-
   son, J. W. A comparison of some physiological and
   psychological effects of propranolol and diazepam
   in normal subjects. Brit. J. clin. Pharmac., 1976,
   3:551-559.

Bauer,  R. H. and Fuster, J. M. Delayed-matching and
   delayed-response deficit from  cooling dorsolateral
   prefrontal cortex in  monkeys. J. comp.  physiol.
   Psychol., 1976, 90:293-302.

Berger, B., Tassin, J. P.,  Blanc, G., Moyne, M. A. and
   Thierry, A. M. Hutochemical confirmation for do-
   paminergic innervation of the rat cerebral cortex
   after destruction of the noradrenergic ascending
   pathways. Brain Res.,  1974, 81:332-337.
 Bockaert, J., Tassin, J. O., Thierry, A. M., Glowinski,
   J. and Fremont, J. Characteristics of dopamine and
   beta-adrenergic-sensitive  adenylate cyclases in the
   frontal cerebrai*cortex of the rat. Brain Res., 1977,
   122:71-86.

 Borda, R. P, Drive and performance related aspects of
   the CNV in reheus monkeys. Electroenceph. clin.
   Neurophysiol., 1970, 29:173-180.

 Braden, W., Stillman, R. C. and Wyatt, R. J. Effects
   of marihuana on contingent negative variation and
   reaction time. Arch. Gen. Psychiat., 1974, 31:537-
   541.

 Bradley, P. B. The central action of certain drugs in
   relation to the reticular  formation of the brain.
   In H. Jasper, L. Proctor, R. Knighton, W. Noshay,
   and  R.  Costello  (Eds.),  Reticular formation  of
   the brain. Little, Brown, Boston,  1958:123-149.

 Bradley, P. B. Dhawan, B. N. and Wolstencroft, J. H.
   Pharmacological properties of cholinoceptive neu-
   rones in the medulla and pons of the cat. J. Phy-
   siol. (Lond.), 1966,183:658-674.

 Brimble, M. J. and Wallis.  D. I. The role of muscarinic
   receptors in synaptic transmission and its modula-
   tion  in the rabbit superior cervical ganglion. Eur.
 .  J.Pharmec., 1974,29:117-132.

 Brown, K. T. and Flaming.  D. G. Instrumentation
   and techniques for beveling fine micopipette elec-
   trodes. Brain Res., 1975, 86:172-180.

 Buchwald, N. A., Hull, C. D. and Trachtenberg, M. C.
   Concomitant behavioral and neural inhibition and
   disinhibition in response to subcortical  stimula-
   tion. Exp. Brain Res., 1967,4:58-72.

 Caspers, H. Changes of cortical d. c. potentials in the
   sleep-wakefulness cycle. In G. E. W. Wolstenholme
   and M. O'Connor (Eds.), Ciba Foundation sympos-
   ium on the nature of sleep. Little, Brown; Boston,
   1961:237-253.

 Caspers, H. Relations of steady potential shifts in the
   cortex to the wakefulness-sleep spectrum.  In M. A.
   B. Brazier (Ed.),  Brain function.  Univ. of Calif.
   Press, Berkley, 1963:177-200.

Castellucci, V.  F. and  Goldring, S. Contribution to
   steady potential  shifts of  slow depolarization  in
   cells  presumed  to be  glia. Electroenceph.  clin.
   Neurophysiol., 1970, 28:109-118.

-------
Electrogenesis references
                                             65
Caton, R. The electric currents of the brain. Br. med.
   J., 1875,2:278.

Celesia, G. G. and Jasper, H. H. Acetylcholine releas-
   ed from the cerebral cortex in  relation to state of
   activation. Neurology, 1966, 16:1053-1063.

Clemente, D. C., Sterman,  M. B.  and Wyrwicka, W.
   Post-reinforcement EEC synchronization during
   alimentary behavior.  Electroenceph.  clin. Neuro-
   physiol., 1964, 16:355-365.

Cohen, M. W. The  contribution by glial cells to sur-
   face recordings from the optic  nerve of an amphi-
   bian. J. Physiol.,  1970, 210:565-580.

Collier, B. and  Mitchell, J. F. The central release of
   acetylcholine during stimulation of the visual path-
   way.  J.  Physiol. (Lond.),  1966, 184:239-254.

Corazza, R. V., Tradardi, V. and Umilta, C. Tonic re-
   sponses to steady diffuse light  illumination of the
   neuronal discharge  in  the cat visual  pathways.
   Brain Res., 1971,27:241-250.

Creutzfeldt, O. D.,  Rosina, A., Ito, M. and Probst, W.
   Visual evoked response of single cells and the EEG
   in the primary visual area of the cat. J. Neurophys-
   iol., 1969,32:127-139.

Curtis, D. and Johnston, G.A.R. Amino acid transmit-
   ters  in  the  mammalian central nervous system.
   Ergebn. Physiol., 1974, 69:97-188.

Dahlstrom, A. and  Fuxe, K. A method  for the dem-
   onstration of monoamine  containing nerve fibers
   in the central nervous system. Acta. physiol. Scand.,
   1964,60:293-295.

Donchin, E., Otto,  D., Gerbrandt, L. K.  and  Pribram,
   K. H. While a monkey waits: Electrocortical events
   recorded during  the foreperiod of a reaction time
   study. Electroenceph. clin.  Neurophysiol., 1971,
   31:115-127.

Dongier, M.  Clinical application  of the CNV: A re-
   view, Electroenceph,  clin. Neurophysiol.,  1973,
   Suppl. 33:307-315.

Dunant, Y. and Dolivo, M. Relations entre les poten-
   tials synaptiques lents et 1'excitabilite' du ganglion
   sympathique chez le rat. J. Physiol., Paris, 1967,
   59:281-294.

Eccles, J. C. The physiology of  synapses, Springer-
   Verlag, Berlin, 1964.

Eccles, R. M. and Libet, B. Origin and blockade of
   the synaptic responses  of  curarized  sympathetic
   ganglia. J. Physiol., Lond.,  1961,157:484-503.
Engberg, I. and Marshall, K. C. Reversal potential for
   noradrenaline-induced hyperpolarization of spinal
   motoneurones of cats. J. gen. Physiol., 1973, 61:
   261.

Evarts, E. V. Motor cortex reflexes associated with
   learned movement.  Science, 1973, 179:501-503.

Evarts, E. V. Pyramidal tract activity associated with
   a conditioned hand  movement in the monkey. J.
   Neurophysiol., 1966, 29:1011-1027.

Ferguson, H. H. and Jasper, H. H. Laminar D. C. Stu-
   dies  of acetylcholine-activated  epileptiform  dis-
   charges in  cerebral  cortex.  Electroenceph. clin.
   Neurophysiol., 1971,30:377-390.

Freeman, W. J. Mass action in the nervous system.
   Academic Press, New York, 1975.

Freeman, W. J. Quantitative patterns of integrated
   neural  activity.  In J. Fentress (Ed.) Simpler net-
   works. Sinauer, Sunderland, Mass., 1976.

Frigyesi, T.  L.  Intracellular recordings from neurons
   in dorsolateral thalamic reticular nucleus  during
   capsular,  basal ganglia and midline thalamic stim-
   ulation. Brain Res., 1972,48:157-172.

Fromm,  G.  H. and Bond, H. W. Slow changes in the
   electrocorticogram and the activity of cortical neu-
   rons.  Electroenceph.  clin.  Neurophysiol,   1964,
   17:520-523.

Fromm,  G.  H. and Bond, H. W. The relationship be-
   tween neuron activity and cortical steady  poten-
   tials.  Electroenceph.  clin.  Neurophysiol.,   1967,
   22:159-166.

Fuster, J. M. Unit activity in prefrontal cortex during
   delayed-response  performance:   Neuronal   corre-
   lates of transient memory. J. Neurophysiol., 1973,
   36:61-78.

Fuster, J. M. and Alexander, G. E. Delayed-response
   deficit by cryogenic depression of frontal cortex,
   Brain Res., 1970,20:85-90.

Fuster, J. M.  and  Alexander, G. E. Neuron activity
   related  to short-term   memory.  Science,  1971,
    173:652-654.

Fuster, J. M.  and Bauer, R. H.  Visual short-term
   memory  deficit from hypothermia of frontal cor-
   tex. Brain Res., 1974, 81:393-400.

Fuxe, K. Distribution of monoamine nerve terminals
   in the central nervous system. Acta Physiol. Scand
    1965,64 Suppl. 247:39-85.

-------
 66
                       Electrogenesis references
 Fuxe,  K., Hamberger, B. and Hokefelt, T.  Distribu-
   tion of noradrenaline nerve terminals in cortical
   areas of the rat. Brain Res., 1968, 8:125-131.

 Fuxe,  K., Hokfelt, T.,  Johansson, O., Lidbrink, G.
   and  Ljungdahl, A. The origin  of  the dopamine
   nerve terminals in limbic and frontal cortex. Evi-
   dence of  mesocortical dopamine neurons.  Brain
   Res., 1974, 82:349-355.

 Gantchev, G. Neuronal activity and motor potentials
   in the sensory-motor cortex  of monkeys during
   spontaneous  movements.  Electroenceph.   clin.
   Neurophysiol., 1974,  37:321.

 Geard, R. W. and Libet, B. The control of normal
   and  "convulsive" brain  potentials. Am. J. Psy-
   chiat., 1940,96:1125-1151.

 German, D. C. and  Bowden,  D. M. Catecholamine
   systems as the neural substrate for intracranial
   self stimulation: A hypothesis. Brain Res.,  1974,
   73:381419.

 Gillespie,  O. A.  Effects of pre-trial  electrocortical
   contingencies on  acquisition of delayed  response
   and delayed matching-to-sample in monkeys. Un-
   published  Doctoral Dissertation, State University
   of New York at Stony Brook, N. Y., 1977.

 Groll-Knapp, E., Wagner, H., Hanek, H. and Haider,
   M. Effects of low carbon monoxide concentrations
   on vigilance and computer-analyzed brain poten-
   tials. Staul-Reinhalt. Luft., 1972, 32:64-68.

Gross,  C- G.  Visual functions of inferotemporal cor-
   tex. In R. Jung (ed.),  Handbook of sensory physi-
   ology. Berlin, Springer, 1972, 7:451-482.

Gumnit, R. J. and Grossman, R. G. Potentials evoked
   by sound in the auditory cortex of the cat. Am. J.
   Physiol., 1961,200:1219-1225.

Hablitz, J. J. Operant conditioning and slow poten-
   tial  changes from  monkey cortex. Electroenceph.
   clin. Neurophysiol., 1973, 31:399-408.

Hablitz, J.  J.  and Borda, R. P. The effects of Dai-
   mane  (flurazepam  HCL) on  the  CNV.  Electro-
   enceph. clin. Neurophysiol., 1973, Suppl. 33:317-
   320.

Hackett,  J. T. and Marczynski,  T.  J. Positive rein-
   forcement and visual evoked potentials in the cat.
   Brain Res., 1971, 26:57-70.

Haider, M., Gangleberger, J. A. and Groll-Knapp, E.
   Computer  analyzed thalamic potentials and  their
   relation to expectancy waves in man. Acta neurol.
   latinoam., 1963, 14:132-137.
 Hayward, J.  N,, Fairchild, M, D. and Stuart, D. G.
   Hypothalamic and cortical D-C potential changes
   induced by stimulation of the mid brain reticular
   formation. Exp.  Brain  Res.,  1966,  1:205-219.

 Herrington,  R.  N.  and Claridge,  G.  S.  Sedation
   threshold  and  Archimedes'  spiral after-effect  in
   early  psychosis. J.  Psychiat.  Res., 1965, 3:159-
   170.

 Hoffer, B.  J. and  Bloom, F.  E. Norepinephrine and
   central neurons. In  Ch. H. Hockman and D. Sie-
   ger (Eds.), Chemical transmission in the mammal-
   ian central nervous  system. University  Park Press,
   Baltimore, 1976:327-348.

 Hoffer, B. J., Siggins, G.  R., Oliver, A. P. and Bloom,
   F. E. Cyclic AMP-mediated adrenergic synapses to
   cerebellar  Purkinje cells. In P. Greengard and G. A.
   Robison (Eds.), Advances  in cyclic nucleotide re-
   search. Vol. 1,  Raven, New York, 1972:411-423.

 Horn, G. The effect of the somaesthetic and photic
   stimuli on the activity of  units in the striate cor-
   tex of  unrestrained  cats. J.  Physio).  (Lond.),
   1965, 179:263-277.

 Hull, C. D., Bernardi, G.  and  Buchwald, N. A. Infra-
   cellular responses of caudate neurons to brain stem
   stimulation. Brain Res., 1970, 22:163-179.

 Irwin,  D. A., Kakolewski, J.  W., Criswell, H. E. and
   Popov.,  A. An injury-induced diffuse slow poten-
   tial from  brain.  Electroenceph.  clin.  Neurophy-
   siol., 1975,38:367-377.

 Irwin,  D, A., Knott, J. R. McAdam, D. W. and Re-
   bert, C.  S. Motivational determinants of the "Con-
   tingent Negative Variation," Electroenceph, Clin.
   Neurophysiol., 1966, 21:538-543.

 Irwin, D. A. and Rebert, C. S. Slow potential changes
   in cat brain during  classical appetitive condition-
   ing of jaw movements using two levels of reward.
   Electroenceph. clin.  Neurophysiol., 1970,28:119-
   126.

Jasper,  H.  Talk  presented  at  Society  for Neuro-
   science, EEC dinner, 1975.

Jasper, H. H. and Koyama, I. Rate of release of amino
   acids from the cerebral cortex in  the cat as affect-
   ed by brainstem and thalamic stimulation. Can. J.
   Physiol. Pharmac.  1969,47:889-905.

 Jonn, E. R.  Walker, P.,  Cawood, D., Rush, M. and
   Gehrmann, J.  Factor analysis of evoked poten-
   tials. Electroenceph.  clin. Neurophysiol., 1973,
   34:33-43.

-------
Electrogenesis references
                                              67
Johnston, G. A. R. Amino acid inhibitory transmit-
   ters in the central nervous system. In Ch. H. Kock-
   man and D. Bieger (Eds.), Chemical transmission
   in  the mammalian central nervous system.  Balti-
   more, University Park, 1976:31-81.

Jones, E. G. and Powell, T.P.S.  An anatomical  study
   of converging  sensory pathways within the cere-
   bral cortex of the  monkey.  Brain,  1970, 93:793-
   820.

Jordan, L. M.  Noradrenergic synapses and the effects
   of micro-electrophoretically applied noradrenaline
   in  the ventral  horn of the cat spinal cord. Abstr.
   Neurosci. Meeting, 1973,426:261.

Jouvet, M. The role of monoamines and acetylcho-
   line-containing neurons in the  regulation of  the
   sleep-waking cycle. In E. Usdin and S. Snyder
   (Eds.), Frontiers in catecholamine research, Per-
   gamon, Oxford, 1973:751-758.

Joyner, R. and Somjen, G.  A model  simulating the
   contribution of glia to sustained potential shifts of
   central nervous tissue.  Electroenceph. din. Neu-
   rophysiol.,  1974,37:417.

Jung, R., Kornhuber, H. H. and  DaFonesca, J. S. Mul-
   tisensory convergence on cortical neurons.  In G.
   Moruzzi, A. Fessard and H. H. Jasper (Eds.) Brain
   mechanisms, progress in brain research, Vol. 1,
   Amsterdam, Elsevier, 1963:207-234.

Kanai, T. and  Szerb, J. C. Mesencephalic  reticular ac-
   tivating system and cortical  acetylcholine output.
   Nature (Lond.), 1965,205:80-82.

Karahashi, Y. and Goldring, S. Intracellular poten-
   tials from  "idle" cells in cerebral  cortex of  cat.
   Electroenceph.  clin.  Neurophysiol.,  1966,  20:
   600-607.

Katchalsky, A. K., Rowland, V. and Blumenthal, R.
   (Eds.), Dynamic patterns of brain cell assemblies.
   Neurosci. Res.  Prog. Bull., 1974, 12:1-187.

Kawamura, H. and Domino, E. F. Differential actions
   of m  and n cholinergic agonists on the brainstem
   activating system.  Int. J. Neuropharmac.,  1969,
   8:105-115.

Kebabian, J. W., Steiner, A. L.  and  Greengard, P.
   Muscarinic  cholinergic regulation of cyclic guano-
   sine  3', 5' -monophoshate in autonomic ganglia:
   possible role in synaptic transmission.  J. Pharmac.
   exp.Ther.,  1975, 193:474488.

Knapp, D. E.  and Domino, E. F. Action of nicotine
   on the ascending reticular activating system.  Int.
   J. Neuropharmac., 1962, 1:333-351.
Knott, J. R. and Irwin, D. A. Anxiety, stress and the
   contingent  negative variation (CNV). Electroen-
   ceph.  clin.  Neurophysiol.,   1968, 24:386-387.

Knott, J. R. and Irwin, D. A. Anxiety, stress and the
   contingent  negative  variation.  Arch. gen.  Psy-
   chiat., 1973,29:538-541.

Kobayashi, H. and Libet, B. Generation of slow post-
   synaptic potentials without increases in ionic con-
   ductance.  Proc. nat.  Acad.   Sci.,  U.S.A., 1968,
   60:1304-1311.

Kobayashi, H. and Libet, B. Actions of noradrenaline
   and acetylcholine on  sympathetic ganglion cells.
   J. Physiol., Lond., 1970, 208:353-372.

Kobayashi, H. and Libet, B. Is inactivation  of potas-
   sium conductance involved in slow postsynaptic
   excitation of sympathetic ganglion cells?  Effects
   of nicotine. Life Sci., 1974, 14:1871-1883.

Kopell, B.  S., Tinklenberg, J. R. and Hollister, L. E.
   Contingent negative variation amplitudes: Mari-
   juana and alcohol.  Arch, gen  Psychiat.,  1972,
   27:809-811.

Kopell, B.  S., Wittner, W. K., Lunde, D. T., Wolcott,
   L. J.  and Tinklenberg, J. R. The effects of meth-
   amphetamine and secobarbital on CNV amplitude.
   Psychopharmacologia (Berl.),  1974, 34:55-62.

Kornhuber, H. H. and Deeke, L. Hirnpotentialander-
   ungen bei wilikurbewegungen und passiven beweg-
   ungen des menschen, Bereitschaftspotential  und
   reafferente potential. Pfliigers Arch.,  1965,284:1-
   17 (English summary).

Kriz,  N.,  Sykova,  E., Ujec. E. and Vyklicky, L.
   Changes of extracellular K+ concentration induced
   by neuronal activity in the spinal cord of the cat.
   J. Physiol., 1974, 238:1-15.

Kriz, N., Sykova. E., and  Vyklicky, L.  Extracellular
   potassium changes in the spinal cord of the cat and
   their  relation  to slow potentials active transport
   and impulse transmission. J.  Physiol., 1975, 249:
   167-182.

Krnjevic, K. Chemical nature of synaptic transmission
   in vertebrates, Physiol. Rev., 1974, 54:418-540.

Krnjevic, K.  and  Morris, M. E. Extracellular accumu-
   lation of  K+ evoked by activity of primary  affer-
   ent fibers in the cuneate nucleus and dorsal horn
   of cats. Can. J. Physiol. Pharmac., 1974,  22:852-
   871.

Krnjevic, K. and Morris,M. E. Extracellular K* activity
   and potential changes in spinal cord and medulla.
   CanJ.Physiol.Pharmac., 1972,50:1214-1217.

-------
  68
                        Electrogenesis references
  Krnjevic, K. and Phillis, J. W. Pharmacological proper-
     ties of acetylcholine-sensitive cells in the cerebral
     cortex. J. Physiol. (Lond.), 1963, 166:328-350.

  Krnjevic, K., Pumain, R. and Renaud. L.  The mech-
     anism of excitation  by acetylcholine in the cere-
     bral  cortex.  J.   Physiol.,  1971,  215:247-268.

  Krnjevic, K. and Schwartz, S. Some properties of un-
     responsive cells in the cerebral cortex.  Exp. Brain
     Res., 1967,3:306-319.

  Kuffler, S. W.  Neuroglial cells: Physiological proper-
     ties and  a potassium  mediated effect of neuronal
     activity on the glial  membrane potential. Pro. R.
     Soc. B., 1967, 168:1-21.

  Kuffler,  S. W. and Nicholls, J. G. The physiology of
    neuroglial cells. Ergebn. Physiol.,  1966,  57:1-90.

  Kuffler,  S.  W., Nicholls, J. G. and Orkand, R.  K.
    Physiological properties of glial cells in  the central
    nervous  system  of amphibia. J. Neurophysiol.,
     1966,29:768-787.

 Kuffler, S. W.  and  Potter, D.  D. Glia in the  leech
    central nervous system: Physiological  properties
    and  neuron-glia  relationship,  J. Neurophysiol.,
    1964,27:290-320.
 Lewis, P. R. and Shute, C.C.D. The cholinergic limbic
   system:  Projections to  hippocampal  formation,
   medial cortex, nuclei of the ascending  cholinergic
   reticular system, and  the  subfornical  organ and
   supraoptic crest. Brain, 1967, 90:521-540.

 Li, C. L. and Salmoiraghi, G. C. Cortical steady po-
   tential changes  and extracellular  microelectrode
   studies. Nature, 1963, 198:858-859.


 Libet, B.  Slow synaptic responses and  excitatory
   changes in sympathetic ganglia. J. Physiol., Lond.,
   1964, 174:1-25.


 Libet, B. Cortical activation  in conscious and uncon-
   scious  experience.  Perspect.  Biol.  Med.,  1965,
   9:77-86.
Libet,  B. Long latent periods and further analysis of
   slow synaptic responses  in  sympathetic ganglia.
   J. Neurophysiol., 1967, 30:494-514.
Libet,  B. Generation of slow inhibitory and excita-
   tory postsynaptic potentials. Fedn. Proc. Fedn.
   Am. Socs. exp. Biol., 1970, 29:1945-1956.
 Libet, B. SIF cell as a functional dopamine-releasing
    interneuron in  the rabbit  superior cervical gangli-
    on. In 0. Eranko (Ed.), SIF cells:  structure  and
    function of  small  intensely fluorescent sympa-
    thetic  cells. U. S. Department of Health, Educa-
    tion and Welfare, Washington, D. C.  Publication
    No. (NIH) 76-942. 1976.

 Libet, B. and Gerard.  R. W. Steady potential fields
    and  neurone  activity.  J.  Neurophysiol.,  1941
    4:438-455.

 Libet, B. and Kobayashi,  H. Adrenergic mediation
    of the slow inhibitory  postsynaptic potential in
    sympathetic ganglia of the frog. J. Neurophysiol.,
    1974,37:805-814.

 Libet, B., Kobayashi,  H. and Tanaka, T.  Synaptic
    coupling into  the  production and storage of a
    neuronal "memory" trace.  Nature, 1975, 258:
    155-157.

 Libet,  B.  and  Owman, Ch. Concomitant  changes in
    formaldehyde-induced fluorescence  of dopamine
    interneurones and in slow inhibitory postsynaptic
    potentials of rabbit superior cervical ganglion, in-
    duced by stimulation of  preganglionic nerve or by
    a muscarinic agent. J. Physiol., Lond.,  1974 237-
    635-662.

 Libet, B., Tanaka,  T. and Tosaka, T.  Different sen-
    sitivities of acetylcholine-induced "after-HP" com-
    pared to dopamine-induced hyperpolarization, to
   ouabain  or to lithium-replacement of sodium, in
   rabbit  sympathetic  ganglia. Life Sci.  1977,  20'
    1863-1869.

 Libet, B.  and Tosaka, T. Slow inhibitory and excita-
   tory postsynaptic responses in single cells of mam-
   malian  sympathetic   ganglia.  J.  Neurophysiol
   1969,32:43-50.

Libet, B. and  Tosaka,  T.  Dopamine as a synaptic
   transmitter and  modulator in sympathetic gang-
   lia: A  different mode of synaptic action. Proc.
   natn. Acad. Sci. U.S.A., 1970,67:667-673.

Lindvall, O., Bjorklund, A., Moore, R. Y. and Stenevi,
   U.  Mesencephalic dopamine neurons projecting to
   neocortex. Brain Res., 1974, 81:325-331.

Longo, V. G. Behavioral and electroencephalographic
   effects  of atropine and related compounds. Phar-
   mac. Revs., 1966, 18:965-996.

Lothman, E., LaManna. J., Cordingley, G., Rosenthal,
   M.  and Somjen, G. Responses of electrical activity
   of  the  cerebral  neocortex of cats. Brain Res.,
   1975,88:15-36.

-------
Electrogenesis references
                                             69
Lothman, E. W., and Somjen, G.  G.  Extracellular
   potassium activity, intracellular and extracellular
   potential responses in the spinal cord. J. Physiol.,
   1975,252:115-136.

Lothman, E. W., and Somjen, G. G. Functions of pri-
   mary afferents, and responses of extracellular K+
   during spinal epileptiform seizures. Electroenceph.
   clin. Neurophysiol., 1976,41:253-267.


Low, M. D., Borda. R. P. and Kellaway, P. "Contin-
   gent Negative Variation" in  rhesus monkeys: An
   EEC sign of a specific  mental process. Percept.
   Mot. Skills, 1966, 22:443446.

Low, M. D., Klonoff, H. and Marcus, A.. The neuro-
   physiological  basis of the marihuana experience.
   Can. Med. Ass. J.,  1973,  108:157-165.

Lux, H. D. and Neher, E. The equilibrium time course
   of [K+]0 in cat cortex. Exp.  Brain Res., 1973, 17:
   190-205.

MacLean, P. D., Yokota, T. and Kinnard, M. A. Pho-
   tically sustained on-responses of units in posterior
   hippocampal  gyrus of awake monkey. J.  Neuro-
   physiol., 1968.31:870-883.

Marczynski, T. J. Cholinergic mechanism determines
   the  occurrence of "Reward Contingent Positive
   Variation" (RCPV) in cat. Brain Res., 1971, 28:
   71-83.

Marczynski, T.  J. Electro physiological correlates of
   positive  reinforcement:   Post-reinforcement syn-
   chronization,  modulation of sensory input, and
   steady potentials.  In A. G. Karczmar and  J. C.
   Eccles (Eds.). Brain  and human behavior. Spring-
   er Verlag, Berlin, Heidelberg, 1972a:303-321.

Marczynski, T.J.  Lysergic acid diethylamide (LSD-
   25)  mimics the  effect  of diffuse light input on
   EEC  correlates of conditioned  operant behavior
   in cats.  Exp. Neurol., 1972b, 34:255-263.


Marczynski, T. J. and Burns, L. L.  Reward Contin-
   gent Positive  Variations (RCPV) and  post-reinf-
   forcement EEC synchronization (PRS)  in the cat:
   Physiological aspects, the  effect of morphine and
   LSD-25, and a new  interpretation of cholinergic
   mechanisms. Gen. Pharmacol., 1976, 7:211-220.
 Marczynski, T.J. and  Hackett,  J.T.  Post-reinforce-
    ment  EEC synchronization and  facilitation of
    cortical  somatosensory  evoked  potentials.  Elec-
    troenceph.  clin.  Neurophysiol.,  1969,  26:41-49.
Marczynski, T.  J., Hackett, J. T., Sherry, C. J. and
   Allen, S. L.  Diffuse light input and quality of re-
   ward determine the occurrence of "Reward Con-
   tingent Positive Variation" (RCPV) in cat. Brain
   Res., 1971a,28:57-70.

Marczynski, T. J. and  Sherry,  C.  J. Induction of
   sleep by  photic stimuli in the cat. Ex per. Neurol.
   1972,34:103-114.

Marczynski, T. J., York. J. L. and Hackett, J. T. Steady
   potential correlates of positive reinforcement: Re-
   ward contingent positive variation. Science, 1969,
   163:301-304.

Marczynski, T. J., York, J.  L.,  Rick, J. H., Allen, S. L.
   and Sherry,  C. J.  Steady  potential correlates of
   positive reinforcement and  sleep onset in cat: Re-
   ward Contingent Positive Variation (RCPV). Brain
   Res., 1971b, 26:305-332.

Marin-Padttla, M. Local circuit neurons in the neo-
   cortex. Neurosci. Res.  Prog. Bull.,  1975, 13:385-
   392.

McAdam, D. W. and  Whitaker, H. A.  Language pro-
   duction:  Electroencephalographic localization in
   the normal human  brain. Science, 1971, 172:499-
   502.

McAfee, D. A. and Greengard, P.  Adenosine 3', 5'-
   monophosphate: Electrophysiological evidence for
   a role in synaptic transmission. Science,  1972,
   178:310-312.

McCallum, W.  C., Papakostopoulos, D., Gombi, R.,
   Winter, A. L., Cooper, R. and  Griffith, H. B. Event
   related  slow potential  changes  in human brain
   stem. Nature, 1973, 242:465-467.

McCance,  I., Phillis,  J. W.  and Westerman, R. A.
   Acetylcholine-sensitivity of thalamic neurones: Its
   relationship  to synptic transmission. Br. J. Phar-
   mac. Chemother., 1968,32:635-651.

McSherry, J. W. Intracortical  origin of the contingent
   negative variation in the rhesus monkey, Doctoral
   dissertation, Baylor College of Medicine, Houston,
   Texas, 1971.

McSherry, J. W. and  Borda,  R.  P. The intracortical
   distribution  of  the CNV in rhesus monkey. Elec-
   troenceph. clin. Neurophysiol.,  1973,  Suppl.  33:
   69-74.


Meltzer, H. Y. and Stahl, S. M. The dopamine hypo-
   thesis of schizophrenia: A review.  Schizophrenia
   Bull.,  1976,2:19-76.

-------
 70
                       Electrogenesis references
 Mishkin,  M. Cortical  visual areas and  their inter-
   action. In A. G. Karczmar and J. C. Eccles (Eds),
   The  brain and human behavior. Berlin, Springer,
   1972:187-208.

 Mitchell, J. F. The spontaneous and evoked release
   of acetylcholine from cerebral  cortex. J. Physiol.
   (Lond.),  1963, 165:98-116.

 Morrell,  F.  Visual system's view  of acoustic space.
   Nature (Lond.), 1972,238:44-48.

 Moskowitz, M.  A.  and Wurtman, R. J. Catecho-
   lamines  and  neurologic diseases. New  Engl.  J.
   Med., 1975,293:274-338.

 Myers, R. D. Handbook of drugs and chemical stimu-
   lation of the brain. New York, Van Nostrand Rein-
   hold, 1974.

 Neal, M.  J., Hemsworth, B. A. and Mitchell, J. F.
   The excitation of central cholinergic mechanisms
   by stimulation of the audiotry pathway. Life Sci.,
   Parti, 1968,7:757-763.

 Niki, H.  Differential activity of prefrontal units dur-
   ing right and  left delayed response trials. Brain
   Res., 1974,70:346-349.

 Nishi, S., Soeda,  H. and Koketsu, K. Unusual nature
   of ganglionic  slow  EPSP  studied  by a  voltage
   clamp method. Life Sci., 1969, 8:3344.

 Nistri, A., Bartolini, A., Deffenu,  G. and Pepeu, G.
   Investigations  into  the  release of acetycholine
   from the cerebral cortex of the cat:  Effects of
   amphetamine,   scopolamine,  and  septal lesions.
   Neuropharmacol., 1972, 11:665-674.


Norton, S.  and Jewett,  R.E.   Alterations by drugs
   of the amplitude  and  variability  of  the  shifts
   during  arousal  in cats.   Med.  Pharmacol. Exp.,
   1967,17:165-174.

O'Leary, J.L. and Goldring, S.  D-C potentials of the
   brain. Physiol. Rev., 1964, 44:91-125.


Orkand,  R.K.,  Nicholls,  J.G. and Kuffler,  S.W.
   Effect of nerve impulses on the membrane poten-
   tial of  glial  cells in the central nervous system of
   amphibia.   J.   Neurophysiol.,  1966, 29:788-806.

Pandya,  D.N.,  Dye, P. and Butters,  N.   Efferent
   cortico-cortical projections  of  the   prefrontal
   cortex  in the  rhesus  monkey.   Brain Res., 1971,
   31:3546.

Pandya, D.N.  and Kuypers, H.G.J.M.   Cortico-cor-
   tical  connections  in  the rhesus monkey.   Brain
   Res., 1969,13:13-36.
Pape, L.G. and Katzman,  R.  Responses of glia in
   cat sensorimotor cortex to increased extracellular
   potassium.  Brain Res.,1972, 38:71-92.

Pedley,  T.A.,  Fisher,  R.S.,  Futamachi, K.J.  and
   Prince,  D.A.  Regulation of extracellular  potas-
   sium  concentration  in  epileptogenesis.   Feder
   Proc., 1976, 35:1254-1259.

Pepeu, G., and Mantegazzini, P.   Midbrain  hemi-
   section:  Effect on  cortical acetylcholine in the
   cat. Science, 1964,145:1069-1070.

Phillis, J.W. Acetylcholine and synaptic transmission
   in the central nervous system. In Ch. H. Hock man
   and D.  Bieger  (Eds.), Chemical  transmission in
   the  mammalian central nervous system.  Univer-
   sity Park, Baltimore, 1976:159-213.

Phillis, J.W.   The pharmacology  of  thalamic  and
   geniculate  neurons.   Int.  Rev.  Neurobiol.  1971
   14:148.

Picton,  T.W. and  Hillyard,  S.A.  Human auditory
   evoked  potentials.    II:    Effects  of attention.
   Electroenceph.  clin.  Neurophysiol., 1974,  36:
   179-190.

Pirch, J.H. and Norton, S. Drugs on cortical d.c. po-
   tentials  and  behavior of  rats with  septal area
   lesions. Physiol. Behav., 1967a, 2:121-125.


Pirch, J.H. and  Norton, S.  Effect of chlorpromazine
   and amphetamine  on conditioned  DC  shifts in
   septal rats.  Int. J. Neuropharmacol., 1967b, 6:125-
   132.

Pirch, J.H. and Osterholm,  K.C. Drug-induced alter-
   ations of slow potential responses in the rat. Phar-
   macologist, 1975,17:189.

Portig, P.J. and Vogt, M. Release into cerebral ven-
   tricles  of substances with possible transmitter
   function in   the  caudate  nucleus.   J.  Physiol.
   (Lond.), 1969, 204:687-715.

Pribram,  K. H. and Luria,  A. R. (Eds.). Psychophy-
   siology  of the frontal lobes. Academic Press, New
   York, 1973.

Prince, D.A., Lux, H.D. and Neher, E.  Measurement
   of extracellular potassium activity  in cat cortex.
   Brain Res., 1973, 50:489495.

Purpura,  D.P.   Comparative physiology of dendrites.
   In G.C. Quarton, T.  Melnechuk and F.O. Schmitt
   (Eds.),  The  neurosciences.  Rockefeller  Univ.
   Press, New York, 1967:372-392.

Purpura,  D.P.,  Frigyesi, T.L., McMurtry,  J.G.  and
   Scarf, T.  Synaptic mechanisms in thalamic regu-
   lation  of cerebello-cortical  projection  activity.
   In D.P.  Purpura and  M.D. Yahr (Eds.), Thalarnus.
   Columbia  Univ. Press, New York, 1966:153-170.

-------
Electrogenesis references
                                              71
Rail, W. and Shepherd, G.M.  Theoretical reconstruc-
   tion of field potentials and dendrodendritic synap-
   tic interactions in olfactory bulb.  J Neurophy-
   siol., 1968, 31:884-915.

Ranson,  B.R.  and Goldring, S.  Ionic  determinants
   of membrane  potential of cells  presumed to be
   glia in cerebral cortex of cat.  J. Neurophysiol.,
   1973a, 36:855-568.

Ranson,  B.R.  and Goldring, S.  Slow depolarization
   in cells  presumed to be glia in cerebral cortex of
   cat. J. Neurophysiol.,1973b, 36:869-878.

Ranson,  B.R.  and Goldring, S.  Slow hyperpolariza-
   tion in cells presumed to be glia in cerebral cortex
   of  cat.    J.  Neurophysiol.,  1973c,  36:879-892.

Reader, T.A.,  De Champlain, J. and  Jasper, H.  Cate-
   cholamines released  from  cerebral  cortex in  the
   cat; decrease during sensory stimulation.   Brain
   Res., 1976,111:95408.

Rebert, C.S.   D.C.  and multiple unit  recording in
   lateral  geniculate  body of the  cat.  Proc. Am.
   psychol. Ass., 1969,77:215-216.
Rebert, C.S.  Cortical and subcortical slow potentials
   in the monkey's brain during a preparatory inter-
   val.  Electroenceph. clin. Neurophysiol., 1972,33:
   389402.
Rebert, C.S. A technique  for simultaneous measure-
   ment of DC and multiple unit responses.  Electro-
   enceph.  clin.  Neurophysiol., 1973a, 34:326-328.


Rebert,  C.S.  Slow potential correlates of neuronal
   population responses in the cat's  lateral genicu-
   late  nucleus.  Electroenceph. clin.  Neurophysiol.,
   1973b, 35:511-515.

Rebert,  C.S.   EEC asymmetry  and  performance:
   Visual detection of words  and patterns  in three
   experiments.    Neurosci.  Abst.,  1976,  2:238.


Rebert,  C.S.  Intracerebral slow potential changes in
   monkeys during the foreperiod of reaction time.
   Progr.  clin.  Neurophysiol., Vol.  1.   Attention
   voluntary  contraction  and  event-related  cerebral
   potentials. Karger, Basel, 1977:242-253.

Rebert,  C.S., Berry,  R. and Merlo, J.  D-C potential
   consequences of induced muscle tension:  Effects
   on contingent negative  variation.  In W.C. McCal-
   lum  and J.R. Knott (Eds.), The responsive  brain.
   Wright, Bristol., 1976-.126-131.

Rebert,  C.S. and  Irwin, D.A.  Simple electrode con-
   figuration for chronic or acute recordings of D-C
   potentials  from  subcortical nuclei of the  brain.
   Electroenceph. clin. Neurophysiol., 1973, 34:440-
   442.
Rebert, C.S. and Irwin,  D.A.  Slow potential changes
   in cat brain during classical appetitive and aversive
   conditioning of  jaw  movement.  Electroenceph.
   clin. Neurophysiol., 1969, 27:152461.

Rebert, C.S., McAdam,  D.W., Knott, J.R. and Irwin,
   D.A.   Slow potential change in human brain re-
   lated  to  level  of motivation.   J. comp. Physiol.
   Psychol., 1967, 63:20-23.

Rebert, C.S., Pryor, G.T. and Schaeffer J.  Slow cor-
   tical potential consequences  of electroconvulsive
   shock in rats.  Physiol. Behav.,  1974, 12:131-134.

Rick, J.H. and Marczynski, T.J.  Dependence of "Re-
   ward  Contingent  Positive Variation" (RCPV) and
   cortical synchronization on visual input in the cat.
   Electroenceph.  clin.  Neurophysiol., 1976,  41:301-
   309.

Rinaldi,  F.  and  Himwich, H.E.   Cholinergic mech-
   anisms involved  in function  of mesodiencephalic
   activating system.  Archs. Neurol, Psychiat. (Chi-
   cago), 1955, 73:396402.

 Roitbak, A.I.  Further  analysis  of slow surface-nega-
   tive potentials, of the cortex:   Action of X-rays
   and analgesias.  Acta Biol. Exp., 1969, 29:125434.

 Rosen,  S. C., Loiselle, D. L., Sandrew, B.  B. and
   Stamm,  J.S.  Conjugate lateral eye  movements
   associated  with  operantly conditioned prefrontal
   steady  potential  shifts in awake monkeys.  Fed.
   Proc., 1974, 33:1260.

 Rosen, S.C. and Stamm, J.S.  Cortical steady poten-
   tial  shifts  during delayed response performance
   by  monkeys.  Electroenceph. clin. Neurophysiol.,
   1969,27:684-685.

 Rosen, S.C. and Stamm,  J.S.   Transcortical polari-
   zation:  Facilitation  of delayed response perform-
   ance  by monkeys.   Exp. Neurol.,  1972,  35:282-
   289.

 Rosenthal,  M. and  Jobsis, F.F.   Intracellular redox
   changes in functioning cerebral  cortex. II.  Effects
   of  direct cortical stimulation.   J. Neurophysiol.,
   1971, 34:750-761.

 Rosenthal, M. and Somjen, G.  Spreading depression,
   sustained potential  shifts and  metabolic activity
   of the cerebral cortex of cats.  J. Neurophysiol.,
   1973,36:739-749.
 Routtenberg,  A.  The two-arousal hypothesis:  Reti-
   cular formation  and  limbic system.   Psychol.
   Rev., 1968, 75:51-80.


 Rowland, V.   Simple non-polarizable  electrode for
   chronic implantation.  Electroenceph. clin. Neuro-
   physiol., 1961,13:290-291.

-------
72
                       Electrogenesis references
 Rowland, V. Cortical steady potential (direct current
   potential) in reinforcement and  learning.   In E.
   Stellar and J.M. Sprague (Eds.), Progress in physio-
   logical psychology, Vol.  3.  Academic Press,  New
   York ,1968:1-77.

 Rowland, V. Steady states and  their shifts.  Neurosci.
   Res. Prog. Bull., 1974,12:109-121.

 Rowland, V. and Anderson, R.  Brain steady  poten-
   tial shifts.  In E. Stellar  and J.M.  Sprague  (Eds.),
   Progress  in   physiological  psychology,  Vol.  4.
   Academic Press, New York, 1971:37-51.

 Rowland, V. and  Dines,  G.  Cortical steady  poten-
   tial shift in  relation  to the rhythmic electrocor-
   ticogram and multiple  unit  activity.   In  R.F.
   Thompson  and  M.M.  Patterson  (Eds.), Bioelec-
   tric recording techniques:   Part  A, cellular  pro-
   cesses  and brain  potentials.   Academic  Press,
   New York, 1973:369-385.

Rowland, V. and  Goldstone, M.  Appetitively  con-
   ditioned  and drive-related   bioelectric  baseline
   shift in cat  cortex.   Electroenceph. clin, Neuro-
   physiol., 1963,15:474485.

Sandrew,  B.B., Stamm,  J.S.  and Rosen, S.C.  Steady
   potential shifts and facilitated learning of delayed
   response in monkeys.  Exp.  Neurol., 1977,  55:43-
   55,

Sasaki,  K., Shimono, T., Oka, H., Yamamoto.T. and
   Matsuda, Y.   Effects of stimulation of the  mid-
   brain reticular  formation upon  thalamo-cortical
   neurons  responsible   for  cortical  recruiting  res-
   ponses.  Exp.  Brain Res., 1976, 26:261-273.

 Schmitt, R.O.,  Dev, P., Smith, B.H.  Electronic pro-
   cessing  of information by  brain  cells.   Science,
   1976,193:114420.

 Shagass, C. An electrophysiological  view of schizo-
   phrenia. Biol. Psychiat., 1976,11:3-30.

 Sheafor, P.J. and Rowland,  V.   Dissociation  of cor-
   tical steady potential shifts  from mass action po-
   tentials in  cats  awaiting  food rewards.  Physiol.
   Psychol., 1974, 2:471-480.

 Sherry. C.J., Marczynski, T.J., Karmos,  G. and  Wei
   Jeanne, Y.   A miniature preamplifier for extra-
   cellular action potentials  in freely moving animals.
   Acta physiol.  Hungarica, 1975,46:83-86.


 Shetty, T.  Photic responses  in hyperkinesis of child-
   hood.  Science, 1971,174:1356-1957.

 Shin, T.M., Khachaturian, Z.S.,  Barry,  III,  H.  and
   Hanin,  I.  Cholinergic mediation of the  inhibi-
   tory effect of  methylphenidate on neuronal ac-
   tivity in the  reticular formation.  Neuropharma-
   col., 1976,15:55-60.
Shute, C.C.D.  and Lewis, P.R.  The ascending cholin-
   ergic reticular system:  Neocortical, olfactory, and
   subcortical  projections.  Brain, 1967, 90:497-520.

Shute, C.C.D.  and Lewis, P.R.  Localization of cho-
   linesterases  in monkey brain.  J. Anat., 1969, 104:
   186-187.

Siegel, J.M. and McGinty, D.J.  Pontine reticular for-
   mation neurons:   Relationship of  discharge  to
   motor  activity.    Science,  1977,  196:678-680.

Sigg, E.B., Drakontides, A.B.  and Day, C. Muscarinic
   inhibition  of  dendritic post-synaptic potentials
   in  cat  cortex.   Int.  J. Neuropharmacol.,  1965,
   4:281-289.

Siggins, G.R.,  Hoffer, B.J. and Bloom,  F.E. Studies
   on norepinephrine-containingafferents to Purkinje
   cells of rat cerebellum.  III.  Evidence for media-
   tion of norepinephrine effects by cyclic 3',5'-aden-
   osine monophosphate.   Brain  Res., 1971, 25:535-
   553.

Singer,  W. and Lux, H.D.   Extracellular potassium
   gradients and  visual receptive fields in the  cat
   striate  cortex.    Brain  Res.,  1975,  96:378-383.


Skinner, I.E. and Yingling, C.D.  Selective regulation
   of thalamic sensory relay nuclei by nucleus reticu-
   laris thalami.  Electroenceph.  clin. Neurophysiol
   1976,41:476482.

Snyder, S.  Amphetamine  psychosis:  A "model" of
   schizophrenia  mediated by catecholamines.  Am.
   J. Psychiat., 1973,130:61-66.

Somjen, G.G.  Sustained  evoked potential changes
   of the  spinal cord. Brain Res., 1969,12:268-272.

Somjen, G.G.  Evoked sustained focal potentials and
   membrane  potential of  neurons and  of unrespon-
   sive  cells of the  spinal cord.  J.  Neurophysiol.,
   1970,33:562-582.

Somjen, G.G.  Electrogenesis of sustained potentials.
   Prog. Neurobiol., 1973,1:199-237.

Somjen, G.G.  Do potassium, glia and neurons  inter-
   act? In C.D. Woody, K. A. Brown, T.J. Crow and
   J.D. Knispel  (Eds.), Cellular  mechanisms subser-
   ving changes  in neuronal  activity.  UCLA  Brain
   Information Service, Los  Angeles, 1974,  3:79-96.

Somjen, G.G.  Electrophysiology of neuroglia.  Ann.
   Rev. Physiol., 1975, 37:163490.

Somjen, G.G.  The role of potassium and of neuroglia
   in  the  generation  of  seizures and in their  treat-
   ment.  In  K.  Penry et al. (Eds.), Mechanisms of
   action  of antiepileptic  drugs.  Raven Press, New
   York, in press.

-------
Electrogenesis references
                                              73
Somjen, G.G. and Lothman, E.W.   Potassium, sus-
   tained  focal potential shifts, and  dorsal root  po-
   tentials of the mammalian spinal cord.  Brain Res,,
   1974,69:151-157.

Somjen, G.G. Rosenthal, M., Cordingley, G., LaMan-
   na, J., and Loth man, E. Potassium, neuroglia and
   oxidative metabolism in central gray matter. Fed.
   Proc., 1976, 35:1266-1271.

Spehlman, R,, Daniels, J.C. and Smathers.C. Acetyl-
   choline and  the  synaptic impulses  to  the  visual
   cortex. Brain, 1971, 94:125438.

Stamm, J.S.  Retardation and facilitation in learning
   by  stimulation  of  frontal cortex  in  monkeys.
   In J.M. Warren  and K.  Akert  (Eds.), The frontal
   granular cortex  and behavior.  McGraw-Hill, New
   York, 1964:102425.

Stamm, J.S. and Rosen,  S.C.  Cortical steady poten-
   tial shifts  and anodal  polarization during delayed
   response  performance.   Acta Neurobiol.  Exp.,
   1972,32:193-209.


Steiner, F.A.  Influence  of microelectrophoretically
   applied  acetylcholine  on  the responsiveness  of
   hippocampal  and  lateral  geniculate  neurones.
   Pflueger's Arch., 1968, 303:173480.

Steinmetz, G.T.  and  Rebert.C.S.  Post-reinforcement
   changes of steady potentials in  premotor cortex
   of monkeys.   Physiol.  Behav., 1973,  9:769-772.

Steriade, M. and Deschenes, M.  Cortical interneurons
   during  sleep and waking in freely moving primates.
   Brain Res., 1973, 50:192499.

Sterman,  M.B. and Wyrwicka, W. EEG correlates of
   sleep:  Evidence for separate  forebrain  substrates.
   Brain Res., 1967, 6:143-163.

Strittmatter,  W.J. and Somjen, G.G.   Depression of
   sustained evoked  potentials and  of glial depolari-
   zation  in the spinal cord by  barbiturates and  by
   diphenylhydantoin.   Brain Res., 1973,  55:333-
   342.

Sugaya, E., Goldring, S. and O'Leary, J. Intracellular
   potentials associated with direct cortical  response
   and  seizure  discharges in cat.    Electroenceph.
   din. Neurophysiol., 1964,17:661-669.

Sutton, S.,  Braren, M.,  Zubin,  J.  and  John, E.R.
   Evoked potential correlates   of  stimulus  uncer-
   tainty. Science, 1965,150:1187-1188.

Szerb, J.C.  Averaged evoked potentials and  choliner-
   gic synapses  in the somatosensory cortex  of the
   cat.    Electroenceph.  clin.  Neurophysiol., 1965,
   18:140446.

Tebecis, A,K. Studies on cholinergic transmission in
   the medial geniculate nucleus.   Br. J. Pharmac.,
   1970,38:138447.
Tecce, J,J. Contingent negative variation (CNV) and
   psychological processes  in  man.  Psychol.  Bull.,
   1972,77:73408.

Tecce, J.J. and  Cole, J.O.  Amphetamine effect in
   man:   Paradoxical drowsiness and  lowered  elec-
   trical brain activity (CNV).  Science, 1974,185:451-
   453.

Tecce, J.J. and Cole, J.O.  The distraction-arousal hy-
   pothesis, CNV, and schizophrenia. In D.I. Mostof-
   sky (Ed.),  Behavior  control and  modification of
   physiological activity.   Prentice-Hall, Englewood
   Cliff s,1976:162-219.
Teece, J. J.,  Cole,  J. O.  and Saviano-Bowman, J.
   Chlorpromazine effects  on brain activity (CNV)
   and reaction time in normal women. Psychophar*
   macologia (Berl.),1975, 43:293-295.


Tennyson, V.M., Heikkila, R., Mytilneou.C., Cote, L.
   and Cohen, G, 54iydroxydopamine "tagged" neu-
   ronal  boutons in rabbit neostriatum:   Interrela-
   tionship between vesicles and axonal membrane.
   Brain  Res.,  1974, 82:341-348.


Terzuolo, C.A. and Bullock, H.  Measurement of im-
   posed voltage gradient adequate to modulate neu-
   ronal  firing. Proc. natn. Acad. Sci., 1956,42:687-
   694.


Ungerstedt,  U.  Stereotaxic  mapping of the mono-
   amine pathways in  the rat brain.  Acta physiol.
   Scand.,1971, Suppl. 367 :M8.


Van Meter,  W.G. and  Karczmar, A.G.  An effect of
   physostigmine on the  central nervous system of
   rabbits, related to brain levels of norepinephrine.
   Neuropharmacol., 1971,10:379-390.

Varon, S. and Somjen, G.  Neuron-glia interactions.
   NRP Bulletin (in press).

Vastola,  E.F.    Steady potential  responses  in  the
   lateral geniculate body.  Electrenceph. clin. Neuro-
   physiol., 1955,7:557-567.

Vaughn,  H.G.  The analysis of scalp  recorded brain
   potentials.  In R.F. Thompson and M.M. Patterson
   (Eds.), Bioelectric recording techniques, Vol. 1-B.
   Academic Press, New York, 1974:158-207.


Vaughn,  H.G.  and Ritter, W.  The source of auditory
   evoked responses  recorded from the human scalp.
   Electroenceph. clin. Neurophysiol., 1970, 28:360-
   367.

Vyklicky, L.,  Sykova,  E. and Krjz, N.  Slow poten-
   tials induced  by  changes  of  extra-cellular potas-
   sium in the spinal  cord of the  cat.  Brain  Res.,
   1975,87:77-80.

-------
 74
                       Electrogenesis references
Vyskocil, N., Kriz, N. and  Bures, J.  Potassium-selec-
   tive microelectrodes used for measuring the extra-
   cellular brain  potassium during spreading depres-
   sion and anoxic depolarization in  rats.  Brain Res.,
   1972,39:255-259.

Walter,  W.G., Cooper, R.,  Aldridge, V.J., McCallum,
   W.C. and Winter, A.L.  Contingent negative varia-
   tion:  An electric sign of sensori-motor association
   and expectancy in the human brain.  Nature, 1964,
   203:380-384.
Walter, W. G.  Slow potential changes in the human
   brain associated  with expectancy, decision and
   intention.  Electroenceph.  clin.  Neurophysiol.,
   1967, Suppl. 26:123-130.

Walter, W.  G.    Expectancy waves  and intention
   waves in the human brain and their application to
   the  direct control of machines.  Electroenceph.
   clin. Neurophysiol., 1966, 21:616.

Weight, F.F. and  Padjen, A. Slow synaptic inhibi-
   tion:  Evidence for synaptic inactivation of sodium
   conductance in sympathetic ganglion  cells.  Brain
   Res., 1973a,55:219-224.

Weight, F.F. and Padjen, A.   Acetylcholine and slow
   synaptic  inhibition in frog sympathetic  ganglion
   cells.  Brain Res., 1973b,55:225-228.
Weight, F. F.  and Votava, J.   Slow synaptic  ex-
   citation in sympathetic ganglion cells:  Evidence
   for synaptic inactivation of potassium conduct-
   ance, Science, 1970, 170:755-758.

Woody, C.D., Marshall, W.H.,  Besson, J.M., Thomp-
   son, H.K.,  Aleonard,  P.  and Albe-Fessard,   D.
   Brain potential shift  with respiratory acidosis in
   the cat  and monkey.  Am.  J. Physiol., 1970 218-
   275-283.

Wurtz,  R.H.   Steady potential shifts during arousal
   and  deep  sleep in the  cat.  Electroenceph. clin.
   Neurophysiol., 1965a, 18:649-662.

Wurtz, R.H.  Steady potential shifts in the rat during
   desynchronized sleep. Electroenceph. clin Neuro-
   physiol., 1965b, 19:521-523.

Wurtz, R.H.  Steady potential fields during sleep and
   wakefulness  in  the  cat.   Exp.   Neurol   1966
   15:274-292.

Yamamoto, C.   Neurotransmitter actions studied  in
   brain slices.  Progr. Brain.Res. (Jap.) 1973 17-64-
   70.

Yingling, C.D. and Skinner, J.E.  Regulation of unit
   activity  in nucleus reticularis thalami by the mes-
   encephalic  reticular  formation and  the  frontal
   granular cortex.  Electroenceph. clin. Neurophy-
   siol., 1975, 39:635-642.

-------
II.  MOTOR CONTROL
    Section Editor:
    Demetrios Papakostopoulos
    Burden Neurological Institute
    Bristol, England

-------
THE  PRESENT STATE  OF BRAIN
MACROPOTENTIALS IN  MOTOR CONTROL
RESEARCH - A SUMMARY OF ISSUES

D. PAPAKOSTOPOULOS
Burden Neurological Institute, Bristol, England
   The  following section is based on preconference
correspondence and  on the Motor Control plenary
session. The objectives of this section are to analyze
and  evaluate  the  contributions  of macropotential
research to the neuropsychophysiology of movement.

Movements as integral formations

    The phenomena encompassed by the term "motor
control" are very broad. The neurologist, neurophysi-
ologist,  neuroanatomist, neurochemist, experimental
psychologist, ethologjst, and cybernetician are in one
way or another studying the organization of move-
ment, its various components, executive organs, and
spatiotemporal  relationships.  Is there need for yet
another approach to  the study  of motor  control
processes?

    There is a vast experimental literature characteriz-
ing the electrical activity of single motor units and
muscle  discharge  patterns. Bernstein  (1967) has
observed, however, that unit discharge patterns are far
removed from movements as integral formations. The
limits of most techniques, such as electrical stimula-
tion, used to study movement have been pointed out.
For example, Phillips (1966) remarks that electrical
stimulation does not evoke the natural functioning of
cortex and that movements are intracerebral processes
which are not simply equated with the contraction of
muscles. Although there is increasing realization that
movements are  "integral formations" which require
conceptualization in terms of intracerebral processes,
it is less clearly understood that such processes extend
beyond the realm of local  brain architectonics. The
concept of movement introduces the need to think in
terms of higher levels of organization than hitherto
accepted forms  of anatomical and physiological evi-
dence permit.

    Mountcastle et al. (1975) summarized the prob-
lem: "Some of our observations suggest that another
mode of organization  is from time to time superim-
posed on the  basically columnar pattern and that this
additional set formation is dynamic and conditional
in nature." The study of such dynamic and condition-
al sets has hitherto been the province of psychological
research. McKay (1966), however, has  pointed out
that "the entities and paths of the psychologist's mod-
els have classically had few pretentions to anatomical
significance, while those of the physiologist, for rea-
sons of sheer complexity, have been correspondingly
vague in their predictions of gross human behavior."
Macropotential  research could supplement classical
methods by offering neurophysiological insight and
precision to those dynamic and conditional intracereb-
ral processes which underlie the highly integrated units
of behavior and internal processing known as move-
ments.

Movement-related brain macropotentials

    Electrical brain activities generally referred to as
motor-related potentials include sustained and phasic
changes time-locked to self-initiated actions or to
actions triggered externally with or without forewarn-
ing. Signal averaging techniques have been widely used
to study the Bereitschaftspotential (BP) and CNV,
although this technique has  diverted interest from
movement-related  brain  macropotentials (MRBMs)
which can be observed  in scalp  recordings without
averaging. The  mu rhythm (Gastaut 1952, Chatrian
et al. 1959) is an example of such an event.

    Fig. 1 illustrates corticographic analogs of the mu
rhythm recorded from sensorimotor cortex. Note that
sensorimotor response is maximal contralateral to the
active hand, while occipital rhythms  are not altered
by movement. The sensorimotor response is less with
iprilateral or sustained contralateral clenching. These
observations suggest both topographical and situation-
al specificity of intrinsic sensorimotor rhythms. The
significance of this specificity and its relationship to
other macropotentials occurring  simultaneously are
important questions for future research.


    The study of evoked potentials while  the organism
remains hi a passive state may also contribute to our
understanding  of the organization  of movement.
Dubrovsky (this section) discusses the significance of

-------
RELAXED
           PRESS-RELEASE L.
                                                       RELAXED
                                                                          PRESS-RELEASE R.
L. ARM  FLEX. EMO


    »   EXTC-N. »
-4
00
DISPLACEMENT     	
              20C)N\
                   isec
           CONTINUOUS PRESS L.
                                                             CONTINUOUS PRESS  R.


flexor* and extensor,
                                     . 6-9). and °"P°f**»                (R PREC.). Precentral and       tral


-------
Macropotentials and Motor Control
                                             79
visual and cutaneous input to the cortex. Investiga-
tions  of somatosensory  evoked potentials,  further-
more, have suggested specific sensory input to precen-
tral  cortex  (Papakostopoulos  et  al.  1974a, 1975;
Papakostopoulos and Crow, in press). Further evidence
concerning afferent pathways to the motor cortex is
contained  in   neurophysiological  (Murphy  et al.
1975)  and  neuroanatomical  (Kalil,   this section)
studies.

     Three factors must be taken into account in any
 attempt to classify movement-related brain macro-
 potentials. The first is  the polymorphism of macro-
 potentials which can be recorded from the same corti-
 cal area during the same period of time. The second
 factor  relates  to the spatial properties  of  specific
 components. The third factor concerns the temporal
 sequence of movement and associated macropotential
 patterns. Macropotentials tend to convey  a deceptive
 static picture of the brain during the organization and
 execution of movement unless the temporal sequence
 of motor and electrical events is carefully  defined.

    These  factors have  been  taken into  account in
the following classification of movement-related brain
macropotentials.
   I.  Resting potentials— potentials related with the
      state of the motor system at rest (e.g., sensori-
      motor rhythm and somatosensory evoked poten-
      tial).
   2.  Preparatory potentials-potentials related with
      functions during the preparatory period  (e.g.,
      BP and CNV).
   3.  Initiation potentials -potentials related  with
      the initiation period (e.g., motor potential).
   4.  Movement  execution  potentials -  potentials
      related with  the execution of the  movement
      (e.g., N2 or motor cortex potential).
   5.  Termination potentials - potentials related with
      the termination of the movement (e.g.,  P2 and
      skilled performance positivity).

     Serial ordering in movement forms the basis of
this  taxonomy  which may be viewed as a five-interval
 time scale. This classification system has the advantage
of accommodating  the fragmented data derived from
movement-related macropotential  studies as  well as
 facts and theories about motor control derived from
other disciplines.

     Factors such as spatial coordination, target deter-
mination, holding, and achievement must  also be con-
sidered  during  the unequal intervals of this time
sequence. At any point in the sequence of operations
involved in any motor act, different parts of the body
are subject  to  different  postural states and displace-
ment forces according to the nature or target of the
movement.  Brain  macropotentials could reflect any
one or combination of these factors.
Functional issues

    Three questions  arise  with  regard to the nature
and significance of changes occurring during each time
interval.  First, do the macropotential changes reflect
factors related to movement alone, or do other non-
motoric  factors contribute?  Second,  to what extent
do peripheral and central  changes involved in every
stage  of  movement influence the macropotential  in
progress? Third, to what extent do movements influ-
ence  the  sensitivity  or selectivity of the organism
toward  environmental changes, and  what  are the
associated effects  on brain macropotentials?

    The  first question, raised briefly by Cohen during
discussion, can  easily diverge to a semantic argument
of what  constitutes a motor act. On the other hand,
this issue underlies the venerable controversy whether
the CNV and  BP are  functionally  equivalent (cf.
McCallum, this section). The excitability of the spinal
monosynaptic   reflex, for example, increases during
the CNV interstimulus interval in the absence of any
overt behavior or EMG activity (Papakostopoulos and
Cooper 1973).

    This  evidence indicates  involvement of motor
structures during  the preparatory process but does not
help to differentiate the task-specific operations which
take  place  during  preparation.  There   is evidence
(Papakostopoulos, this section) that the context within
which similar  motor actions are executed is reflected
in the amplitude and wave form of event-related mac-
ropotentials.  The evidence suggests that  preparation
involves   motor  structures, even at  the  spinal  cord
level, and  that cognitive  or contextual elements of
action are reflected in macropotential configurations.

    It is probably as misleading to attempt to distin-
guish a  particular movement  from the purpose for
which it  has been organized as it is to suggest a clear-
cut separation of sensory and motor  elements within
the nervous system.  Magendie  pointed out  in 1824
that the  separation of nerves of feeling and nerves of
motion is arbitrary and of no practical value.

    The  second  question, relating to feedback, was
discussed with  reference to the motor potential (MP)
of Xornhuber and Deecke (1965), the N2 component
of Ulden et al.  (1966) and Vaughan et al. (1968),
and the  positivities which follow it known as reaffer-
ent potentials  (Vaughan et al.  1968, Deecke et  al.
1969). The first two  terms are generally considered to
be  interchangeable  (cf.  McAdam 1973 and  Buser
1976). However, Deecke made explicit  the  position
of the Ulm group that the MP is not equivalent to N2.
Deecke described the MP as an  additional negativity
appearing at the precentral cortex contralateral to the
moving finger 54 msec before EMG onset. This com-
ponent is superimposed on  an already asymmetrical

-------
 80
                               Papakostopoulos
 BP. The problem is how to distinguish between the
 two entities if the BP itself is asymmetrical.

    An alternative hypothesis is that the N2 compon-
 ent is an index of reafferent activities (Papakostopou-
 los et  al.  1974a, 1975). Evidence of specific sensory
 input  to precentrai  cortex  has  been found in man
 (Goldring and Ratcheson 1972) and animal (Kalil,
 this section).  The results  of Gerbrandt et al. (1973)
 and Papakostopoulos et al.  (1975) indicate, further-
 more,  that N2  follows movement onset.  This evi-
 dence   supports  a  reafferent  view of the  N2
 component.

     In  discussing macropotentials  associated  with
 afferent input to the cortex, account must be taken
 of  the  state  of the  nervous sytem preceding input.
 For instance,  the  configuration  of a sensory evoked
 response differs when a subject passively receives a
 stimulus compared to when  the subject is triggered to
 action  by the  stimulus or when the stimulus occurs as
 a result of motor action by the subject. Precentrai
 and postcentral cortical areas process afferent input
 in different modes, depending on  the motor contin-
 gencies of the  situation.  Hazemann (this  section)
 reviews evidence concerning the effect of movement
 on sensory evoked potentials.

    Vaughan  et al.(1968) and Deecke et al. (1969,
 1976)  suggested  that the positive  potentials which
 follow  N2 reflect response reafferent activity. The
 response reafferent hypothesis, however, is not consis-
 tent with the results of several studies reported in this
 section (see Abraham et al., Delaunoy et al., Otto et
 al., and Papakostopoulos) which  suggest that these
 potentials could reflect neurophysiological concomi-
 tants of internal afferent  feedback or efferent feed
 forward processes, possibilities which require experi-
 mental verification.

    Another related issue is  the homogeneity of post-
 movement positivities and other late positive compon-
 ents described in the ERP literature. For example, is
 the P300 functionally related to positive potentials
 associated with  visual detection of infrequent events
 (Cooper et al.  1977) or to skilled performance positiv-
 ity (Papakostopoulos, this section)?

 Theoretical and experimental integration

    Kornhuber and his colleagues (Kornhuber 1971;
 Deecke et al.  1973,  1976) have  proposed a compre-
 hensive theory of motor function which integrates
 data from several disciplines including event-related
 potential research. According to this theory-, prepara-
 tory functions are executed prior  to movement in
 several  regions of the brain such as the cerebellum,
basal ganglia and precentrai and parietal cortex.  The
 specific loci involved in  preparatory functions  vary
with the type  of anticipated movement. This theory
 provides an excellent example of how macropotential
 research can  contribute to  clinical knowledge  and
 stimulate further experimentation in the area of motor
 control.

     Another  approach seeks to integrate data from
 the  central, peripheral, and  autonomic nervous  sys-
 tems during  the preparation for  and execution of
 movement. This approach, exemplified by the work
 of Lacey and Lacey (1970), Ingvar (1977),  McCal-
 lum et al. (1973, 1976), Cooper et al. (1975),  and
 Papakostopoulos  and Cooper (1973, 1976, this  sec-
 tion),  is more experimental than theoretical, but has
 also contributed substantially to the understanding of
 motor  control processes.  These investigators  have
 studied changes in cortical 02a, heart rate, and spinal
 reflexes during specific phases of movement.

 Need    for   technological   and  conceptual
 innovation

     Three methodological issues were discussed: (1)
 the effect of experimental artifacts on  the reliability
 of data; (2) the need to study integrated motor sequen-
 ces in  place of isolated,  purposeless  actions; and (3)
 the limits imposed by current technology.

     Extracranial artifacts such  as the large electrical
 potentials generated by  eye movements (rotation of
 the  comeofundal dipole) pose serious  problems in
 most areas of ERP research including motor control.
 Despite the ubiquity of the problem  and the keen
 awareness of most investigators, no foolproof method
 has yet been devised to entirely eliminate eye-move-
 ment artifact  from  ERP recordings in all subjects.
 Rosen   (this  section) presents  disquieting evidence
 from corticographic recordings in monkeys that oculo-
 motor  processes associated with saccadic eye move-
 ments  may significantly  alter cortical evoked poten-
 tials even when the  eye  is  fixated. The contribution
 of frontal eye fields to  and  the effects of different
 oculomotor control  strategies on ERPs require  further
 study.
    The  reports of McCallum and Papakostopoulos
(this section) illustrate  the  need for refinement of
experimental procedures.  Both experiments demon-
strated that the amplitude and spatial distribution of
the CNV and BP were task dependent. Task demands
were found to be  a more important determinant of
macropotential configuration than were individual or
group stereotypes. Individual differences do exist, as
discussed by Deecke et al. (1976 and this section),
but the present data suggest that such differences will
be more effectively revealed in experiments which
involve  complex  patterns of interaction with  the
environment where the pattern of response is flexible
rather than stereotyped.

-------
Macropotentials and Motor Control
                                              81
    Naturally, such developments will make Increased
demands on technology in order to study the neuro-
physiological substrate of complex, integrated move-
ment. Otherwise, we will be limited to the study of
simple,  stereotyped responses. Improved  technology
will not only provide stronger evidence for the existing
range of phenomena,  but will  reveal more  clearly
where fundamental discrepancies exist. For example,
Gerbrandt's attempt (this volume) to determine the
smallest number of scalp electrodes necessary to study
motor processes revealed problems in previous concep-
tualization 01 potentials related to self-paced move-
ments. His data suggest that the BP is far from being
a unitary phenomenon and that its waveform varies
with time in different ways for different areas.

    Such approaches underline the complexity of the
issues we face and challenge our traditional methodol-
ogies. Although there may be agreement that the BP
is a preparatory potential, one is obliged to ask how
many types or variants  of preparation exist. We are
only at the threshold of a constructive neurophysiolog-
leal analysis of preparatory processes  and ideational
components of movement organization in man. The
solution probably lies in  understanding the spatiotem-
poral organization of macrostates. The number of
macrostates that  can  be seen with macroelectrodes
will  be determined by  the spatial  constant (i.e., the
degree of attenuation over distance) of the brain which
is  yet unknown  for cortical  or scalp  recordings.
(Methodological details and problems of spatial averag-
ing are discussed extensively in  Section IX). Perhaps
spatial sampling techniques already within the bounds
of our technology will  enable  us  to elucidate what
Katchalsky  et al. (1974) have called the dynamic
pattern of brain macrostates.

-------
 SPINAL CORD STIMULATION AND
 EVENT-RELATED POTENTIALS1
 P. ABRAHAM, T. DOCHERTY, AND S. SPENCER

 Royal Victoria Hospital, Netley, Southampton, England
 A. COOK AND A. OYGAR
 State University, New York, NY, U.S.A.

 L. ILLIS AND M. SEDGWICK
 Wessex Neurological Centre, Southampton, England
     Cook  (1976)  reported insertion of electrodes
 percutaneously in the thoracic epidural space of the
 spinal canal in 185 patients with multiple sclerosis
 (MS). Following a  period of continuous inductively
 coupled electrical  stimulation,  69 patients  showed
 significant improvement in some neurological function
 such as speech, bladder function, or locomotor perfor-
 mance. Electrodes  were implanted permanently  in
 these  patients and  stimulation  was maintained;
 improvement persisted in 51 cases.

     The mechanism by which electrical activity in-
 duces this benefit is unknown. Symptomatic improve-
 ment could be the direct consequence of electrical
 stimulation of nonspecific projection systems of the
 brain via the midbrain reticular formation. If so, evok-
 ed and slow potential changes might provide useful
 measures from which to infer the underlying neuro-
 physiological mechanism. McCallum et al. (1973), for
 instance, found that  the shape and duration of SP
 changes recorded from intracerebral MRF  electrodes
 in humans  closely  resembled the vertex contingent
 negative variation (CNV).
    Improvement might also be a secondary effect of
 the emotional impact of the implantation and stimula-
 tion procedure. The drama of surgery and experimen-
 tation, the novelty of the techniques, the use of an
 electronic device, the encouragement of well-wishers,
 and the new-found hope of relief from the distressing
 symptoms  of progressive disease may all play  a part.
 If recovery is dependent on a novel state of sustained
 excitement,  it should be reflected in a postoperative
 change-presumably an enhancement-in CNV ampli-
 tude, at least when the apparatus is switched, on and
 the patient is aware of it.

 Methods

 Subjects

     This  preliminary report concerns data obtained
 from limited experimentation with five patients who
 had spinal electrodes temporarily implanted. Three
 patients (RW, CJP, and DHS), aged 34, 36, and 41,
 were male and two (SE and EFM), aged 43 and 23
 were female. Four had multiple sclerosis and the fifthi
 RW, had motor neurone disease. Lesions were judged
 clinically  to be infratentorial in all patients. In some
 cases, the nature and severity of the lesions  were
 demonstrated by H-reflex recovery  curves and audi-
 tory evoked response recording. Detailed clinical and
 physiological information concerning  these patients
 appears in Dlis et al. (1976).


 Surgical procedures and results

    Two electrodes were inserted under X-ray control
in the midthoracic  region one or two vertebral widths
apart on the dura surrounding the spinal cord. Elec-
trodes  were  made of insulated steel with  exposed
platinum tips. A battery-operated stimulator/transmit-
ter at the patient's side generated 100-jusec pulses  at a
rate of 30  to 40/sec. A receiver located on  the skin
surface was attached to the opposite end of the elec-
trodes.  The voltage of the  pulses could  be controlled
by the patient.
'This work was carried out with the support of the Army Department, Ministry of Defense (British).

-------
Spinal Cord Stimulation and ERPs
                                             83
    At normal stimulation levels, the patients reported
paraesthesia throughout the body up to the dermatome
level corresponding to the site of the electrodes. In
view of the electrode positions adjacent to the dorsum
of the dura, it is probable that the  dorsal columns
were principally stimulated. Increasing voltage could
induce a painful sensation and  muscle fasciculation,
suggesting that  stimulation was not confined to the
dorsal columns, at least at high voltage levels.

    After a day of continuous stimulation, improve-
ment in function was clinically  discernible. Improve-
ment continued with  further stimulation over the
next 2 to 7 days, involving extensive areas of the body
and  additional  functions. Patients also consistently
reported a carry-over effect, i.e., they continued to
feel benefit hours after prolonged stimulation ceased.
Clinical and physiological change are detailed in Dlis
et al. (1976).

Experiment I

    To gain insight into  the mechanism  by  which
electrical activity induces the benefits reported and to
evaluate the utility of evoked-response techniques for
such an application, CNV was recorded both pre- and
postoperatively. Postoperative recordings were obtain-
ed under  both  stimulator-on and stimulator-off con-
ditions. Ag/AgCl electrodes were employed at Fpz,
Fz, Cz, and Pz, with linked earlobes as reference. The
Cz derivation was compensated for eye movements as
described by McCallum  and Walter (1968). Skin sites
were abraded with a blunt needle, giving an impedance
of less than 1 k-ohm. EEC was recorded with a 10-sec
time constant. Subjects  were seated in a large, quiet,
darkened  room kept at a comfortable temperature. In
view of the carry-over effect mentioned above, sub-
jects were asked to switch off the stimulator IS hours
prior to postoperative recording; three of them did so.

Experimental paradigm

     A simple CNV paradigm consisting of a distinct
warning click through earphones followed  1 sec later
by a train of light flashes (16/sec) from a stroboscope
25 cm in  front of the subject was employed. Subjects
were instructed to terminate the flashes as quickly as
possible by pressing a button. The intertrial interval
varied irregularly between 5 and 15 sec. Performance
was encouraged during short pauses between blocks
of 12 trials. Eight blocks were presented in the follow-
ing sequence: three acquisition, two distraction, one
resolution,  and two with eyes  open and  fixated on
the pupil reflected  in a mirror attached to the strobo-
scope.  During the  first  six blocks, subjects were
instructed to keep their eyes closed and still. CNVs
were derived (using a POP 12) from  averages of 8 of
the  12 most artifact-free trials. Baseline was computed
from mean EEC activity during the 2.1-sec epoch pre-
ceding SI, and amplitude from the 200-msec epoch
preceding S2.
   Distraction trials were included in order to evalu-
ate the possible distracting effect of electrical stimula-
tion  (McCallum  and Walter 1968). An intermittent
tone was presented through the earphones between
trials during this phase.
Results

    Recordings made  prior to electrode implantation
showed the form,  amplitude, distribution of CNV
components, and effect of distraction to be similar to
those found in healthy  subjects. Mean drop in Cz
amplitude during distraction compared to acquisition
was 7 ju V(standard deviation = 3.9).

     Averages of pre- and postoperative vertex CNVs
are shown in Fig. 1. Waveforms are indistinguishable
except for a slight increase in postimperative negativi-
ty and a slight increase in the P300 component after
surgery. There was no consistent change in the P300
component, and 75% of the difference was derived
from one subject (RW). Postoperative CNVs with the
stimulator on and off are compared in  Fig. 2. Even less
difference  is apparent  between  these  two average
waveforms. Neither pre- nor postoperatively  were
there  large individual   differences  that  might be
concealed by averaging.  The largest difference was a
5-uV increase in amplitude after surgery in the case of
EFM, who failed to benefit  from the procedure. The
mean computed difference in maximal pre-82 CNV
amplitudes in both  Fig. 1 and 2 was 1 /iV. The effect
of distraction  after  surgery,  with or without  the
stimulator, was essentially the  same as before (6 M'V,
SD=3.9).
Discussion

    The absence of any difference between CNVs with
the stimulator on  or  off could be explained by  a
balance between the  stimulator's distraction (CNV
reducing) and stimulation (CNV enhancing) effects.
On the other hand,  the experience  of other CNV
experimenters  that  a  continuous stimulus  that does
not intrude on the subject's awareness does not have
a CNV-reducing effect suggests that there was no dis-
traction  effect and no compensating CNV enhance-
ment.

    It had also been suggested that the drama of the
operation and experimentation and the hope of bene-
ficial treatment might be reflected in a  postoperative
increase  in CNV in either the stimulator-off or the
stimulator-on conditions. The absence of any substan-
tial  change suggests  that the benefit  derived from
stimulation was not a placebo effect dependent on
the general excitement of the patient. However,  it a
possible  that the subjects were  so highly  motivated
that no further increase in CNV was  possible (ceiling

-------
  84
                                       Abraham et al.
                    BEFORE SURGERY
                    .AFTER SURGERY
                                                                        10/uV
                                                                                     1 sec
            F/& ./. Combined average of CNVs from five subjects after surgery compared with CNVi
            from the same subjects before surgery. (Each trace is derived from 320 trials.)
                    • STIMULATOR OFF

                    .STIMULATOR ON
                                                                         10 MV
                                                                                    1 sec
                           S1
S2
           Fig. 2.  Combined average  of CNVs from five subjects receiving spinal cord stimulation
           compared with CNVs from the same subjects not receiving stimulation.  (Each trace is
           derived from 160 trials.)
hypothesis,  Knott and Irwin 1973). The brief op-
portunity available  for  experimentation  did  not
permit the rejection of this hypothesis  until  later
(Abraham et al. 1978 ).

    It appears possible that  the impulses, although
not  susceptible  to end-organ  habituation  or peri-
pheral  inhibition  since they were generated  by a
stimulator in the spinal cord, did not reach the cortex
in   sufficient  strength to affect  the CNV.  It  also
appears possible,  if CNV generation  has a primary
subthalamic source (McCallum et al.  1973), that the
spinal  cord  stimulator,  in  activating  the  dorsal
      columnar  pathway which has few collaterals  to the
      reticular formation, could virtually bypass the mid-
      brain reticular  formation and  thus  have  minimal
      influence on the CNV. An additional experiment was
      carried out to evaluate these possibilities.

      Experiment II

          The four multiple sclerosis patients took  part in
      this  experiment. A mechanical vibrator was strapped
      to the skin overlying the  lower end of the right kid-
      ney. The purpose of this was to provide a stimulus
      comparable to  that which the electrical stimulator

-------
Spinal Cord Stimulation and ERPs
                                              85
was supposed to generate. The kidney site was chosen
so that it  did  not cause  discomfort to the seated
patient, who  could  nevertheless  feel the  vibrator
when it was switched on in spite of loss of sensation
in other areas. The patient could not see  or hear the
vibrator, since the sound was  effectively masked by
70 dB white noise  transmitted  through earphones.
The current in the vibrator was adjusted to induce the
minimum  possible  vibration  when the  circuit was
closed. The patient then adjusted  the voltage of the
spinal cord stimulator so that both pieces of apparatus
delivered sensations  of roughly equivalent intensity.
This  generally involved a reduction in stimulator volt-
age and an accompanying reduction of the paraesthc-
siae,  sometimes to  a band around the chest. The
current to vibrator and stimulator was controlled by a
PDP  1 2 computer which  delivered a 250-msec burst of
stimulation via  one or the other apparatus. This burst
was  incorporated into the CNV paradigm, replacing
the flashes at S2 for 40 trials and  then replacing the
click  at SI for  40 trials. The vibrator and stimulator
were alternated after each block of 10 trials. Patients'
eyes were  open and fixed on  the mirror throughout
this phase of the experiment, and a rest was allowed
after every 20 trials.

Results

    Vertex CNVs  observed  during stimulator  and
vibrator conditions  are superimposed  in Fig.  3 (S2
substitution) and  Fig.  4  (SI substitution).  CNVs
recorded in S2 substitution trials differed from Experi-
ment I waveforms only  in the  marked attenuation of
the S2 evoked potential. CNV resolution was equally
rapid  in both  experiments. In  the SI substitution
conditions  fast and slow evoked potentials appeared
earlier and were larger when generated by the stimu-
lator than  by the  vibrator. The  increased magnitude
may be an electrical artifact from the stimulator since
a sharp rise appeared at SI onset. The latency differ-
ence is attributable to a mechanical delay in vibrator
response, a tremor transducer attached to the vibrator
indicated a slow rise  time to peak.

Discussion

     Spinal cord  stimulation  did  not produce  any
observable change  in CNV amplitude or shape in
Experiment I. The results of Experiment II rule out
the  possibility  that  patients did not  perceive  spinal
cord stimulation since motor responses were perform-
ed correctly in both S2 substitution  conditions. It
should  be  noted,  moreover, that the voltage of the
stimulator had  to be reduced in  Experiment  II to
match the  perceived intensity of the vibrator. Percep-
tion of the electrical stimulus should therefore have
been more pronounced in Experiment I than in Exper-
iment II.  It is  improbable  that impulses  from the
stimulator were too weak to affect the CNV.
    The  apparent differences in evoked  and slow
potential patterns evoked by electrical and vibratory
stimuli in SI substitution  conditions were probably
artifactual. The slow mechanical rise time of the vibra-
tor presumably induced a time lag before the stimulus
level reached the threshold of the sensory end-organ.
Differences in axon length from the sites of vibratory
and  electrical stimulation  cannot account  for  the
observed time lag. If latency and amplitude differences
are dismissed as  artifact, then one may conclude that
comparable  CNVs were  elicited by both  types  of
warning  cues. This  finding implies either  that  the
CNV may be initiated by stimulation  of the dorsal
column-medial   lemniscus-specific   thalamic  nuclei
pathway, or that the stimulator activated other path-
ways.
 Conclusions

    The absence of pre- and postoperative differences
 in CNV found in this study contrasts with other physi-
 ological findings (Illis  et al.  1976). A return toward
 normal of the  H-reflex recovery  curve and the fifth
 component of the auditory evoked response in some
 patients suggested that physiological change had taken
 place. Results of these preliminary experiments  sug-
 gest that the beneficial effects-psychological or physi-
 ological—of spinal cord stimulation are not mediated
 via cortical excitement as measured by the CNV.  The
 data, however,  do not preclude  the possibility  that
 CNV amplitude reached a ceiling level prior to spinal
 cord stimulation in highly motivated patients, though
 subsequent experimentation  (Abraham  et  al. 1978)
 did  so. The similarity of CNV waveforms with  and
 without dorsal  column stimulation suggests, more-
 over, that reafferent activity,  presumably mediated
 by  this pathway, does not contribute to the CNV.
     Finally, the ability to bypass peripheral nerves
 and end-organs provided by spinal cord stimulation
 offers  a unique opportunity  to  those investigating
 event-related potentials in situations easily confound-
 ed by peripheral factors.

 Summary

     Clinical improvement has been observed in multi-
 ple sclerosis patients following spinal cord stimulation,
 although the mechanism underlying this effect is not
 known. Two brief CNV experiments were undertaken
 to investigate  possible  changes  in psychological or
 neurophysiological  state  associated  with  clinical
 improvement. Spinal cord stimulation did not produce
 any discernible change  in the CNV,  suggesting that
 therapeutic  effects are not mediated  by  cortical
 mechanisms.

-------
86
                                                                                 Abraham et al.
      S2
-VIBRATOR

-STIMULATOR
           Fig. 3. Combined average of CNVs from three subjects in whom spinal cord stimulation
           was used as the imperative stimulus compared with CNVs from the same subjects in whom
           external somatosensory stimulation  was used as the imperative stimulus. (Each  trace is
           derived from 48 trials.)
        SI
 -VIBRATOR

 -STIMULATOR
             Fig. 4.  Combined average of CNVs from three subjects in whom spinal cord stimulation
             was used as the warning stimulus compared with CNVs from the same subjects in whom
             external somatosensory stimulation  was used as the warning stimulus. (Each  trace is
             derived from 48 trials.)

-------
 FUNCTIONAL SIGNIFICANCE OF  CEREBRAL
 POTENTIALS  PRECEDING VOLUNTARY
 MOVEMENT

 L. DEECKE
 Department of Neurology, University of Ulm, West Germany
    It has been established that cerebral activity pre-
ceding voluntary self-paced  movement (e.g., index
finger flexion) can be recorded through the intact hu-
man scalp (Kornhuber and Deeckc 1964, 1965). Three
different potentials occur: a slow potential, called the
Bereitschaftspotential (BP)or readiness potential, and
two faster potentials, referred to  as the premotion
positivity (PMP),  and the motor potential (MP). Cer-
tain methodological requirements are  necessary to
record these three  potentials. Amplifiers  with long
time constants are necessary to record BP. To pick up
the faster potentials  and to achieve the necessary
signal-to-noise ratio, exact temporal triggering condi-
tions  and a large number of trials  are required. Best
results are obtained if the subject  is trained to relax
the agonist muscle to the extent of complete silence
m the intramuscular electromyogram (EMG) and then
perform an abrupt and  rapid movement. A sensitive
triggering level must be used to ensure triggering from
the very first intramuscular action potential. Deecke
et al. (1976) have shown with index finger  flexion
that the earliest activity  occurs in the agonist muscle
(M. flexor digitorum communis,  pars  indicts),  al-
though accompanying muscular activity is recorded in
many other arm and neck muscles. Additional meth-
odological requirements  are  the  exclusion  of eye
movements (e.g., by gaze fixation) and careful editing
of every trial.
    If these precautions are taken,  three different
types of subjects are found (Fig. 1). Type A (15%)
shows increasing negativity until movement onset in
all precentral and parietal leads with a steep rise about
60 msec prior to EMG onset in the contralateral pre-
central lead. Type B (41%) shows increasing negativ-
ity  until movement onset in  the  contralateral pre-
central lead but a positive deflection about 90 msec
prior to movement onset in other  precentral and
parietal leads. Type C (44%) shows a  positive deflec-
tion m all precentral and parietal leads, but a less steep
stope in  the contraiateral precentral lead. Both Type
B and Type C  (85%) exhibit PMP. Fig. 1 illustrates
that, in the final 150 msec prior to EMG onset, the
waveform  is complicated by the superimposition of
the three  potentials mentioned  above, differing in
polarity and topographical  distribution and perhaps
morphology between subjects.
                                       Type A
                                      Type  B
                                      Type C
           750
'0  msec
Fig. 1. Three  different Types of subjects. Schematic
diagram of the potential course in left and right pre-
central and midparietat leads  preceding right-sided
finger movement. (Deecke et al., 19 76/

-------
  88
 Bereitschaftspotential

     The slow premovement potential (Bereitschafts-
 potential, readiness potential, or Nl) is a negative shift
 of the cortical dc potential, as is the contingent nega-
 tive variation (CNV). The principal characteristics of
 BP are early onset  (commencing up to 1.5 sec  or
 more-mean 0.8 see-before movement), gradual neg-
 ative increase, and bilateral distribution over the pari-
 etal and precentral  cortex of the two hemispheres.
 The BP is bilateral-even if the movement is only uni-
 lateral. These characteristics suggest that the early BP
 cannot  reflect  motor command processes, since  a
 potential reflecting motor command would have to
 be faster than motor reaction time.  The term early
 preparatory process  is suggested for BP, meaning a
 thalamocortical facilitation process which selectively
 excites those cortical areas involved  in the intended
 movement  and inhibits or  does not affect other areas.
 A negative shift of cortical dc potential is caused by
 an increase in synaptic drive in (upper) cortical layers,
 since intracellular recordings of cortical neurons reveal
 a simultaneous decrease of membrane  potential and
 an increase in excitatory postsynaptic potential (EPSP)
 rate (Caspers and Speckmann, in press). The early
 preparatory process of the BP does not reflect simply
 general arousal because it can be modified by the
 experimental situation. The BPis, therefore, a valuable
 indicator for the loci of cortical activity or inactivity.

     The literature concerning BP laterality is contro-
 versial. The CNV appears  to  be  symmetrical even in
 split brain  patients responding with their right hand
 to stimuli flashed into their right visual field (Gazzani-
 ga and Hillyard 1973). Lateralization of the CNV has
 been observed only  when higher lateralized cortical
 functions such as speech (Lowet al. 1976) or numeric
 operations  (Butler and Glass  1974) are addressed in
 the  experimental situation. When we first  recorded
 the  brain  activity  preceding voluntary  movement
 (Kornhuber and Deecke 1964), we expected that  a
 unilateral movement  would be preceded by unilateral
 (i.e., contralateral) cortical activity as  suggested by
 classical concepts based on stimulation  of the motor
 cortex. Finding a bilateral negativity preceding uni-
 lateral  movement was, therefore,  a surprise. Bilateral-
 ity does not necessarily imply bilateral symmetry, but
 this was the case in  the initial part of  the readiness
 potential shown In bipolar recording! of left vs right
 precentral and left vs right parietal leads. Only pre-
 centrally  did a  slight contralateral  preponderance
occur,  commencing about  400 msec  prior  to EMG
onset and  averaging  0.9  juV  at  a time  ISO msec
prior to movement.

    In order to extract the three different potentials,
superimposed on each other  in the final ISO msec
 prior to EMG onset, certain measurement rules were
established (Fig. 2). BP was measured  150 msec prior
                          -150            0 msec
 Fig. 2. Rules for measuring the three different poten-
 tials.  The  Bereitschaftspotential  (BP) is  measured
 (with respect  to  a prepotential baseline)  ISO msec
 prior  to  EMG  onset (BP15Q) and at  its kinking (if
 present)  to positivity (BPkn); PQ is  the amplitude at
 EMG onset. For measurements of the premotion posi-
 tivity  (PMP), two differences were taken: P0-BPjtft
 possible in all the graphs and P0-BPkn (PMP proper)
 only possible in graphs with a kink. The motor poten-
 tial (MP)-dotted line-was measured in bipolar record-
 ings, contralateral vs ipsilateral precentral or contra-
 lateral precental vs midparietal,  to extract the  two
 bilateral potentials, BP and PMP. The MP represents
 the additional negativity over the contralateral motor
 cortex, which  occurs  on  average  54 msec  before
 EMG onset  (Deecke etal, 1976.)

 to EMG onset to avoid contamination by other poten-
 tials.  The contralateral  precentral preponderance is
 equivalent to Gerbrandt's  (1977) asymmetrical  Nl
 component. Pane tally, however, the  BP is absolutely
 symmetrical  150 msec before EMG onset (Fig. 3). In
 conclusion, whether the  BP is symmetrical or lateral-
 ized depends on the location and time of measure-
 ment.  Fig. 3, a three dimensional plot  of averages
 from 39 subjects, illustrates BP lateralization.


     The  precentral contralateral preponderance of
 the BP and the MP correlates with handedness (Deecke
 et. al. 1973). Twenty-five right-handed subjects show-
 ed  significantly  more negativity (2 p<05) recorded
 over the dominant motor cortex than the minor one,
 both being contralateral  to the movement. In 13 left-
 handed subjects, an analogous trend  (not statistically
 significant)  toward  lateralization  was observed, In
 accord with sinistrals being more ambidextrous (less
 cortlcally lateralized) than right-handed subjects. Simi-
 lar  results were obtained by Kutas  and Donchin
 (1974). The BP  is subject to  psychological factors in-
 cluding motivation, and  gradually declines with age
 beyond the fourth decade of life (Deecke et al.,  this
volume).

 Premotion Positivity

    In Type  B and C subject! (85% of the test popu-
 lation), a positive deflection (premotion positlvity,
 PMP, or PI) was observed in the final ISO msec prior
 to EMG onset.  PMP was maximum over the anterior

-------
Potentials Preceding Movement
                                             89
                                                                                   160 maec
             PRECENTRAL
                                                                PARIETAL
Fig. 3. Laterally of the readiness potential. Three-dimensional plot of the distribution of negativity preceding
right-sided finger movement (grand averages across 39 subjects), X-axis; electrode positions over the left, mid, and
right precentral and parietal regions.  Y-axis: negativity in microvolts. Z-axls: time prior to EMG onset. In precen-
tral leads (left sub figure),  the BP is slightly lateralized (contralateral preponderance atBPjso and BP%n). At P0,
pronounced lateralization due to motor potential is  seen. In parietal leads (right sub figure), the BP is exactly
bilaterally symmetrical (BPjso an(i BPkn)- Tne sliSnt (insignificant) lateralization atP0 may be due to spread of
the motor potential.
parietal region, Pz (cf. Fig. 1 in Deecke et al.,  this
volume). PMP was measured as the difference Po-BPjm
or P0-BPj5o (Fig. 2). Since the mean onset time of
PMP is 87 msec prior to EMG onset, which is shorter
than mean motor reaction time, this potential  might
reflect cerebral  activity associated with the motor
command.

    The PMP has also been described by Vaughan et
al. (1968), Shibasaki and Kato (1975), and Cerbrandt
(1977). At present, there seems to be agreement about
PMP phenomenology, but major controversy about its
functional significance. The Ulm group considers the
PMP to be an expression of movement initiation or
motor command, which originates in the parietal area.
Vaughan et al. (1970) suggested that the PMP is associ-
ated with pyramidal tract activity. Shibaski and Kato
(1975) proposed that the PMP  is the expression of
unilateral inhibition, unilateral movement  being the
result  of unilaterally  inhibited  bilateral movement.
Gerbrandt (1977) has confirmed the existence  of the
PMP, but speculates that it may be an epiphenomenon
of movement occurring toward the end of long experi-
ments.  The  Ulm group  demonstrated  at  Bristol
(McCallum and Knott 1976), however, that the PMP
can be  seen  after the first few trials (Fig. 4). The
reason for the variable appearance of the PMP in sub-
jects is not known (see discussion above of different
types of subjects), but we have found that a subject
maintains his type in repeated experiments, i.e., there
is high intra-subject constancy of the PMP.

    The  hypothesis that the  PMP  reflects a motor
command from the parietal cortex is supported by
several considerations: (l)The PMP  is not  simply the
resolution of the BP since there is no correlation
between PMP and BP amplitudes (Deecke et al. 1976).
(2) The PMP preceding tactually guided finger move-
ments is maximal  over the anterior  parietal region of
area 5, which is the sensory association cortex of the
somatosensory  modality. The  parietal area, whose
destruction causes apraxia, is classically considered to
be the  key structure for learned  tactually guided
movement  and skills. The PMP  may, therefore,  re-

-------
 90
                                         Deecke
  0052
                                                            EMG  ONSET
                                            E06
                                                                                      38
                   L front
                                                                      R front
                                               Vertex
                  L par
                                                                      R par
                                              EMG
                                                                                          20
                                                                                      1  sec
                                                                                                   i
Fig. 4.  Premotion positivity in the beginning of an experiment.  Typical PMP preceding right-sided finger move-
ment, which is discernible here after 38 trials. Maximum at the vertex.
fleet the activation of a movement program stored in
the parietal cortex. This spatial  code must then be
converted to temporal code for  the exact timing of
movement  parameters. According to  the theory of
Komhuber (1971, 1974),  conversion is achieved by
subcortical motor function generators of cerebellum
(fast,  preprogrammed movements) and basal ganglia
(slow,  smooth movements).  (3)  The 33-msec  dif-
ference   between  PMP and MP  mean onset times
permits  information transfer and processing via  a
cortico-cerebello-motor  cortical   loop.  (4)  Thach
(1970) has confirmed in animals  that the cerebellum
is activated prior to the motor cortex. The onset time
of frequency changes in cerebellar dentate cells (70
msec before EMG) would fit well between PMP onset
time (87 msec) and MP onset time (54 msec). Thach
(1975) reported  that neuronal  discharge patterns of
dentate  cells changed significantly earlier than cells in
motor cortex  (simultaneous recordings in the same
animal). In the thalamic  relay of the dentato-cortical
path (ventrolateral nucleus), activity also changes well
in  advance of movement onset (Jasper and Bertrand
1966, Evarts 1970). Finally, Mountcastleet al. (1975)
have demonstrated that  the anterior parietal region

-------
Potentials Preceding Movement
                                              91
contains so-called "command neurons" that are active
prior to movement

Motor potential

     What  we  call the motor  potential  (MP)  is the
additional negativity that occurs over the contralateral
motor cortex immediately prior to EMG onset. Where-
as the  PIVEP and BP are bilateral, the MP is the only
unilateral  potential preceding voluntary  movements
of one side. A bipolar montage was used to extract
this  additional negativity  from  the BP  i.e., contra-
lateral  vs  ipsilateral precentral and contralateral pre-
cenlral vs midparietal  (Fig.2).  A  definite  upward
kinking of the waveform, arising from the slow neg-
ative deflection of the  contralateral preponderance,
was  discernihle in about 76% of the subjects. Onset
of this kink averaged 54 msec prior to the first intra-
muscular  action  potential  in  the agonist  muscle.

     It  is impossible to execute an index finger move-
ment in isolation. There  are  always  accompanying
movements in  other hand, arm, and even neck mus-
cles. However, it has been  shown  that with  index
finger  flection  the earliest activity occurred in the
agonist muscle, from which  the  trigger pulse was
derived (cf. Deecke et al. 1976, Fig. 2B). Thus, we are
certain that the motor potential recorded in bipolar
leads is a premovement potential.

     It  should be stressed that  the MPis not identical
with the monopolarly recorded  N2 component, the
premovement nature of which  has been questioned in
epicortical  recordings (Papakostopoulos ct al.!974a).
Indeed, this component also occurs with passive move-
ments (Kornhuberand Deecke 1965) and shows phase
reversal between precentral recordings (negative polar-
ity)  and postccntral recordings (positive polarity,  cf.
Kornhuber and Deecke 1965, Fig.7C).These findings,
obtained with  subgaleal  needle electrodes, have now
been confirmed in epicortical recordings by Papakos-
topoulos et al.  (I974a).

     Spatial and temporal characteristics  of the MP
immediately suggest its  functional significance:  its
location is restricted (with unilateral finger  movement)
to the hand area of the contralateral motor cortex
and  its onset is very close to EMG onset (54 msec on
the average). The MP, therefore, is probably associated
with activity of the motor cortex initiating  the des-
cending volley  in  the  pyramidal tract.  The motor
cortex  is  a highly specialized structure,  but by  no
means  the origin of all  voluntary  movement,  as
assumed in classical concepts. According to current
theory (Kornhuber 1971,  1974),  only those  move-
ments  that  require the highly sophisticated tactile
analysis provided by  the  precentral and  postcentral
gyri  have ascended to motor cortex during phylogen-
esis.  Generation of movements which do  not  require
 tactile guidance,  such as eye movement, remains in
 subcortical structures. Therefore, no motor potential
 is found  preceding  eye movements  (Becker et  al.
 1972).  Movements dependent  on  an intact motor
 cortex in the primate  are fine finger, toe, lip, and
 tongue movements, and  manual skills requiring tactile
 control. Thesomatotopic representation of these parts
 is thus disproportionally large in the motor (and soma-
 tosensory) homunculus.

     Further  experiments  with  different  types  of
 movements  are  necessary. Movements can differ in
 purpose, speed, context of action, internal needs, and
 modality  of  mediating stimuli, as well as the part of
 the body involved in movement. The classical assump-
 tion  that movements originate in  motor cortex  is
 therefore  false.  The motor cortex  comes  into play
 later in the course of movement preparation and only
 with certain movements.  The willful initiation  of
 movement can arise in  many cortical areas of the
 so-called  "sensory association" fields, depending  on
 the special type of movement: factually guided move-
 ments in the somatosensory association cortex, visual-
 ly guided movements in the  visual  association areas,
 speech  movements in  the speech  centers, writing,
 music playing,  etc.,  in  their respective centers. It
 appears that the entire cortex has immediate access to
 movement by means of known omnicortical projec-
 tions to subcortical motor function generators of the
 cerebellum and basal ganglia.

 Reafferent potentials

    This term has been  suggested  for  potentials that
occur after movement onset (Kornhuber and Deeeke
 1965) because the sequence of positive and negative
components after EMG onset closely resembles the
average  evoked potential  to  somatosensory stimuli.
 Furthermore, the same  components, although some-
what larger  in amplitude, occur  after passive move-
 ments (Kornhuber and Deecke 1965,Papakostopoulos
 et al. 1975). The functional significance of  this po-
 tential complex is probably reafferent activity evoked
 by the movement. Thus, this complex could be called
 a proprioceptive evoked response. It seems reasonable
 to assume that reafferent potentials originate mainly
 from peripheral  receptors such  as cutaneous, deep
 somcsthetic, and muscle spindle  receptors, but may
 also come from feedbackor reentrant activity of lower
 motor  centers  in  the   brain stem,  cerebellum,  or
 spinal cord.  As  a result, the reafferent potentials
 would not be completely  abolished after rhizotomy
 (Vaughan et  al.  1970). The findings of Vaughan et al.
 are  therefore not entirely  contrary to the reafferent
 conception   of  potentials after  movement  onset.
 The  notion  that passive  movements evoke larger
 potentials than similar active movements is of interest
 in view of theoretical concepts such as Effernzkopie
 or corollary  discharge (Hoist ;md Mittelstedt 1950).

-------
 EXPERIMENTAL  MANIPULATION OF MOTOR
 POSITIVITY:   A PILOT STUDY*

 J. DELAUNOY, A. GERONO, AND J. ROUSSEAU
 Department of Medical Psychology and Psychosomatic Medicine, University of Liege,
 Liege, Belgium
    The slow positive wave that  follows movement
 onset has generally been attributed to proprioceptive
 or kinesthetic afference  (cf. Kornhuber and Deecke
 1965), although this hypothesis has never been con-
 firmed experimentally. Several observations do not
 support this hypothesis: (1) postresponse positivity
 persists in monkeys after dorsal rhizotomy (Vauahan
 et al. 1970), (2) the positivity persists longer 0>500
 msec) than the predicted duration of proprioceptive
 discharge, and (3) the waveform is frequently absent
 in psychopathological patients in  whom there is no
 apparent proprioceptive deficit.

    An  alternative  hypothesis  is  that postresponse
 positivity may be analogous to the P300 wave (Sutton
 et al. 1965) and may reflect the uncertainty or proba-
 bility  of occurrence  of the motor act. The present
 experiment was designed to test this hypothesis.

 Method
    Three normal adults (two men and one woman,
mean age 35 years) were instructed to extend the right
forefinger at regular intervals against a given resistance
of 50g (Rl) or 200g (R2).  Each subject completed
two sequences of ISO trials each in which Rl (or R2)
was presented randomly on 1/3 of the  trials. The
probabilities were reversed during the second sequence.

    EEC was  recorded at the vertex with chlorided
silver electrodes referred to the left earlobe and with
a 5-sec amplifier time constant. EOGwas also recorded
for rejection of trials with eye movement artifact.
Averages were triggered from  the response mechano-
gram. The averaging epoch was 4 sec (1.5 sec before
the beginning of movement and 2.5 sec after). Two
measurements were made on the averaged waveforms:
(1) the readiness potential (RP) computed as the volt-
age difference between the potential at movement
onset defined by mechanogram, and baseline (initial
500 msec of the average) and (2) the positive wave
measured as the voltage difference between baseline
and  the maximum positivity  following movement
onset.

Results and comments

    RP did not vary as a function of anticipated or
unexpected resistance. The positive wave did not vary
as a function of the expected (2/3) resistance, but
was larger for the unexpected (1/3) resistance, irrespec-
tive of actual level, in two subjects.

    These results provide limited support for the pro-
posed hypothesis that postresponse positivity may be
analogous to P300 and may reflect the certainty or
probability of a salient feature of the motor act. Per-
haps this positivity reflects the operation of a central
comparator between  a motor program elaborated
before movement  and a neural representation of the
executed movement. Observations in eight additional
subjects run under the same experimental conditions
confirm the preliminary results reported here.
 This work was supported in part by Grant 20397 - FRSM.

-------
TELERECEPTIVE, PROPRIOCEPTIVE, AND
CUTANEOUS INFORMATION  IN  MOTOR
CONTROL1
 B. DUBROVSKY

 The  Allan Memorial Institute,  Departments  of Psychiatry and Physiology, McGill  University,
 Montreal, P.Q., Canada
    Two broad views can be distinguished in the inter-
 pretation of mammalian brains. In one, articulated by
 Craik (1943), the "fundamental  feature  of neural
 machinery" is the "power to parallel or model exter-
 nal events." This model is matched against input from
 daily life in order to select the most appropriate line
 of action (Young 1971). In contrast to this perspective
 which  emphasizes  perceptual  aspects of the central
 nervous system, other views emphasize motor control
 as the central determinant in the evolution and func-
 tion of the brain. This conception  can be traced to
 S. C.  Pierce's pragmatic view  that "the evolutionary
 increase in man's  capacity for  perception,  feeling,
 ideation, imagination, and the like, may be regarded,
 not so much as an  end in itself but as something that
 has enabled us to behave, to act, more wisely and
 efficiently" (Sperry 1952). Lashley (1951) encom-
 passed both positions: "the neurological problem is in
 large part, if not entirely, the translation of the affer-
 ent pattern of impulses into  the efferent pattern."
 This review will summarize recent work on the contri-
 bution of proprioceptive, tele receptive, and cutaneous
 stimuli in the  development of forces and the spatio-
 temporal pattern necessary for movements directed
 toward the outside world.
    The  distance  receptors  of vision, audition, and
 olfaction are  axial in the development and construc-
 tion of the nervous system (Sherrington 1906). Eyes,
 by  signaling distant events, can provide for the  plan-
 ning of behavior—not merely reflex responses, but
 complex  activity in which cortical integration plays a
 fundamental role. In recent experimental work, the
 relevance of midbrain structures in visually guided
 behaviour has been clearly established. Phylogenetic
 and physiological data, as well as ablation and depri-
 vation experiments,led to the notion that visual local-
 ization is primarily  associated with activity in the
 superior  colliculus, whereas visual identification in-
 volves  the geniculostriate  pathway (see review in
 Dubrovsky and Garcia-Rill 1971). It is clear, however,

 1 Supported by the Medical Research Council of  Canada.
that  perception of a single  unified object or event
includes attributes  of quality, intensity, position in
space, and duration in time. The integration of the
output of parallel channels  of intermediate sensory
processing is then a necessary condition to the prin-
ciple of parallel processing (Davis  1956).

    Phillips (1966)  has advanced the concept that
corticofugal neurons in the motor cortex are "common
paths" leading out of the cortex where many differen-
tiated peripheral inputs are integrated. Since planning
of extrapersonal motor strategies requires a properly
defined spatial context,  we decided to explore the
possibility of convergence and interaction (a prereq-
uisite for integration, Davis 1956) in the motor cortex
of two distinct  neuronal groups related  to visual
perception-the superior  colliculus and the occipital
cortex.
    Experiments with anesthetized cats revealed that
the motor-sensory cortex receives stimuli both from
the visual cortex and from the superior colliculus.
Almost half of the  100 cells studied in this cortical
zone  received  convergent stimuli from both central
areas. Inhibitory effects, studied by stimulating against
a background  of activity induced by iontophoretic
release of glutamate, were  observed in 51% of the
neuronal population studied. Analysis of latencies,
duration of inhibitory effects, and following frequen-
cies, indicated  that the superior colliculus and occipi-
tal cortex stimuli arrive at the pericruciate cortex by
independent pathways (Dubrovsky and Garcia-Rill
1971).

    After establishing that  the visual cortex  and
superior colliculus  may  mediate visual information
reaching motor cortex, the response characteristics of
motor-sensory  cortex neurons to natural visual stim-
ulation were investigated (Garcia-Rill and Dubrovsky
1971, 1973,1974). Briefly, responses to stimuli with-
in visual receptive fields were of the "on," "off," and
"on-off' type. Most receptive fields were rectangular
and very large.  About 75% of the 203 neurons studied

-------
 94
                                      Dubrovsky
 had receptive fields averaging 72° x 83° and impinged
 on both nasal and temporal regions of the visual field;
 the rest were limited to the nasal side and averaged
 38°x49°.

     Latency of excitatory responses to both the onset
 and disappearance of stimuli averaged 45 msec. Inhib-
 itory  effects analogous to the excitatory activity
 evoked  by  discrete   retinal stimulation  were also
 found. The latency for inhibitory responses was long-
 er, averaging 75 msec to  an "on" stimulus, and  65
 msec  to  an "off" stimulus. A characteristic of the
 response  of motor cortex neurons to visual  stimuli
 was the  inability to  follow recurrent presentations
 beyond I/sec.

     A majority of units (56%) having visual receptive
 fields  displayed  movement sensitivity. Only 19% of
 movement-sensitive cells had a  preferred direction for
 visual  stimuli.

     Finally, in  examining the relationship between
 visual  receptive fields and cutaneous receptive fields
 of the same units, a  topographical distribution  of
 visual  input to motor-sensory cortex became evident.
 Essentially, more colonies of  cells with  input from
 the trunk received visual information than those colo-
 nies with input from  the distal forelimbs. Only 15%
 of cells with exclusive input from the distal extremi-
 ties received visual input,  whereas 75%  of neurons
 with exclusive and convergent input from proximal
 areas received visual information.

    The topographical organization of visual afferents
 in motor cortex, the large size of visual  receptive
 fields,  the  long  latency of response, and  the poor
 frequency-following characteristics of motor cortex
 neurons suggest  that  control  of accuracy of  move-
 ments  is  not mediated by visual input  to  motor
 cortex. The data  do suggest,  however  that  visual
 afferents  to motor cortex may be related to spatial
 aspects of perceptual-motor function. This hypothesis
 is further supported by the binocular convergence of
 visual  pathways in motor cortex. According to Young
 (1962), bilateral representation is the basis of "a sys-
 tem for representing features of the environment that
 are signaled by various receptors in their correct spa-
 tial relations."

    Although neurons in the motor cortex are char-
acteristically polymodal  for  cutaneous  and  deep
modality  receptors, visual  and vestibular afferents
(spatial information)  converge preferentially  onto
areas corresponding to the body axis and proximal
limb zones. The latter areas are also richly endowed
with callosal connections,  while areas corresponding
to distal  extremities   are almost devoid  of callosal
projections  (see  Garcia-Rill  and Dubrovsky 1973).
 Results reviewed here are also consistent with the
 differential organization of neural systems controlling
 axial musculature of the limbs on one side, and distal
 on the other (Kuypers 1963). These  data, clinical
 observations on the role of vision in postural mechan-
 isms  for  normal  and  pathological  conditions, and
 developmental  studies emphasize the formation of a
 body-centered  spatial  framework as essential condi-
 tions for the development of visually guided behavior
 (Garcia-Rill and  Dubrovsky  1974) and suggest that
 visual  afferents  arriving at the motor cortex are' re-
 lated to tracking  functions involved in defining the
 spatial context in which movement will take place.
    We  further proposed that integration of propri-
 oceptive, somatic, spatial,  and  callosal  information
 converging  onto  areas of  the motor cortex which
 correspond  to the midline  of the body  is part of a
 neural  process related to construction of a reference
 axis for limb movement. This axis would form part of
 a system of coordinates, essential for the proper local-
 ization  of physical objects in space. Poincare (1923),
 in his classical analysis of the concept of space, recog-
 nized that we locate external objects in reference to
 our own body and that we apprehend spatial relations
 between objects  only in  relation to our own body.
 Interference with  the establishment of this referential
 body axis results in selective impairment of motor
 behavior, as we have shown (Dubrovsky et al.  1974).

    Concomitant with the electrophysiological study
 of the organization of telereceptive stimuli to motor
 cortex,  we  are analyzing the role of dorsal column
 afferents (a proprioceptive and somesthetic path) in a
 sequentially organized  motor act (Dubrovsky  et  al.
 1971,  Dubrovsky  and  Garcia-Rill  1973).  Cats  are
 trained  to jump up to release a piece of raw chicken
 liver attached to a vertically oriented, revolving wheel.
 The wheel is rotated by a motor of variable speed and
 testing  is done with constant speed for  all animals.
 The cats jump from a  force transducer platform.
 Components of force are recorded in three axes  on
 magnetic  tape and converted for digital  computer
 processing. The time of jumping in relation  to the
 position of the rotating wheel is evaluated by attach-
 ing  to  the  wheel  a small magnet  that  activates a
 switch attached to the rotating axis. Each revolution
 is  then  marked by a pulse  generated by the  switch
 and recorded  with  the  three components of force.
 The biomechanical investigation  is  complemented
 by high-speed cinematographic analysis of the move-
 ment.

    Behaviorally different  sequences  of  the act are
 independently  analyzed by studying the  following
 parameters: efficiency, accuracy, tracking, and search-
 ing index. Efficiency is evaluated as the percentage of
 successful  releases  of liver, without  taking  into
account  the  precision  of  execution. Accuracy  is

-------
Peripheral Input in Motor Control
                                                                 95
expressed as the frequency with which an animal hits
the holder instead of the target; i.e., an animal is less
accurate  the more it hits the holder. Tracking of the
released liver in space is assessed in terms of the abili-
ty of the animal to localize it on the floor within 3
sec after it lands. Finally, the searching index  is the
frequency with which an animal attempts to localize
a  piece  of  liver after  an unsuccessful  attempt  to
release it.

    After section of the dorsal column above the Cl
level,  significant impairment in all behavioral indices
was observed. The following parameters, directly relat-
ed to force, were also significantly decreased (Table
1): (1) height of jump,(2) time in the air, (3) maxi-
mum  resolved  force, and (4) peak to mean force ratio
(an indication of the speed at which the cat developed
the maximum muscular force).

    Film analysis showed that intact cats consistently
extended their limbs in a smooth and progressive way
towards  the liver in order to release it. The distance
between  the tips of forelimbs remained more or less
constant during flight. Postoperatively, the extension
of the limbs was interrupted  by  fast flexion  move-
ments and the distance between tips increased signifi-
cantly during flight.

    Impairment in  force  development after surgery
may  relate, in part,  to sectioning fibers  from skin
mechanoreceptors conveying  information  on limb
loading conditions.  Marsden et al. (1972) showed the
importance of cutaneous  afferents for load compen-
sation in proportion to force  and suggested a trans-
cortical mechanism  for its development. Furthermore,
we think that section of the dorsal column suppresses
an important path for cutaneous afferents  involved
in  supraspinal mechanisms  of  synchronization  of
                    muscle activity (Milner-Brown  et al. 1975). These
                    authors suggested that the mechanisms for exertion of
                    large, brief forces  (e.g., jumping)  may result from
                    strengthening reflex pathways involving  a  fast, lem-
                    niscal  route  to the cortex.

                        Before  surgery, the  animals jumped up with a
                    consistent stereotypic pattern including the direction
                    of takeoff.  After dorsal column section, the initiation
                    of jumping  was delayed and more variable in relation
                    to preoperative timing, and the cats showed a signifi-
                    cant change in the direction of takeoff, suggesting a
                    change in strategy of the animal while on the ground.

                        The postsurgjcal impairment in reaching the target
                    may be associated with the fact that the dorsal col-
                    umns are the exclusive afferent path to brain centers
                    for muscle spindles and low threshold joint afferents
                    from the forelimbs. Muscle spindles convey informa-
                    tion on the length and speed of changes in length of
                    muscles. Low threshold joint receptors yield informa-
                    tion on joint position.

                        Since position sense  is necessary for knowledge
                    of both the spatial and temporal parameters of move-
                    ment initiation, section of the  dorsal columns  may
                    impair  the ability  of an animal to readily initiate the
                    sequence  by interrupting fibers carrying impulses
                    from muscle spindles and low threshold joint afferents
                    from the forelimbs.  Both types of information are
                    involved in  kinesthesis. Absence of information from
                    forelimb  musculature  on length, and  changes in
                    length, of the muscles can also seriously impair the
                    accuracy  of a motor sequence  involving the use of
                    these   extremities.   Decreased    cutaneous  signals
                    should further reduce the  accuracy of forelimb move-
                    ment because skin sensation  is essential for refined
                    motor acts (Granit 1975).
            Table  1.  Effect of  Dorsal  Column Section on Mechanical  Performance
         Behavioral Index
Preoperative
Postoperative
                                                                                           Probability
  Height of jump, meters
    H-  -
    H'2g m
  Time  in the air, sec
  Work, joules
    W=mgH
  Maximum resolved force, Newtons
 Peak/mean force rate
   0.3118

   0.6139

   9.721


  82.15

   2.417
    0.1469

    0.5613

    4.582



   46.914

    1.500
p<.001

p<.02

p<001



p<.001

p<.002
 ' Results Irom one cat representative of a group of three. H * height of jump; 1 = impulse at takeoff; g= acceleration of grav-
  ity, 9.81 meters/sec^, x,y,z = axes of rectangular coordinate system (z vertical); m = mass of cat; W = work; F = force.

-------
 96
                                     Dubrovsky
    Tracking deficits, we believe, are due to the inter-
 ruption  of proprioceptive  signals  from  the  dorsal
neck  region ascending through  the dorsal columns.
Low threshold muscle afferents from deep and super-
ficial dorsal neck muscles project to zones correspond-
ing to the frontal eye fields of the cat brain (Dubrov-
sky 1974,  Dubrovsky and  Barbas 1975).  These sig-
nals  transmitted through  the dorsal  column may
play  an  important  role in proper coordination of
eye-head-movements  in complex  motor acts  re-
quiring movement of the head relative to the entire
body  while the body is  in motion.  For instance, we
recently  found that  afferents from  extraocular mus-
cles also project to  frontal eye field areas in the
feline (Dubrovsky and Barbas 1977).

    The  bizarre behavior of lesioned cats-frequently
searching for a target that has not been released—may
be a manifestation of the disruption in the organiza-
tion of  an integrated serial movement that occurs
when  essential information for programming of the
act does  not reach higher brain centers. The disturb-
ances  in timing of jump behavior after surgery may be
partially  related to proprioceptive  deficits. "Time"
appears to be a derived concept in the central nervous
system (Piaget 1970).   In light of current  knowl-
edge of muscle spindle physiology (Matthews 1972),
we suggest  that central  processes for the  construc-
tion of precise  timing strategies in  movements are
based, at  least in part, on information conveyed by
the spindles to  telencephalic areas.  This would in-
clude  information  concerning  changes in  muscle
length (distance  traversed  by  the body) and  the
speed at which  this change in position took place.
Time can  then  be derived from this information.
    Results of experiments reviewed here support the
preponderant role played by movements of the body
and extremities in the genesis of the concept of space,
as originally proposed  by Poincare (1923) and later
developed and  extended to the concept of time by
Piaget  (1970),  With respect to the function of tele-
proprioceptive  and cutaneous input in motor con-
trol, our investigations  suggest  that  visual  signals
arriving  at the motor  cortex  mainly in zones cor-
responding with axial and proximal body regions,
are probably   related  to  tracking  functions, i.e.,
bringing  the limbs to a target.  Control of accuracy,
precise timing, and  force  to  be  developed  during
movement  appear to be related to information origi-
nating  from proprioceptive and cutaneous receptors.

-------
METHODOLOGICAL CRITERIA FOR  THE VALIDATION
OF MOVEMENT-RELATED  POTENTIALS
L. GERBRANDT

Department of Psychology, California State University, Northridge, CA, U.S.A.
   Several scalp-recorded macropotentials, of distinct
functional origins, can become synchronized tempo-
rally with the occurrence of abrupt voluntary move-
ments  of the  hands or fingers.  These  movement-
related potentials (MRPs)  also superimpose spatially
across  the Rolandic region even  though most peak
elsewhere. It is not surprising, then, that tight meth-
odological control and extreme caution in measure-
ment  are essential  to  differentiate  optimally this
spatiotemporal mixture of MRPs into functionally
distinct components. Moreover, most studies of MRP
and CNV research are probably uninterpretable with
regard  to which component was manipulated and
measured, simply because investigators have not used
methodologies shown to be essential for effective dif-
ferentiation of MRP components (Deecke et al.  1969,
Gerbrandt et al. 1973, Deecke etal. 1976,Gerbrandt
1977). Thus, when investigators compare the effects
of an Independent variable on "the readiness poten-
tial," it is uncertain whether a readiness potential or a
compound potential of several functional origins is
being affected  by the variable under study. In order
to  put known considerations about the possible
sources of "conceptual confounding" into a useful
perspective,  this review will  be  used to derive a
methodological checklist for evaluating MRP results.

   Investigators usually assume that there is a unitary
readiness potential (Nl) that can be recorded In isola-
tion, even though MRP research  indicates that this
negative sustained potential (cf. Fig.  1, top trace) is
easily  contaminated  by  other movement-adjacent
activities  (e.g., PI, N2, N3, P2, and  artifacts). This
review  will describe procedures regarded to be essen-
tial In  distinguishing Nl from these other activities.
Because procedures that are helpful in isolating  Nl
from other components include methods for Isolating
these confounding activities, this review  should also
be useful to  Investigators interested in MRP compo-
nents other than Nl. The  methodological considera-
tions  discussed here also  apply  to  CNV research
because voluntary movements are Involved in most
CNV paradigms, Evidence suggests, for instance, that
the CNV is  enhanced by requiring an overt motor
response  (Donchin  et  al.  1973,  Irwin et al. 1966).
Late stimulus (Sl)-llnked SPs are  also confounded by
superposition with response-linked  negative  SPs pre-
ceding movement (Rohrbaugh et al.  1976, Rohrbaugh
et al., in press).

   What factors,  then, need  to be considered to iso-
late Nl, functionally, from confounding components
such as the CNV and other MRPs?

Technical Factors

   The history of MRP research  began with failures
by several investigators to  observe any potentials pre-
ceding self-paced, voluntary movements (Bates 1951,
Caspers et al. 1963, Low  et al.  1966, Donchin and
Undsley  1967).  The first two failures are  attribut-
able, in part, to the fact that an  averaging computer
was not used. Since  these  potentials average about 5
/uV,  while background EEC often varies  between
10-50 MV (S.D. = 10 nV), a clear and reliable visual
resolution (Signal-to-Notse Ratio = 4) requires at least
64 averaged trials.
                        FS.N.R. x S.Dl *
                        [  N1 ampi. ]
        vi   u    rtJ  i
        Number of trials
To obtain quantitative data, such as onset latencies
and topographic voltage gradients, some investigators
recommend using several hundred (Vaughan  et al.
1968), or even a thousand trials (Deecke et al. 1976).
Notice in Pigs. 6 and 8 of Deecke et al. (1976), for
example, that 0.5-1.0 pV of EEC noise still remains
after averaging 630-1082 trials. The inescapable con-
clusion is that, in some subjects, a 10*tV noise esti-
mate Is far too conservative. Therefore, more than 64
trials need to be averaged  to resolve, properly, many
of the MRP componenti that may superpose upon Nl
and  contaminate  its measurement  and functional
Interpretation.

  Inadequacies in averaging, however,  cannot entire-
ly explain these early  failures, since  premovement

-------
98
                              P2
                                    512 msec
                  EMG ONSET
Fig.  1. Top trace: schematic of MRP waveform indi-
cating temporal relationship of each component to
EMG onset (vertical line), using abrupt hyperexten-
sion of the dominant index finger. Traces 2-4: actual
MRPs of a single subject  simultaneously recorded in
three amplifiers with different time  constants. Each
trace is an average of 64 movements (dominant index
finger hyperextension). Recording site located over
presumed motor cortex  (0% site},  contralateral to
movement, as shown in Fig. 3.
potentials such as Nl  or  PI  under some conditions
can  be visualized in single trials  (Gerbrandt  et  al.
1973, cf. Fig. 3; Gerbrandt 1977, pi82). Failures to
observe  Nl  have also been  reported when  even
smaller signals were  apparent in  averages (Low et al.
1966,  Donchin and  Lindsley 1967). The problem in
the latter study may have been the time constant of
the recording system. As Deecke et al. (1976) noted,
if it  is  assumed that  Nl onset averages about 0.8 sec
before the start of EMG activity, Nl amplitudes will
                                       Gerbrandt
 be reduced by about one third when a time constant
 of 1.2 sec is used:
                                                        True amplitude = Recorded amplitude x

                                                                           [Nl duration of 0.8 sec
                                                                             2 • TC of 1.2 sec
                                                        Fig. 1 (traces 2-4) compares Nl activities as a func-
                                                        tion of recording dc versus TCs of 0.80 and 0,45 sec.
                                                        At least 22% and 67% of the Nl  signal are lost, re-
                                                        spectively, by using these  shorter time constants. In
                                                        this case, Nl  onset (in dc recordings) occurred  550
                                                        msec before EMG onset so that the attenuation of Nl
                                                        by shorter TCs was not as severe as it would  have
                                                        been with Nls of more typical durations. Neverthe-
                                                        less, with a TC of 0.45 sec, this short-duration Nl  was
                                                        reduced  near the EEC noise level.  Since the ratio of
                                                        Nl to other components of shorter duration becomes
                                                        smaller,  it is more  heavily contaminated by other
                                                        components when short TCs are used.

                                                          Another problem  that  could develop when  RC
                                                        coupling is used is  capacitative unloading of Nl be-
                                                        fore or during movement onset. Spuriously, "Pis" or
                                                        "P2s"  would then tend to correlate with Nl in ampli-
                                                        tudes,  topography,  and latencies of onset. Variables
                                                        that affect the slope of Nl, such as duration of re-
                                                        sponse  (Deecke  et al. 1976), may  differentially
                                                        change   the  temporal  locus  of  this  capacitative
                                                        unloading.
 Instructional Factors

   In the study by Low et al. (1966), neither a lack
 of averaging nor a  problem with time  constants ac-
 counts for the  failure to  record Nl  in  advance  of
 self-paced button-pressing. What  appears to bean Nl
 potential  was, however,  recorded in  an  average  of
 only 12 trials when subjects pressed a button to avoid
 a loud buzzer (Sidman avoidance tasks). Although
 this  negative  wave  may  be  conceptualized as an
 "expectancy"  of the buzzer, subjects were so success-
 ful in avoiding this stimulus that they should rarely
 have expected  to hear it.


   Nl-like activities have  also been recorded in nu-
 merous studies where button-pressing was not explic-
 itly reinforced (e.g., McAdam and Seals 1969,  Rohr-
 baugh  et  al. 1976, Otto et  al. 1977). What is more
 likely,  then, is that the Sidman avoidance schedule
 highly  involved the subjects in button-pressing, that
 subsequent button-pressing alone constituted an
extinction of this involvement, and that involvement
in the  action is essential  for the appearance of Nl.
Without strong intentional participation (Kornhuber
and Deecke 1965) or incentive-motivation for correct
responding (McAdam and Scales 1969), Nl amplitude
may decrease by more than half.

-------
Criteria For Nl Validation
                                               99
   In experiments where  Nl activities were reliably
recorded (e.g., Vaughan et al.  1968,  Deecke et al.
1969, Gerbrandt et al. 1973), even though reinforce-
ments or instructional sets were not separately manip-
ulated  as experimental variables, subjects were explic-
itly  instructed what  to move,  when to  move, how
rapidly, and how long to hold the movement. Base-
line  EMG  activities were  continuously  sampled to
ensure  that subjects followed instructions. This type
of instructional set may demand the involvement of
subjects and thus enhance Nl because the response
form is specified in advance by  the voluntary plan of
action, rather  than  by triggering nonspecific motiva-
tional  processes (arousal). Nl   onset is  earlier  and
hemispheric asymmetry  possibly greater when  a
longer  duration response  is required  (Deecke  and
Kornhuber 1977).  Papakostopoulos  and  Cooper
(1976) have also shown that, during  the preparatory
interval before a response,  the background reflex field
is specifically repaItemed  to assist the impending vol-
untary  movement. This reduction of stretch reflexes,
which  would   otherwise  antagonize  the  intended
action, may play an  important role in smoothing and
accelerating action sequences.

   It appears  then, that Nl  reflects  a  process func-
tionally significant in readying the central and periph-
eral  nervous systems according to the specific actions
that are intended. Experimenters must  ensure, there-
fore, that, during the period of preparation for re-
sponding,  subjects  attend only to  the  exact para-
meters of  response; e.g.,  type  and  extent of move-
ment,  force, speed, duration and interval of primary
response, and  specific sets to eliminate adventitious
responses. When a movement in MRP or CNV research
is  described simply  as  "button-pressing," too little
attention is probably being paid to motor control by
either the subject or experimenters to know whether
or not Nl  functions are being  reliably engaged  and
sampled.

Spatiotemporal and Triggering Factors

   Few MRP  investigators  have provided evidence
that appropriate steps have been taken to minimize
the Spatiotemporal  superposition of Nl  and  artifact-
ual or  other MRP components. A crucial consideration
is whether the  Nl distribution is distinct from the dis-
tributions of other  components known to encroach
on the  spatial  and temporal domain of Nl. Since Nl
seems  to  reflect  an  action-smoothing function via
specific readiness  processes  antecedent  to  specific
actions, Nl asymmetries  are differentially  focused
over the pyramidal motor  system contralateral to the
moving hand in right-handed subjects (Vaughan et al.
1968,  Deecke et al. 1969, Gerbrandt et al.  1973,
Deecke et al. 1976).

   This topographical feature is illustrated in Fig. 2,
where  the  asymmetry of Nl is 35% less over  pre-
+ 75 percent
                                          1.25/LtV
                   EMG ONSET
                                 512 msec
Fig.  2. Bipolar derivations (contralateral minus  ipsi-
lateral to dominant index finger hyperextension) are
shown representing five paired recording sites.  The
waveforms at each recording location are grand aver-
ages (N = 352) over five subjects. The earliest average
EMG onset is indicated by a vertical line. Recording
sites shown in Fig. 3.


sumed sornatosensory cortex  (-25% site) relative  to
motor cortex (0% site). Electrode locations are de-
fined in  Fig. 3. The  attenuation of Nl asymmetries
away from motor cortex occurs so rapidly  that, in
monopolar comparisons of MRPs  contralateral and
ipsilateral  to movement, significant asymmetries are
observed  only over this pre-Rolandic site (Gerbrandt
et al.  1973, Deecke  et al.  1976).  It is quite likely,
then, that the  C3 and  C4 sites used by many MRP

-------
100
               S.ORB
S.ORB
1 em
                                              1 cm
 Fig.  3.   The scalp-recording sites used for the data
 presented in Fig. 1, 2 and 4 are shown in relationship
 to the 10-20 system.

 and CNV investigators are too far posterior for maxi-
 mal sensitivity to Nl asymmetries since these sites are
 closer to the Rolandlc line on the declining gradient
 of Nl. At the same time, these more posterior loca-
 tions maximize the chance  of mixing attenuated Nl
 activities with activities of different functional origin,
 as  discussed below. Depending upon the method used
 to locate the Rolandlc  line (Vaughan  et  al. 1968,
 Deecke et al.  1969, Gerbrandt et al.  1973), the pre-
 Rolandic electrode should be situated 1-2 cm anterior
 to  the C3/C4 position.

    Even  when electrodes are  placed optimally  for
 sampling Nl  asymmetries,  it may  be difficult  to re-
 cognize them. Although the recording ipsilateral  to
 movement was about 60% smaller than contralateral-
 ly in the study by Vaughan et al. (1968) and 30%
 smaller In the study by Gerbrandt et al. (1973), it was
 only about  10% smaller (and not statistically signifi-
 cant) In the  work of Deecke et al. (1969) and 20%
 smaller in  subsequent work  (Deecke et al. 1976).
 Indeed, very large  asymmetries may  Indicate trie
 superposition of asymmetrical MRP activities with an
 origin  other  than Nl. Deecke  et al. (1976)  showed
 that the Ipsilateral  Nl  was 55% smaller than contra-
 lateral only  when  measured  from baseline to the
 point  of  EMC onset.  They note that  the added
 asymmetry is largely due to contamination of the Nl
 measure  by  N2 activities  that occur 20-100  msec
 before EMG  onset  and are more  focused than Nl
 over the motor cortex contralateral to movement.
 They note further that, If one triggers MRP averages
 from mechanograms, photocells, or even skin-surface
                                     Gerbrandt

EMG electrodes (rather than the first intramuscular
EMG activity of the principal effector), N2 activity is
more likely  to be confounded with the Nl measure-
ment because (1) the earliest EMG activity is not de-
tected and (2) there is greater temporal jitter between
N2  and  movement-estimated  onsets  compared  to
EMG onsets. If total  EMG quiescence is not achieved
prior to abrupt  movement (documented by multiple-
EMG recordings), and if the most sensitive EMG  re-
placement is not used, then N2 activities could begin
as soon as the subject begins to prepare (i.e., as early
as 1.5 sec before the actual movement).

   This criticism by Deecke et al. is so lethal that the
validity  of most MRP and CNV research  is suspect.
Perhaps no research group has succeeded in measuring
Nl free from N2 contamination since it is unlikely
that the same motor  units sampled from a given elec-
trode are  always the first motor units to fire. Pilot
work by the Kornhuber group shows, however, that
the  magnitude  of this functional rotation  among
motor units and  muscle groups  is  probably  not
greater than 100 msec  (cf. Fig. 2B of Deecke et al.
1976), at least  in  the  Ulm experiments  where the
movement Is simple and  abrupt and subjects are all
highly practiced  In  silencing  intramuscular motor
activity across several muscle groups.

   Given that the  average N2  wave commences  54
msec before the  average EMG onset, It may  be
assumed that N2 would not be appreciably confound-
ed  with  Nl  more  than 154 msec (100 + 54 msec)
before the average EMG  onset. No other group has
studied the  topography of Nl while  systematically
triggering MRPs  from  the onset  of  intramuscular
EMG activity, and no other group has estimated the
onset latencies of N2 compared to EMG in a topogra-
phic study. Since only the Ulm group has determined
how early N2 can occur prior to  an intramuscular
trigger, and  since  the extent of functional rotation
has not been estimated, the extent  of N2 contamina-
tion elsewhere Is as yet unknown,


   Vaughan  et  al.  (1968) triggered averages  from a
skin-surface  EMG location, a procedure which raises
doubts about the sensitivity of detecting the eurUe&t
EMG onset. However,  they used  a  simple, abrupt
movement (hyperextension of fingers or hand) with a
superficially located  principal  effector (extensor
dlgttorum communis},  Deecke et al,  (1976)  found
that estimates  of  N2  onset latency  and amplitude
obtained from intramuscular EMG, skin-surface EMG,
and mechanogram  triggering were sufficiently similar
with a simple, abrupt flexion of the  Index finger to
conclude  that Nl  and N2 were  not  confounded in
this case. Using a  simple, abrupt hyperextension of
the Index finger, Gerbrandt (1977) also reported that
intramuscular triggering from the extensor dtgltorum
communts did not change the onset latencies of the

-------
Criteria For N1 Validation

 N3 component from earlier estimates obtained with a
 photocell  trigger  and skin-surface  EMG electrode
 (Gerbrandt et al. 1973). If it can be assumed that the
 functional rotation among motor units is similar fora
 simple, abrupt  hyperextension and flexion of  the
 index finger, and that EMG silence and sensitivity to
 earliest onset  were achieved,  then the rotation-plus-
 N2 onset safety margin  of 154 msec estimated by
 Deecke et al. (1976) can be used for measuring Nl in
 the studies ofVaughan et al. (1968) and Gerbrandt et
 al.  (1973).  Vaughan et  al. did  not terminate  Nl
 measurement systematically at 154 msec or more be-
 fore EMG onset. Contamination of Nl by N2,  there-
 fore, probably accounts for the large Nl asymmetries
 reported  (62%  smaller ipsilaterally). Nl  measure-
 ments  in the Gerbrandt et ah work are probably  not
 contaminated  by N2 since measurements  were begun
 150 msec before  EMG onset  and the earliest EMG
 onset was determined for each subject. Although we
 did not report onset latencies for the premovement
 N2, reanalysis of  the data using  bipolar  derivations
 clearly shows  an N2 component arising 46 msec be-
 fore EMG onset (cf. Fig. 2,  0% site),  whereas  Nl
 onset occurs at  an average of 794 msec before EMG
 onset.

 Task Factors

   Most other Nl experiments  (and probably all CNV
 experiments involving a terminal motor response) are
 uninterpretable  with regard to Nl  contamination by
 EMG-adjacent activities. Complex flexion responses
 such as balloon-squeezing, telegraph-keying, button-
 pressing,  or  squeezing a  hand  dynamometer have
 typically been used. Even though some investigators
 attempted to sample  EMG activities associated with
 the movement,  an amazing number of principal ef-
 fectors  are  involved in complex responses. Kinesio-
 logy tests offer little suggestion as to which effectors
 might  have  the  earliest onset.  Often, principal  effec-
 tors are deep  muscles that are inadequately sampled
 by  skin-surface  EMG electrodes.  Arm  and  body
 positions  required to maintain  contact with  the
 apparatus - "ready" to respond — make it difficult
 to achieve EMG quiescence. Some devices require the
 application of large  forces (1-20 kg) that probably
 involve functional rotation across an extensive muscle
 field.

   Subjects  must  also expect and  encounter  extra-
 kinesic and tactile resistance  against the apparatus,
 factors that  undoubtedly introduce potentials that
 would not  occur in simple   unimpeded  movement
 situations. The use of oscilloscopes or other on-line
 feedback signals may similarly  introduce confounding
 nonmotoric expectancies. These problems were com-
 pounded in  many studies by  the use of short inter-
 movement intervals  (34  sec). It is difficult for both
 subjects and experimenters to ensure EMG quiescence
                                            101

and stability in such a short interval after movement.
These kinesiological questions require extensive docu-
mentation to  show that the  earliest  EMG activities
have been sampled and that EMG quiescence has been
achieved. Since N2 may be the scalp-recorded integral
of specific commands or initiations of every muscle
group cortically activated during movements,  these
methodological errors may introduce N2 confounding
more than  150 msec prior to the arbitrarily deter-
mined EMG onset. The  150-msec estimate obtained
in  early work with   simple,  abrupt  movements
(Deecke et al.  1969, Gerbrandt et al. 1973) probably
does not apply to these more complex movements.

Adventitious Movement Factors

   Even when  the  methodologies seem adequate for
attaining good  estimates  of EMG quiescence and earli-
est  EMG onset for the responding limb, adventitious
or occult events (not primarily involved in self-paced
or  planned  movements) may  occur  that  produce
potentials that summate with  Nl.  It is essential,
therefore, to place  electrodes  over all  recording loca-
tions where possible contaminating (superimposed)
sources of activity  appear differentially compared to
Nl.  Otherwise, one cannot verify that the observed
component is  behaving  as  a  function of Nl  rather
than extraneous factors. Routine placement of elec-
trodes  for  detection of vertical eye  movements
(EOG),  for  example, is  essential for  ruling out this
source of artifact (Hillyard and Galambos 1970). Re-
quiring eye fixation on an external reference, denying
subjects view of their own movements, and pretrain-
ing of subjects by  having them  watch their own eye
movements  reduce, but do  not entirely eliminate,
artifactual eye movements (Papakostopoulos et al.
1973). Routine monitoring of eye movements is es-
sential (Wasman et al. 1970).

   Although horizontal eye movements do not contri-
bute artifactwlly to  scalp recordings when a linked-
earlobe  reference is  used, potentials  preparatory to
eye movements may appear over Rolandic and frontal
regions even when  averaged EOG recordings may not
indicate  artifactual  contamination (Becker  et  al.
1972, Syndulko and Lindsley 1977, Rosen et al. this
volume). These potentials may be functionally related
to Nl in the sense of involving readiness processes,
but  not  specifically associated with voluntary hand
movement and may thus confound measurements of
the symmetry  and topography of components related
to hand movement. Both horizontal and vertical eye
movements  should, therefore,  be  monitored  at all
times.  Any trials  showing  orbital-related activities
should be deleted before averaging.

   Investigators should  also  demonstrate that other
extracranial sources of  Nl  contamination (e.g., glos-
sopharyngeal and neck EMG artifact) have been elimi-
nated. The neutrality  of the  reference electrode

-------
 102
                                                                 Gerbrandt
         Table 1.  Electrode Sites and Waveforms Needed to Distinguish N1 Spatiotemporally
                                        from other Components*
    Active electrode sites
No.
Waveform derivations
                                                                       Wo.
Purposes of derivations
 External canthus
Supra-orbital(C&l)
+25%(C&I) or F(3&4)
0%(C&I) or modified C(3&4)
-25%(C&l&Z)orP(3&4)
  1    EC-LE
 2   SO(C&I)-LE, SO(C)-SO(I)
     +25%(C&I)-LE,
             or               or
               F(3&4)-LE,     F3-F4
     0%(C&I)-LE,
     or             or
     MC(3&4)-LE, MC3-MC4
     -25%(C&I&Z)-LE, -
     or
      P(3&4&2)-LE,      P(C)-
                                                                 P4
                               1   Monitor  horizontal  eye  move-
                                  ments, differentiate  eye  move-
                                  ments  from "asymmetrical  and
                                  reversed N1"

                               3   Monitor vertical eye movements.
                                  confirm differential presence of
                                  "reversed  and  symmetrical IM1,"
                                  confirm narrowness of N1  distri-
                                  bution

                               3   Confirm  presence  of N3, con-
                                  firm narrowness  of N1 distribu-
                                  tion and difference from N3 dis-
                                  tribution,  confirm  narrowness
                                  and difference in N2  vs. N1  and
                                  N3 distributions

                              3   Confirm focus of  asymmetrical
                                  N1, confirm focus of  N2
                                  Confirm  focus  of  symmetrical
                                  PI, confirm  posterior  distribu-
                                  tion of symmetrical INI1, confirm
                                  posterior distribution of P2, con-
                                  firm "reversed"  N3 postcentrally
•Abbreviations:  C = Contralateral; EC ;= External Canthus; F = Frontal; I = Ipsilateral; LE = Linked Ear Reference; MC = Motor
 Cortex; P - Parietal; SO - Supra-orbital; Z = Midline. Electrode sites are illustrated in Fig. 3.
 should also he demonstrated with a noncephalic indif-
 ferent electrode.

 Electrode Derivation Factors

    When all these procedures are employed, the proof
 of their effectiveness still rests in the demonstration
 that  Nl  is  spatiotemporally  focused  appropriately
 and  is  distinguishable from  other MRP components
 (PI,  N2, N3,  and P2) capable of occupying the same
 domain.  A list  ol  electrode derivations considered
 necessary  to  analyze  Nl  is shown in Table  1. Note
 (hat  bipolar derivations (contralateral - ipsilateral) are
 also  essential  in  most subjects to isolate N2 from
 superposition  with symmetrical PI activities (Deecke
 et a!. 1060, compare the 0% sites in Fig. 2 vs. Fig. 4).
 Jjipobr derivations are also helpful In distinguishing
 Nl from N2 activities because  N2  is more symmetri-
 cal and differs temporally in slope of amplitude (Fig.
 2). Notice, in  addition, that Nl has a different distri-
 bution  with bipolar vs. monopolar derivations. With
 monopolar  recordings (Fig.  4), Nl   is largest post-
                          centrally, whereas in bipolar derivations (Fig. 2) Nl is
                          35% larger (at  150 msec before EMG onset) over the
                          motor cortex.

                             An even greater contrast between monopolar and
                          bipolar distributions is obtained when the postcentral
                          electrode is  moved posteriorly  over the P3/P4 sites
                          (Deecke et al.  1076), where Nl  shows no asymmetry
                          at all. The bipolar distribution  seems  to correspond
                          more closely to specific readiness processes preceding
                          specific actions (i.e.,  it is asymmetrical and focused
                          over the  motor cortex). The  fact that motivational
                          variables may differentially  affect symmetrical com-
                          pared  to  asymmetrical  Nl  activities (MeAdam  and
                          Scales  1069, Kutas and Donchin 1077) suggests that
                          these are functionally  different types of readiness
                          that should be separately measured.

                            Finally  investigators  should  score  records blind
                          with regard to electrode site,  condition, and subject
                          to avoid experimenter bias about the existence of a

-------
Criteria For N1 Validation
VLr'A/V-vSA^'
+ 75 percent

'
                                   512 msec
  Fig. 4.  Monopolarly-recorded MRPs (linked ear refer-
  ence)  for five  pairs of electrodes  (contralateral to
  dominant index finger movement = solid line; ipsi-
  lateral  = broken line).  Waveforms are grand averages
  (N =  352) over five subjects. Earliest average EMC
  onset   indicated  by vertical  line.  Recording  sites
  located as shown in Fig. 3.
                                             103

component.  For instance,  random fluctuations in
EEC  voltage may inadvertently be identified as Nl
onset, or "PI" or "N2" bumps, even when such com-
ponents may  not be present (cf. possible "N2" at
-25%  site, Fig. 2).  Peak-seeking  algorithms  which
automatically select a maximum or minimum voltage
point within a specified time window, for instance,
will always provide an output value, regardless of the
quality or nature of  the input. Since all MRP analysis
techniques are  more or less susceptible to biases or
limits of  this  type,  investigators should explicitly
state  the  set  of measurement rules  used (cf. Ger-
brandt et al. 1973, Deecke et al. 1976).

Conclusions

   The foregoing methodological  considerations are
summarized in  Table 2 which is organized as a check-
list for use  in  planning and evaluating MRP  studies.
The specific procedures listed in Table 2 can be re-
duced to three  basic criteria for optimizing Nl resolu-
tion:  (1)  demonstrate clear  visual resolution of Nl
(signal-to-noise  ratio  =  4);  (2)  demonstrate  clear
spatiotemporal resolution of N1; and (3) demonstrate
clear functional resolution of Nl.

   If  previous  MRP  and CNV research is evaluated
rigorously in accordance with this checklist, it may be
concluded that Nl has rarely been recorded with suf-
ficient methodological  controls  to  ensure  uncon-
founded  measurement and interpretation. That is, in
most  studies the sustained negative potential associ-
ated with preparation for voluntary movement can-
not be distinguished from other potentials of differ-
ent functional  origin.  Indeed, so  much data are
brought into question that one must ask whether the
criteria are too stringent and why so few investigators
have applied them.

   An analogy from the field of neuropharmacology
 may be helpful in answering these questions. A neuro-
 transmitter is a chemical substance by which informa-
 tion  is transferred  from neuron to neuron.  Three
 criteria must be met to conclude with certainty that a
 particular substance functions as  a neurotransmitter
 (Julien  1978). These criteria are: (1) the compound
 must be contained in, and released from, specific pre-
 synaptic nerve endings; (2) control over the rate of
 release of the  putative  compound must mimic post-
 synaptic actions (excitation, inhibition) that follow
 from presynaptic nerve  stimulation; and (3) there
 must be  a  mechanism  of inactivation of the com-
 pound that shapes the  time course of the "natural"
 transmitter action.

   When  students first hear how difficult these  cri-
 teria  are  to satisfy  in  the  central compared to  the
 peripheral nervous system, they question  whether it  is
 "fair" to apply  peripherally-devised  criteria  to  the
 CNS. In the PNS or CNS, the purpose of the criteria

-------
 104
                                     Gerbrandt
             Table 2.  Methodological Checklist of Procedures to Optimize Spatiotemporal
                                    and Functional Resolution of IM1


 1.   EEC Electrode and Recording Factors
     (a) Use appropriate electrode montages, including unipolar and bipolar derivations as shown in Table 1 and Fig.
        3, to maximize Spatiotemporal resolution.
     (b) Use long time constants (> 0.8 sec).
     (c) Determine that reference electrode is neutral by use of noncephalic indifferent technique.

2.   EMG  Electrode and Triggering Factors
     (a) Use appropriate recording methods, preferably needle electrodes  inserted in principal effector muscles, to
        trigger averages from the earliest motor unit discharge.
     (b) Determine degree of functional rotation among relevant motor units.
     (c) Achieve stable periods of EMG silence  (> 4 sec) between successive movements.

3.   Task and Instructional Factors
     (a) Use a simple, abrupt, unimpeded movement that requires minimal  support.
     (b) Use long intermovement intervals (> 10 sec).
     (c) Use right-handed and left-handed responses (in right-handed subjects) to show lateralized Nl focus.
     (d) Carefully instruct subjects to focus attention on  relevant parameters of the required response and to inhibit
        all other movement. Monitor EMG to ensure compliance.

4.  Signal A veraging Factors
     (a) Average 64-1082 trials as necessary to achieve an EEG signal-to-noise ratio = 4
     (b) Reject high-noise trials and high-noise subjects.
     (c) Demonstrate that  waveforms  are not confounded by art if actual or adventitious potentials that  are not
        primarily involved in the self-paced movement (e.g., vertical or horizontal eye movements, glossopharyngeal
        or neck EMG activity, nonmotoric expectancies).

5.  Measurement Factors
    (a) Explicitly state the set of measurement rules used in data analysis.
    (b) Measure  each  MRP component at the  electrode location and with the particular derivation that maximally
        differentiates the component from other components (cf. Table 1 for N1).
    (c) Determine the onset latency of N2. Measure  N1  from  the N2-free terminal  point (> longest  rota-
        tional-plus-N2 onset point).
    (d) Score records "blind" to electrode site, condition, and subject to avoid  experimenter bias.
 is to ensure  that the compound in question, rather
 than another, is the  one that is  involved in trans-
 neuronal information transfer, rather than some other
 function.  Application of these criteria in the CNS is
 very difficult, but positive identification of a neuro-
 transmitter can  only be claimed when these  criteria
 have been met.

    This checklist, like the neurotransmitter criteria, is
 meant to  ensure that a particular functional compo-
nent (Nl) has been demonstrated in isolation. Short-
cuts in applying Nl criteria simply may lead to uncer-
tainty about what is being demonstrated. Hopefully,
this review provides a systematic framework for eval-
uating past Nl research and will encourage a broader
application  and  further  refinement  of criteria for
demonstrating Nl and other component functions in
subsequent MRP studies.

-------
EFFECTS  OF MOVEMENT ON SENSORY INPUT


P. HAZEMANN
Laboratory of Occupational Physiology, Pitie - Salpetriere Hospital, Paris, France
    Sensory input is generally diminished when move-
ment coincides with sensory stimulation if the two
events are not related. This effect can be demonstrated
by the study of evoked potentials in animal and man.
Evidence related to the effects of movement on audi-
tory and somatosensory evoked potentials is reviewed
here.  Visual evoked potentials are not considered
because visual inputs are modified by ocular move-
ments and pupillary contraction.
Animal Studies

    Starr (1964) studied the influence of motor activ-
ity on click-evoked responses in the auditory path-
way of the awake cat. Responses recorded from the
round window, cochlear nucleus, inferior colliculus,
medial geniculate body, and auditory cortex were of
smaller amplitude when animals were  moving. The
extent of attenuation was roughly  related  to the
extent of movement. When tendons of the middle ear
muscles were sectioned bilaterally, responses recorded
from  subcortlcal stations  did not  decrease  during
movement, whereas cortical responses were still atten-
uated. Amplitude of cortical responses to an electric
shock,  applied through bipolar electrodes located
along subcortical stations of the auditory pathways,
decreased during motor activity if the stimulation was
subthalamlc, while it  was unmodified by   medial
geniculate  stimulation.  This finding  suggests  that
mechanisms modifying cortical responses do not influ-
ence the cortex directly, but act at intermediate sites.

    Chez  and  Lenzl (1971) studied changes in the
transmission of somatosensory volleys to the medial
lemniscus during a conditioned voluntary movement.
Cats were trained to lift the right paw, press a lever at
the onset of a tone, and then replace the paw on the
ground. The right superficial radial nerve was stimu-
lated, and responses were recorded in the medial lem-
niscui. Responses to stimuli were significantly smaller
during active  movement,  with depression occurring
between 100 and 200 msec prior to the lifting move-
ment.  Responses returned to control levels  within
100 to 200 msec after the cat replaced its paw on the
ground. During  the  period of postural  fixation,
responses did not differ from control values.

    Ghez and Pisa (1972) attempted to identify the
parameters  of movement related  to  the change in
lemniscal transmission. They found a negative linear
correlation  between  the  amplitude  of lemniscal
response and the logarithm of the velocity of the bar,
determined  at the instant the stimulus was delivered.
Neither force exerted nor passive  movement of the
upper extremity  produced noticeable changes. Ghez
and Pisa attributed the attenuation of lemniscal poten-
tials to central influences impinging on the cuneate
nucleus.

    Coulter (1974) showed that responses evoked in
the medial lemniscus by stimulation of the contralater-
al forelimb in the awake cat were reduced in amplitude
during gross motor activity. During discrete  move-
ments, depression of the lemniscal potential could be
clearly related to the occurrence  and duration of
bursts of muscle activity. This depression was observed
about 100 msec prior to EMG activity. Neither move-
ment of the forelimb opposite to stimulation nor pas-
sive movements resulted in depression of the lemniscal
potential.
Human studies
    Broughton et al. (1964) noted a decrease in soma>
tosensory evoked potentials (SEPs) during fist clench-
ing, Giblin (1964) also  observed a decrease when
fingers were  moved  either  actively  or passively.
Coquery (1971) and Coquery et al. (1972) showed
that SEPs elicited by electrical stimulation of a finger
increased during the 200 miec before  the beginning
of EMG activity and decreased  during contraction
when  flexion was performed by the stimulated hand.
SEPs increased when flexion  was performed  by the
nonstimulated hand, but decreased during active and
passive plantar flexion of both feet. Coquery conclud-
ed that both facilltatory and inhibitory influences act
on the somatosensory pathway and that at least part
of the inhibitory influence has peripheral origin.

-------
106
                                     Hazemann
    Lee and White (1974) studied the effect of repeti-
tive flexion and extension  of the  fingers on SEPs
elicited by stimulation  of the fingers.  The most
obvious change during movements of the stimulated
hand was an increase in amplitude and a slight increase
in latency of a late negative component. The enhance-
ment of evoked potentials was specific to movements
occurring in  the immediate vicinity of stimulation,
but was not limited to the contralateral area. Enhance-
ment was maximal at the vertex and almost as promi-
nent  over  the  ipsilateral as the  contralateral hemis-
phere. Lee and White suggest that the change in evoked
potentials  reflects  an  interaction  between  central
efferent  and afferent  systems,  a cumulative effect
that probably occurs at several levels.
    Papakostopoulos et al. (1975) compared cortical
potentials following passive or externally paced dis-
placement (EPD) to those related to similar but self-
paced  voluntary  displacement  (SPD) of the  index
finger. Secondly, evoked responses to brief electrical
stimuli applied to the medial nerve at the wrist during
SPD were compared with similar responses elicited in
subjects at rest. After  EPD, a clear evoked response
could be seen from p re frontal, precentral, and post-
central areas, whereas, after SPD, only a diminished
response could  be seen in the precentral area. In the
same way, electrical stimulation of the median nerve
evoked clear responses in precentral and postcentral
cortex when the limb was at rest. When stimuli were
timed to occur  during self-paced displacement of the
index finger, responses were diminished in the post-
central area, but remained relatively unaffected in the
precentral area. The evoked  potentials  returned  to
resting values 700 msec after the movement onset,
even if the  displacement was sustained. Papakostop-
oulos et al. interpreted  the results as  a movement-
related gating action that selectively affects sensory
input to somatosensory and motor cortex, the latter
being less inhibited.

    Hazemann et al. (1975) investigated the temporal
relationship  between self-paced movement and pre-
sentation of  test stimuli. Evoked potentials were
averaged in 10 successive epochs, extending from 880
msec  before  to  2.5 sec after movement.  Auditory
evoked potentials were attenuated in all epochs, with
the greatest  decrease  appearing in the  epoch just
following movement.  Somatosensory evoked poten-
tials were similarly attenuated when movements were
performed by the hand contralateral to stimulation.
In the ipsilateral case, SEP amplitude was attenuated
only when stimulation was administered close to the
active muscle. It appears from these results that move-
ment  induces a  generalized  central modulation  of
sensory  evoked potentials that  reflect the degree  of
occupation  of  a  single  limited-capacity  channel
(Broadbent 1971). The same channel appears  to  be
used selectively for the transmission of either afferent
input or efferent output signals.

Summary

    The  amplitude of somatosensory evoked poten-
tials has almost always been  found to diminish when
the stimulated part of the limb was moved. Generally,
the decrease  commenced  prior to movement onset.
The effects of contralateral and passive movements
are not  as clear. When studied, the auditory evoked
potentials have also been found to diminish during
movement in both cat and man. Explanations vary,
but most authors suggest  that mechanisms of inter-
action between voluntary motor activity and sensory
input may operate at different levels of  the central
nervous system.

-------
INFLUENCE  OF FORCE,  SPEED  AND
DURATION  OF ISOMETRIC CONTRACTION
UPON SLOW  CORTICAL POTENTIALS IN MAN

P. HAZEMANN, S. METRAL, AND F. LILLE
Laboratory of Occupational Physiology, Pitie-Salpetriere Hospital, Paris, France
    The first studies  on  the human cortical motor
potential described the type of movement performed,
but did not take into account the fact that movements
may vary during the  experiment.  Since that time,
studies have been completed that consider the physi-
cal  parameters of movement, although these investi-
gations generally examined only one parameter: force
(Ford et al. 1972, Wilke and Lansing 1973, Kutas and
Donchin 1977) or speed (Becker et al. 1976). Their
results  are difficult to compare since, for example,
the movements are performed by segments of differ-
ent limbs.

    The present investigation  seeks to examine the
systematic influence of force, speed of force variation,
and duration  of maintained force on  the cortical
motor potentials associated with isometric contrac-
tions of the index finger while a subject regulates his
performance through visual control of the mechano-
gram.

Methods

    Thirteen subjects (8 male, 5 female, 2040 years
old) participated in this study. Subjects were comfort-
ably seated in an armchair with forearm support and
index finger immobilized at the last phalango-phalan-
gian joint by a large  plastic ring, rigidly fixed to a
strain gauge. The subject was told to make an index
dorsiflexion against this gauge. The apparatus allowed
contraction to occur under practically isometric con-
ditions.

    In the three situations defined below, the subject
controlled the mechanograms of his contractions on
an oscilloscope screen 1 meter in front of him. Con-
tractions, timed by the appearance of the oscilloscope
spot, were performed  every 4  or 6 sec by the domi-
nant hand and were repeated 300 times for each con-
dition. In the "force" situation, contractions perform-
ed with a great force (GF) (about 500 g) were compar-
ed to contractions of small force (SF) (about 200 g),
achieved in  the same  time. In the "speed" situation
(speed of force variation dF/dt), high-speed contrac-
tions (HS),  done with 500-g  force and achieved in
about 160 msec, were compared to low-speed contrac-
tions (LS) in which the same force was achieved in
660 msec. For  the  "time maintained"  situation,
contractions involving 200-g force, achieved in 160
msec, were sustained for 500 msec (BS) or 1000 msec
(CS).

    A separate  recording session was held for each
situation. Ten subjects participated in the force and
speed situations, and six in the sustained contraction
situation. EEC was recorded from  the rolandic area
(C3 or C4) contralateral to movement, with a scalp
electrode referenced to linked ears (Al, A2). The time
constant was 0.7 sec. EMG was recorded by means of
surface electrodes over the extensor digitorum com-
munis muscle.  EEC, EMG and mechanograms were
recorded simultaneously on paper and magnetic tape.
The averaged motor  potentials of 300  successive
movements were obtained from a CAT 400 with an
analysis  time of 4 sec. The averager was triggered
from the mechanogram, with a Schmitt trigger used
to obtain a stable threshold. The lag between reading
and recording heads  of the magnetic  recorder  al-
lowed a 1760-msec delay of the EEC. Motor poten-
tials were displayed on an  XY plotter. For each con-
dition, mechanograms were averaged in groups of 100
and 300 to ensure that instructions were observed.

   Two subjects  participated in supplementary
recordings during which ocular and visual phenomena
were studied. Contractions were performed under HS
and LS conditions with and without visual feedback
from the mechanogram on the  oscilloscope in order
to check for ocular tracking movements and visual
afferents. Contractions were averaged, with all trials
preceding,  following,  or simultaneous with  ocular
movement or blinking eliminated, as a test for elec-
trooculogram diffusion. For these experiments, EEC
was recorded from the contralateral Rolandic (C3 or
C4), frontal (Fl or F2), coronal (F3 or F4), and oc-
cipital (Ol or 02) areas, and EOG was recorded with
silver electrodes  from the superciliary arch and the
lower rim of the orbit.

    The cortical  motor potential was analysed in
terms of component latency and amplitude. Latencies
were measured from the beginning of the mechano-
gram to inflexion  of the baseline  for Nl and peak
latency  for all  the other components. Amplitudes

-------
 108
                                 Hazemann et al.
 were measured from the baseline, 200 msec before
 the medianogram for Nl, at peak for N2, P2, andP'2.
 For PI and P'l, amplitudes were measured from the
 preceding peak. Latency and  amplitude  values for
 each condition were compared  for individual subjects
 and for the entire subject  population. Mean values
 and standard  deviations were compared by student's
 test for group data.

 Results

     Analysis of group data showed that force exerted
 (determined by the amplitude  of the mechanogram)
 was 495 ± 147 g for the GF  condition and 222 ± 41 g
 for SF, Speed of force variation  (measured as rise time
 of the mechanogram) was 192 ±46 msec in the HS
 condition, 851 ± 106 msec for LS, 200 ± 47 msec for
 BS, and 276 ± 58 msec for CS. For conditions BS and
 CS, maintenance time (defined  as the duration of the
 plateau) was 506 ± 122 msec  and 956 ±211 msec,
 respectively. Therefore, subjects were able to follow
 Instructions for force and duration maintenance, but
 speed of rise time seemed difficult to control. In every
 situation, speed of force achievement was slower  than
 requested. There was an evident prolongation of rise
 time when movements were  performed at slow speed
 or with sustained movement.

     With visual analysis of motor potentials, positive
 components were characterized that differed  from
 those classically described (Fig. 1). The classical motor
 potential (using the nomenclature of Gilden et al.
 1966) has  a premotor negativity Nl, an inconsistent
 premotor positivity PI, a negativity  N2 culminating
 after movement, followed by a positivity P2. As PI
 precedes an increase in  the slope of negativity (Nl),
 the slope inflection was considered to correspond to
 PI even  when a distinct PI  was missing.  A second
 minor positivity P'l  was noted after movement and
 preceding or following the N2  peak. P2 was usually
 followed by another positivity  P'2 and,  although P2
 was at  times a mere  notch  within P'2,  usually  two
 individual peaks were noted,  either of which might be
 greater. A double-peaked configuration of late positiv-
 ity is also mentioned by Arezzo  and Vaughan (1975).

    Amplitudes  and latencies of components were
 measured for the six experimental conditions (Fig. 2).
 Mean values and standard deviations for the latencies
 are given in Table 1.

    Nl latency ranged from -1600 msec to -320 msec,
 with an average of -980 msec.  Nl was seen in all
 subjects for conditions HS, BS and CS, in 8 out of 10
subjects for condition  GF, and in only 5 subjects for
condition SF. Its amplitude did not vary significantly
from one condition to another; however, its latency
was  significantly  longer  for  sustained  contractions
(<.02) and contractions done with little force (<.02).
                 ONSET OF THE
                  MOVEMENT
       100 muo
 Fig.  1, Averaged cortical motor potential. Continuous
 line: components described by Gilden et al. (1966);
 dotted line: supplementary components described in
 this study.

     Latency of PI  was from  440 msec to -20 msec,
 with an average of -204 msec. PI was seen in only
 half  the subjects for conditions GF, SF, HS, and  LS
 while it  appeared  reliably and significantly earlier
 (<001) for conditions BS and CS.

     P'l latency averaged 85 msec and ranged from 0
 to 180 msec. P'l was absent in LS, but appeared fairly
 consistently  in  all other conditions.  Amplitude and
 latency  were not  systematically  altered by  other
 manipulations.

     N2 latency was from -100 to +320 msec, with an
 average of +100 msec.  N2 was consistently found for
 conditions HS, BS, and CS and was seen nine times for
 condition LS, eight times forGF, and seven times for
 SF. Its average latency was significantly longer (<.05)
 for condition LS compared to HS. N2 amplitude was
 significantly smaller (<.05) for condition SF compar-
ed to GF. For all other conditions, there was no signifi-
 cant difference in latency or amplitude.


    P2 latency ranged from 220 to 500 msec, with an
 average of 367 msec. P2 was constant for conditions
 GF, SF, and HS, but was missing twice for condition
 LS and once for conditions BS and CS. Amplitude for
 the three latter conditions was significantly lower
 (<.001) than for other conditions.


    P'2 was observed in nearly all recordings, but was
 missing in one subject for conditions SF and  LS. P'2
amplitude was significantly higher (<01) for condi-
tion HS compared to LS and for situation HS (<,05)
compared to GF and SF. Latency  was linearly corre-
lated  (0.92) with the duration of  the mechanogram.

    Because of the precocity of Pi in the sustained
situation, the  length of time between peak Pi  and

-------
 Physical Parameters of Contraction
                                             109
Fig.  2. Cortical motor potential (C3-A1A2) recorded in two subjects under six experimental conditions.  From
top to bottom: HS = high speed, LS = low ipeed, GF= great force, SF= small force, BS = briefly sustained, and
CS » considerably sustained.
 peak N2 was significantly longer (<.01) in the sustain-
 ed situation compared to all other situations.

     No significant difference in slow cortical poten-
 tials recorded from the rolandic area (Fig. 3 A), or in
 EEC recordings averaged without ocular artifacts (Pig.
 3B) was apparent whether or not visual control  of
 the  mechanogram was provided.  These  two tests
 excluded any contamination of the motor potential
 by  extracerebral phenomena  or by  visual evoked
 potentials.

 Discussion

    Most authors who have studied  the cortical slow
phenomena associated with movement in man report
high interindividual  variability, whereas intraindi-
vidual phenomena for one kind of motor movement
appear stable (Gilden et al. 1966;Deecke et al. 1969,
1973; Gerbrandt  et  al.  1973).  The present findings
and  high standard deviations support these  observa-
tions.  Only Castaigne et al. (1969) report high intra-
individual variability.

    Modifications of slow potentials  in relation to
the force of movement have been studied previously
by Ford et al. (1972), Wilke and Lansing (1973), and
Kutas and  Donchin (1977). In  these three investiga-
tions, the amplitude  of the motor potential increased
with force exerted, thus affecting either the amplitude
of the readiness potential (Ford et al., Donchin and
Kutas)  or  negative-positive postmotor deflection
(Wilke and  Lansing). Furthermore, Wilke and Lansing
observed the maximum negativity point at a constant
latency from the beginning of EMC activity for a given
subject and suggested that this measure might contain

-------
 110
Hazemann et al.
  Table 1.  Means (m), Standard Deviation (a), and Number (n) of Observations for the Latencies
       (msec) of Cortical Motor Potential Components as a Function of Physical Parameters
                                        of Concentration

N1
P1
P'1
N2
P2
P'2
m
0
n
m
a
n
m
a
n
m
a
n
m
a
n
m
a
n
Force
500 g
-858
488
8
-120
87
5
110
47
8
80
97
8
370
62
10
662
133
10
200 g
-1016
454
5
•137
87
5
104
50
9
57
90
7
376
62
10
637
107
9
Speed
160 msec
-855
423
10
-224
144
5
660 msec
-830
298
8
•110
48
5
77
72
8
98
97
10
348
34
10
594
81
10
194
117
9
407
98
8
1104
266
9
Duration
500 msec
-1080
511
6
-326
148
6
70
47
4
106
65
6
336
55
5
1203
268
6
1000msec
-1156
266
6
-288
79
R
64
51
5
108
48
6
356
88
5
1745
296
6
Fig.  3.  A : Averaged cortical motor potential (C3-A1A2) for condition HS (left) and condition LS (right), with
(dotted line) and without (continuous line) visual control.  B: Averaged cortical motor potential (C3-AIA2) for
condition HS (left) and condition LS (right), from the raw EEC recording (dotted line) and EEC recording with-
out blinking and ocular movement (continuous line).

-------
Physical Parameters of Contraction
                                             111
information  about the  force exerted during a move-
ment. The only significant differences between GF
and SF noted in the present work concern the ampli-
tude of N2 (p<.05).

    The  influence  of movement  speed on  motor
potentials has  been studied by Becker et al. (1976).
Slow potential modifications affected Nl; the latency
was 800 msec for rapid and 1300 msec for slow move-
ments and the amplitude was greater for slow move-
ments. In the present study, no significant difference
in latency or in amplitude  was found, although HS
and LS  results  tended  to corroborate the  preceding
authors. The significant N2  latency changes were not
described by previous researchers. The absence of PI
in low-speed movements, although difficult to inter-
pret, is  perhaps evidence in agreement with the con-
clusion of Becker et al. that the organi/.ation of slow
movements is different from that of rapid movements.

    Gilden et  al. (1966) mentioned the influence of
movement  duration  on the motor potential. They
observed a positive wave beginning 50 to  150 msec
after the start  of EMG; the wave persisted, or dimin-
ished slowly, with sustained contraction. The behavior
of this positive wave is similar to the wave P'2 describ-
ed herein.  Otto et al,(1977  and  this  section) have
noted a prolonged positive variation during sustained
motor response. Are/.zoand Vaughan(1975) described
a positive component, P3, at 265  msec, which might
encode information on  body position.

    There  appear to  be  no major  contradictions
between  the results of this study and those of authors
cited. Supplementary  data, however, have been pre-
sented, allowing the definition of hypotheses and lend-
ing support to the hypotheses of others.  The most
important data concern the mobility of PI latency,
the relative stability of N1-P1  duration, the modifica-
tions of P1-N2 duration, and P'2 latency variations.

    Apparently, the limits  of variation in the motor
potential are influenced by  physical characteristics of
contraction  and depend upon the interrelations of its
various components. The stable N1-P1 suggests that  a
general  preparation mechanism is necessary for the
performance of any contraction.  The latency of PI
seems to be related to the type of motor contraction.
Gerbrandt et al. (1973) hypothesized that PI alone
might be of motor origin.  Deecke et al. (1973) pro-
posed that it might be an electrophysiological expres-
sion of movement initiation by parietal and, perhaps,
other association  areas.  Groll-Knapp et al.  (1977)
attributed PI to processes of thalamocortical control
of the efferent message.

    The  results tend to relate P1-N2 to the initiation
and contraction control period.  In the case of force
control, the period is short but lasts longer when con-
 trol affects speed (particularly the low-speed situation)
 or both speed of movement initiation and anticipation
 of its sustained nature.
     Despite the difficulty of relating an animal model
 to the present research on humans, animal experimen-
 tation shows variations in cell activity according to
 force, speed  of force  variation (dF/dt), and sustained
 movement. Evarts (1968, 1969), using the monkey,
 demonstrated a  relationship between  the activity of
 certain cells in the pyramidal tract and the magnitude
 and rate of change in force, as well as the direction
 of displacement. Working with the same species, Smith
 et al. (1975) differentiated three cell types in the pre-
 central cortex during performance of a sustained grip
 between the thumb and forefinger. Some cells behaved
 in  a dynamic manner (increase  or decrease  in the
 frequency of action potentials related to movement).
 Their  participation preceded  movement  by  500  to
 750 msec; their activity  rate could be related to the
 speed  of force variation and, less frequently, to  force
 alone. Other cells were static and came into play 0 to
 400 msec  after  the  beginning  of movement;  their
 activity vaguely increased with force and  remained
 constant until the end ofmovement. Finally, a greater
 number of cells were of a mixed type, dynamic and
 static at the same time.
     The stable temporal  link  between mechanogram
 commencement  and P2  peak  suggests that this wave
 is independent of contraction programming and  varia-
 bles of performance. This wave may represent  the late
 component of the somatosensory evoked response, or
 the homolog of Sutton's uncertainty wave, as  suggest-
 ed  by Gerbrandt  (1977). In contrast, P2, which
 occurs at the end ofmovement regardless of its physi-
 cal characteristics, varies in latency and becomes  larger
 with brief contraction. Central and  peripheral  feed-
 back mechanisms have often been proposed to explain
 electrical phenomena coming  after movement onset.
 Papakostopoulos et  al.  (1974a), studying  cortical
 potentials  related to passive  movements, attributed
 P2  to articular  afferences. Megirian et al.  (1974)
 observed an amplitude increase in rat motor potentials
 after novocainization of the moving paw and deduced
 that motor behavior  is controlled by peripheral and
 central sensory feedback mechanisms. The hypothesis
 of  a purely  peripheral origin seems further weakened
 by the findings  of Vaughan  et  al. (1970) that the
 ovfdl configuration of  the motor potential in the
 monkey is unchanged after deafferentation  of the
 moving limb.'

     In conclusion,  the   physical characteristics  of
 contraction induce  changes  in  PI latency in response
 to  temporal  program complexity. P'l  is absent and
 N2 latency  increases  during low speed contractions.
 Lastly, P2 latency changes in  relation to contraction
 duration.
'Note  Deecke's  counterargument elsewhere  in  this
 section.

-------
 NEUROANATOMICAL  ORGANIZATION  OF THE
 PRIMATE  MOTOR  SYSTEM: AFFERENT AND
 EFFERENT  CONNECTIONS OF THE  VENTRAL
 THALAMIC  NUCLEI.

 K.  KALIL
 Department of Anatomy, University of Wisconsin, Madison, WI, U.S.A.
     The ventralis anterior (VA) and ventralis lateralis
 (VL) nuclei comprise a major territory of the mam-
 malian ventral thalamus, extending forward from the
 ventroposterolateral nucleus (VPL) to the rostral pole
 of the thalamus. Although the ventral tier nuclei of
 the primate thalamus have been carefully parcellated
 into cytoarchitecturally distinct sub regions (Olszewski
 1952), some boundaries, particularly that between
 the VL and the rostral region of the VPL, remain diffi-
 cult to define.

     It has been demonstrated anatomically and physi-
 ologically that the VA and VL nuclei form an import-
 ant subcortical region of convergence for a number of
 structures related to the motor system. Afferent fibers
 to  the VA-VL arise from the deep cerebellar nuclei
 (Angaut 1970, Mehler 1971, Kievet and Kuypers
 1975, Rinvik  and  Grafova 1974, Chan-Palay  1977,
 Kalil  1977), from  the  globus pallidus  (Kuo and
 Carpenter  1973),  and  substantia nigra (Carpenter
 et al. 1976). Intracellular recordings show that a single
 VL neuron can be influenced by synaptic inputs from
 several of these afferent fiber systems (Purpura et al.
 1966, Sakata et al. 1966), and there is also consider-
 able physiological evidence that VA and VL exert a
 strong monosynaptic influence on those  neurons of
 the motor cortex that give rise to the pyramidal tract.
 Amassian and Weiner (1966) reported that pyramidal
 tract neurons could be monosynaptically excited by
 inputs from both VA and VL (see also Yoshida et al.
 1966). Furthermore, by combining antidromic activa-
 tion of VL neurons with stimulation of the cerebel-
 lum, Sakata et al. (1966) were  able to demonstrate
 that fibers arising  from  the deep cerebellar nuclei
 synapse directly with VL neurons that, in turn, project
 to the motor cortex.

    Despite the  fact that the  VL nucleus has long
been recognized as the principal source of thalamic
affercnts to the primate motor cortex (Walker 1934,
 1944), little it known  about either the precise top-
ography of thalamocortical connections arising from
the ventral lateral nucleus or their distribution within
 cortical laminae. Still less is known about the cortical
 projections of VA. While it has been suggested that
 VA projects to area 6 (the premotor cortex) (Mettler
 1947), large  numbers of VA neurons survive hemi-
 decortication (Powell  1952),  which was thought to
 imply that VA sends  relatively few axons directly to
 the cortex (see review by Carpenter 1967).

    The strong influence of VA and VL on pyramidal
 tract neurons points to the importance of these thala-
 mocortical pathways for the initiation and control of
 movement (Evarts and Thach 1969). Thus, it is sur-
 prising that so little is known about  the detailed
 anatomy of these pathways. One of the major reasons
 for the paucity of studies on thalamocortical connec-
 tions of the monkey  VA and VL  nuclei is the  fact
 that,  until  recently,  appropriate  neuroanatomical
 methods for such studies were lacking. Almost all of
 our knowledge of the cortical  connections of the VA
 and VL is derived from retrograde cell degeneration
 studies (Walker 1934, Chow and Pribram 1956), but
 these studies give little detailed information about the
 topography  of this  thalamocortical system. More
 recent  studies (Kievet  and Kuypers 1975, Strick
 1976b) employed the  horseradish peroxidase method
 and reported  retrograde transport of the enzyme to
 the VA, VL, and rostral VPL after injections into the
 motor cortex of the rhesus monkey. Though a more
 precise topography can be obtained with this method,
 it cannot elucidate the laminar organization of thala-
 mocortical fibers within the cortex.

   Similar problems  of methodology have also beset
 the detailed study of afferent input to  the VA and
 VL nuclei, and thus little is known about the precise
 organization of cerebellar inputs  to the  thalamus.
 Fortunately, however, the autoradiographic method
 for tracing neuronal pathways in the central nervous
system provides a means of studying the connections
of structures deep within the brain without the prob-
lems  of damage to fibers of passage  in  anterograde
studies or resorting to large  lesions in retrograde
studies. In this technique, small amounts of radioactive

-------
Neuroanatomical Organization of Motor System
                                             113
amino acids are injected into the brain and incorp-
orated into proteins by neurons in the vicinity of the
injection site. Labeled proteins are transported soma-
tofugally to axon terminals, where they may be sub-
sequently localized by autoradiographic procedures.
Since axons do not synthesize protein in significant
amounts (Droz and Koenig 1970, Lasek 1970), label-
ed amino acids are not incorporated by fibers passing
through the injection site (Cowan et al. 1972, Cross-
land et al. 1973). Thus, in the present study the auto-
radiographic method was used to trace the afferent
cerebellar connections of the VA and VL thalamic
nuclei and their   efferent cortical  projections. The
results of these experiments will be  described  and
related to the  functional organization of the primate
motor system.

Methods

     Rhesus monkeys were  anesthetized with sodium
pentothal and  mounted in a Kopf stereotaxic head
holder. The location of target areas within the cere-
bellar nuclei and the VA and  VL were determined by
the atlas of Snider and Lee (1961) and by location of
cortical surface landmarks visualized after removal of
a large bone flap and dural incision. Injections of the
gracile and  cuneate nuclei were  made under direct
vision by  incising the  a tl an to-occipital  membrane.
Small amounts of tritiated proline (0.1 to O.Sml) in a
concentration  of 20 to 30 AiCi/pl were injected into
the thalamus,  dorsal column nuclei, and deep cerebel-
lar nuclei through a 26-gauge needle, fitted to a 1 jul
Hamilton syringe. In order to prevent cellular damage
and  to restrict the size of the injection site, the (3H)
proline was delivered slowly  by means of a Harvard
infusion pump. After the injection, the needle was left
in place for at least 20 minutes to avoid contamination
of the cortex when the needle was withdrawn.

     The animals were sacrificed at survival times rang-
ing  from 48 hours to  1 week in order to visualize
labeled material in both axon terminals and in axons.
The  monkeys were  reanesthetized   and  perfused
through the heart with  10% formol saline. After fur-
ther hardening in formalin, the head and intact brain
were placed in the stereotaxic apparatus, and the brain
blocked transversely  in  the  Horsley-Clark plane or
sagittally. The blocks were immersed In sucrose form-
alin  and then  encased in a gelatin-albumin mixture.
Frozen sections (cut at 30pim) of the entire brainstem,
thalamus, and  cortex were mounted on gelatinized
slides, defatted in xylene and individually coated with
NTB-2 (Kodak) emulsion. The slides were exposed at
4°C in light-proof boxes for 6 to 8 weeks, developed
in D-19 at  1S°C, stained through the emulsion with
cresyl violet, and examined with the light microscope
under bright-field  and dark-field illumination. Chart-
ings of closely spaced individual sections through the
injection sites and labeled areas of the brain were made
with an overhead projector. Anterior-posterior levels
of the brain sections were numbered to  correspond
with the atlas  of Olszewski  (1952). Surface recon-
structions of the labeled cortical areas were made on
enlarged photographs of the cortex.

Results

Afferent  connections of the ventral thalamic
nuclei

    Cerebellar  Input:  Injections of the dentate  and
interpositus nuclei of the cerebellum were made in
five  monkeys.  The preliminary  results will be sum-
marized by describing one case in detail. As Illustrated
in Fig. 1, the injection site comprises  a  2-mm wide
strip in the center of the nuclei; it encompasses all
but  the   caudal  pole of the  dentate-interpositus
complex  and extends dorsoventrally throughout the
entire width of the nuclei. Rostrally, after crossing in
the decussation of the superior  cerebellar peduncle,
labeled axons enter the contralateral thalamus at levels
shown in the illustration. Two sharply defined "puffs"
of label are located in the central lateral nucleus (CL),
and  their consistent location  throughout  serial thala-
mic  sections indicates a precise topography between
the lateral cerebellar nuclei and the central lateral nu-
cleus. By contrast, almost no label is found in the
centromedian  nucleus (CM). Cerebellar axons enter
the  ventral thalamic  nuclei  and terminate in VLc
(ventrolateral nucleus, pars caudalls of Olszewski), in
VPLo (ventroposterolateral nucleus, pars oralis),  and
in VLo (ventrolateral nucleus, pars oralis). As shown
in Fig. 1, these axonal terminations are concentrated
in a band approximately  2 mm wide and located in
the most lateral part of the  ventral nuclei.  This ar-
rangement strongly suggests a  precise  mediolateral
topography in  the  cerebello-thalamic  projections.
There are no projections, for example, to the medial-
ly located VA or nucleus X. Moreover, cerebellar ter-
minations are not distributed evenly in  this narrow
strip of thalamus. Rather, silver grains are arrayed in
striking bands  and semicircular shapes that surround
the deeply staining cell clusters of the VL nucleus.

     It is also noteworthy that the VPLo receives some
of the densest cerebellar terminations. In view of the
fact that  this thalamic region projects to the  motor
rather than the sensory cortex (see results in next sec-
tion), it is important that this rostral part ofVPLbe
further identified as a motor region of the thalamus
by virtue of its afferent input from the cerebellum.

     Dorsal-column input:  In several monkeys, injec-
 tions of the dorsal column nuclei were made to deter-
 mine the precise border between the  cerebellar and
 lemnlscal recipient regions of the thalamus and to
 answer the  question  of  whether  these  motor and

-------
 114
                                             Kalil
                                                                                                      A4.5
                             P9.0                                 P10.0                                 MO. 5
Fig. 1. Series ofchartings of coronal sections through the brain of a rhesus monkey (21), following an injection of
(3/f) proline into the dentate and interpositus nuclei of the cerebellum. Center of injection  site represented by
black area; periphery represented by shaded  area. Labeled axons represented by  broken lines; labeled axonal
terminations by stippling. Note horizontal bands of labeled terminals in the  VL and VPLo nuclei. Abbreviations
used in the figures are as follows: Ace. CUM., n. cuneatus accessorius; AD, n. anterior dorsalis; AM, n. anterior
medialis; AV, n. anterior ventralis; CL, n. centralis lateralis; CM, n. centrum medianum; Cun., n. cuneatus;Den, n.
dentatus;Fas, n. fastigii; GM, n. geniculus medialis; Grac., n. gracilis;Int., n. interpositus; LD, n. lateralis dorsalis;
LP, n. lateralis posterior; MD., n. medialis dorsalis; Pul., n. pulvinaris; R, n. reticularis; S.G., n. suprageniculatus;
VA, n. ventralis anterior; VAmc, n.  ventralis anterior, pars magnocellularis;  VLc, n. ventralis lateralis, pars cau-
dalis; VLo, n. ventralis lateralis, parsoralis; VPI, n.  ventralis posterior inferior; VPLc, n. ventralis posterior latera-
lis, pars caudalis; VPLo, n. ventralis posterior lateralis, pars oralis; VPM, n. ventralis posterior medialis; X, area X.
sensory  thalamic  regions overlapped to any  extent.      Efferent cortical connections  of  the ventral
One such case will be described.                           thalamic nuclei

     As shown in Fig.  2, the injection site  includes
parts of both the gracile and cuneate nuclei. Lemniscal
fibers are heavily  labeled and their terminations are
easily identified  in  the thalamus. In this and other
cases, the lemniscal terminations in the ventral nuclear
group are confined  exclusively to the caudal part of
VPL (VPLc of Olszewski). No silver grains are found
beyond  the  VPLc-VPLo border region. Thus, there
appears to be no overlap in the cerebellar and lemnis-
cal thalamic inputs.
    Thalamo-cortical projects of the VA: A number
of injections of (-^H) proline were made into regions
of the ventral anterior nucleus (VA). Two represent-
ative  cases will be illustrated. In  case VA  9 R, the
injection site (represented by dark shading) is centered
in the VA nucleus (Fig. 3). As shown by the series of
chartings, the silver grains representing labeled axonal
terminals are found in those regions of the  premotor
cortex (area 6)  occupying  the medial  wall of the
hemisphere  and  the  dorsal bank of the  cingulate

-------
Neuroanatomical Organization of Motor System
115
                                                                                 Ace. Cun.
                                                                                 N. Cun.
                                                                                 N. Gruc.
                 PlO
         Fig. 2. Series of charting* of coronal sections through the brain of a rhesus monkey (CB 1 Rj
         following an injection of f^ff) proline into the dorsal column nuclei. Injection site represented
         by shaded area. Labeled lemniscal fibers shown by broken lines; labeled axon terminations in
         VPLc represented by stippling.

-------
116
      c                           C
Fig 3 Series of charting! of coronal sections through
the brain of a rhesus monkey (VA 9 R) following
injection  of (3H) proline  into the ventral  anterior
nucleus (VA).  Note terminal  labeling in area 6. occu-
pying the medial  wall of the hemisphere and the dor-
sal bank of the cingulate sulcus.
sulcus. (This cortical recipient zone will be reconstruct-
ed in Fig. 5.) No fibers from VA terminate in area 4
(,t the motor cortex. These thalamic afferents to area
6 terminate massively in layer 3. Layers  1 and 6 also
contain significant numbers of silver grains.

    In another case, VA 3 L, the injection site occu-
pies the VA but  also spreads posteriorly into VLc.
As shown by the shaded areas in Fig. 4,  there is only
slight spread oflabel into VLo. The resultant axoplas-
mic  transport reveals a dense projection  to area 6 on
the medial wall of the hemisphere and to the cortical
surface dorsal to the arcuate sulcus (see reconstruction
in Fie  5)  The very sparse projections to area 4 most
probably result from the slight labeling of VLo.

     When  these two cases are compared with other
injections of VA, it appears  that this thalamic  region
projects exclusively upon the premotor cortical area 6
   a rough topography such  that more anterior levels
 ol  the VA project to more anterior regions of area 6.
 Moreover, on the basis of its cortical projections, VLc
 appears to be a caudal continuation of VA rather
 than  part of VL, whose cortical association  is with
 area 4 of the motor cortex.
                                            Kalil

    Thalamo-cortical projections of the VL and VPL:
In a large number of rhesus monkey brains, localized
injections were placed in  different areas of the  VL
and VPL nuclei. In one such case, Monkey VL  1 L
(Fig. 6),  the center of the injection site was localized
in VPLo. Labeled  axons stream out of the internal
capsule  and terminate heavily  in the motor cortex
(area 4).  When these projections are reconstructed on
an actual photograph of the brain (as illustrated later
in Fig. 9), it is apparent that, although there is some
labeling of the somatosensory cortex, the overwhelm-
ing amount of label is localized in area 4 of the motor
cortex. According toWoolsey's map (1958), the label-
ed regions of the motor cortex correspond primarily
to the face, head, tongue, and a small part of the fore-
limb areas.

    In another case, the injection site is concentrated
in VPLo  but also extends rostrally into VLo. As shown
in Fig. 7, heavy  labeling is found in the motor cortex,
but despite the fact that a considerable portion of the
VPL was labeled by the injection, the somatosensory
cortex contains almost no labeled axons. The surface
reconstruction (Fig. 8) shows that the  labeled areas
extend over a broad area of the right precentral  motor
cortex (area 4), corresponding to  the  head, face,
trunk, and forelimb  regions (Woolsey 1958). A care-
ful comparison  of reconstructions  of a number of
VL-VPL injections reveals a topographic projection to
the  motor cortex such  that medial  regions of the
VLo-VPLo  project  upon  the arm  and  face areas
and lateral regions project upon the leg area.

     Precise information was also obtained regarding
the projections  of the border zones between various
regions  of the VL and  VPL nuclei. The results of
many  experiments in  which small  volumes of (3H)
proline  were injected  into  the thalamus  reveal  that
injections of rostral VPLo and  VLo result in labeling
of motor cortex alone, while those centered in VPLc
and caudal VPLo reveal a dense projection to sensory
 areas exclusively. The border zone between these
motor and sensory regions is localized in the posterior
 region of  VPLo and the  rostral extremity of VPLc.
 When an  extremely  small volume  (approximately a
 1-mm sphere) of this transitionary thalamic region is
 injected, labeled axons arising from the injection site
 terminate exclusively upon  area 3a in the  floor of the
 central  sulcus.  This cortex, which  receives  muscle
 spindle  information (Phillips et al. 1971), is thought
 to be transitionary between the sensory and motor
 cortex.
     The autoradiographic technique gives extremely
 precise information regarding the mode of termination
 of thalamocortical fibers, particularly when the auto-
 radiographs are viewed in dark field illumination. By
 far the  densest termination of thalamic  fibers  is in
 lamina  3, a layer containing many  medium-sized
 pyramidal cells and the apical dendrites from the larger

-------
Neuroanatomical Organization of Motor System
                                            117

 Fig. 4. Charting! of a second case in which proline was injected into VA (monkey VA 3 L).  The resultant transport
 of label to area 6 on the medial wall of the hemisphere and to the cortical surface dorsal to the arcuate sulcus is
 shown by stippling.
pyramids of layer 5.  However, there is also a substan-
tial projection to layer 6. The termination  of thala-
mocortical fibers in layer 1 of the cortex is more varia-
ble.  In some cases, particularly  in the  premotor
regions, this termination is quite dense, but in other
cases (Fig. 9 and 10) there is only sparse labeling of
layer 1. Moreover, there is variability in the pattern of
thalamocortical axon termination  within layer 3. The
dark field photo-montage  in Fig. 9 shows that  the
distribution  of label in layer 3 of the arm area motor
cortex  forms a continuous dense band. By  contrast,
Fig. 10 shows a striking series of columns or patches
of label in layer 3 of the face area motor cortex. The
bands are about  1 mm in width and are separated by
relatively grain-free spaces,  about  0.5 to 1 mm wide.
These columns also extend into layer 6 of the cortex.
Discussion
    The results reported in these experiments reveal
that the motor and sensory regions of the monkey's
ventral thalamus are separate and discrete. The sensory
fibers arising from the dorsal column nuclei terminate
primarily upon VPLc  and do not extend beyond the
VPLc-VPLo border region, which sends efferent fibers
to area  3a. This thalamic region corresponds precisely
to the areas of VPL in which Poggio and Mountcastle
(1963)  recorded classical lemniscal responses follow-
ing  stimulation  of the periphery. Rostral to  this
lemniscal recipient zone (i.e., in the region designated
by  Olzsewski as  VPLo),  Poggio and  Mountcastle
described neurons with properties very different from

-------
118
                                           Kalil
   VA9
                                           VA3
Fig. 5. Surface reconstructions of the medial wall of
the hemisphere to show areas of terminal labeling in
area 6 of cases charted in Fig. 3 and 4.

the modality-specific lemniscal cells, with their precise-
ly localized peripheral receptive fields. These VPLo
neurons were described as  having very large or  even
bilateral receptive fields and were activated only by a
heavy or prolonged stimulation  of deep tissues. Our
experiments  reveal  a  dense topographic termination
of efferent fibers from  the deep cerebellar nuclei in
VPLo, VLc and VLo. These thalamic areas thus repre-
sent the  motor region of the  ventral nuclei. Recent
physiological  evidence (Strick 1976a)  confirms this
view since some of these neurons, which are related
to  arm movements performed by awake  monkeys,
could not be driven  readily by peripheral afferent
input.  Thus,  the present anatomical  findings show
that motor and sensory inputs to the thalamus do not
overlap, and  these results are  strongly  supported by
physiological  experiments.

    Similarly, the efferent thalamocortical  connec-
tions of  the  ventral nuclear complex reveal a topog-
raphy and clear-cut demarcation among (1) VA and
VLc projecting to the premotor cortex (area 6), (2)
VLo and  rostral VPLo projecting to the motor cortex
(area 4),  (3) a VPLo-VPLc  border zone projecting to
area 3a, and (4) caudal VPLo-VPLc projecting to the
somatosensory cortex.  Again, there is no overlap  in
the motor and sensory  thalamocortical fiber systems.
These results are supported, in part, by the retrograde
horseradish-peroxidase method used by Strick (1976b),
who also found that a large rostral zone of VPL(i.e.,
VPLo) projects not to sensory but to motor cortex.
This finding raises the question as to whether this
region  of the  thalamus is actually part  of the VPL
receiving peripheral afferent input or  whether it is, in
fact, a caudal  region of VL that has heretofore not
been identified correctly. The issue  of whether the
motor  cortex  receives  somatosensory input directly
from the thalamus is especially significant in view  of
the recent observations by Rosen and Asanuma (1972)
that cortical efferent zones in  the monkey's  motor
cortex,  which influences  distal  forelimb  muscles,
receive a projection from cutaneous afferents. These
authors concluded  that  cooling of  the postcentral
gyrus showed that sensory cortex was not involved in
transmitting peripheral input to the motor cortex and
that "the thalamic region of origin for the fibers sub-
serving peripheral  afferent input to the motor cortex
is  obscure." Our  anatomical  results clearly  show,
however, that VPLo does not receive  lemniscal input
from the dorsal column nuclei but is, instead, a cere-
bellar recipient zone. Thus, physiologically demon-
strated peripheral  afferent input to the motor cortex
must travel by routes other than the classical lemniscal
pathway.

    More recently,  Lemon and Porter (1976) and
Lemon et al.(1976) have reinvestigated the correlation
between  peripheral afferent  input  and movement-
related neurons in the motor cortex of awake behav-
ing monkeys, rather than in the anesthetized prepar-
ations used by Rosen and Asanuma. Lemon and Porter
found that the most  powerful peripheral input to the
motor cortex was generated by joint movement. This
input probably originates in muscle, tendon, or joint
receptors, but  the receptors could not be specifically
identified in these experiments. The  brevity  of the
latencies reported in these studies argues for a fairly
direct  projection  from the  periphery to the  motor
cortex, but the authors conclude that neurons in VL
are unresponsive to peripheral inputs or have poorly
localized responses and that VPL and its fast afferent
pathways are a more likely route for peripheral affer-
ent input to the motor cortex. This conclusion accords
well with the anatomical data reported here, i.e., that
the VPLo receives a massive cerebellar input  that is
then relayed directly to the motor cortex. However,
the "fast afferent  pathways" to VPL must come from
the cerebellar nuclei and not the dorsal column nuclei.

    Regarding the  columnar organization  of the
motor  cortex,  Rosen and Asanuma (1972) suggested
that afferent input is topographically  organized such
that each motor cortical "column" receives a specific

-------
A 13
                                                                                                                             V
                                                                                                                                 U
                                   A TO 5
              c                               c                                  c                                   c


    Fig. 6 A series of charting! of coronal sections through the brain  of a rhesus monkey (VL 1  LJ following infections of f H) proline

    into VPLo of the thalamus. Terminal labeling (represented by stippling) is heaviest in the motor cortex, area 4, and is concentrated i

    layer 3. Labeled axons also  terminate in the somatosensory cortex.
                                                                                                                                                 I
                                                                                                                                                 SL

                                                                                                                                                 |

                                                                                                                                                 1'
                                                                                                                                                 -

                                                                                                                                                 I
                                                                                                                                                 c
                                                                                                                                                 c
                                                                                                                                                 —*
                                                                                                                                                 c
                                                                                                                                                 I

-------
                                                                                                                                           ~
Fig. 7. Charting* of coronal sections of a brain with injection of(3H) proline into VPLo and VLo. Heavy labeling is found in the motor
cortex but almost none in the somatosensory cortex.

-------
Neuroanatomical Organization of Motor System
121
                                                      peripheral  input  from the same anatomical zone to
                                                      which the  column projects. By  contrast, the experi-
                                                      ments of Lemon and Porter (1976) on awake behaving
                                                      monkeys suggest a less rigid arrangement of the affer-
                                                      ent projection since a local region of the motor cortex
                                                      may  receive inputs from entirely different joints.
                                                      These authors conclude that "there is no highly speci-
                                                      fic columnar organization of cells within the motor
                                                      cortex, divided and  separated one column from the
                                                      other, either from the  point of view of the afferent
                                                      input to these cells, or from the point of view of the
                                                      motor activities with which the discharges of the cells
                                                      are associated."

                                                         On the basis of the physiological data  of Lemon
                                                      and Porter, it  seems unlikely  that "puffs" of label
                                                      found in layer 3  of the face area of the motor cortex
                                                      after   injection  of the  VPLo represent  physiological
                                                      "columns." It seems more likely that "puffs" of high
                                                      grain density represent the terminations of thalamo-
                                                      cortical projection fibers and that the grain-free spaces
                                                      are filled  in by  commissural  fibers  from  the  other
                                                      hemisphere. If this interpretation is correct, the  corti-
                                                      cal areas representing the arm  and leg receive a con-
                                                      tinuous  band of thalamocortical fibers since  cortical
                                                      regions serving distal extremities  would not receive
                                                      commissural fibers.
Fig. 8. Surface reconstruction of cases plotted in Fig.
6 and 7. Dots on brain photographs show labeled areas
extending over broad areas of the precentral motor
cortex (area 4),  with relatively little labeling of the
somatosemory cortex from injections of the VL-VPL
nuclei.

-------
  122
                                                                                                   Kalil
Fig.   9.  Dark field photo-montage shows distribution of label in layer 3 of the arm area motor cortex.  Note
that the label forms a continuous dense band, which appears white in dark-field.

-------
                                                                                                                                                       f

                                                                                                                                                       §
                                                                                                                                                       H
                                                                                                                                                       B
                                                                                                                                                       •
                                                                                                                                                       P4
                                                                                                                                                       Q


                                                                                                                                                       o'
                                                                                                                                                       B.
                                                                                                                                                       S'
                                                                                                                                                       H

                                                                                                                                                       tt
                                                                                                                                                       £


                                                                                                                                                       C
                                                                                                                                                       2
FJ^.  10. Dark field photo-montage shows series of columns or patches of label in layer 3 of the face area motor cortex

in width and are separated by relatively grain-free spaces about 0.5 to 1 \im wide.
The bands are about 1mm

-------
  RELATIONSHIPS BETWEEN
  BEREITSCHAFTSPOTENTIAL AND
  CONTINGENT NEGATIVE VARIATION

  W. C. McCALLUM
  Burden Neurological Institute, Bristol, England
     The  Bereitschaftspotential  (BP),  or  readiness
 potential, and contingent negative  variation (CNV)
 were discovered independently, and first reported in
 1964. The term BP is generally reserved for cerebral
 slow potential changes preceding a voluntary action,
 while the term CNV is most frequently applied to slow
 potential changes occurring in the interstimulus inter-
 val of a foreperiod reaction time situation. Both are
 electrically negative phenomena at the surface of the
 cerebral cortex, over which they are widely  distrib-
 uted, and  both appear in circumstances  involving
 preparation for action or decision. The tendency in
 the past to speak  of the CNV as if it were a simple
 unitary phenomenon may have been misleading. Two
 or more relatively independent components emerge in
 certain circumstances.  Loveless and Sanford  (1973)
 and Loveless (1976) suggest that the late negativity of
 the CNV-i.e., that part  which immediately precedes
 the imperative stimulus-may be identical with the BP
 and  distinct from  earlier negativity associated with
 the warning stimulus, a view supported by Rohrbaugh
 et al. (1976). Others suggest that the two effects may
 be manifestations of activity  in the same underlying
 systems. Before such issues can be  resolved, it is neces-
 sary  to take a closer look at what relationships can be
 observed between the BP and CNV.

    As a result of collaborative work between the
 groups responsible  for the discovery of both potential
 changes, an experimental  paradigm  and computer
 program's  were evolved  to examine the form  and
 distribution of the two potentials separately and joint-
 ly. This procedure  was demonstrated by Deecke and
 his colleagues at Bristol (McCallum and Knott  1976).
 This  paradigm has been modified and extended in the
 present experiment to  further investigate the rela-
 tionship involved.

Method

   Scalp recordings were made from 12 normal sub-
jects  (8 female, 4  male) using Ag/AgCl electrodes.
Similar cortical recordings were made from 12 patients
  (9 female, 3 male) in whom gold electrodes had been
  implanted in connection with clinical procedures.

     Recording was by a modified  16-channel Elema
 Schonander Mingograph linked to a PDF-12 com-
 puter. Time constants of 5 sec were used for normal
 controls and, where possible, for the patient group In
 the latter group,  however,  it was necessary in rive
 cases to reduce the time constant  to 2.5 sec and in
 five cases to 1.2  sec to offset  the effects of high-
 amplitude infraslow activity, to  which the intracere-
 bral electrodes become increasingly sensitive. Scalp
 electrodes were located at  2 cm above the nasion
 F3, F4, C3, Cz, C4, P3, and P4, and referred to a
 common  reference consisting  of  the  linked  mas-
 toids.  Cz  was compensated for eye movements as
 described by McCallum and Walter (1968). Individual
 trials  contaminated by  eye  movement  or other
 artefacts were rejected on-line. Recording in  the
 patient group was  from eight selected subdural
 gold electrodes, each 150 jum in diameter and 4 mm
 long, extending in an  anterior-posterior  line  over
 the  right  cerebral cortex from the prefrontal re-
 gion  to  the  postcentral gyrus. Surface  Ag/AgCl
 EMG electrodes were located over the flexor muscles
 of the right and left forearms in normal  subjects and
 patients.

    The basic task consisted of pressing,  with the
 index finger, a  1-cm-diameter metal stud, housing a
 pressure transducer, set  in a flat wooden  base. All
 subjects reclined on a bed with eyes open  and fixated.
 The experimental procedure consisted of a BP condi-
 tion in which subjects made brief, voluntary finger
 presses with a minimum interval of about  10 sec
 between presses. Twenty-four presses were sampled
 from the left hand and 24 from the right. For all con-
 ditions, the epoch sampled was 4 sec at a rate of 128
 Hz. Two seconds were sampled before and 2 sec fol-
 lowing each trigger pulse. The trigger pulse was initiat-
ed from the output of the pressure transducer. The
second, or CNV, condition consisted of a foreperiod
reaction time (RT) situation in which a single warning

-------
BP and CNV
                                            125
click, delivered 2 sec after the start of the sampled
epoch  through headphones at approximately 70 dB,
was followed 1.5 sec later by a 700-Hz tone terminated
as rapidly as possible by pressing the metal stud with
the index  finger of the preferred hand (or the left
hand in the case of the patient group). RT was auto-
matically recorded by  the computer. Finally, in  a
BP/CNV condition, the two situations were combined,
the click stimulus  being  removed. Subjects made  a
series of voluntary button presses with the preferred
hand (normals) or left hand (patients). Each press was
followed  1.5  sec later  by the tone which subjects
were required to terminate as rapidly as possible by a
second button press, with the same  finger. BP and
CNV amplitudes were  measured  with respect to  a
baseline calculated as the mean amplitude of activity
recorded during the first 300  msec of the sampled
epoch.

Results
    Both BPs and CNVs were present in  all subjects
and patients and, superficially, showed a similarity of
distribution.  Both were largest at the vertex, the
mean scalp amplitude for Cz being - 12.5 juV for CNV
and -5.0 juV for BP. As in previous studies, scalp CNV
showed no consistent asymmetry in its distribution.
Individual differences between bilateral pairs of elec-
trodes could be up to 3/uV, but these were not consist-
ent in  direction across the group, nor did they appear
to be  related to handedness  or to the task involved.
Scalp BP showed a small asymmetry, amplitude values
over the hemisphere contralateral to  the hand used
for pressing being slightly  higher than those over the
ipsilateral hemisphere (Table  1). Differences for left-
hand presses failed to reach significance, but those for
right-hand presses were significant for central (p<.05)
and parietal (p<.01) locations using a t-test for corre-
lated samples.

    Both BP and CNVs were  smallest over the frontal
and pre frontal regions, and the BP frequently  could
not be observed at all in these areas. The CNV, how-
ever,  was almost invariably observable  frontally, if
only at low amplitude. In the combined BP/CNV con-
dition (Fig.  1), scalp CNV diminished by up to 30%
over central and 20%  over  parietal areas compared
with the  standard CNV condition (Table  1), but
remained  relatively unchanged over frontal areas. In
the same condition (BP/CNV),  scalp  BP  showed
increases of 30% or  higher over all areas (including
frontal) compared with the simple voluntary press
condition. Fig. 2 presents cross-subject vertex averages
for the four conditions.

    Anterior-posterior CNV distribution over the right
cortex of patients was similar to scalp recordings in
normal  subjects with the largest amplitudes found
precentrally. In some patients the most anterior elec-
trodes showed no CNV, while in others the CNV was
quite prominent there, but varied somewhat in form.
BP amplitudes were  markedly diminished  when the
ipsilateral hand was used for pressing. Fig. 3 shows
this asymmetry in one patient. In the one left-handed
patient, this difference between contralateral and ipsi-
lateral hand presses was even more pronounced. BP
distribution also spread more anteriorly in this patient.
Fig. 4,  which illustrates waveforms  from  the eight
selected electrodes averaged across patients  shows the
consistency of the asymmetry. Fig. 5 provides similar
across-patient  averages  for  the CNV and  BP/CNV
conditions. The locations of electrodes varied to some
extent from patient to patient, although  the most an-
terior electrode was  always over prefrontal cortex; the
second electrode was always over frontal cortex; and
the remaining electrodes extended posteriorly from
precentral cortex, the most posterior electrode usual-
ly being over the postcentral gyrus. The distance of
electrodes from  the midline  also varied considerably.
This variability and  time constant  differences are ig-
nored in Fig. 4 and 5 to obtaining a broad picture of
hemispheric differences. Therefore, these summary
averages should  not  be taken as a reliable  indication
of localized differences of  waveform or amplitude,
although individual  records clearly indicate that such
differences exist.
               Table 1. Mean Amplitude (/uV) of the Bereitschaftspotential  and CNV
                                 on the Scalp in  Different Tasks


Measure (task)
BP deft)
BP (right)
BP (combined)
CNV (simple)
CNV (combined)

Above
nasion
-1.82
0
-1.1
-0.7
-1.3


F3
-1.55
•0.91
-2.9
-4.1
-4.3


F4
-1.36
+0.45
-1.7
-4.5
-4.0
Electrode

C3
-3.18
-3.64
-4.8
-8.2
-6.7
position

C4
-3.45
-1.45
-2.7
-8.2
-6.4


Cz
-5.27
-4.55
•6.0
-12.5
-8.5


P3
-2.0
-1.3
-2.6
-6.1
-5.4


P4
•2.09
0
-1.4
-6.1
-5.0

-------
  126                                                                                      McCaDum

               VOLUNTARY BUTTON PRESS (RIGHT INDEX FINGER)   TONE (TERMINATED BY BUTTON)

 -20 u
            1 sec
                           »w
                           '   U
                                                             EYE
V^^4AU^»    J
    F3
Ffc /. 5ca/p distribution of the combined BP/CNV condition from one normal subject. Averages of 24 trials in
which subject made a voluntary right index  finger press, followed 1. 5sec later by a tone, terminated by a second
press using the same finger. •
    Mean RT to the imperative stimulus in the stand-
ard CNV condition  was 160 msec (SD 40) in normals
and  185 msec (SD  50) in  patients. In  the BP/CNV
condition, these values increased to 185 msec (SD 52)
tor normals and 224  msec (SD 71) for patients. Differ-
ences between  the  conditions  were significant for
both groups on at-test forcorrelated samples (p<.05).

Discussion
    Results indicate that BP is maximal at the vertex
and is weakly asymmetrical, showing increased nega-
tivity «ver tne hemisphere  contralateral  to the opera-
tive musics, the largest values occurring prerolandical-
ly.  The asymmetry  appears most marked when the
dominant hemisphere is the one  primarily involved in
the movement. This finding is consistent with that of
Kutas and Donchin (1974)  and is in general agreement
with that of Deecke et al. (1973), although the latter
liiilhuis wish to differentiate between several separate
components in  the  activity preceding motor move-


     Intracerebral  data confirm the  distribution  pat-
tern that emerges from scalp data. The fact that corti-
cal electrodes in any one patient were  restricted to
one hemisphere prevented bilateral comparisons, but
the differences seen in response to left- and right-hand
pressing reveal the  contralateral predominance of the
BP more clearly than scalp data.
    As indicated in Table 1, mean CNV amplitudes
for the normal  group  were remarkably symmetrical.
This finding is consistent with those of McCallum and
Cummins (1973), McCallum (1976), and Marsh and
Thomson (1973). Isolated reports of CNV asymme-
tries under special conditions  have appeared, which
might seem to contradict the  established view (Low
et al. 1966, Cohen  1969) that CNVs are  bilaterally
symmetrical. Otto and Leifer(1973b)reported larger
negative shifts over the motor  cortex contralateral to
the movement, but the measure used was total CNV
area and the differences only reached significance
when pooled to include a condition in which a motor
response was initiated during the S1-S2 interval. Butler
and Glass (1974) have  observed larger CNVs over the
dominant hemisphere during numeric operations, but
the findings have  not been  replicated. Low  et al.

-------
 BP and CNV
                                                                                                 127
                                                                   VOLUNTARY BUTTON PRESS
                                                          RIGHT HAND
                                                                                    LEFT HAND
                                                                            12
 D

-10
                         t
                       PRESS
                   (OR SI-CLICK)
                                         S2
                                       (TONE)
           1 sec
Fig.  2.  Averages (12 subjects) of vertex (Cz) channel
for  each  of the four experimental conditions. A:
Voluntary press,  left  index  finger;  B:  Voluntary
press, right index finger; C: CNV-click, tone, button
press with index finger of preferred hand; D: BP/CNV
(as in Fig. 1).
                                                      ^
                                                                                                    -
(1976) also  reported mean CNV area  asymmetries
related to hemispheric dominance for language pro-
duction  in patients  and to handedness in normals.
The extent to which they apply to "late" negativity,
however, is not clear. So far, the balance of evidence
does not substantiate any claim for consistent asym-
metry  of CNV  amplitude measured  prior  to  the
imperative stimulus.

    Although superficially similar, the BP and CNV
are independent  in  terms of  distribution and  the
balance between the two distributions varies with the
situation. Results of the present study further  suggest
that, as task involvement increases, frontal areas
make  a  relatively greater contribution, as reflected
in  the  distribution  of  both types  of  slow  poten-
tial  change.  In   planning  future  ERP  experiments
to examine the role of different cerebral  regions in
the context of planned motor  actions,  the value of
adding levels of task complexity or involvement to
the simple press or warned foreperiod situation should
thus be  considered.  (See also Papakostopoulos,  this
section.)
                                                                                         t
                                                                 -20
                                                                              2 sec
                                                     Fig. 3. Anterior-posterior cortical distribution of BP
                                                     over the right hemisphere of one patient, recorded
                                                     from  gold subdural electrodes extending from pre-
                                                     frontat (electrode  12) to postcentral (electrode 6)
                                                     regions. Averages of 24 trials.
                                                         While this study underlines the inappropriateness
                                                     of speaking of the CNV as if it were  immutable in
                                                     form and distribution, it lends little support to those
                                                     who, rejecting a "unitary phenomenon"  theory of

-------
 128
                     McCallum
                RIGHT-HAND PRESS
LEFT-HAND PRESS
                       20 pV
                                 1 sec
                                                  (h)
Fig.  4-  Averages (eight patients} from an anterior-posterior line of gold subdural electrodes over the right hemis-
phere for BP conditions involving left and right voluntary finger presses. Note: The diagram showing electrode
placement gives only  a  general indication of their distribution.  Exact locations of equivalent electrodes varied
from patient to patient.  These differences have been ignored in averaging.

-------
 BP and CNV
                    129
                   CNV
BP/CNV
                                         ELECTRODE

                                             EMG
      jKr^^V^^v-^^jlv          Vr
       20 vV
              1 sec
                                              (h!
Fig. 5. Averages (eight patients) from an anterior-posterior line of gold subdural electrodes over the right hemis-
phere for CNV and BP/CNV conditions. See note on Fig. 4.

-------
  130
                                                                                             McCallum
  CNV, seek to replace it with a two-, three-, or multi-
  component theory. To equate early CNV negativity
  with an orienting response when it can be seen to
  occur-albeit slightly diminished-when SI is replaced
  by a voluntary button press, is at best an oversimplifi-
  cation. To equate late CNV negativity with the BP is
  also an oversimplification because a clearly lateralized
  BP and symmetrical CNV can be  demonstrated at
  the same electrode sites.

     Mounting evidence  points  to  slow  potential
 changes as a function of an extensive mosaic of corti-
 cal neuronal domains. Which domains are active during
 a particular process of preparation for action or deci-
 sion  and at what time during the  course of that pre-
 paration they are active will depend upon innumerable
 factors inherent in both the individual and the situa-
 tion. Intracerebral recording reveals some of the result-
 ant subtle changes and local differences. Few of these
 subtleties, however, survive the spatial  averaging that
 takes place in  the transition through skull  to scalp.
 Variations  in  the  distribution of waveforms seen at
 the scalp can hopefully provide information on broad
 areas of function and can indicate cortical regions
 primarily involved at given points in time. However, it
seems unwise to be lured into a process of component
 labelling  on the basis of sparsely sampled scalp data
'p.d. note:  See D. Papakostopoulos, "Macropotentials as a
 Source for Brain Models." Section X, this volume.
 acquired in a limited range of circumstances. Hastily
 applied  labels, particularly those derived from broad
 psychological constructs, add little to our understand-
 ing of brain events and may  even impose constraints
 on our thinking and interpretation. It would be wiser
 to review some  of the constructs concerned  in the
 light of recent event-related potential data than to
 squeeze  the physiological data into ill-fitting psycho-
 logical or behavioural categories.1

 Summary

     Bereitschaftspotential and  contingent  negative
 variation were recorded independently and in associa-
 tion with one another from multiple scalp electrodes
 in a group of normal subjects and from subdural elec-
 trodes over the right hemisphere in a group of patients.
 The distribution  of the two phenomena was similar,
 but not  identical. CNV was slightly more prominent
 anteriorly than BP, and BP showed some asymmetry-
 largest over the hemisphere contralateral to the hand
 used  in  pressing.  In  the  situation in  which BP and
 CNV  were combined,  BP appeared frontally and
 increased in amplitude at all electrodes, whereas CNV
 decreased over central and parietal areas.

 Acknowledgment

    The  help of Philip Newton  in recording data and
Philip Pocock in  development of the computer pro-
gram is gratefully acknowledged.

-------
SLOW POSITIVE SHIFTS  DURING
SUSTAINED MOTOR  ACTIVITY IN HUMANS
D. OTTO AND V. BENIGNUS

U. S. Environmental Protection Agency, Research Triangle Park, and Department of
Psychology, University of North Carolina, Chapel Hill, NC, U.S.A.
    Bates (1951) was the first to report slow cortical
potentials associated with movement onset. He describ-
ed a negative shift commencing about 20 msec after
EMG onset and attributed the wave to the "arrival of
afferent impulses from the periphery." Bates failed to
detect  any cortical potentials preceding movement.
Kornhuber and Deecke(1965), using improved averag-
ing techniques, observed a slow negative shift before
movement, followed by a positive shift after move-
ment onset. Vaughan et al. (1968) designated this late
positive wave as the P2 component of the motor poten-
tial. Since the late positive wave appears to follow the
initiation of motor activity, Komhuber and Vaughan
both hypothesized a peripheral afferent origin. The
evidence, however, is inconclusive.
    Otto et al. (1977) observed a prolonged positive
shift, maximal postcentrally, during sustained motor
response. In that study, subjects were required to esti-
mate the duration of a 1-sec holding interval without
external cues. It is possible, therefore, that the positive
shift  reflected a  central timing  mechanism, rather
than  afferent  input from  peripheral muscles.  This
hypothesis may be tested by estimating time without
sustained motor response. On the other hand, if the
positive  shift represents  incoming afference,  one
would expect to see larger shifts over the sensorimotor
region corttralateral to response. Further, one would
expect to  find the amplitude of positivity directly
proportional to the force of contraction. The present
study was undertaken to test these hypotheses.
nize signal averages. Each subject's maximal pressing
force, with the dominant hand, was determined with
a Jamar PC5033 hand dynamometer. Each subject
then completed a series of 3-sec contractions at force
levels  of 10, 25, and 50% maximum, as monitored
visually  on the gauge of the dynamometer. Subjects
were instructed to initiate and terminate contractions
briskly and to relax responding muscles as completely
as possible between presses.

    Two additional conditions were run to determine
the effect of estimating time without tonic contraction
and the  corollary effect of tonic contraction without
estimating time.  The estimation  task  consisted of
squeezing the dynamometer briefly to indicate the
lapse of 3-sec after presentation of a tone pip. In the
nonestimation task, subjects initiated and maintained
contraction until presentation of a tone pip.


Results

    Prolonged positive shifts were observed during
hand pressing in  most subjects in most conditions.
Averaged waveforms  recorded  in six subjects during
the 50% condition are shown in Fig. 1. Positive shifts
during pressing were easily discernible in the raw data
of many subjects, as shown in Fig. 2. This sample
illustrates, however, that the positive shift is not invari-
ably present, nor is the onset and termination clearly
time-locked to activity in the brachioradialis muscle.
Method

    EEC  recordings, referred  to linked ears, were
obtained from nine young adult male subjects at C3,
C4, P3, and P4 using amplifiers with 2.25-sec time
constants. EOG and rectified EMG from the brachio-
radialis muscle were also recorded. A Schmitt trigger,
adjusted to fire at EMG onset, was used to synchro-
    Fig. 3, averaged across subjects, illustrates that
the magnitude of positivity is not directly related to
pressing force at C3 and C4. Larger shifts were observ-
ed during the 10% and 50% conditions  than during
the intermediate 25%  force level. Only at P3 was
there a suggestion of  a linear relationship between
amplitude and force. No consistent lateraliry differ-
ences were observed in any condition.

-------
  132
                                              Otto and Benignus
             DY01
DY03
  EMG
            DY07
DY09
 EMG
 EOG
     .3  O  J  4  6  8
       TIME (SECONDS)
Fig I.  Slow potential shifts observed in six subjects
during 50% isometric contraction. Averages of 6 to
11 trials were triggered from onset of brachioradialis
muscle EMG and digitized at 48 msec/point by means
of a PDP-12 computer. EOG was recorded diagonally
from above the  inner canthus to below the outer can-
thus of  the right  eye. Baseline  was  computed from
the initial 400-msec segment in each average.
 Fig 2-  Single trials from one subject during 50% con-
 tractions. Note  that positive shifts are not present in
 all trial* "r clearly time-locked to the firing onset or
 • i/fifi »f the brachioradialis muscle. Trial 5 was trig-
 gered fram  a transient KMG spike.  Trials with im-
 proper  F.MG triggering, or EOG artifact, during or
 preceding contraction were excluded from averages.
                                                         EMG
                                                          C3
                                                          C4
                                                         EOG
                                                                                    .:
                                              -rEEG:IO/uV
                                                 EOG: SO/iV
                                                RELATIVE FORCE
                                                OF CONTRACTION
                                                    -10%
                                                      25%
                                                  	50%
                                                                               8  SECONDS
Fig 3. Composite averages from seven subjects during
10%, 25%,  and 50% contractions. Positive shifts do
not appear to be proportional to force of contraction.

    Results of the delayed response task 0abeled SR)
provide evidence that the positive shift during sustain-
ed  motor activity  is not the  reflection of a central
timing mechanism. Time estimation in the absence of
tonic  motor response yielded a slowly incrementing
negative  shift, similar to CNV, as shown in Fig. 4.
When subjects initiated a hand press voluntarily,  but
released upon presentation of a tone pip (condition
HS), no  consistent  positive or negative shifts were
observed (Fig. 4).
EOG
SLJ
64


][ l

                                                  HS
                                                         EEG: 10ATV
                                                         ~
                                                         EOG: 50;UV
                                                       EMG
                                                        •101   234
                                                           TIME (SECONDS)
                                                                                         	i
                                         -101234
                                           TIME (SECONDS)
                  Fig. 4.   Summary averages  (eight subjects) during
                  time estimation without contraction (SRj and con-
                  traction without time estimation (HS). Negative shifts
                  were observed during  SR, but no consistent shifts
                  were found during HS.

                  Discussion

                      Results of this experiment confirm our previous
                  report (Otto  et al. 1977) of a  prolonged  positive
                  potential  over the  centroparietal  region of the brain
                  during tonic  motor activity.  Similar motor-related
                  positive shifts have been noted previously  (Gilden et

-------
Positive Shifts during Motor Activity
                                             133
al.  1966, Jones and Beck  1975), but  never studied
systematically.
   Other  investigators  report  conflicting  evidence.
Vaughan et al.  (1970) observed a prolonged negative
shift  in  epidural  recordings over  the  contralateral
motor cortex of monkeys during sustained wrist con-
tractions. Robert et al. ( 1976) also observed sustained
negative  shifts  at  the vertex in humans during tonic
weight pulling. Rebert's experimental design was con-
founded, however, by external stimuli and a secondary
reaction-time task which bracketed the weight-pulling
within an extended CNV interval.

    Ottoet al.  (1977)have shown that negative antici-
patory  and positive  response-related processes sum
linearly  on the scalp when tonic  motor activity is
superimposed on a CNV-eliciting interval.  Since  the
negative potential tends to be larger than the positive
potential in this situation,  the observed effect of sus-
tained key-pressing is a  reduction in CNV amplitude.
Papakostopoulos and Cooper (this volume) have con-
firmed this phenomenon.

     Further evidence of the interaction on  the scalp
of negative and positive slow potentials occurs in this
study. When subjects were asked  to briefly squeeze
the dynamometer 3 sec after presentation of a tone
pip  (time  estimation  without tonic  contraction),
a slow negative  shift was  observed consistent with
McAdam's (1966) report.When subjects were instruct-
ed to initiate and maintain contraction until presenta-
tion  of the tone pip, however, no consistent positive
or negative slow potentials were observed. The sim-
plest  explanation  of this finding  is  that negative
anticipatory and  positive  response-related potentials
 summed to zero at the scalp recording sites.

     Is the slow positive shift during motor response
an extracerebral artifact? All trials  containing signifi-
cant eye movements as indicated in EOG tracings were
carefully excluded from averages. Picton and Hillyard
 (1972)  and Corby et  al.  (1974) have described a
cephalic  skin potential artifact which could conceiv-
 ably contribute to the observed waveforms. This arti-
 fact  was not observed, however, by Picton or Corby
 when linked ear references were used, as in the present
 study. The small amplitude of positive shifts (5 to 10
     also is inconsistent with a cephalic skin  potential.
     Another possibility is that the baseline is biased
 in a negative direction by the readiness potential that
 precedes movement. The baseline measure was com-
 puted from  the initial 400 msec of data in each aver-
 age. The baseline epoch terminated  2.3 sec prior to
 the EMG trigger and appeared to precede the onset of
 the readiness potential in most cases. The precise onset
 of  the  readiness potential,  is,  however, difficult to
 determine since  it is an internally generated event not
 necessarily time-locked to the initiation of movement.
    A closely  related problem concerns the selection
of an appropriate trigger event.  EMG  onset of the
brachioradialis  muscle was  chosen  since this  large
superficial muscle discharges prominently and reliably
during fist clenching. Many other muscles of the fore-
arm also participate  in this complex  motor response.
A transducer attached to the handle of the dynamom-
eter might have provided a more reliable trigger index
of movement onset.

    The question of  the functional significance of the
prolonged positive shift remains unanswered. Results
of the  delayed  response-time estimation  task argue
against  the central timing hypothesis. If the positivity
reflects  afferent input from peripheral  muscles, one
would expect  the amplitude to increase directly with
the force of pressing. One would also expect the wave-
form  to be larger over the somatosensory region con-
tralateral to response. The fact that neither effect was
observed  casts  dcubt on the peripheral afference
hypothesis.

    Considerable  evidence  reviewed by  Hazemann
(this section) indicates that afferent input is inhibited
during movement. Marczynski (this volume) concludes
from  an extensive review of behavioral, electrophysio-
logjcal,  and pharmacological data  that  slow surface-
positive brain potentials,  including the P300, RCPV,
and  other  motor-related  positivities, reflect cortico-
fugal   suppression (inhibitory  modulation) of the
ARAS.  Papakostopoulos  and Cooper (this section)
propose a similar explanation of the CNV reduction
observed during the  Jendrassik maneuver. Slow posi-
tive  shifts  described in   this study  ate presumably
related, functionally  and  neurophysiologically,  to
these other ERP phenomena. Further research is need-
ed to evaluate the corticofugal inhibitory hypothesis
of Marcyznski and  Papakostopoulos,  and to clarify
the  relationship  of  the   expanding  family of slow
positive ERPs.

Summary

    A prolonged positive shift was observed at central
and  postcentral recording sites during tonic contrac-
tion  of the hand. No consistent  hemispheric differ-
ences or linear relationship with the  force of contrac-
tion  were found.  It appears unlikely,  therefore, that
this  waveform reflects proprioceptive  or kinesthetic
feedback from responding muscles. CNV-like negativi-
ty  during  simple time estimation  suggests, further-
more,  that  the positive  waveform does not reflect a
central timing mechanism.

Acknowledgment

     The authors thank J. H. Knelson for support and
L. Ryan and J. Bedrick for technical  assistance.

-------
  ELECTRICAL  ACTIVITY OF THE  BRAIN
  ASSOCIATED  WITH SKILLED  PERFORMANCE
  D. PAPAKOSTOPOULOS
  Burden Neurological Institute, Bristol, England
     Self-paced and externally  triggered movements
 are preceded  by sustained  negative brain potential
 changes known respectively as Bereitschaftspotential
 (BP) (Deecke et  al. 1969)  and contingent negative
 variation (CNV) (Walter et al. 1964). The movements
 employed are usually simple brisk flexions or exten-
 sions of a  finger  with no other demand imposed on
 the subject. In the case of the CNV, reaction time to
 the imperative signal is usually measured and a con-
 troversial literature has developed  (McCallum and
 Knott  1973,  Papakostopoulos  and  Fenelon 1975).
 When  studied in  isolation,  reaction time, limb dis-
 placement, or development  of muscular force, how-
 ever important they may be for movement, contrib-
 ute little to  our  understanding  of the organization
 and development of skillful action. An experimental
 paradigm was devised in  which brain activity was re-
 corded  during a  manipulative task  demanding pre-
 cision and improvement of performance by providing
 the subject with  real-time  information on the  out-
 come of his  action.  Comparison  was made with
 more  conventional  situations where  only  a  brisk
 motor action was required.

 Method

    Seven subjects, aged  18 to 34 and including five
 male and two female, were investigated. Four were
 left-handed and three were right-handed.

    The subject sat in a comfortable chair facing the
 10-cm  screen  of a cathode ray  oscilloscope  (CRO).
 In each  hand, he held a plastic ball 3 cm in diameter
 fitted  with a  button that could be pressed.  The
 excursion of the button was Smm. Pressing with the
 left thumb initiated a single sweep of the oscilloscope
 spot;  pressing with the right thumb  stopped  the
 sweep.  The spot velocity was I  m/sec.  The  subject
 was told to  initiate  the sweep using the left hand
 and stop it with the right within ±1 cm of the center
 of the CRO screen, i.e., within a 40- to 60-msec in-
 terval  following sweep initiation.  The above pro-
cedure will be referred to as the skilled performance
  test (SPT).  A PDP-12 computer was programmed to
  acquire  EEC and  EMG data for 1.2  sees preceding
  and 0.8 sees after the initiation of the trace.  The
  sampling rate  was 250 points/sec.  The latency be-
  tween left and right-hand press was also computed.

     For each trial, six EEC and three  EMG channels
  were recorded and stored on digital magnetic tape.
  EEC electrodes were  placed  at Fpz,  Cz, and 5 cm
  lateral to the midline 2 cm anterior (precentral) and
  2 cm posterior (postcentral) to  the central sulcus of
  both hemispheres.  EMG was recorded  from left fore-
  arm flexors and extensors and right forearm flexors.
 The  time constant  and  high-frequency  response
 (-3 dB) were 5 sec and 70 Hz for EEC and 0 03 sec
 and 3000 Hz for EMG.

     The experimental procedure included 24 trials
 with left-hand  press  (LHP), 24  trials with  right-
 hand press (RHP), 24 trials with  both-hand press
 (BHP),  and 100 trials of  the  skilled performance
 test, all self-paced.

     Stored data were analysed either  as single trials
 or  averages. Averaging was done with two different
 criteria according to the temporal order or precision
 of  responses.  In order to examine temporal effects
 three averages  of 24  trials of LHP, RHP, and BHP
 conditions and four averages of 24 successive trials
 of SPT were computed. To classify data in terms of
 performance, the time following initiation of the spot
 by  the left-hand press was separated into successive
 20-msec intervals.  Correct performance was defined
 as right-hand press  in  the interval between  40 and
 59  msec.  Trials in which the right-hand press was
 early (20 to 39  msec) were called  -wrong (-W) and
 late  (60  to 79  msec) were  called +wrong (+W).

    Mean amplitude measurements were computed
 for  three different  200-msec epochs for individual
 trials and averages as follows:  (1) immediately pre-
ceding left EMG onset-termed Bereitschaftspotential
(BP),  (2) immediately following left  EMG onset-

-------
Skilled Performance Positivity
                                             135
 termed motor  cortex  potential  (MCP)  as suggested
 by Papakostopoulos et al. (1975), and (3) commenc-
 ing 350 msec after left EMG oaset. The baseline for
 measurements was the mean amplitude of the initial
 200  msec  of data.  The Wilcoxon  matched-pairs
 signed-ranks test  and  the paired t-test were  used to
 evaluate findings.
  Results

      Subjects averaged 42.4 correct trials (range: 23-
  57).   The  700 total  skilled  performance trials in-
  cluded  155  early (-W), 297 correct, and 131 late (+W)
  trials;  117  trials  occurred  outside  these  intervals.
  The longest latency to right-hand press was 150 msec.
  Early   presses  significantly  decreased  (N=7,   t=5,
  p < .05) while  late presses  significantly decreased
  (na7, t=0, p=.02) in the latter part of the experiment.
  The frequency of correct presses  increased slightly,
  but not significantly, across trials.
      Measurements  of brain  potentials, eye  artifact
 (Fpz) and EMG averaged across subjects for each ex-
 perimental condition are shown  in Table 1.  Note
 that vertex HP and MCP measures for LHP, RHP, and
 BHP conditions were very  similar,  although lateral
 measurements  were differentially   affected  by the
 responding hand.   For  all four lateral electrodes for
 all subjects  combined, the difference between the
 measures using the contralateral hand and those using
 the  ipsilateral hand was significant both for the BP
 (n=14,  t=4.64, p  < .001)  and for the MCP (n=14,
 t=4.85, p<.001). This result suggests lateralization of
 both the BP and MCP  components. Table 1  also in-
 dicates   different  antero-posterior  distributions of
 these components. BP  was maximal at  central and
 postcentral  locations, while the MCP was maximal
 precentrally. Values of the third measure were small
 and  inconsistent  at all recording sites in conditions
 LHP, RHP, and BHP. Typical waveforms  from one
 subject  are shown in Fig. 1A.
      Engagement  in the skilled performance  test pro-
 duced marked  increases in the amplitude of BP and
 MCP components and  the emergence of a large, slow
 positive  component peaking about  400 msec after
 left EMG onset.  This  positive component was maxi-
 mal at the vertex and  averaged 14.2 juV (range: 4.5
 to  21 t*V) across subjects.  Since this large positive
 component  appeared  only  during the SPT, it was
 called skilled performance  positivity  (SPP).  This
 waveform is illustrated for one  subject in  Fig. IB.
      In order to assess the electrophysiological effects
  of performance, BP, MCP and SPP  measurements
were pooled  across all electrode sites  to derive the
mean voltage of correct  and wrong (±W combined)
trials for each subject. BP (n=7, t=2.55, p< .05) and
MCP (n=7, t=2.2,  p <.l) measures were greater for
correct  than  wrong trials. SPP and EMG amplitude
did not  vary consistently as a function of correct/
wrong response (Fig. 2).

    No   significant temporal order effect  was  ob-
served in any electrophysiological measure.

Discussion

    EEG, EMG, and behavioral  data were  collected
during  two types  (skilled  and  unskilled) of self-
paced motor activities.  Unskilled actions  consisted
of initiating the sweep of an oscilloscope beam with
either or both  hands.  The outcome of this action
was  invariant.  Skilled  actions entailed the  initiation
and  termination of the  sweep within  very narrow
limits (40 to 60 msec), a paradigm that permits the
study  of movement-related brain macropotentials
(MRBMs) during skilled  manipulative  performance.

    Considerable changes were  observed in MRBM
patterns  when  self-paced  movements  were direct-
ed toward a specific objective as opposed  to stereo-
typed  movements  without any  apparent  purpose.
Both the Bereitschaftspotential (BP) and motor cor-
tex potential (MCP) were larger in amplitude during
skilled  compared  to unskilled  performance.    The
major  finding,  however  was the  emergence during
skilled actions  of  a broad positive component  that
peaked   about  400 msec after  EMG  onset.   This
component  was observed  only in the skilled  task
and, therefore, has been  termed skilled  perform-
ance positivity (SPP).

     Since  scalp data  reflect a  spatial average of
underlying brain activity (Cooper et al. 1965), the
increase  in BP  amplitude during skilled performance
could  reflect  either increased  cortical synchroni-
zation  or  increased cortical  negativity.   The sig-
nificance of  this finding is not  clear,  although the
BP  ratio of skilled/unskilled tasks could conceiv-
ably provide a useful  electrophysiological  index of
brain efficiency.

     Kutas and  Donchin  (1974) have shown that BP
varies as a function of contractile force.   BP  differ-
ences observed in this  study cannot  be   attributed
simply to the force applied  in the two tasks because
significant differences in  BP amplitude  were observed
when correct and  wrong trials were selectively aver-
aged. Rectified EMGs were the same for both  sets of
trials.

     The MCP has been  interpreted elsewhere  (Papa-
kostopoulos  et  al. 1975) as an index  of  peripheral

-------
136
                                                                                     Papakostopoulos
              Table 1.  Mean Voltage (and S.D.) in pV of Bereitschaftspotential (BP),
             Motor Cortex Potential  (MCP), and Skilled Performance Positivity (SPP)
                                       During Self-paced Tasks
Left-Hand Press
Electrode
Fpz
Cz
Prec L
Prec R
Postc L
Postc R
EMG FL
EMG FR
BP
+0.4(3.7)
-3.1(0.9)
- 1.0(1.3)
-3.4(1.9)
-2.1(1.2)
- 3.9(0.9)
+ 2.8(3.6)
0
MCP
+ 0.4(2.7)
- 6.0(2.4)
- 3.7(2.2)
- 6.9(1.6)
- 3.6(2.0)
- 5.9(2.5)
+26.4(21.2)
0
SPP
+2.1(6.5)
+1.3(2.0)
+1.1(2.5)
+0.6(2.7)
+1.3(1.9)
+0.4(2.5)
+1.2(2.0)
0
Both-Hands Press
Fpz
Cz
Prec L
Prec R
Postc L
Postc R
EMG FL
EMG FR
+3.0(2.8)
-3.0(4.1)
- 2.0(3.6)
-3.2(3.5)
-4.7(2.6)
-4.7(2.6)
+6.0(5.2)
+4.8(3.6)
+ 3.5(4.2)
- 7.2(3.9)
- 6.0(4.8)
- 8.0(2.2)
- 8.0(3.7)
- 7.5(2.5)
+68.8(47.6)
+ 50.8(15.2)
+0.2(8.6)
+2.5(2.5)
+0.2(2.6)
0.0(3.0)
-2.0(3.4)
-0.7(8.5)
+4.0(5.6)
+6.8(11.6)
Right-Hand Press
BP
+ 0.0(3.0)
- 3.4(4.0)
- 3.9(2.8)
- 2.9(2.5)
- 5.0(3.5)
- 3.1(2.3)
0
+3.6(6.0)
MCP
+0.0(2.6)
-6.3(4.1)
-7.3(3.4)
-4.3(2.5)
-6.7(3.4)
-3.3(2.2)
0
+41.6(20.8)
Skilled Performance
+1.4(3.2)
- 10.2(4.7)
-6.9(4.0)
-8.9(3.2)
-8.8(3.7)
-9.7(4.1)
+3.2(3.6)
+2.8(3.6)
+0.5(4.2)
-14.9(5.6)
-12.7(5.7)
-15.2(4.1)
-13.1(5.0)
-13.6(5.3)
+63.2(46.4)
+54.6(22.0)
SPP
+ 0.6(2.8)
+ 3.0(3.7)
+ 0.4(3.0)
+ 1.4(3.1)
+ 0.6(3.5)
+ 2.1(3.5)
0
+1.2(2.8)
Test
+ 3.7(3.6)
+ 14.2(6.0)
+10.9(4.0)
+ 9.7(4.9)
+ 10.4(4.5)
+ 9.5(5.7)
+3.6(2.8)
+7.2(9.6)
reafferent  activity  from  skin,  joint, and  muscle
receptors during movement.  The change in MCP am-
plitude  observed in this  study  could  reflect  an in-
crease in cortical excitation and, therefore, increased
responsivity to peripheral input during skilled actions.
Such  increased  excitability has been  observed in a
CN'V  task  (Papakostopoulos et al. 1970).  Alterna-
tively, the  MCP increase could reflect  relaxed  gating
(disinhibition) or selective subcortical  facilitation of
peripheral  input (Frigyes et al.  1972; Papakosto-
poulos et al.  1975; Ha/emann, this section) during
skilled performance.

    The SPP is a new component in the constellation
ol movement-related macropotentials.  It  is distinct
in waveform, amplitude,  and latency  from the P2
component that occurs in averages of unskilled press-
es (Fig. 1).   Since  there is little apparent difference
in EMG patterns for the both-hand press and skilled
performance  conditions, the SPP cannot be account-
ed  for  in  terms  of increased  reafferent activity.
                                                       The time course of the SPP suggests that it may
                                                   coincide with  the  realization  by the  subject that
                                                   his actions  succeeded  or failed, i.e.,  when informa-
                                                   tion concerning the consequences of  performance is
                                                   being processed.  Although the task  was self-paced,
                                                   instantaneous visual feedback was  available from the
                                                   oscilloscope  to evaluate performance.  It  is likely,
                                                   therefore, that  the  SPP is related to the P300 wave'
                                                   which  has  been associated  with similar  concepts
                                                   of  information  delivery,  resolution of uncertainty,
                                                   and  feedback (see  Tueting, this volume, for an ex-
                                                   tensive review of the P300 literature).
                                                       Few  data  are  available  concerning the neural
                                                   substrate of the SPP or P300. Marczynski (1972 and
                                                   this  volume)   has  described  similar slow  positive
                                                   shifts in cats associated with the delivery of appetitive
                                                   reinforcement.   Otto et  al. (1977 and  this  section)
                                                   have  described  prolonged  positive  shifts associated
                                                   with sustained motor responses.  The association and

-------
Skilled Performance Positivity

         BpMCP
                                                                                                137
                                                                              MCP
Fig. AX. Superimposed averages of brain and EMG
potentials during left-, right-, and both-hand presses
of a button.  B. Data from the same subject and lo-
cations during the skilled performance test. Horizon-
tal bars indicate measurement epochs.

significance  of  these  various  movement- and rein-
forcement-related  slow positive macropotentials re-
main to be elaborated in future studies.

Summary

    Movement-related  brain  macropotentials
(MRBMs) were compared during skilled and unskilled
self-paced motor tasks. Engagement in a task that re-
quired  precisely coordinated  movements  of both
hands yielded  increased  amplitude  of the  BP  and
MCP and the emergence of a new component  called
skilled  performance  positivity  (SPP). The BP, MCP,
                                                      EYE
                                                      Fig. 2. Comparison between wrong (-W) and correct
                                                      (C) trials during the skilled performance test. Hori-
                                                      zontal bars as in Fig. 1.

                                                      and SPP can be differentiated in terms of topograph-
                                                      ical distribution and functional significance.  The BP
                                                      seems  to  be related to the organization of prepro-
                                                      grammed  action, the  MCP to  movement-generated
                                                      reafference, and the SPP  to the evaluation of per-
                                                      formance outcome.

                                                      Acknowledgments

                                                          The author  wishes to gratefully  acknowledge
                                                      the assistance of P. Pocock in computer programming
                                                      and statistical analysis  and of P. Newton in experi-
                                                      mental support.

-------
  THE ELECTROMYOGRAM, H  REFLEX,
  AUTONOMIC FUNCTION,  AND
  CORTICAL POTENTIAL CHANGES
  DURING  THE JENDRASSIK  MANEUVER

  D. PAPAKOSTOPOULOS AND R. COOPER
  Burden Neurological Institute, Bristol, England
     During the foreperiod of a simple reaction time
 (RT) experiment, when sustained potentials known as
 the contingent negative  variation (CNV)  develop
 (Walter et al. 1964), the excitability  of the  spinal
 monosynaptic reflex  increases (Papakostopoulos and
 Cooper  1973, 1976).  However,  the  relationship
 between the  development of cortical sustained nega-
 tivities and increase in the excitability of the spinal
 monosynaptic reflexes, as measured by the H reflex
 technique (Paillard 1955), is not a simple one as the
 two phenomena show a different time course. It has
 been proposed, therefore, that these two phenomena,
 and heart  rate deceleration which accompanies them,
 are not causally related but are mediated through a
 common mechanism in the brain stem that integrates
 the whole body reaction to the stimulus and response
 situation (Papakostopoulos and Cooper  1973). To
 test this hypothesis further, we decided to study the
 time course of H reflex excitability and CNV develop-
 ment  in two situations with similar stimulus condi-
 tions, but  different response requirements. The first
 situation was the simple RT experiment (situation 1),
 the second was the Jendrassik maneuver (situation 2).

    This  maneuver is a classical clinical  procedure
 designed to increase spinal monosynaptic reflexes. It
 involves clenching the fist to an external command
 and relaxing  it to  another. By  replacing  the verbal
 commands with two stimuli, 2 sec apart, and appropri-
 ate instructions to  press a button to the first stimulus
 and release it to the second, the situation is similar to
 a simple RT experiment (situation 1) with the addition
 of initiation and sustainment of a continuous motor
 output during the foreperiod.


    Previous  data   by   Donchin  et  al.  (1973),
McCallum and Papakostopoulos (1972), and  Otto et
al. (l''73a,  1977) where motor action has been sus-
 tained during the foreperiod, suggest that the CNV
during  such  a procedure should  be smaller in com-
parison with the CNV to similar external cues in the
sample RT experiment.
 Method

     Five female and four male volunteers (mean age,
 24 years) were paid to participate in the study. Mono-
 synaptic reflexes (H reflex) were electrically elicited
 at various times during and after the 2-sec foreperiod
 of a simple RT experiment with a click as SI and a
 tone as S2 and an average intertrial interval of 29+ 11
 sec. The H reflex was elicited once per trial in pseudo-
 random order 0.2, 1,1.5,1.8,1.9 sec after SI and 0.2,
 0.5, 2,  5, 10 sec  after S2 by a 0.5-msec electrical'
 pulse applied to the tibial nerve in the popliteal fossa.

    Five sequences each of 10 trials  were presented
 in which the  subject responded to S2 by pressing a
 button with his left thumb (task 1) and five sequences
 in which he responded to SI by a press that had to be
 maintained  until release after the presentation of S2
 (task 2). Either of these  actions to  S2  caused the
 cessation of the tone. The occurrence of stimuli for
 eliciting the CNV and H reflex was controlled by the
 PDP-12 computer which was also sampling and display-
 ing data. RT was measured and printed out on-line.
 Before  initiating the trigger pulse  for each trial, the
 computer measured and stored the R to R interval for
 5 heart  beats. The  fifth beat  initiated the  stimulus
 sequence and SI appeared  1 sec later. This procedure
 was used to establish EEC, EMG, and heart rate base-
 line before SI. EMG of forearm flexors,  gastrocnemius
 and tibial anterior muscles from surface electrodes
 were each sampled at 1-msec intervals until 1  sec after
 S2. Respiration (chest expansion) and heart beat inter-
 vals continued to be stored for 6 and 10 sec after S2,
 respectively. Electrodermal  resistance (EDR) was also
 monitored.

    For the H reflex, the total sampling time was 50
msec. The CNV, examined during a 4-sec epoch, was
recorded from the vertex, C3, and C4 referred to'link-
ed mastoid process. Anasion + 2-cm electrode referred
to the same  reference was used to monitor eye move-
ments and also served for compensation of the vertex
channel.  A  modified 16-channel  Elema-Schonander

-------
Cortical Potential during Jendrassik Maneuver
                                                             139
electroencephalograph acted as the main amplification
system. For EEC, respiration, and EDR, a time con-
stant  of 5  sec was used. Special Ag/AgCl electrodes,
selected to have less than 1-mV potential between any
pair in distilled water, were used. Separate amplifiers
with  10-Hz bandwidth  were used for  recording the
H  reflex. Individual trial data were stored for further
off-line analysis.

Results

    RTs to S2  in tasks  1  and 2 were very similar
(267  ± 60 msec and 255 ± 54 msec, respectively).
RTs to SI  in task 2 were consistently slower (391 ±
91 msec) than those to S2 in both situations.

    EMG activity from  the forearm muscles in task 1
was absent or just detectable during the S1-S2 period.
After S2 the phasic EMG activity leading to the action
of pressing was  recorded (Fig. 1 left).  The EMG in
situation  2 was  characterized by  an initial phasic
increase followed by a diminished and sustained activi-
ty until after S2, when a new phasic increase appeared
(Fig.  1 right). The latency of the phasic EMG increase
after  S2 in both situations was similar. This latency
was always shorter than the latency  of the phasic
EMG increase following SI in task 2. No EMG activity
was recorded from leg muscles at any stage of task 1
or 2. The consistent appearance of phasic EMG activi-
ty  after S2 in task 2 suggests  that the button release
was an active manipulative action that involved flexor
and extensor forearm muscle groups. In other words,
the button release was not a passive concomitant of
the cessation of action by the forearm flexors.
     CNV from the vertex and two central areas for
 all  subjects  tested was always  smaller during  the
 press-wait-ielease situation (task 2) compared with
 task 1. This is illustrated in Fig. 2 where  the grand
 averages obtained from all subjects for each condition
 at the vertex and central areas are shown. The possi-
 bility  of differences  due to eye movements can be
 excluded as the activity at Fpz during the interstim-
 ulus period was  very similar for the two situations.
                0.2 mV
                          1 sec
  EMG1
                    The H reflex elicited during the S1-S2 period was
                larger than H reflexes elicited 2 to 10 sec after S2 for
                both  situations. This increase during the foreperiod
                was larger in task 2 than in task 1.  This is shown for
                the whole group in Fig. 2 and for one subject in Fig.
                3. A  further increase of the H  reflex was recorded
                200 msec after  S2 in both situations and 200 msec
                after SI  in task  2. This  additional increase coincided
                in time with the phasic EMG increase leading to press
                or release after S2 and to press after SI in task 2 (Fig.
                2). No consistent changes in electrodermal potentials
                were recorded in the two situations.

                    The heart  rate during the foreperiod of task 1
                decreased in all subjects; this decrease was not observ-
                ed in task  2. The  deceleration during the foreperiod
                in task 1 and the absence of it in task 2 for all subjects
                tested is graphically represented in Fig. 2.

                Discussion

                    Data from  task  1  confirmed  previous  findings
                (Papakostopoulos  and   Cooper   1973,   1976)  that
                increased excitability of spinal monosynaptic reflexes
                and heart rate deceleration accompany CNV develop-
                ment during the foreperiod of a simple RT experiment.
                A very different pattern emerges in task 2, where a
                further increase  of H  reflex excitability occurs in
                association with diminished CNV and no heart rate
                deceleration. RT to the  second stimulus is similar in
                both tasks.

                    The EMG and H reflex data provide clues con-
                cerning the possible mechanism and significance of
                the neurophysiological  difference observed between
                the two tasks. Two basic types of motor action were
                involved in these tasks, phasic pressing or releasing in
                responses  to stimuli, and  tonic  pressing during the
                interstimulus interval of task 2.  Of all  the variables
                measured,  only the H  reflex from the  leg extensor
                muscles paralleled  the  EMG  changes in   forearm
                flexors. That is, phasic increases in H reflex excitability
                were  observed  with each  phasic action of  forearm
                muscles, while a tonic increase in monosynaptic excit-
                ability occurred with sustained pressing.
                                                       EMG 2
             SI (CLICK)
S2  (TONE)
PRESS
                                                                t
SI
PRESS
                                                   t
                                                                                         82
                                                                                          RELEASE
 Fig. 1. Averaged rectified EMG from the forearm flexors in task 1 (left) and task 2 (right) from one subject.

-------
  140
                    Papakostopoulos and Cooper
                                                                        C4
                                    EMG
                                                                        0.2mV
                                          SI
   t
   S2
                                                                                   1 sac
 RESP.
                                                                  S2+0*2 0*5
 Fig. 2.  Grand averages (nine subfects) of the CNV, rectified EMG of the left forearm flexors (EMG), respiration
 (RESPI, H reflex (HI, H2), and heart rate changes. The CNV, EMG, and RESP of task 1 are shown as thin traces,
 those of task 2 as thick traces. The H reflex (upper graphs) and heart rate (lower graphs) of task 1 are shown as
 continuous  lines and  these of task 2 as lines with squares. Measures were obtained at points indicated by  dots.
 Time calibration in RESP, 1 sec. Arrows indicate  the time of SI and S2 presentation.  Time after S2 is nonlinear
 for H reflex and heart  rate.
    Gottlieb et al. (1970)and Pierrot-Dcseilligny et al.
(1971) attribute phasic increases in H reflex excitabil-
ity during arm flexion to involvement  of the cortico-
splnal  or  pyramidal  system. Phasic actions in this
itudy were presumably mediated by pyramidal path-
ways. Other observed cortical and autonomic changes
may also relate  to pyramidal  function. The  CNV
decrease   associated  with  sustained  pressing,  for
example, could result from direct corticofugaJ action
on  bruinitem structures. It is conceivable  that  the
corticabulbar  component  of the  pyramidal system
selectively excites descending tracts of the  reticulo-
spinal system and simultaneously  inhibits ascending
retkular pathways. Selective Inhibition and facilitation
of subcortical pathways  by corticofugal action have
been demonstrated In  animals (Frlgyesi  et al. 1972;
Skinner,  this volume) and probably in  man (Papako-
stopmilos ct al. 1975). Moreover Steriade (1969) has
shown that reactivation of motor-sensory (precentral)
cortex, in contrast to other cortical regions, diminishes
with reticular activation. Diminished cortical activa-
tion,  therefore,  is not necessarily inconsistent with
Increased motor output.

    If this is the  case, CNV reduction in task 2 should
be selective for motor sensory areas. Otto et al.(1973b,
1977)  have shown,  however, that  sustained motor
activity during the S1-S2 interval yields a generalized
decrease in CNV over anterior (Fz) and posterior (Pz)
areas, as well as central regions.

    Dimished CNV could be  attributed alternatively
to reafferent feedback from peripheral receptors that
discharge during tonic motor activity (of. Matthews
1972). Several strands of  evidence are inconsistent
with this hypothesis. McCallum and  Papakostopoulos

-------
Cortical Potential during Jendrassik Maneuver

    1                                                          2
                                            141
                                                   IN
                                              2mV
                                                  OUT
                                                    10 msec

 Fig. 3.  Superimposed single H reflexes in task I  (left) and task 2 (right) elicited 5 and 10 sec after S2 (lower
 traces-Out) and during the period when the CNV develops (upper traces-In). Note that H reflexes during CNV
 (upper  traces) are larger than those elicited out of the CNV period. Yet Ms H increase is consistently larger for
 task 2 (upper right) than for task 1 (upper left).
(1972) have shown that CNV amplitude is not affected
if the button is depressed continuously and released
momentarily in response to S2. This finding suggests
that the initiation of action at the beginning of the
interval, rather  than sustained response, is the critical
factor. This initiation should be followed by feedback
from phasic receptors in the periphery. If such feed-
back can be seen with scalp electrodes, it should appear
as short duration negativity (Wilke and Lansing 1973;
Jones and  Beck 1975; Papakostopoulos et al. 1975;
Papakostopoulos, this volume). One  might  expect
such negativity  to summate with the CNV and  appear
as an increase of the sustained potential. The  results
of this  experiment are clearly contrary  to  this argu-
ment. Two other papers in this section (Hazemann et
al., Otto and Benignus) also present evidence incon-
sistent with the reafference hypothesis.

     Another approach is to assume that the  reduction
in CNV during tonic motor activity is phenomenal
rather  than real.  For instance, if another positive
potential occurred during  the foreperiod in parallel
with  the CNV, the  true  CNV amplitude  might be
masked. The existence of a prolonged positive shift
during  sustained  contractions has, in  fact,  been
demonstrated by Otto et al. (1977 and this section).
Prolonged  positivity  following skilled  performance
has  also  been  described  (Papakostopoulos,   this
section). Otto et al.  (1977) showed, moreover, that
the decrease in  CNV amplitude when sustained motor
activity was superimposed could be accounted for by
the simple linear combination of anticipatory negative
and response-related positive components.


    The neuronal generators and functional signifi-
cance of such positivities are as yet unknown. If these
potentials are related to the reinforcement contingent
positive  variation  (RCPV) reported by Marczynski
(1972),  increased rhythmicity  of the sensorimotor
areas during task 2 could be expected. This possibility
seems remote in terms of present data, which suggest
that motor  activity, phasic or tonic, is accompanied
by suppression  of sensorimotor rhythms (Chatrian
1959; Papakostopoulos, this volume).

Summary

    During  the  2-sec  interval  between  a  warning
stimulus SI  and a second stimulus S2 that requires a
motor action by the subject, the brain develops sus-
tained potential changes, known  as the contingent
negative variation (CNV), the  H reflex increases in
amplitude, and heart rate decreases. During the same
period,  EMG of relevant muscles is  silent  or only
moderately  active. To assess the significance of myo-
graphic  activity during the S1-S2 period, nine subjects
pressed a switch with the left thumb when SI occurred
and released it at S2.

    In both situations, the reaction times for releasing
or pressing  the switch to S2 were similar. However,
the CNV was significantly smaller and the heart  rate

-------
 142                                                                    Papakostopoulos and Cooper

deceleration during the S1-S2 interval did not occur      spinal  excitability extending to  autonomic function
when motor action was required during this interval.      and behavioral output. The effect of the motor action
At  the same time, there was a significant increase of      on subcortical centres that underlie cortical negativity
the H reflex amplitude beyond that normally seen in      can be seen as a result either of reafferent activities
the simple S1-S2 preparatory situation. A dissociation      generated by the action itself or, most probably, as a
has  thus been demonstrated  between cortical  and      direct inhibition by cortical efferents.

-------
 OCULOMOTOR COMPONENTS OF
 EVENT-RELATED ELECTROCORTICAL
 POTENTIAL IN MONKEYS.
 S. ROSEN, J. ROBINSON, AND D. LOISELLE
 Department of Psychology, State University of New York at Stony Brook, Stony Brook, NY,
 U.S.A.
    This report concerns the extent to which event-
related  cortical macropotentials, particularly those
time-locked to sensory stimuli,  reflect oculomotor
processes. Cortical structures have long been implicat-
ed in oculomotoi functions as a consequence of the
findings of electrocortical stimulation and single unit
recording studies.  Stimulation of  the  frontal and
occipital eye fields in the monkey (Brodmann's area 8,
18, and 19) result in eye deviations generally contra-
lateral  to the hemisphere stimulated  (Crosby et al.
1952, Ferrier  1875, Wagman 1964). Moreover, brief
pulse stimulation of the awake monkey's frontal eye
field (FEF) elicits single saccades of specific amplitude
and direction with latencies of 15 to 25 msec (Robin-
son and Fuchs 1969). The latter finding strongly sug-
gests that the FEF serves an oculomotor command
function. This view is challenged, however, by reports
indicating that the vast majority of FEF cells, which
indeed discharge in relation to saccadic and  pursuit
eye movements, fire during, or following, the move-
ments, but never before them (Bizzi 1968, Bizzi and
Schiller 1970,Mohleret al. 1973).

    Despite the evidence of cortical involvement in
the control  of eye movements, event-related cortical
macropotentials have rarely been examined in relation
to the  oculomotor processes  they may reflect. This
has been due in part to the difficulty in dissociating
between potentials of cortical origin related to eye
movements  and those  artifactually generated poten-
tials arising  as a consequence of  the rotation of the
corneofundal dipole with eye movements (Hillyard
and Galambos 1970, Peters 1967).  Furthermore, in
animal studies in which eye movement artifacts can
be reduced by means of bipolar recording electrodes
placed  directly on the cortex equidistant from  the
eyes   and use of differential amplifiers  affording
common mode rejection of eye potentials, great diffi-
culty is encountered in the behavioral control of eye
fixations and  the elicitation  of eye movements of
specifiable amplitude  and  direction.  The present
report presents the results of investigations in which
many of these difficulties have geen successfully over-
come.  Our  findings indicate that significant com-
ponents  of  prefrontal  electrocortical  potentials
evoked by visual stimuli upon which a monkey must
fixate  reflect  cortical mechanisms of oculomotor
control that both precede and follow accompanying
saccadic eye movements.

Methods

Subjects

    Nonpolarizable Ag/AgCl electrodes were chroni-
cally implanted in prefrontal (bilateral), precentral,
and occipital cortex and also subcutaneously  across
the eyes for horizontal electrooculogram (EOG) re-
cording in two stumptail monkeys.  Cortical elec-
trodes were implanted in pairs with one electrode of
each pair on the pial surface and the other in subjacent
white matter (5- to 10-mm tip separation). The pre-
frontal surface electrode was located on  cortex within
the posterior  half of the principal sulcus, and the
depth  reference was in  white matter  of the  lateral
bank of the sulcus. EOG electrodes, of the type de-
scribed by Bond and Ho (1970), were cemented to
the nasal  and  temporal sides  of the bony orbit. All
electrode  leads were  soldered to the Amphenol con-
nectors, which were then fixed to the skull with stain-
less steel screws and dental  cement.  Embedded in
the resulting cement cap were  two parallel brass tubes
(oriented perpendicular to the saggital plane) for head
immobilization in the testing apparatus.

Apparatus

    During  testing,  the monkey was  seated in  a
restraining chair containing a response lever within
easy reach. The lever had a 5-mm stroke, and 120  g
force was required to depress  it and activate a micro-
switch. The chair  was positioned in an electrically
shielded soundproof chamber so  that the monkey's
head was between two parallel plates rigidly fixed to
the chamber. The plates contained slots which permit-
ted attachment of metal  rods that were inserted
through the tubes  in the monkey's skull cap  to the
plates.  Once the monkey's head was immobilized in
this manner, the tip  of an adjustable brass tube for
injection  of  liquid rewards was brought to the

-------
 144
                                     Rosen et al.
 monkey's mouth. For the presentation of visual stim-
 uli, a perimeter  containing five light-emitting diodes
 (1-1.l>s) 20°apart was situated at eye level in front of
 the monkey at a distance of 1 meter. Onset arid dura-
 tion  of IJJJ illumination (equated for intensity and
 color), as well as the sequence of light presentations,
 was controlled by an external program  panel.

     The  chamber also contained connectors and low-
 noise  cables  for  recording EOG  and  electrocortico-
 gram  (hCoG) events.  Inputs were led  to low-level
 preamplifiers outside the chamber, and amplified sig-
 nals  were stored on FM  magnetic tape  for off-line
 analysis with a PDP-12A computer. A videotape sys-
 tem also  permitted simultaneous recording of electro-
 graphic and eye movement events.


 Procedures

     In order to  obtain steady fixations and saccadic
 eye movements  of known amplitude  and direction,
 monkeys were trained on  a difficult reaction time
 (R I) task. A single  fruit juice reward was delivered if
 the animal  pressed  the lever twice, once during the
 0.75-sec dimming of a central (0°) fixation light and
 again  during  the dimming  of a peripheral  test light
 that was  illuminated immediately following the first
 press.  Both  fixation and test lights were of variable
 duration  (1.5 sec  minimum), and the order of test
 light  presentations was random  from trial  to trial. A
 DHL schedule was  used  to  eliminate  presses  during
 the light-on, but not the dimming, periods. Animals
 were tested daily until their performance resulted in
 consistent patterns of lateral eye movements.

     In  another experiment involving operant  condi-
 tioning of unilateral cortical slow potential (SP) shift,
 rewards were made contingent upon on-line computer
 detection  of 3-sec surface-negative SP shifts of 50- to
 100- iiV amplitude. Conditioned SP events were then
 correlated with records of eye movements.


 Data analysis

    During testing, hori/ontal EOGs and ECoGs from
 left and  right prefrontal, left  precentral,  and left
occipital  cortex  were  recorded with  dc amplifiers
and the data were stored on FM magnetic tape. Corti-
cal  evoked potentials and HOG events resulting from
test-light     presentations were  computer-averaged
off-line, with  separate averages  obtained  from each
area for each of the five test lights. Each average repre-
sented the data from 40 similar stimulus presentations
with a time  base  of 2 sec. Pcak-to-peak amplitudes of
 the evoked cortical potentials and peak  latencies were
computed. These in turn were related to the ampli-
 tude, duration,  latency,  and direction of  averaged
 laieral eye movements.
 Results

 Stimulus-triggered eye movements

     EOG and videotape recordings made during suc-
 cessful RT  task performance by  two monkeys (i.e.,
 when  90%  or more of the monkey's paired presses
 were rewarded) indicated that the animal's gaze was
 directed towards the Fixation light  when presented
 and that appropriately  directed saccadic eye  move-
 ments followed test  light onset.  Computer-averaged
 EOG activity time-locked to the  different  test light
 presentations confirmed the existence of characteris-
 tic patterns  of  eye  movements  to each of the test
 lights (Fig.  1). The averaged recordings appear similar
 to  records  of individual  saccadic movements, with
 mean latencies of the averaged movements ranging be-
 tween 240 and 3 18 msec (Table 1). The amplitude of
 the  averaged saccades (both initial  and  final) were
 linearly  related  to  the  angular displacement of the
 visual targets (Fig. 2A).

 Potentials evoked by test stimuli

    Prefrontal:  Components of averaged electrocor-
 tical  potentials  evoked  in prefrontal cortex by test
 light presentations were observed  22  to 750 msec fol-
 lowing test  light onset.  Generally, two negative (Nl,
 N2) and two positive (PI, P2) peaks could be identi-
 fied in a N1-P1-N2-P2 sequence (Fig. l).TheNl,Pl
 and N2  components  preceded initiation  of saccadic
 eye movements  to the test target  by  20 to 150 msec.
 The rise of the P2 component coincided with the sac-
 cadic movement, although its peak amplitude was at-
 tained some  100 to 150 msec after the initial saccade
 was completed.  Prefrontal components could only be
 consistently identified in the hemisphere contralateral
 to  the direction of  the  stimulus-triggered  saccade
 i.e., right  prefrontal  EP  components  were reliably
 seen with  left  movements to  the 2Qo and 40° left
 targets, whereas  the left  prefrontal components were
 seen with  right movements elicited  by the corre-
 sponding right targets (Fig. 1). Amplitudes  were sig-
 nificantly larger  for  more  peripheral  targets  (Fie
 2B and 2C).
    Precentral: As many as eight peaks could be iden-
tified in averages of precentral events time-locked to
onset of the test target. Unlike prefrontal components,
these could be identified in  averages associated with
each test light  presentation.  The amplitude of these
components, some of which preceded (Nl and  Pi)
and the remainder which followed saccadic eye move-
ments, did  not vary systematically with test lights in
different locations. Mean amplitudes of the first four
components (Nl,  PI, N2, and P2) associated with
each test light location were not significantly different
(Fig. 2D). These results were not surprising in light of

-------
Oculomotor ERP Components
                                             145
                  HORIZ. EOG
   4O°L
 Table  1. Mean Latencies3 (msec) of Averaged
    Saccadic Eye Movements and Prefrontal
    Evoked-Potential Components for Two
                    Monkeys
    4O*«
 Fig.  1. Averaged horizontal EOG and left and right
 prefrontal  ECoG  events  time-locked  to test light
 onset.  Individual traces represent the average  of 40
 responses evoked  by each test light during a  single
 testing session  with  Monkey 297, Left (L) and right
 {R}  test  lights are indicated to the left of the tracei.
 The  time base equals 2048 msec and calibration bars
 show 100 i*V.  The minimum test light duration  was
 1.5 sec, indicated by the horizontal bar at the bottom
 of the figure.
Averaged
waveform b
Monkey 297
HEOG
RFpNI
P1
N2
P2
LFp N1
P1
N2
P2
Monkey 293
HEOG
RFp N1
P1
N2
?2
LFp P1
N1
P2
N2
Test light
40° L

257.6
143.6
176.8
219.2
499.6





292.4
85.6
161.2
221.6
401.2




20° L

240.8
144.0
177.6
221.2
509.2





318.4
106.8
158.4
210.4
382.8




0°











293.2








20° R

285.2









296.0








40° R

297.2




60.0
103.2
187.1
339.2

312.4




22.4
101.2
316.8
747.6
aLatencies were obtained for only those tett lights that pro-
 duced consistently identifiable waveforms. The mean latencies
 are comprised of 8 to IS measures of the time intervals  be-
 tween test light onset and the initial averaged EOG deflection
 (for saccadic tye movements),  and  the peak amplitude of
 each of the averaged EP components.
* HEOG • horizontal electro-oculognm; RFp • right perfrontal
  cortex; LFp - left prefrontat cortex.

 the dear implication of precentral areas in the initia-
 tion of  skeletal motor responses.  Since  the  motor
 response, namely  the  lever press, Is  the  same  re-
 gardless  of which test  light is presented, no  differ-
 ences  in  averaged  potential  would  be expected.

      Occipital:  Evoked-potential components  from
  occipital cortex were observed 40 to 1500 msec after
  test light onset. The rise of an initial negative com-
  ponent  was observed some 40 to 60 msec after target
 onset, and it  attained peak amplitude at about ISO
  to 200  msec.  This was followed by somewhat more
  variable positive and negative waves with peak laten-
  cies of about  300 to 900 msec. The amplitudes of the
  initial negative wave (Nl) and the final positive wave
  (P2)were greater for the centrally located test targets,
  i.e., the 0  (fixation) test target and the 20° left  and
  the 20°  right test lights  (Fig. 2E). However, the differ-
  ences  in the  mean amplitudes of these  components
  were not statistically significant.

-------
  146
                                                                                   Rosen et al.
  U

  u

    J

  -0.2

  u  ;



  ISO

  100

  H

    I
tso
_J

<25


1°


  100

   A

  SO

  26



 200

 ISO

 100
#297
0.4

0.2

 0

0.2
                     ISO

                     100

                      so

                       I


                    I

                      75

                      SO

                      25

                       0



                    )
                     300

                     22S

                     ISO


                      75


                    E

                      SO

                      N

                      40

                      20
        K
   •A
                      TEST LIGHT
 Fig. 2. Mean amplitudes of averaged EOG and ECoG
 components as functions of test light location. Data
 for  Monkeys 297 and  293 are shown.  Amplitude
 measures are of (A) horizontal EOG, (B) right pre-
 frontal,  (C) left prefrontal, (D) left  precentral, and
 (E) left occipital averaged responses. Mean values rep-
 resent measurements of components seen in eight to
 fifteen 40-trial averages obtained during consecutive
days of testing.  Amplitudes are in microvolts except
for the EOG, which is in millivolts.
 Operantly conditioned slow potential shifts

     Reinforcement of unilateral surface negative SP
 shifts in prefrontal cortex resulted in an increase in
 the incidence and amplitude of these spontaneously
 occurring cortical events. Reinforced shifts  of 50 to
 100 juV amplitude and 2.5  to 3.0 sec duration were
 accompanied by a series of contralateral saccades and
 sustained conjugate eye deviations  until reward was
 delivered (Fig. 3). Delivery of reward resulted in  rapid
 centering eye movements, but  no marked change in
 the cortical  SP.  Contralateral  prefrontal SP shifts
 were  completely  unrelated  to eye  movements or
 reward delivery.

 Discussion

     This  investigation  provides  clear  evidence  of
 electrocortical  potentials   in  prefrontal  cortex  of
 monkeys  that  encode  amplitude  and  direction  of
 stimulus-triggered  saccadic  eye movements. These
 potentials are time-locked to the onset of a visual
 stimulus that  the animal  is required  to fixate upon
 and appear in the hemisphere contralateral to the
 direction  of  the  saccade. Moreover, components of
 these potentials both precede and arise coincidentally
 with the initiation of horizontal saccades. The dura-
 tion of the preceding components is less than  100
 msec,  whereas that  of the following  components
 often exceeds several seconds. The recording methods
 employed do not permit specification with any degree
 of certainty that the potentials arise from the posteri-
 or  half of the principal sulcus, or more posteriorly
 from the traditionally  defined  FEF (area  8)  as a
 consequence of volume conduction to the recording
 electrodes in  the  principal sulcus.  However, these
 potentials are clearly specific to the prefrontal region
 since averaged recordings  of precentral and occipital
 cortical activity were  not  consistently related to eye
 movements. Moreover, it is unlikely that  the observed
 potentials reflect eye movement artifacts. Clearly the
 issue pertains not to  the components preceding the
 eye  movements  but  to those arising concomitantly
 with eye movements. The  finding that the latter com-
 ponents appear only in the hemisphere  contralateral
 to the direction of the saccade is inconsistent with an
 eye  movement artifact interpretation since saccadic
 movements involve both eyes and  significant poten-
 tials of opposite  polarity  would  be  expected  at
 homologous locations in the other hemisphere. These
 were not  observed. Furthermore, in  the  procedure
 involving operant conditioning of unilateral prefrontal
 SP  shifts, concomitant contralateral  saccades were
observed during the increase of cortical surface nega-
tivity, but subsequent centering eye movements were
completely independent of cortical SP shifts, a find-
 ing that could not be explained if rotation of the
 corneofundal dipole was the source of the potentials.

    With regard to the functional significance of pre-
 frontal potentials  related to eye movements, it would

-------
Oculomotor ERP Components

                   1                    2
                                           147
       ^^A/$\!^^
                                                                     4. MEG
                          J
                          1sec
Fig. 3. An operantly conditioned surface-negative cortical steady potential shift in left prefrontal (LFp) cortex
and concomitant right (Rt) oculomotor responses indicated in both horizontal EOG and drawings of stop-action
videotape recording of eye position.  The vertical bar indicates time of food reward. Vertical lines indicate times
during bioelectric recordings when eye positions were photographed. Note the lack of response in the homo-
logous right prefrontal (RFp) ECoG recording.
appear that those  accompanying and following sac-
cades reflect the sustained activity  of FEF neurons,
which reportedly discharge during saccades and steady
fixations (Bizzi 1968, Mohler et al.  1973). Insofar as
the components preceding eye movements are con-
cerned, the finding that the amplitude of these com-
ponents increases with more peripherally placed visual
targets argues  against the notion that they  simply
reflect a visual  sensory input to the prefrontal area via
known cortical-cortical afferents from areas 18 and
19  (Jones  and Powell  1970).  Mohler et al.  (1973)
have reported  that FEF  cells  with visual  receptive
fields have  extremely large fields, and individual cells
tend to discharge minimumly,  not  maximally, when
spots of light  are presented in the  periphery of the
field. It is possible, however, that prefrontal  poten-
tials preceding  saccadic eye movements reflect an en-
hanced sensory response  of FEF neurons  that has
been observed  when the visual target for an impend-
ing  saccade falls within the receptive fields  of a sub-
population  of  FEF cells (Wurtz and  Mohler  1976).

    Alternative explanations include  the possibility
that presaccadic potentials reflect  attempted head
movements and possible feedback from neck muscles.
These views are suggested by the findings of FEF cells
that discharge prior to head turning (Bizzi and Schiller
1970) and FEF neurons in the cat that respond, with
short  latencies, to stimulation of the neck  muscles
(Dubrovsky  and  Barbas  1975,  Mandl and  Guitton
1975). Another possibility is that the observed poten-
tials reflect a corollary discharge mechanism  (Teuber
1964)  or  a  motor command system for voluntary
saccadic movements, perhaps related to search behav-
ior or attention.
    It should be noted that the present data are not
the only indication of cortical events  preceding eye
movements. Lynch et  al. (1977) have recently found
cells in parietal cortex (area 7) that discharge prior to
stimulus-triggered  saccades and  appear to encode
direction of movement, and Schlaget al. (1971) have
found cells  in the internal medullary lamina of the
cat that similarly discharge preceding eye movements.
This region  is known  to receive input from the FEF
(Orem and Schlag 1971).

-------
148
                                    Rosen et al.
    In conclusion,  the results of  this investigation
strongly suggest that significant components of corti-
cal  potentials evoked by sensory  stimuli  can reflect
oculomotor processes,  particularly if attention  to
these  stimuli by the subject involves saccadic eye
movements and eye  fixations.

Summary

    In order to examine the extent to which poten-
tials evoked in monkeys' dorsolateral prefrontal cortex
by  visual stimuli reflect oculomotor functions, mon-
keys with chronically Implanted nonpolarizable elec-
trodes in prefrontal, precentral, and occipital cortex,
and also subcutaneously across the eyes, were trained
on  a  reaction time  task for liquid rewards. The task
required  that the monkey  detect the dimming of a
centrally located fixation light and then fixate upon
one of five test lights in order to  detect its dimming.
The  test  lights  were  perimetrically  presented  in
random order at 20° and 40C to the left and right of
center. Computer averages of cortical potentials time-
locked to 40 presentations of each test light, within a
daily  testing session, revealed prefrontal components
that preceded and followed saccadic eye movements.
These components were  of maximal  amplitude for
the most peripherally located test stimuli in the con-
tralateral visual field. By contrast, the largest occipital
evoked potential components were observed with cen-
tral stimuli,  and precentral potentials showed no vari-
ations in amplitude with different test light presenta-
tions. The data indicate the evoked potentials in pre-
frontal cortex elicited by peripherally presented visual
stimuli, upon which an animal must fixate, reflect
oculomotor  processes related to  saccadic  eye move-
ments.

Acknowledgments

    This research was supported by National Science
Foundation  Research Grant GB 35735X1 to Dr. John
S. Stamm. The authors wish to  thank Richard Reeder
for technical assistance, and Michele Parker for help
In preparation of the manuscript.

-------
MOTOR CONTROL  REFERENCES
Abraham P., Docherty, T. and Spencer,  S. Cerebral
    event-related potentials and spinal cord stimula-
    tion. Electroenceph.  clin.  Neurophysiol.,  1978,
    44:132.

Amassian, V. E. and Weiner, H. Monosynaptic and
    polysynaptic activation of pyramidal tract neu-
    rons by thalamic stimulations. In D. P. Purpura
    and M. D. Yahr (Eds.), The thalamus. Columbia
    Univ. Press, New York, 1966: 255-282.

Angaut, P.  The ascending projections of the nucleus
    interpositus of the cat cerebellum:  An experi-
    mental study using silver impregnation methods.
    Brain Res., 1970, 24:377-394.

Arezzo, J. and Vaughan, H. G., Jr. Cortical potentials
    association with voluntary  movements  in  the
    monkey. Brain Res., 1975, 88:99-104.

Bates, J. A. V. Electrical activity of the cortex accom-
    panying movement.  J.  Physiol.,  Lond.,  1951,
    113:240-257.

Becker, W., Hoehne, O., Iwase, K., and Kornhuber, H.
    H.  Bereitschaftspotential,  pramotorische  Posi-
    tivierung und andere Hirnpotentiale  bei sakkadi-
    chen  Augenbegwegungen. Vision Res.,  1972,
    12:421-436.

Becker, W,, Iwase, K., Jurgens, R., and Kornhuber, H,
    H.   Bereitschaftspotential  preceding  voluntary
    slow and rapid hand movements. In W. C. McCal-
    lum and J. R. Knott(Eds.),The responsive brain.
    Wright, Bristol, 1976: 99-102.

Bernstein,  N.  The coordination  and regulation of
    movements. Pergamon  Press, Elmsford,  1967.

Bizzi, E. Discharge of frontal eye field neurons during
    saccadic and following eye movements in unanes-
    thetized monkeys. Expl. Brain Res., 1968,6:69-
    80.

Bizzi, E. and Schiller, P.  H. Single unit activity in the
    frontal eye fields of unanesthetized monkeys
    during eye and head movement. Expl. Brain Res.,
    1970,  10:151-158.

Bond, H. W. and Ho, P.  Solid miniature silver-silver
    chloride electrodes   for  chronic implantation.
    Electroenceph.   clin.  Neurophysiol.,   1970,
    28:206-208.
Broadbent, D. O. Decision and stress. Academic Press,
    Lond., 1971.

Broughton,  R., Regis, H. and Gastaut, H. Modifica-
    tions du potentiel somesthesique evoque pendant
    une bouffee  de  rythme mu et/ou la contraction
    des poings. Rev. Neurol.,  1964, 111:331-332.

Buser,  P. Higher functions of the nervous system.
    Ann. Rev. Physiol., 1976,38:217-245.

Butler, S.R.  and Glass, A.  Asymmetries  in the CNV
    over left and right  hemispheres while subjects
    await numeric information. Biol. Psychol., 1974,
    2:1-16.

Carpenter, M. B. Ventral tier thalamic nuclei. In  D.
    Williams  (Ed.),  Modern trends in neurology.
    Appleton-Century-Crofts, New York, 1967:1-20.

Carpenter, M. B., Nakano, K. and Kim, R. Nigrothala-
    mic projections in the monkey demonstrated by
    autoradiographic technics. J. com. Neurol., 1976,
    4:401-416.

Caspers, H. and Speckmann, E. J. Cerebral pO2 and
    pH: Changes during convulsive activity and their
    significance for  spontaneous  arrest  of seizures.
    Epilepsia, 1972,  13:699-725.

Caspers, H.  and  Speckmann, E.  J. DC-potentials of
    the cerebral  cortex.  Rev.  Physiol.  Biochem.
    Pharmacol. (in press).


Castaigne, P., Cathala, H., Pierrot-Desseiligny, E. and
    Bussel, E. A preliminary study of the evoked re-
    sponse at the time  of voluntary movement in
    normal man. Electroenceph.  clin. Neurophysiol.,
    1969, 27: Extra number 684.

Chan-Palay,  V. Cerebellar  dentate nucleus. Springer-
    Verlag, Berlin-Heidelberg, 1977.

Chatrian, G.E., Peterson, M. C. and Lazarte, J. A. The
    blocking of the rolandic wicket rhythm and some
    central  changes related to movement.  Electro-
    enceph. clin. Neurophysiol.,  1959,  11:497-510.

Chow, K.L.  and Pribram, K. H. Cortical projection of
    the thalamic ventrolateral nuclear group  in mon-
    keys. J. comp. Neurol., 1956, 104:57-75.

-------
  150
                       Motor Control References
 Cohen, J. Very slow brain potentials related to expec-
     tancy: The CNV. In E. Don chin and D. B. Linds-
     ley (Eds.), Average evoked potentials. Methods,
     results and evaluations, NASA SP-191, U.  S.
     Govt.  Printing  Office, Washington, D.C., 1969,
     143-198.

 Cook, A.W. Electrical stimulation in multiple sclerosis.
     Hospital Practice, 1976, 11:1-58.

 Cooper, R., Winter, A. L., Crow, H. J. and Walter, W.
     Comparison of subcortical, cortical  and  scalp
     activity using chronically indwelling electrodes in
     man.  Electroenceph. clin. Neurophysiol., 1965,
     18:217-228.

 Cooper,  R.,  Papakostopoulos,  D.  and Crow, H.  J.
     Changes of cortical oxygen associated with motor
     and cognitive function in man. In M. Harper,  B.
     Jennett,  D. Miller and  J. Rowan (Eds.), Blood
     flow  and  metabolism  in  the  brain.  Churchill,
     Livingstone, 1975.


 Cooper, R., McCallurn, W. C, Newton, P., Papakos-
     topoulos,  D., Pocock,  P. V. and Warren, W.  J.
     Cortical potentials associated with the detection
     of  visual  events. Science, 1977,   196:74-77.


 Coquery, J. M. Changes in somesthetic evoked poten-
     tials during movement. Brain Res., 1971, 31:375.

 Coquery,  J. M., Coulmance, M. and Leron,  M. C.
     Modifications des  potentiels evoques corticaux
     somesthesiques  durant des mouvements actifs et
     passifs chez rhomme. Electroenceph. clin. Neuro-
     physiol., 1972,33:269-276.

 Corby, J. C., Roth, W. C. and Kopell, B. S. Prevalence
     and methods of  control of the cephalic skin
     potential  EEC artifact. Psychophysiology, 1974,
     11:350-360.

 Coulter, J. D. Sensory transmission through lemniscal
     pathway during voluntary movement  in the cat.
     J. Neurophysiol., 1974, 37:831-845.

 Cowan, W. M., Gottlieb,  D.  I.,  Hendrickson, A. E.,
     Price,  J. L. and Woolsey, T. A. The  autoradio-
     graphic demonstration of axonal connections in
     the central nervous system. Brain  Res., 1972,
     37:21-51.


Craik,  K. The nature of explanation. Cambridge Univ.
     Press, Cambridge, 1943.

Crosby, E. C., Yoss,  R. E. and Henderson, J. W. The
     mammalian mid brain and isthmus regions. Part II,
    The fiber  connections.  D.  The pattern for eye
     movements of the frontal eye field  and the dis-
    charge of  specific portions of this field to and
     through midbrain levels. J. comp. Neurol., 1952,
     97:357-383.
  Crossland, W. J., Cowan, W. M. and Kelly, J. P. Obser-
     vations on the transport of radioactively labeled
     proteins in the visual system of the chick. Brain
     Res., 1973,56:77-105.

  Davis, H. Space and time in the central nervous system.
     Electroenceph. clin. Neurophysiol., 1956  8'185-
      191.

  Deecke,  L., Scheid, P. and Kornhuber, H. H. Distribu-
     tion of readiness potential, pre-motion  positivity,
     and  motor potential of the human cerebral cortex'
     preceding  voluntary  finger movements.  Expl
     Brain Res., 1969,7:158-168.

  Deecke,  L., Becker, W., Grozinger, B., Schied, P. and
     Kornhuber, H. H. Human brain potentials preced-
     ing  voluntary limb movements.  Electroenceph.
     clin.  Neurophysiol.,  1973,  suppl.   33:67-94.

  Deecke, L., Grozinger, B. and Kornhuber, H.  H. Vol-
     untary finger movement in man. Cerebral poten-
     tials  and theory.  Biol. Cybernetics,  1976, 23-99-
     119.

  Donchin, E., Otto, D., Gerbrandt, L. K. and Pribram,
     K.  H. While  a  monkey  waits:  Electrocortical
     events recorded  during the foreperiod of a reac-
     tion  time  study. Electroenceph.  clin.  Neuro-
     physiol., 1971,31:115-127.

  Donchin, E., Otto, D., Gerbrandt, L. K. and Pribram,
     K. Studies in  the physiology of CNV. In W. c!
     McCallurn and J. R. Knott (Eds.), Electroenceph.
     clin.   Neurophysiol.,  1973, suppl.  33:257-261.

  Droz, B.  and Koenig, H. L. Localization of  protein in
     neurons. In A. Lajtha (Ed.), Protein metabolism
     of the  nervous   system.  Plenum,  New  York
     1970:93-108.

 Dubrovsky, B.  Cerebral  and  cerebellar projections
     from neck muscles in cats.  Proceedings of the
     International  Union  of Physiological  Sciences
     New Delhi, 1974,  11:545.

 Dubrovsky, B.  and Barbas. H.  Frontal projections of
     dorsal  neck  and extraocular  muscles.  Exper
     Neurol., 1977, 55:680-693.

 Dubrovsky, B. and Garcia-Rill, E. Convergence of tec-
     tal and visual input in cat motorsensory cortex
     Expl. Neurol., 1971, 33:475-484.

Dubrovsky, B. and Garcia-Rill, E. Role of dorsal col-
    umns in sequential motor  acts requiring precise
    forelimb projection.   Expl.  Brain  Res.  1973
    18:165-177.

Dubrovsky, B.,  Davelaar, E. and Garcia-Rill, E. The
    role of dorsal columns in serial order acts.  Expl
    Neurol., 1971,33:92-102.

Dubrovsky, B., Garcia-Rill, E. and Surkes, M. Effects
    of discrete  precruciate cortex lesions on motor
    behavior. Brain Res., 1974, 82:328-333.

-------
Motor Control References
                                             151
Evarts, E. V. Relation of discharge frequency to con-
    duction velocity  in pyramidal  tract neurons.
    J. Neurophysiol., 1965, 28:216-228.

Evarts, E. V. Pyramidal tract activity  associated with
    a conditoned hand  movement in the monkey.
    J. Neurophysiol., 1966, 29:1011-1027.

Evarts, E. V. Relation of pyramidal tract activity to
    force  exerted   during  voluntary  movement.
    J. Neurophysiol., 1968,31:14-27.

Evarts, E. V. Activity of pyramidal tract neurons dur-
    ing postural  fixation. J. Neurophysiol.,  1969,
    32:375-385.

Evarts, E. V. Activity of ventralis lateralis neurons
    prior to movement in  the monkey. Physiologist,
    1970,13:191.


Evarts, E. V. and  Thach, N. T. Motor mechanisms of
    the CNS: Cerebrocerebellar interrelations. Ann.
    Rev. Physiol., 1969,  21:451-498.


Ferrier,  D. Experiments on the brain of monkeys.
    Proc. R. Soc. B., 1975, 23:409-430.

Ford, J.  M., MacPherson, L.  and Kopell,  B. J. Differ-
    ences in readiness potential associated with push-
    button construction. Psychophysiol. 1972, 9:564-
    567.

Freeman, W. J. Mass action in the nervous system.
    Academic Press, New York,  1975.

Frigyesi, T. L., Rinvik, E. and Yahr, M.  D. (Eds.),
    Corticothalamic  projections  and sensorimotor
    activities. Raven Press, New York, 1972.

Garcia-Rill,  E.  and Dubrovsky.  B. Organization  of
    visual input  in cat motorsensory cortex. Expl.
    Neurol., 1971,33:597-606.

Garcia-Rill,  E.  and  Dubrovsky,  B.  Topographical
    organization  of visual input to precruciate cortex
    of cat. Brain Res., 1973,56:151-163.

Garcia-Rill, E. and Dubrovsky, B. Responses of motor
    cortex cells to visual stimuli. Brain  Res., 1974,
    82:185-194.

Gastaut, H. Etude electrocorticographique de la react-
    ivite des   rythmes   rolandiques. Rev. Neurol.,
    1952,87:176-182.

Gazzaniga, M. S.  and Hiliyard, S. A. Attention mech-
    anisms following brain bisection. In S. Kornblum
    (Ed.), Attention and performance IV. Academic
    Press, New York, 1973:221-238.

Gerbrandt, L.  K. Analysis  of movement  potential
    components. In J. E. Desmedt (Ed.), Prog. clin.
    Neurophysiol.,  1977, 1:174-189.
Gerbrandt, L. K., Goff, W. R. and Smith, D. B. Distri-
    bution of the  human average movement poten-
    tial.  Electroenceph.  clin. Neurophysiol., 1973,
    34:461-474.

Ghez, C.  and Lenzi,  G.  L.  Modulation of sensory
    transmission in cat lemniscal system during volun-
    tary  movement. Pflugers Arch., 1971,323:273-
    278.

Ghez, C.  and Pisa,  M.  Inhibition of afferent transmis-
    sion  in cuneate nucleus  during voluntary move-
    ment in the cat.  Brain Res., 1972, 40:145-151.

Giblin, D. R. Somatosensory  evoked potentials in
    healthy  subjects and in patients with lesions of
    the nervous system. Ann. N.Y. Acad. Sci., 1964,
    112:93-142.

Gilden, L., Vaughan, H.G., Jr. and  Costa, L. D.  Sum-
    mated human EEG potentials with voluntary
    movement. Electroenceph.  clin. Neurophysiol.,
    1966,20:433-438.


Goldring, S. and Ratcheson, R. Human motor cortex:
    Sensory input data from single neuron recordings.
    Science, 1972, 175:1493-1495.

Gottlieb, G.  L., Agarwal,  G. C. and Stark, L. Interac-
    tions between voluntary and postural mechanisms
    of the human motor system.  J. Neurophysiol.,
    1970,33:365-381.


Granit, R. The functional role of muscle spindles:
    Facts and speculation. Brain, 1975, 98:540-558.


Groll-Knapp, E., Ganglberger, J. A. and Haider, M.
    Voluntary movement related slow  potentials in
    cortex and thalamus in  man. In J. E. Desmedt
    (Ed.), Prog. clin. Neurophysiol., 1977, 1:164-173.

Hazemann, P. and Lille, F. Somatosensory  evoked
    potentials, attention  and voluntary self-paced
    movements. In W. C. McCallum and J. R. Knott
    (Eds.),  The responsive  brain.  Wright,  Bristol,
    1976:111-113.
Hazemann, P., Audin, G. and Lille, F. Effect of volun-
    tary self-paced  movements  upon auditory and
    Somatosensory evoked potentials in man. Electro-
    enceph. clin. Neurophysiol.,  1975, 39:247-254.

Hiliyard, S. A. and Galambos, R. Eye movement arti-
    fact in the CNV. Electroenceph. clin. Neurophys-
    iol., 1970,28:173-182.

Hoist, E. von and Mittelstaedt, H. Das reaffevenzprin-
    zip  (Wechselwirkungen  zwischen zentralnerven-
    system  und Peripherie).  Natur wiss.  1950, 37:
    464-476.

-------
  152
                       Motor Control References
  Illis,  L.S., Sedgwick, E.M.  and Oygar, A.E. Dorsal
      column  stimulation  in  the  rehabilitation  of
      patients with multiple  sclerosis.  Lancet,  1976:
      1383-1386.

  Ingvar, D.H. and Philipson, L. Distribution of cerebral
     blood flow in the dominant hemisphere during
     motor ideation  and  motor performance. Ann.
     Neurol., 1977, 2:230-237.

  Jasper, H. H. and Bertrand, G. Thalamic units involved
     in somatic sensation and voluntary and involun-
     tary movements in man. In D. Purpura and M.
     Yahr (Eds.), The Thalamus. Columbia Univ. Press,
     New York, 1966:365-390.

  Jones, E. G. and Powell, T. P. S. An anatomical study
     of converging sensory pathways within the cere-
     bral cortex of the monkey. Brain, 1970,93:793-
     820.

 Jones, J. G. and Beck, C. H. Motor potentials and tim-
     ing of  muscular activity.  Electroenceph.  clin.
     Neurophysiol., 1975,38:273-279.

 Kalil,  K.  Topographic  organization  of projections
     from the dentate and interpositus nuclei in the
     rhesus  monkey:  An  autoradiograpnic   study.
     Society  for Neuroscience, 7th Annual Meeting,
     1977:56.

 Katchalsky,  A.  K., Rowland, V. and Blumental, R.
     Dynamic patterns of brain cell assemblies. Neuro-
     sciences, 1974, 12:1-187.

 Kievet, J. and Kuypers, H. G. J. M. Fastigial cerebellar
     projections to the  ventrolateral nucleus of the
     thalamus and the organization of the descend-
     ing pathways. In T. Frigyes, E. Rinvik and M.D.
     Yahr  (Eds.). Corticothalamic  projections and
     sensorimotor activities. Raven Press, New York,
     1972:91-124.

 Kievet, J. and Kuypers, H. G. J.  M. Subcortical affer-
     ents   to  the frontal lobe in the rhesus monkey
     studied by means of retrograde horseradish perox-
     idase transport. Brain Res., 1975, 85:261-267.

Knott, J.R. and Irwin, D.A. Anxiety, stress, and the
     contingent   negative  variation.   Arch.   Gen.
     Psychiat., 1973, 29:538-541.


Kornhuber, H. H. Motor functions of cerebellum and
     basal ganglia: The cerebello-cortical saccadic (bal-
     listic) clock, the cerebello-nuclear hold  regulator
     and the basal ganglia ramp (voluntary speed
     smooth movement) generator. Cybernetics, 1971,
     8:157-162.

Kornhuber, H.  H. Cerebral cortex, cerebellum and
     basal  ganglia: An introduction to  their  motor
     functions. In  F. O.  Schmitt (Ed.), The Neuro-
     •cience*. Ill Study  Program. MIT Press, Cam-
     bridge, 1974:267-280.
  Kornhuber, H. H. and Deecke, L. Hirnpotentialan-
      derungen beim Menschen vor und nach Willkurb-
      ewegungen, dargestellt mit Mangetbandspeicher-
      ung und Ruckwartsanalyse. Pflugers Arch  ges
      Physiol., 1964,281: R 52.

  Kornhuber: H. H. and Deecke, L. Hirnpotentialander-
      ungen  bei   Willkurbewegungen  und  passiven
      Bewegungen des Menschen: Bereitschaftspotential
      and reafferente  Potentiale.  Pflugers Arch  ees
      Physiol. 1965,284:1-17.                 '     '

  Kuo, J. S. and Carpenter, M. B. Organization of palli-
     dothalamic projections  in  the  rhesus monkey
     J. comp. Neurol., 1973, 151:201-236.

 Kutas,  M.  and Donchin,  E. Studies  of squeezing:
     Handedness,  responding hand,  response  force
     and asymmetry of readiness potential Science'
     1974, 186:545-548.

 Kutas,  M.  and Donchin, E. The effect of handedness,
     of responding hand, and of response force on the'
     contralateral dominance of the  readiness poten-
     tial. In J. E. Desmedt (Ed.), Prog. clin. Neurophy-
     siol., 1977, 1:189-210.

 Kuypers, H. G. J. M. The organization of the motor
     system. Int. J. Neurol., 1963, 4:78-99.

 Lacey,  J. I. and Lacey, B. C. Some autonomic-central
     nervous  system interrelationships.  In  P.  Black
     (Ed.),  Physiological correlates of emotion. Aca-
     demic Press, New York,  1970:205-228.

 Lasek,  R. J. Protein transport in neurons  Int Rev
     Neurobiol., 1970, 13:389-424.

 Lashley, K. S. The problem of serial order in behavior
     In  L. A. Jeffress (Ed.), Cerebral mechanisms in
     behavior. Wiley,  New  York,  1951:112-136.

 Lee, R. C. and White, D. G, Modification of the human
     somatosensory evoked response during voluntary
     movement.  Electroenceph.  clin. Neurophysiol
     1974,36:53-62.                             "

 Lemon, R.N. and Porter, R. Afferent input to move-
     ment-related  precentral  neurones  in  conscious
     monkeys  Proc. R.  Soc.  Lond. B., 1976  194-
     313-339.

 Lemon, R.N., Hanby, J.A. and Porter, R. Relation-
    ship between the activity of precentral neurones
    during active and passive movements in conscious
    monkeys. Proc. R.  Soc.  Lond. B., 1976  194-
    341-373.

Loveless, N. E. Distribution of response to non-signal
    stimuli. In W.C. McCallum and J. R. Knott (Eds.),
    The responsive brain. Wright, Bristol, 1976:26-29!
Loveless, N. E. and Sanford, A. J. The CNV baseline:
    Considerations of internal consistency  of  data.
    Electroenceph. clin. Neurophysiol., 1973, suppL
    •5 3119~23.

-------
Motor Control References
                                            153
Low, M. D., Borda, R. P., Frost, J. D., Jr., and Kella-
    way,  P.  Surface  negative,  slow-potential  shift
    associated with conditioning in man. Neurology,
    Minneap., 1966, 16:771-782.

Low, M. D., Wade, J. A. and Fox, M. Electroencepha-
    lographic localization of conative aspects of lan-
    guage production  in the human brain. In W. C.
    McCallum and J. R. Knott (Eds.), The responsive
    brain. Wright, Bristol, 1976:165-168.

Lynch, J.C.,Mountcastle,V.B.,Talbot,W.H. and Yin,
    T.C.T. Parietal  lobe mechanisms  for directed
    visual  attention.  J. Neurophysiol,  1977, 40:
    362-389.

Magendie, R. An  elementary compendium of Physi-
    ology. James Webster, Philadelphia (Transl. by E.
    Milligan), 1824.

Mandl, G.  andGuitton, D. Eye movements and frontal
    eye fields: Electrical stimulation and unit record-
    ing in the awake  cat. Neuroscience Abs., 1975,
    1:230.

Marczysnski,  T. J. Electro physiological  correlates of
    positive reinforcement:  post-reinforcement syn-
    chronization,  modulation of sensory  input, and
    steady potentials. In A. G.  Karcimar and  J. C.
    Eccles (Eds.), Brain and human behavior. Spring-
    er-Verlag, New York, 1972:303-321.

Marsden, C. D., Morton, P. A. and Morton, H. B. Servo
    action in human  voluntary movement. Nature,
    London, 1972,238:140-143.

Marsh, G.R. and  Thompson, L.W. Effects of verbal
    and  non-verbal psychological set on hemispheric
    asymmetries  in the CNv,  Electroenceph,  clin,
    Neurophysiol., 1973, suppl. 33:195-200.

Matthews, P. B. C. Mammalian muscle receptors and
    their  central actions.  Arnold, London, 1972.

McAdam, D. W. Slow potential changes recorded from
    human brain  during learning of a temporal inter-
    val. Piychon,  Sci,, 1966, 6:435-436.

McAdam, D. W, Physiological mechanisms: A review.
    Electroenceph. clin. Neurophysiol., 1973, suppl.
    33:79-86.

McCallum, W. C. Brain ilow potential changes and
    motor response in a vigilance situation. In W. C,
    McCallum and J. R. Knott (Eds.), The responsive
    brain. Wright, Bristol, 1976:46-50,

McCallum, W. C. and Cummins, B. The effects of brain
    lesions on the contingent negative variation in
    neuroiurgical  patients.  Electroenceph.  clin.
    Neurophysiol., 1973, 35:449-456.

McCallum, W. C. and Knott, J. R. (Eds.),  Event related
    slow  potentials of the brain: Their relations to
    behavior. Electroenceph.  clin.  Neurophysiol.,
    1973 wppl. 33:390.
McCallum, W.C. and Knott, J.R. (Eds.). The respon-
    sive brain. 1976, Wright, Bristol.

McCallum, W. C. and Papakostopoulos, D. The effects
    of sustained motor activity on the contingent
    negative  variation. Electroenceph.  clin. Neuro-
    physiol.,  1972,33:446.

McCallum, W. C., and Walter, W. G. The effects of
    attention and distraction on the contingent nega-
    tive variation in normal and neurotic subjects.
    Electroenceph.   clin.   Neurophysiol.,   1968,
    25:319-329.

McCallum, W. C., Papakostopoulos, E., Gombi, R.,
    Winter, A. L,  Cooper,  R.  and Griffith,  H. B.
    Event related slow potentials in human brain
    stem.  Nature, 1973, 242:465-467.

McCallum, W. C., Papakostopoulos, D. and Griffith,
    H. B.  Distribution of CNV and other slow poten-
    tial changes in human brainstem structures. In
    W.C.  McCallum and J.R.   Knott  (Eds.),  The
    responsive brain. Wright, Bristol,  1976:205-211.

McKay, D. M. Cerebral organization and the conscious
    control of action. In J. C. Eccles (Ed.), Brain and
    conscious experience. Springer-Verlag, New York,
    1966:422-445.

Megirian,  D.,  Buresova, O., Bures, J. and Dimond, S.
    ELectTOphysiologkal  correlates of discrete fore-
    limb  movements  in  rats.  Electroenceph.  clin.
    Neurophysiol., 1974, 36:131-139.

Mehler, W. R. Idea of anew anatomy of the thalamus.
    J. Psychiat. Res., 1971, 3:203-217.

Mettler, F. A. Extracortical connections of the primate
    frontal cerebral cortex.  I. Thalamo-cortical  con-
    nections. J. comp. NeuroL, 1947,  86:95.

Milner-Brown, H.S., Stein, R. B. and Lee, R. G.  Syn-
    chronization of human motor units: Possible roles
    of exercise  and supraspinal  reflexes. Electro-
    enceph.  clin. Neurophysiol.,  1975, 38:245-254.

Mohler, C. W., Goldberg, M. E.  and Wurtz, R. H. Vis-
    ual receptive fields of frontal eye field neurons.
    Brain Res.,  1973,61:385-389.


Mountcutle, V. B., Lynch,  J. C., Georgopoulos, A.,
    Sakata, H. and Acuna.C, Posterior parietal associ-
    ation cortex of the monkey: Command functions
    for operations  within  extrapenonal  space.  J.
    Neurophysiol., 1975,38:871-908.

Murphy, J. T., Wong, V. C. and Kwan, H, C. Afferent-
    efferent  linkages in motor cortex for single  fore-
    limb  muscles,  J, Neurophysiol., 1975, 38:990-
     1014.


 Olszewiki, J. The thalamus of Macaca Mulatta. Karger,
     Basel, 1952.

-------
  154
                      Motor Control References
 Orem, J. and Schlag, J. Direct projections from cat
     frontal eye field to internal medullary lamina of
     the  thalamus. Expl. Neurol., 1971, 33:509-517.

 Otto, D. A. and Leifer, L. J,  Effect of varying the
     magnitude, duration and speed of motor response
     on the CNV. Electroenceph. clin, Neurophysiol,,
     1973a,34:695P.

 Otto, D. A. and Leifer, L.  J. The effect of modifying
     response and performance feedback  parameters
     on  the CNV in  humans.  Electroenceph. clin.
     Neurophysiol., 1973b, suppl. 33:29-37.

 Otto,  D. A., Benignus, V. A.,  Ryan, L. and Leifer,
     L.  J.  Slow potential  components of stimulus,
     response, and preparatory processes in man. In
     J. E.  Desmedt  (Ed.),  Prog. clin. Neurophysiol.,
     1977, 1:211-230.

 Paillard,  J. Reflexes et regulations d'origine proprio-
     ceptive chez  1'homme. Arnette,  Paris,  1955.

 Papakostopoulos, D. and Cooper, R. The contingent
     negative variation and the excitability of the spin-
     al monosynaptic reflex. J.  Neurol. Neurosurg.
     Psychiat, 1973,36:1003-1010.

 Papakostopoulos, D. and Cooper, R. Brain, spinal cord
     and  autonomic changes before, during and after a
     planned motor action in man. In W. C. McCallum
     and  J. R.  Knott  (Eds.), The responsive  brain.
     Wright, Bristol,  1976:114-119.

 Papakostopoulos, D. and Crow,  H.  J.  The cortical
     somatosensory evoked  responses and their chang-
     es with stimulus, sensory or motor context. Inter-
     national symposium on cerebral evoked potentials
     in man. Brussels (in press).

 Papakostopoulos, D. and Fenelon, B. Spatial distribu-
     tion of the contingent negative variation (CNV)
     and  the relationship between CNV and reaction
     time. Psychophysiology, 1975, 12:74-78.

 Papakostopoulos, D., Walter, W. G. and Crow, H. J.
     The  effect of stimulus context on evoked respon-
     ses in human frontal cortex. Electroenceph. clin.
     Neurophysiol., 1970, 29:327-328.

Papakostopoulos, D., Cooper,  R. and Crow,  H.  J.
    Cortical potentials evoked by finger displacement
     in man. Nature,  1974a, 252:582-584.


Papakostopoulos, D., Cooper, R.  and Warren,  W.  J.
    Measurement of the temporal and systemic elec-
    trophysiological changes of behavior in man. Elec-
     troenceph.  clin.  Neurophysiol., 1974b, 37:102-
     103.

Papakostopoulos, D., Cooper,  R. and Crow,  H.J.
    Inhibition  of cortical  evoked  potentials  and
    sensation  by self-initiated  movement  in  man.
    Nature, 1975,258:321-324.
 Peters, J. F. Surface electrical field generated by eye
     movements.  Am.   J.  EEG   Technol.,  1967
     1:27-40.

 Phillips, C.  G,  Changing concepts  of the precentral
     motor  area. In J. C. Eccles (Ed.), Brain and con-
     scious  experiment.  Springer-Verlag, New York
     1966:389-421.

 Phillips, C.G., Powell, T.P.S.  and Wiesendanger,  M.
     Projection from low-threshold muscle afferents of
     hand and forearm to area 3 a of baboon's cortex
     J. Physiol. (Lond.), 1971, 217:419-446.

 Piaget, J.  Genetic  epistemology.  Columbia Univ
     Press. New York, 1970.

 Picton, W.T. and Hillyard, S.A. Cephalic skin  po-
     tentials  in  Electroencephalography.  Electroen-
     ceph.  clin. Neurophysiol.,  1972,  33:419-424.

 Pierrot-Deseilligny, E., Lacert, P. and Cathala, H. P.
     Amplitude  et variabilite des reflexes monosyn-
     aptiques avant  un mouvement volontaire. Physi-
     ology and Behavior, 1971, 7:495-508.

 Poggio, G. F. and Mountcastle, V. B. The functional
     properties of ventrobasal thalamic neurons stud-
     ied in unanesthetized monkeys. J. Neurophysiol
     1963,26:775-806.

 Poincare,  M.  Science and hypothesis.  Dover, New
     York, 1923.

 Powell, T. P. S.  Residual neurons in the human thala-
     mus  following  hemicortication.  Brain  1952
     75:571-584.

 Purpura, D.. P., Frigyesi, T. L, Mac Murtry, J. G.and
     Scraff, T. Synaptic mechanisms in thalamic regu-
     lation of cerebello-cortical projection activity.  In
     D. P. Purpura and M. D. Yahr (Eds.), The thal-
     mus. Columbia Univ. Press, New York, 1966-153-
     172.

 Rebert, C.,  Berry, R. and Merlo, J.   D.C. potentials
     consequences of induced muscle tension: Effects
     on contingent negative variation. In W. C. McCall-
     um and J. R. Knott (Eds.), The responsive brain
     Wright, Bristol, 1976:126-131.

Rinvik,  E. and Grofova,  I. Cerebellar projections to
     the nucleus ventralis lateralis and ventralis anteri-
    or thalami: Experimental electron miscroscopical
    and light microscopical studies in the cat. Anat
    Embryol., 1974, 146:95-111.

Robinson, D.  A. and Fuchs, A.  F.  Eye  movements
    evoked  by  stimulation  of  frontal   eye  fields
    J. Neurophysiol., 1969, 32:637-648.

Rohrbaugh, J. W., Syndulko, K. and Lindsley, D. B.
    Brain wave components  of the contingent nega-
    tive variation in  humans.  Science,  1976.  191-
     1055-1057.

-------
Motor Control References
                                           155
Rosen, I. and Asanuma, H. Peripheral afferent inputs
    to the forelimb area of the monkey motor cortex:
    Input-output  relations.  Expl. Brain Res., 1972,
    14:257-273.

Sakata, H.,  Ishijima, T. and Toyoda, Y. Single  unit
    studies  on the ventrolateral nucleus of the thala-
    mus  of the cat:  Its relation  to the cerebellum,
    motor  cortex and basal ganglia. Jap. J. Physiol.,
    1966, 16:42-60.

Schlag, J., Lehtinen, I. and Schlag-Ray, M. Neuronal
    activity b^fr-re  and  during  eye  movements  in
    thalamic internal  medullary  lamina of  the cat.
    J. Neurophysiol, 1971, 37:982-995.

Sherrington, C. S. The integrative action of the ner-
    vous system. Yale Univ. Press, New Haven, 1947.

Shibasaki, H. and Kato, M. Movement-associated corti-
    cal potentials with unilateral and  bilateral simul-
    taneous hand movement.  J. Neurol., 1975,  208:
    191-199.

Smith, A. M.,  Hepp-Reymond, M. C. andWyss, U. R.
    Relation of activity in piecentral cortical neurons
    to force and rate of force change during isometric
    contractions of finger muscles. Expl. Brain  Res.,
    1975,23:315-332.

Snider, R.W.  and Lee J. A stereotaxic atlas of the
    monkey brain. Univ. of Chicago Press, Chicago,
    1961.
Sperry, R.  W. Neurology and the mind-brain problem.
    Am. Scient.,  1952,40:291-312.

Starr, A. Influence of motor activity on click evoked
    responses in the auditory pathway of waking cats.
    Expl. Neurol., 1964, 10:191-104.

Steriade, M.  Ascending control of  motor cortex
    responsiveness. Electroenceph. clin. Neurophys-
    iol., 1969, 26:2540.

Sutton,  S., Braren,  M., Zubin, J. and John, E.R.
    Evoked potential  correlates of stimulus uncer-
    tainty. Science, 1965, 150:1187-1188.

Strick, P. L. Activity of ventrolateral thalamic neurons
    during arm  movement. J. Neurophysiol., 1976a,
    39:1032-1044.

Strick, P. L. Anatomical analysis  of ventrolateral thal-
    amic input to primate motor cortex. J. Neuro-
     physiol., 1976b, 39:1020-1031.

Teuber,  H. L. The riddle of frontal lobe function in
    man.  In J.M. Warren and K. Akert (Eds.). The
    frontal granular cortex and behavior.  McGraw-
    Hill, New York, 1964:410-441.

Thach, W.  T.  Discharge of cerebellar neurons related
    to two maintained postures and two prompt
    movements. I. Nuclear cell output. J. Neurophysi-
     ol.,  1970,33:527-536.
Thach, W. T. Timing of activity in cerebellar dentate
    nucleus and cerebral motor cortex during prompt
    volitional movement. Brain Res., 1975, 88:233-
    241.

Vanderwolf, C. H,  Limbic-diencephalic mechanisms
    of voluntary movement. Psychol.  Rev.,  1971,
    78:83-113.

Vaughan, H. G., Jr., Costa, L. D. and Ritter, W. Top-
    ography of the human motor potential. Electro-
    enceph.  clin.  Neurophysiol.,  1968, 25:1-10.

Vaughan, H. G., Jr., Gross, E. G. and Bossom, J. Corti-
    cal motor potential in monkey  before and after
    upper limb deafferentation. Expl. Neurol., 1970,
    26:253-262.

Wagman,  I.  H. Eye movements  induced  by electric
    stimulation  of  cerebrum in  monkeys and their
    relationship  to bodily movements.  In  M.  B.
    Bender  (Ed.), The oculomotor  system. Hoeber,
    New York, 1964:18-39.

Walker, A. E. The thalamic projection to  the central
    gyri in Macacus rhesus. J. comp. Neurol., 1934,
    60:161-184.

Walker, A. E.  Afferent connections. In P. H.  Buey
    (Ed.), The precentral motor  cortex. Univ. of Illi-
    nois Press, Urbana, 1944:112-13 2.

Walter, W. G., Cooper, R., Aldridge,  V. J., McCallum,
    W. C. and Winter, A. L. Contingent negative vari-
    ation: An electric sign of,sensorimotor association
    and  expectancy  in  the human brain. Nature,
     1964,203:380-384.

Wilke, J. T. and Lansing, R. W. Variation in the motor
    potential  with force exerted during voluntary
    arm  movements  in man.  Electroenceph. clin.
    Neurophysiol., 1973,35:259-265.

Woolsey,  C. N. Organization of somatic sensory and
    motor areas of  the  cerebral cortex. In  H.  F.
    Harlow and C.  N. Woolsey (Eds.), Biological and
    biochemical bases of behavior. Univ. of Wisconsin
    Press, Madison, 1958:63-81.

Yoshida, M., Yajima, K. and Uno, M. Different activa-
    tion  of  the two types of the pyramidal tract
    neurones  through the cerebello-thalamocortical
    pathway. Experientia, 1966,  22:331-332.

 Young, J. Z. Why  do we have two brains? In V.  B.
    Mountcastle (Ed.),  Interhemispheric  relations
     and  cerebral dominance.  Johns Hopkins  Press,
    Baltimore, 1962:7-24.

 Young,  J. Z. An introduction to the study of man.
     Oxford Univ. Press, London,  1971.

 Wurtz, R.H. and Mohler, C.W. Enhancement of visual
     responses  in monkey  striate cortex  and  frontal
     eye fields. J. Neurophysiol., 1976, 39:766-772.

-------
III. INFORMATION PROCESSING AND COGNITION
     Section Editor:

     Patricia Tueting
     Maryland Psychiatric Research Center
     University of Maryland Medical School
     Baltimore, Maryland, U.S.A.

-------
EVENT-RELATED  POTENTIALS,  COGNITIVE
EVENTS,  AND INFORMATION PROCESSING1


P. TUETING2
Department of Psychophysiology, New York State Psychiatric Institute, New York, NY, U.S.A
   This review is based on pre-Conference correspon-
dence among the members of the EPIC IV Information
Processing Panel;  additional comments from discus-
sion at the conference have also been included.  Con-
tributors  were:  P, Tueting (chairman),  New York
State Psychiatric Institute, New York, NY; E.  Don-
chin, University of Illinois, Champaign, IL; J. Ford,
Stanford  University, Palo  Alto,  CA, D. Friedman,
New York State Psychiatric Institute, New York,NY;
M.R. Harter, University  of North Carolina, Greens-
boro, NC; S. Hillyard (co-chairman), University of
California, San Diego, CA, W. Ritter, Lehman Col-
lege, Bronx, NY; J. Rohrbaugh, University of Califor-
nia, Los Angeles, CA, W.T. Roth, Stanford  Universi-
ty, Palo Alto, CA; D. Ruchkin, University  of Mary-
land, Baltimore, MD, K. Squires, University of Illi-
nois, Champaign, IL, N, Squires, University of Illi-
nois, Champaign, IL, S. Sutton, New York State Psy-
chiatric Institute, New York, NY, and R.T. Wilkin-
son,  Medical Research Council, Cambridge,  England.

    The papers in this  section are divided  into two
categories: (I) issue-oriented  reviews and  (2) data
papers. The review papers developed out of the corre-
spondence and discussion at the conference and are
presented first. Roth considers  the issue of how many
late  positive waves can  be defined. Ritter  discusses
the relationship of  ERPs  to  decision latency, and
Ruchkin explains the importance of the  concept of
equivocation in interpreting ERP data. The relation-
ship  of ERPs to orienting is discussed by Friedman.
Ford presents template and match/mismatch theory.
Sutton and  his colleagues document P300 studies
relevant to feedback issues. The research reports that
follow the review papers relate to one or more of these
issues and are representative of current research in the
area.   Research  reports  germane  to  information
processing  were  submitted by  R.  Cooper, Burden
Neurological Institute, Bristol, England; N. Lesevre,
Salpetriere  Hospital, Paris, France; and JJ. Tecce,
Tufts University School of Medicine, Boston, MA, as
well as by members  of the Information Processing
Panel.
Background

    New  evoked  potential  (EP)  components  are
emerging so rapidly that it is difficult to determine
their validity, let alone their function in information
processing. Temporal sequences of components have
traditionally been evaluated within a "stages of infor-
mation processing" framework. However, it is becom-
ing increasingly  clear  that EPs also reflect parallel
processing. Serial decisions may be accompanied by
parallel memory storage, memory retrieval, set adjust-
ments, and motor processing. Parallel processing is
indicated by the occurrence of components overlapped
in time at a single electrode location and by observa-
tion  of independent components  occurring  at  the
same time but with different scalp distributions.

    High-frequency far-field components, thought to
originate in auditory nerves and various levels of brain-
stem, can be recorded at vertex for an auditory stim-
ulus. These potentials originate mainly from the high-
frequency  end of the cochlea and carry information
concerning stimulus  intensity.  Woods and Hillyard
(this volume) failed to find evidence of peripheral gat-
ing in brainstem potentials evoked by probe stimuli in
a dichotic listening task demanding selective attention.
There are,  however, animal EP data that suggest effer-
ent gating at the brainstem level.

    Components related to the initial registration of
stimulus information at the cortex have been identified
 lThis project received support from NIE Grant No. 6-74-0042 awarded to Dr. Samuel Sutton. The assistance
  of Dr. Steven Hillyard and Dr. Sutton is gratefully acknowledged.

 ^Dr. Tueting is  now at the Maryland Psychiatric Research Center, Department of Psychiatry, University of Mary-
  land Medical School, Baltimore,MD. Dr. Tueting organized the correspondence, wrote the narrative summary,
  and edited this section. She was, however, unable to attend the Congress. Dr. Hillyard chaired the panel discus-
  sion in her absence.

-------
160
                                        Tueting
 in  the somatosensory and visual modalities.  These
 components also carry basic information related to
 stimulus parameters  such as intensity, wavelength,
 and spatial location,  and are affected very little  by
 the attention state  of the organism, by the arousal-
 sleep continuum, or by decisions regarding stimulus
 input.

    The primary registration of information at cortex
 is only the beginning of the perceptual-cognitive eval-
 uation involved in the  ultimate use, nonuse, or storage
 of stimulus information. The task of the Information
 Processing  group of the EPIC IV Congress was to con-
 sider some  of these  later information processing deci-
 sions in a realm that is often referred to as cognitive.

 Early components and selective filtering

    At the conference, Hillyard defined a selective
 attention experiment  as one  involving a situation in
 which there are at least two stimulus categories. Selec-
 tivity is demonstrated when the subject processes one
 of these stimulus categories more effectively than the
 other, as indicated by a behavioral measure. For
 example, selective  attention is  demonstrated when
 the subject makes  more  accurate and more refined
 discriminations in  one  stimulus  category  than  in
 another, reacts faster to one category than to another,
 or is able to retrieve from memory one stimulus cate-
 gory better than another. McCallum pointed to the
 unfortunate tendency to use the term selective atten-
 tion without defining a specific operational context.
 Behavioral  validation  in EP experiments of selective
 attention is often nonexistent or inadequate.

    Hillyard and his colleagues are now using concur-
 rent signal  detection  measures  of attention  in their
 dichotic listening tasks. In a  dichotic listening task,
 the amplitude of N100 is larger  for relevant stimuli
 possessing  easily identifiable  characteristics  such  as
 pitch  or spatial position  toward which the subject's
 attention has been  directed-usually by instructions
 and task.  N100 amplitude is correspondingly smaller
 for stimuli  occurring  in irrelevant, rejected channels
 (Hillyard et  al. 1973, Schwent and Hillyard 1975).
These  findings have been  most clearly demonstrated
when channels are separated by both pitch and spatial
 attributes (Schwent et al.  1976)  and when subjects
are exposed to information overload, i.e., when a sub-
ject is placed in a situation where there is difficulty in
processing all of the input effectively in the time avail-
 able.

    Based  on a study by Wilkinson and Lee (1972)
 involving  auditory  channels differing  in  stimulus
 frequency,  Wilkinson has expressed the hypothesis
 that N100, and possibly N100-P200, of the auditory
 evoked response reflects  selective filtering only for
 bade stimulus attributes like modality and location in
 space and time, but not for other attributes like pitch
 and linguistic distinctions. According to the hypo-
 thesis, contingent negative variation (CNV) resolution,
 rather than N100 itself, may reflect these latter cate-
 gories of stimulus selection. However, Schwent et al.
 (1976) reported that selective effects on N100 could
 be obtained with either pitch or spatial location. Larg-
 er effects  were found when both pitch and spatial
 cues were used. These results do not seem to be inter-
 pretable in terms of CNV resolution.

    The conclusion that N1 is correlated with between
 channel (i.e. stimulus set) selection is based primarily
 on  the work discussed above on the auditory vertex
 N100. At the conference, Hillyard proposed that the
 vertex Nl may index a cortical process related to an
 initial selection of a source of information for further
 processing regardless of modality-auditory, visual, or
 somatosensory. Hillyard's proposal  assumes that an
 Nl  component with a frontocentral distribution is
 elicited by stimuli in all three modalities. Measurement
 of Nl across modalities is complicated, however, by
 latency changes and component overlap.  It is unclear
 whether auditory Nl has a frontal source, a temporal
 source, or both a  frontal and  a  temporal  source
 (Arezzo et al. 1975). Vertex Nl may be represented
in occipital recordings at reduced  amplitude, but Nl
of the auditory evoked potential (recorded from ver-
tex) and Nl of the visual evoked potential (recorded
from occipital) may  not be directly comparable. In
the somatosensory modality, there  is evidence  that
there may be an overlap of components originating in
primary somatosensory cortex and in association cor-
tex in recordings over either the vertex or the somato-
sensory region (Donald  1976).

    Harter described selective attention in the visual
modality.  In the Eason et al. (1969) study, selective
attention  to spatially separate stimuli (light flashes
presented  to either the left or right visual field) was
found to influence the amplitude of the first negative
deflection of the occipital EP beginning 120 to 130
msec after the  eliciting stimulus.  These findings are
consistent with Hillyard and Wilkinson's proposal that
selective  filtering is reflected by  early components
only when easily identifiable stimulus attributes are
being discriminated. Harter and Salmon (1972), how-
ever, have reported selective attention effects on the
earliest components of their visual EPs when the rele-
vant and irrelevant stimuli were different colors with
identical spatial attributes (but not when  the relevant
and irrelevant stimuli were in different orientations or
were  different  shapes).  Color is represented neuro-
physiologically in the peripheral visual system, while
orientation and shape are represented at higher levels.
Harter therefore suggested that it is the  presence of
peripheral representation of stimulus attributes, rather
than the complexity of the attributes per se, that is a
prerequisite to selective filtering.

-------
ERPs and Information Processing
                                                                                                 161
    A vertex negative component elicited by somato-
sensory stimulation has been described by Debecker
and  Desmedt (1971)  and Desmedt and  Debecker
(1978). The  findings with respect to the somatosen-
sory vertex Nl are similar to the findings described
above for the auditory and visual Nl.

The P300 component

    By far the most controversy and discussion in
relating psychological variables to EPs has been gener-
ated by the P300 component, first described by Sutton
et al. (1965). The P300 component is not modality
specific and has been shown to be endogenous by the
fact that it can be elicited in the absence of stimulation
by an  apparently  internal trigger if (1) stimulus
absence is significant and (2) stimulus absence is time-
locked either by  a prior stimulus or by some  other
time-marking aid.

    A modality-specific negative component preced-
ing the emitted P300 is elicited when stimulus omis-
sions are rare in comparison to stimulus presentations
(Simson et al. 1976),  Therefore,  this component is
also an  endogenous component. Mystery  surrounds
the question  of whether the stimulus-elicited P300 is
always  preceded  by a  negative component or not-
N2?, N190?,  N250?, NX?-but it is clearly  evident in
the waveforms of Harter and Salmon (1972), N.
Squires et al. (1975) and others, and has emerged in
principal component analysis (K. Squires et al. 1977).
Part of the ephemeral nature of the negative compon-
ent may be due to the fact that it is overlapped with
other components known to occur in the same latency
range when  a stimulus is actually presented. If the
latency variability of the negative  component is the
same as the latency variability for P300, then it would
not be seen as easily as P300 because of the negative
component's   relatively small amplitude (Ritter, this
volume). It is also possible that the negative com-
ponent does not always accompany P300 but that the
exact  variables,  subject or otherwise, necessary and
sufficient for its  elicitation have not been specified.
So far, however, the N190 component seems  to be
affected by probability in the same way as P300 is, at
least  when both components  are  clearly  elicited in
target-identification  situations (cf. K.  Squires et al.
 1977).

     Karlin (1970) expressed the  opinion  that P300
represents a nonspecific reactive change of state sub-
sequent to cognitive evaluation of significant stimuli.
This opinion was originally prompted by the fact that
P300  components tend to be  elicited in situations
where a slow negative expectancy process (CNV) pre-
cedes stimulus  presentation.  The  CNV  frequently
resolves into a slow positive-going poststimulus process
with recording, and this CNV resolution appears simi-
lar in some respects to P300. It was argued that any
nonspecific state prior to the stimulus, a state indexed
by  EEC  desynchronization (Naatanen 1967), or by
nothing recordable from the scalp, could be associated
with a nonspecific change of  state subsequent to
stimulus presentation and reflected in P300 (see also
Naatanen 1975).

    The existence of an independent P300 related to
specific cognition  and  stimulus identification  is no
longer the issue. P300 components have been elicited
under conditions  of uncertainty when .the  stimulus
delivers feedback concerning the accuracy of a guess
or of a judgment, in situations where the subject is
required to make a choice response as soon after stim-
ulus presentation as possible, and in situations where
low probability targets are presented against a back-
ground  of more  probable nonsignals. It has been
shown by factorial experimental design and by scalp
topography (e.g., Donchin et al. 1975) that CNV and
P300 are dissociable components. However, the issue,
in slightly different form, remains. It is probable that
in some situations a distinct P300 and a positive CNV
resolution (or "resident potential resolution") may be
overlapped. Wilkinson suggests  that baseline-to-peak
measurements of P300 would ordinarily be confound-
ed in these cases of overlap. (Random presentation of
experimental conditions, although it does not elimin-
ate  overlap,  does preclude  differential prestimulus
states, and should always be used in experiments aimed
at poststimulus processing.)
How many "PSOOs"?

     It would be more parsimonious if one unifying
psychological construct could be found for the P300
component. A search for such a construct began sever-
al years ago after it was observed that P300 was elici-
ted in a wide variety of situations. Uncertainty, infor-
mation delivery, significance, salience, orienting, inhi-
bition, selective recognition involving  response  set,
and awareness have all been postulated as candidates
for the one unifying construct for P300. (See Sutton
et al., this volume). Donchin has postulated that P300
reflects the activity of a specific processor that can be
invoked in a wide variety of situations. In this formu-
lation (and  most others),  P300  latency would  be
determined by  the time at which the processor is in-
voked (see Kutas and Donchin, this volume).

     The trend  now, however, is to search for more
than one  P300. Two is the current vogue in some cir-
cles, but  three or more (P3a, P3b, P3 visual frontal,
P3 no-go, P3 vertex) have been proposed. It is difficult
to say at  this point whether the seeming multiplicity
of P300&  simplifies the search for cognitive correlates
of P300 or increases the complexity. Donchin, at the
conference,  argued for stricter criteria for admission
 to the P300 "club."

-------
 162
                                         Tueting
      The criteria that have been used for isolating
 P300 components  are scalp distribution and wave-
 form characteristics such as latency  and amplitude.
 Separation has been made on the basis of differences
 obtained in these measures for different psychological
 situations and  for different modalities of stimuli. A
 P300, "P3a" is elicited by  low-probability  auditory
 stimuli presented while subjects are engaged in another
 task (e.g.,  reading) and "ignoring" the stimuli (Roth
 1973, Ford et al.  1976, N. Squires et al. 1975, K.
 Squires et  al. 1977). "P3a"  has an early latency, rela-
 tively small amplitude, and a  frontocentral scalp distri-
 bution. A  P300, "P3b," elicited by low-probability
 signals in a vigilance task is  later in latency, larger in
 amplitude,  and  has a  centroparietal  distribution.
 "P3b" has  a virtually identical scalp distribution in a
 wide  variety  of  target-detection  tasks involving
 threshold-level  stimuli, omitted stimuli, and auditory
 and visual  discriminations (Hillyard et al. 1976; Pic-
 ton  and   Hillyard  1974;  Ritter et  al., in press).
 Courchesne et al. (1975) reported a P300  to  novel
 visual stimuli  that were not task-related. This fron-
 tal P300, which  had a relatively  long latency, habit-
 uated quickly  as novelty decreased.   Although this
 P300 was also elicited by low-probability stimuli un-
 related  to  the  subject's main task and had a fronto-
 central distribution, it was not necessarily considered
 to  be equivalent  to the auditory  "P3a."  The "P3b"
 may or may not  be identical to the original P300 de-
 scribed  by Sutton et al. (1965), which is elicited in
 guessing situations  and tends,  if anything, to be a
 little larger at  vertex than at parietal loci. The  "de-
 tection  potential" described by  Cooper  et al. (this
 volume) is also  similar to "P3b."

     There  is speculation that the  frontal P300 to
 both  auditory and  visual irrelevant stimuli may be
 related to orienting. The idea is that subjects  will
 orient to low-probability stimuli even though they are
 told to ignore them  and are involved in a task with
 other stimuli. The parietal P300 on the other hand
 has  been related to  delivery  of relevant information,
 "response set," and  decision.  It  is elicited by  low-
 probability  target stimuli to which a specific response
 such as counting has to be made. Ford has distinguish-
 ed  the two P300s within an active-attention versus
 passive-attention framework.  The term "passive atten-
 tion" might apply to the frontal P300 since attention
 is drawn involuntarily by a low-probability stimulus
 that is irrelevant to the subject's main task. The parie-
tal P300, however, seems to require active  attention,
 being obtained mainly when  subjects are actively en-
 gaged in a task in relation to the eliciting stimulus.

    A frontal P300, which may be related to response
 inhibition  rather  than orienting, was  proposed by
 Papakostopoulous (Donchin  1976) for the Bristol
 Congress on Event Related  Slow Potentials. A P300
 with a frontocentral distribution similar  to that of
 "P3a" (but much longer in latency) is elicited to the
 no-go stimulus in a go/no-go reaction time task (Tuet-
 ing and Sutton 1976; Ritter et al., in press; McCallum
 1976).  The P300 elicited by the go stimulus has a
 more parietal  distribution, reminiscent  of  the  task
 related  "P3b." In a choice-reaction time (RT) task,
 the subject is prepared to react as fast as possible to a
 stimulus, and the no-go stimulus can be considered to
 set up a red flag, i.e., an inhibition of the go. However,
 extending the construct of inhibition for P300 beyond
 choice-RT  situations requires viewing motor function-
 ing as a largely  covert activity that is closely intertwin-
 ed with cognitive processing  (cf. Sokolov  1972 or
 McGuigan and Schoonover 1973).

    The group generally agreed that latency by itself
 should not be  used  to  dissociate P300 components.
 More information about the range  of responsiveness
 of the P300 component to experimental variables and
 about the implications of component overlap at vari-
 ous electrode locations is needed before latency alone
 can be used to dissociate P300 components. Some of
 the problems involved in interpreting scalp distribution
 data are reviewed by  the Scalp Distribution group
 (Donchin, this volume).

    Moreover,  it is still unclear whether several inde-
 pendent P300  components truly exist, and the issue
 of one versus several P300s was a focal point of discus-
 sion at the conference. The issue is  reviewed in detail
 by Roth (this  volume). The concept of only one P300
 is still viable. Some of the findings interpreted as indi-
 cating multiple P300s could be interpreted as indicat-
 ing one P300 whose anterior-posterior amplitude dis-
 tribution varies with task demand. In addition,  the
 amplitude of "P3a" is often small, and K. Squires et
 al. (1977) reported  that "P3a" is an elusive compon-
 ent relative to "P3b." A further  problem for  the
 multiple P300  theory is that differences in anterior-
 posterior distribution could be a result of differences
in the overlapping slow wave component as suggested
 by N. Squires in the correspondence. The slow wave
has been reported to be negative-going frontally  and
 to become  progressively more  positive-going pane tal-
ly. Conceivably, P300 latency could  also be affected-
P300 could be truncated in latency frontally by  the
negative slow wave and lengthened posteriorally by
the positive slow wave.

    A note of  caution should be mentioned in com-
paring observations of  P300 latency from different
laboratories. A  lower cutoff frequency may influence
P300 latency.  In the  correspondence,  Rohrbaugh
pointed out that P300 peak latency is  sometimes
shortened appreciably with a  1-Hz  cutoff compared
 to a dc recording, reflecting phase  shifts at frequen-
 cies near the low  edge of the bandpass; its amplitude
 may also be reduced. However,  it is unclear whether
 raising the  low-frequency cutoff decreases the appar-

-------
ERPs and Information Processing
                                             163
ent P300 latency because of phase shifts of the P300
itself or because the concurrent slow wave is  not
passed by the filter.

P300 and orienting

    The construct most frequently  associated with
the frontal  P300 at present is orienting (see review by
Friedman, this volume). The  frontal P300 that follows
novel visual stimuli  may be  a stronger candidate for
an orienting correlate than the auditory "P3a" since
it  increases with  the complexity/nonrecognizability
of the stimulus and habituates rapidly (Courchesne et
al. 1975). Friedman pointed out in the correspondence
that orienting is reduced or absent in patients with
frontal  lobe damage  (Luna 1966), and this is consis-
tent with an orienting construct for the frontal P300.

    One way to assess the validity of this hypothesis
will  be  to  record traditional autonomic measures of
orienting such as heart  rate, galvanic skin response,
vasodilation of the skin, pupil,  etc.,  in conjunction
with EPs.  Friedman et al. (1973) have studied event-
related  potentials (ERPs) and pupil  dilation concur-
rently.  They found situations in which dilation  and
ERPs correlated in response to psychological variables.
For example, both vertex P300 and peak dilation of
the pupil were monotonic inverse functions of stim-
ulus and guessing probability. Roth et al. (this volume)
have  recorded auditory EPs and skin conductance
concurrently, but their results were equivocal.

     A significant methodological problem in record-
ing ERPs  concurrently  with autonomic measures is
that autonomic measures have a much longer latency
than  most ERPs. Only certain  paradigms involving
long  interstimulus intervals may be  used. Another
problem is that the relationship of autonomic variables
to orienting is not completely clear, and complexities,
particularly variations in recording techniques  and
individual differences in autonomic responsivity, arise.
The data  overload  and complexities are increased
even further when ERPs (latency, amplitude, and scalp
distribution measures  of several components) are
added to the picture.

P300 and decision latency

     One conceptual framework implicates the P300
and decision-making. Unfortunately, the exact nature
of the  presumed decision is not thoroughly worked
out, but a common thread running through the differ-
ent tasks  that elicit P300  seems to indicate that a
comparison of stimulus input against representations
in memory (templates) is involved either directly or
indirectly. P300 is not directly involved in the decision
regarding the selection of a specific motor response to
a stimulus, as the large  P300 obtained in  choice-RT
studies might imply, since large P300s are routinely
obtained in  a guessing task where  subjects are not
required to choose one overt response over another in
the immediate poststimulus period. Similarly, specific
poststimulus response selection is not involved when
P300s  are obtained with stimuli delivering feedback
concerning accuracy in a discrimination trial or with
non-task-related low-probability  or  novel stimuli.
However, Hillyard pointed out in the correspondence
that a theory relating decision-making to P300 must
take into account  the  finding of Karlin  and Martz
(1973) that low-probability responses, as well as low-
probability  stimuli, are associated with larger P300s.

     The appeal of the term "decision" is its general-
ity, but a useful feature is that a decision is made at a
precise point in time that is measurable, at least theo-
retically.  Both N. Squires et  al.  (1975) and  Roth
(1973) found that P300 evoked in  an orienting
response situation occurs earlier than the P300 evoked
in  a signal-detection paradigm. It  seems reasonable
that the decision to orient is easier and faster than the
decision  that the  target  signal  just occurred, and
P300  latency could be reflecting that  difference  in
decision time, a possibility suggested by Ford in the
correspondence. However, since there is no voluntary
intention when P300s are elicited by low-probability
events  presented  while  subjects are reading,  Roth
pointed out that it might be clearer semantically in
this case to use the term "reaction" instead of "deci-
sion." Thus, the term decision may be more appropri-
ate for the parietal task-related P300 than for a frontal
orienting P300.

     In the correspondence, Donchin pointed to find-
ings relating P300 latency to an internal trigger, which
varies with stimulus evaluation time (e.g., Kutas and
Donchin, this section). The fact that P300 can be elici-
ted without a stimulus also supports the idea that the
latency of P300 is related to the latency of a cognitive
event. Assuming that the trigger for P300 is basically
internal,  considerable variability  in  P300  latency
across conditions in averages time-locked to stimulus
presentation should be expected, and has been observ-
ed. Variability from trial-to-trial within a single con-
dition, and variability among subjects in P300 peak
latency distribution has also been noted. "Inexplica-
ble" results when  conditions that are stimulus time-
locked (e.g., "hits" in threshold signal-detection para-
digms) are compared to conditions that are not time-
locked (e.g.,  "false alarms" in signal-detection para-
digms) are also to be expected. Threshold signal-
detection data become more interpretable when time
marking aids are used (K. Squires et al. 1975).

      Validity of latency measures: A first consideration
in relating P300 latency to decision latency should be
whether a difference in P300 latency is a valid measure.
It is relatively simple to say that  at a given latency
 there  is an amplitude difference, but more assump-
tions may be involved in concluding that a component

-------
 164
                                         Tueting
 of assumed amplitude, shape, and approximate latency
 shifts peak latency as a function of condition or elec-
 trode location.  Donchin has repeatedly pointed out
 that  it may be  misleading to identify components
 with the appearance of visible peaks in the waveform,
 and it may  be preferable  to define components in
 terms of the effects of experimental variables on volt-
 age measurements.

     Most of the group felt  that, despite considerable
 problems,  latency can  be  a valid  measure. In fact,
 amplitude  measures may also be invalid or confound-
 ed. For example, a condition with greater variance in
 peak latency distribution may result in smaller aver-
 age amplitude because of high latency variability and
 not because of a real decrease in amplitude (Ruchkin
 and Sutton,  1978). One solution  to the problem of
 the  validity  of amplitude  and latency  is  to  obtain
 amplitude  distributions and latency distributions on a
 triaJ-by-trial basis. This, although more difficult to do
 for small components, is a possibility for P300, which
 can be fairly large on individual trials in some subjects.

     The technical problems involved in defining P300
 latency in  individual  trials  are considerable however
 (see correspondence summary on Alternatives to Sig-
 nal Averaging,  by Weinberg).  The problems develop
 from the low signal-to-noise (S/N) ratio for P300 at
 the scalp, which may be so low for some subjects and
 for some experimental  conditions as to make single
 trial analysis an impossibility.

    Typical single-trial analysis procedure for obtain-
 ing P300  peak  latency  and amplitude distributions
 involves setting a latency search window for a comput-
 er instructed to select the most positive peak within
 the time span specified. The selection  of the search
 window is critical-too  wide  a search  window may
 result in loss of the peak in some trials. A painstaking
 visual inspection of the data (and often of individual
 trials) before selecting a latency search window  is
 often  necessary. Ruchkin is now using Woody filter
 analysis (Woody  1967), which  requires an iterative
 correlation procedure as a basis for realigning  single
 trials before latency-corrected averaging. This proced-
 ure  is  more quantitative and  objective than  visual
 inspection  or peak selection  within search windows.

    Other  single-trial techniques do not yield peak
latency information,  although  the latency at  which
there is an amplitude difference may  be identified
and  experimental trials  classified. For example, K.
Squires and Donchin are  using the value of a discrimi-
nant score  on each trial. Researchers unable to deal
with  single-trial analysis may  nevertheless obtain
tome information from additional measurements like
breadth of P300, area of P300, rise and fall of P300,
and onset latency ofP300. Specific post-hoc clustering
 of the data may be helpful, e.g., clustering EP trials
 on the basis of differing ranges of reaction time.

     Despite problems involved in single-trial analysis,
 the initial findings are intriguing. Ritter et al. (1972)
 found that variation in parietal P300 latency from
 trial to trial correlated  with trial-to-trial variation in
 RT. Ruchkin and Sutton (1978) found that P300
 varied in latency from trial to trial in a guessing task.
 In  their study,  latency  variability was considerably
 greater when stimulus omission  rather than stimulus
 presence delivered information  concerning the sub-
 ject's guess. Presumably, the point in time at which
 the subject decides whether a guess is correct or not
 (and hence P300 latency) is more dependent upon
 time-estimation  ability  in the omitted-stimulus case
 than in the present-stimulus case.

     Validation of decision latency: Donchin pointed
 to  a significant  problem, namely that  a method has
 not been  developed  to  independently determine the
 time of invocation of the processor that P300 presum-
 ably represents.  For example, Ruchkin found a ten-
 dency, not  statistically  significant, for emitted P300
 latencies to be longer for incorrectly guessed single-
 click trials than for correctly guessed single-click trials.
 (Subjects  guessed whether a single or a double click
 would be presented on each trial, and a P300 was
 emitted at the point in time of omission of the second
 click in a single-click trial.) One  possible inference is
 that subjects may wait  longer  to acknowledge the
 absence of a second click when they have predicted
 that it would be present. Testing this plausible infer-
 ence directly is a  difficult problem, as Ruchkin pointed
 out. (A similarly delayed P300 for feedback that dis-
 confirmed a prior judgment was found by K. Squires
 et al. 1973a.)

     Psychologists have  traditionally made inferences
 about  cognitive  events  via behavioral measures, and
 the traditional behavioral measure of decision latency
 has been  RT. The correspondents all agreed to take
 seriously the caution to record physiological and beha-
 vioral data in the same  set of trials; Sutton  (1969)
 outlined the very good reasons for doing this. A major
 problem arises, however, because potentials preceding
 and following the response may  be overlapped (espe-
 cially in the centroparietal area) with potentials related
 to  stimulus  evaluation  (decision). These  "motor"
 potentials consist of  a slow premotor negative poten-
 tial, a higher frequency complex  associated more
 directly with response initiation, and then a large and
 slow postresponse positive wave peaking about ISO
 msec after response (thought by some to be related to
somatosensory feedback from muscle and joint recep-
 tors). InanRT task, premotor potentials overlap evok-
ed potential components up to the point of the reac-
 tion, and for fast RT, the postresponse positive wave
could overlap P300, N350, and P400.

-------
ERPs and Information Processing
                                            165
    A further complication emphasized by Roth and
by Wilkinson is that motor involvement increases the
size of the CNV, and  they point out that the timing
of decision may be more closely associated with the
latency of CNV resolution than with the latency of
P300.  In any case, it is likely that in some instances
the latency and amplitude of CNV resolution and the
latency and amplitude of P300 will be confounded by
the overlap of these two components.

    Some correspondents  emphasized that it may
prove impossible to disentangle decision-related com-
ponents like P300 from motor components associated
with a required reaction. Simple linear addition of
motor components when  an RT  task is added to a
situation is unlikely. However, the advantage of having
an independent behavioral measure of decision latency
is so important that investigators are compelled  to use
a number of  strategies that converge on this issue.
Several techniques considered in this context follow:

  1. Choice of response measure in  RT studies. The
judicious choice of response measure in RT studies
may help in reducing motor potentials and movement
artifact. In a study of the  relationshp of EPs and RT
to uncertainty,  Tueting  and  Sutton (1976)  found
that a  lift no-lift response  task gave cleaner response-
evoked potentials than a  key press response.  Rohr-
baugh  suggested that systematically varying force
and excursion of the responding member may  aid in
partialling out motor  responses,  and that, in  many
experiments, keeping   force and  distance at a mini-
mum may reduce motor potentials. Stretching this
strategy  even further, it  may be  possible to train
subjects  to  make  an  extremely small response that
can still be  picked up by EMG  recording. Covert
verbal  reactions such as counting to onself should be
throughly investigated to  see  if minute muscle or
laryngeal responses can be picked up reliably, since
counting should be considered a motor response and
may be accompanied by motor potentials and  motor
artifact. Whatever the  response elected, it should be
absolutely silent and should produce as little tactual
and visual sensation as possible to avoid inadvertent-
ly adding sensory EPs to the already complex picture.

  2. Differential topography. In  come cases, motor
potentials can be  separated from stimulus-evoked
responses by  precise  delineation  of  topography
(Vaughan et al  1965, Vaughan et  al. 1968) and some
motor potentials have been identified by their asym-
metry  (Kutas  and Donchin  1974).  Possibly a  factor
analysis approach using data from  several recording
sites could separate "motor" and "cognitive" effects,
but factor analysis assumes an underlying linearity,
which  according to Ruchkin  may  not exist in this
situation

  3. Delayed  response control. A  delayed-response
task (or a task requiring no motor response) can be
compared to an otherwise identical RT task (Vaughan
et al. 196S,Karlinand Martz 1973, Picton et al. 1974,
Courchesne 1975). However,  Roth pointed out that
delaying or omitting motor  responses  is  likely  to
change the nature of the decision and other cognitive
features  of the task.  A  delayed-response  control
would  be better than a no-task control, as long as
anticipatory reactions are prevented.
    4.  Self-paced motor response control. Another
strategy is  to collect self-paced motor responses in a
separate control condition.  Potentials obtained in the
self-paced condition can be inspected and inferences
drawn (Ritter et al. 1972), or the motor potentials
obtained in the self-paced task can actually be subtrac-
ted from the potentials  obtained in the RT task. Pre-
sumably, subtraction would leave potentials  related
to stimulus evaluation (decision) without motor con-
founding. The subtraction  procedure  assumes, how-
ever,  that the motor potentials in RT situations are
similar  to  motor potentials  recorded in self-paced
response tasks.  Several  correspondents  pointed  out
that self-paced tasks may differ in a number of ways
from  the task for which they  are being used  as con-
trols.  For example,  Ruchkin  et al. (1977)  used self-
paced motor responses as a control in a time-estima-
tion study.

   5.  Comparison of stimulus  versus response-locked
potentials.  Averages time-locked to the  response can
be compared to averages time-locked to the stimulus
(Ritter et al. 1972, Karlin et al. 1971).  Presumably,
decision time would be  more closely related to actual
response time  that to  stimulus  onset  time; there-
fore P300  should be more precisely time-locked to
response time. The experimental findings to date are
equivocal (Karlin et al.  1971), indicating that P300
may actually be  smaller in response-locked averages
than  in  stimulus-locked averages. However,  motor
potentials are more closely time-locked to response
time,  and motor potentials can be seen  more clearly
in response-locked  averages than in  stimulus-locked
averages.

   6.  The  go/no-go reaction time  task.  Still another
technique is to use a go/no-go  choice-RT task, and to
counterbalance conditions  (usually within subjects)
so that every  condition has a replicate  (one  go and
one  no-go). The assumption  is  that no-go  evoked
responses are not confounded with motor potentials
since  a motor response  is  not required. However, it
seems likely that there  is motor  involvement in the
no-go case in  a choice-RT  task. For example, motor
response inhibition in the no-go case could be reflected
in a  motor  potential related to motor inhibition.
Specifically,  frontal P300  is  a candidate  (Donchin
1976; McCallum 1976; Ritter  etal.in press; Tueting
and Sutton 1976).

-------
 166
                                          Tueting
     P300 latency and RT: If P300 occurs in precise
 temporal relation to a cognitive decision, an indepen-
 dent behavioral  measure  of decision latency such as
 motor RT ought to correlate with  P300  latency.
 Hillyard, for the correspondence, reviewed studies in
 which  P300  and  RT  were  recorded concurrently.
 Only two of the studies indicated that RT variations
 paralleled changes in P300 latency (Ritter et al.  1972,
 Picton et al. 1974). A few other studies provided less
 direct support for a  positive P300 latency-RT corre-
 lation (Bostock  and Jarvis 1970, Rohrbaugh  et  al.
 1974, Routh et al. 1975, Posner et al. 1973). On the
 other hand, many  studies showed a  dissociation  of
 P300 latency and  RT (Karlin et al. 1970, Karlin et
 al.  1971,  Karlin and Martz  1973, Donchin et  al.
 1973,  Donchin  et   al.   1975, Parasuraman  and
 Davies 1975,Courchesne  1975, Ford 1975).


     Admittedly,  these studies suffer from problems
 discussed above, such as confounding from component
 overlap, and from problems in comparing mean RT  to
 peak latencies obtained from evoked potentials averag-
 ed over trials (rather than from single trials). Despite
 these shortcomings, the inconsistencies are too  great
 to dismiss. The  dissociation  between  RT and P300
 latency seems to be separable into the  two following
 categories.


   1. P300-reaction time  correlations  in relation  to
 discrimin ability and equivocation.  Ruchkin and Sut-
 ton (this section) have formulated  the  concept  of
 equivocation in relation to P300. Equivocation refers
 to a loss  of information related to the subject's
 a  posteriori uncertainty,  and  can be  thought of  in
 terms  of the  "immediacy and ease"  with  which a
 decision  can be  made. An increase  in equivocation
 leads to a decrease in P300 amplitude. The equivoca-
 tion concept is in some ways similar to  the "confi-
 dence  of the  decision"  concept proposed by K.
 Squires etal. (1973b).
     Dissociation between P300 latency and RT seems
to occur  when  discriminability, or equivocation, is
the variable.  Both P300  latency and  RT increase as
equivocation  increases, but  the  magnitude  of the
change appears to be three  to five times greater for
RT than for  P300 latency. For example, Ford et al.
(1976) varied the  degree to which rare target tones
differed in frequency (pitch) from  more  probable
background  stimuli.  As  the  frequency  difference
decreased  (increasing  equivocation),  P300  latency
increased by  an average of 26 msec while RT increas-
ed by an average of 81 msec. Parasuraman and Davies
(1975) found that P300 latency and RT were greater
for false alarms as compared to hits in a vigilance task,
but the magnitude of the difference was greater for
RT than for P300 latency.
   2.  P300-RT correlations in relation to uncertainty
 and  probability. The second kind of dissociation is
 more serious for the hypothesis that P300 latency is a
 measure of decision latency. This kind of dissociation
 seems to come about whenever equivocation is low
 but uncertainty  or probability is the  variable. (In in-
 formation theory,  a low-probability  stimulus  has
 greater uncertainty than a high-probability stimulus.)
 As uncertainty increases,  P300 latency remains  the
 same (or it may even decrease), while  RT increases by
 a considerable amount (Hyman 1953). The situation
 is further  characterized by the fact that the increase
 in  RT often is accompanied by an increase in P300
 amplitude. For example, when comparing a condition
 where the subject does not know in advance what
 stimulus will occur next to a condition where the sub-
 ject does know, RT  increases considerably with  the
 added uncertainty, but P300 latency  does not. P300
 amplitude may increase (Tueting and Sutton 1976) or
 not  (Donchin et al. 1973), apparently  depending
 upon other  factors  such  as pressure for  fast RTs.
 Hillyard suggested that these dissociations between
 RT and P300 latency when uncertainty (or probabil-
 ity) is the variable  could represent a disruption of
 motor processing as  a result of the  emotional con-
 comitants of the added uncertainty or by some other
 confounding of decision latency with probability.

    The theory  that P300 is  directly related to a
 decision  involving response selection encounters still
 another obstacle. It has been reported that the peak
 of the parietal-occipital P300 component in a partic-
 ular trial can occur, under certain circumstances, after
 the subject's  reaction in that trial (Eason et al. 1969,
 Ritter et al. 1972, Rohrbaugh 1973). The finding that
 P300 latency can occur after the  reaction indicates
 that the correlation  between parietal P300 latency
 and decision latency  is not a causal one. Ritter (this
 section) concludes  that even P300 onset cannot  be
 causal to reaction when the time between the initia-
 tion of motor processes in the brain and RT is precise-
 ly considered.

    In summary, the theory that P300 latency is
 directly related to  an independent RT measure  of
decision latency runs into trouble. There are reports
 that P300  latency and RT can be dissociated, that
response-locked averaging can result in a smaller P300
than stimulus-locked averaging, and that the peak of
P300 can occur after the reaction.

    A casual relationship between  P300 and reaction
 can  probably  be excluded, but P300 may still  be
 indirectly related to response selection. Hillyard sug-
gested that  RT-P300  latency  dissociatons   could
 be explained  if P300 reflects an early decision such as
 "the stimulus belongs to  a relevant class" rather than
 to a later decision such as "which stimulus it  is and
 what  response  is required"; subjects may hold off a

-------
ERPs and Information Processing
                                             167
"one  of a relevant class decision"  until a target is
detected with low-probability targets, but not hold
off with frequently occurring targets. Alternatively,
the P300 could simply represent a phasic arousal pro-
cess related  to  the  consequences of the  decision.
Simson  et al. (1976) have proposed  that P300 could
be related to "conscious awareness of the outcome of
the decision." Finally, P300 could reflect adjustments
of memory representations  and  set  that are conse-
quent to the decision  on the current trial but may be
reflected in  decisions on future trials (see section
below on Feedback and P300).

    Ritter concluded that P300  does not reflect the
selection of a particular target within a class of stimuli
in vigilance tasks because (1) P300 follows rather than
precedes reaction  time  and  (2) stimulus  probability
appears to be a more  potent variable than whether or
not the stimulus is a "target."  Ritter (this section)
proposes that the earlier negative component may be
more directly related to decision latency than isP300.

P300 and template match/mismatch

    Decision processes are no doubt different for the
different  tasks that elicit P300  (guessing, feedback,
detection). An underlying feature, however, could be
a  match/mismatch judgment  (Ritter and  Vaughan
1969, K. Squires  et  al. 1973a,  Posner  et  al. 1973,
Thatcher  1977) involving comparison with a template
stored in memory. One proposal is that mismatch
involves  extra processing time compared to match.
RT is shorter for match than for mismatch judgments,
and Posner et al. (1973) have reported that P300 for
match judgments  occurs earlier  than P300  for mis-
match judgments. A  second  proposal is  that match
judgments are accompanied by a larger P300 because
the stimulus fits a representation stored in memory.
The fact that P300 amplitude is larger for match than
for mismatch judgments in auditory threshold tasks
(Hillyard et al. 1971) and in semantic tasks (Thatcher
1976) supports this latter proposal.

    The problem is that the above  match/mismatch
premises do not hold when probability is the variable.
RT results are consistent in that RT is shorter for a
high-probability stimulus (match) than for a low-prob-
ability stimulus  (mismatch).  However, P300 results
do not  flt-P300  components recorded  for a high-
probability  stimulus  (match) in identical  tasks are
smaller in amplitude and equal, or even later, in laten-
cy  than for a low-probability stimulus  (mismatch)
(Tueting  1968,  Friedman et al.  1973,  Ritter et al.
1968, Ford 1975).

    According to orienting response theory, the tem-
plate is inferred to be for a high-probability stimulus
because  of its greater frequency and familiarity (Lynn
1966). In an orienting  response  framework, a low-
probability stimulus is viewed as  a mismatch because
it is unexpected, and Ritter and Vaughan (1969) have
proposed that P300  represents additional perceptual
and cognitive processing called in to evaluate the sig-
nificance of.a mismatch.  Mismatches in orienting
response theory usually involve low-probability stim-
uli that are also novel. With  the exception of Cour-
chesne et al. (1975), however, low-probability stimuli
in EP studies have not necessarily been novel.

    In summary,  RT results are consistent in that RT
is longer for  mismatch, which corresponds  to the
hypothesis  of longer  processing time for a mismatch.
However, there seems to be a paradox for P300 ampli-
tude and latency.  For threshold and semantic-meaning
tasks, a match is  related to a larger amplitude, ear-
lier latency centroparietal P300. However,  for proba-
bility situations, a match is related to smaller ampli-
tude and equivalent  (or later) latency P300 (frontal
and parietal).

    Donchin proposed that some of the critical ques-
tions concerning  match/mismatch  theory  are "How
do we infer what the template is?" and "Why are some
templates singled  out for association with P300 while
others are  not?"  One way  out of  the paradox is  to
postulate two different match/mismatch processes for
P300, one for probability (frontal P300?) and one for
more complex target-detection tasks (parietal P300?).
In  addition, low-probability events are inherently
more significant to the organism than high-probability,
familiar events, even if the task does not specify them
as targets.  Some  developmental theorists  have pro-
posed that  infants learn  by attending more to events
that are lower in probability and slightly discrepant
(e.g., Kagan 1972). In a certain sense then, low-prob-
ability events are  target-like, and it is clear that P300
is elicited by target events. The inference of a template
for low-probability events on the basis of their inher-
ent target-like nature would have the advantage of fit-
ting all  of the P300 results  into the same match/mis-
match logic. Note, however, that  RT results would
not fit  the notion of a  template for low-probability
stimuli-RT  is  longer for low-than  for high-proba-
bility events.

    Because inferring what the template is,  or how
many templates there  are, may be difficult, as the
above  discussion  demonstrates, progress   may be
made in using experimental designs specifically aimed
at memory search (cf. Roth et al. 1975, Posner et al.
1973, Thatcher 1976, Poonetal. 1976). Roth's design
involves storage of specific templates with subsequent
probing of memory for  the template. An analysis  of
the relationship between P300 and template match/
mismatch has been made by Ford (this section).

P300 and feedback

    Dissociation between P300 latency and RT, the
finding  that P300 can occur  after  the response, and

-------
 168
                                         Tueting
 the fact that response-locked averages can attenuate
 P300 bring into question the generality of the decision
 construct for P300.  The  decision concept can be
 reconstructed, as Donchin has done, if the argument
 is  made that P300 may not be so much related to
 decisions of response  selection in the current trial as
 to modifications that  will be reflected in decisions in
 future trials.  This view of  P300 is supported by the
 fact that large PSOOs are obtained to stimuli delivering
 feedback concerning guessing outcome  or discrimina-
 tion-task accuracy, and to low-probability or novel
 stimuli requiring no task whatsoever. Response selec-
 tion is not required immediately following stimulus
 presentation in these situations.  However, stimulus
 information in these cases may influence future gues-
 ses, future judgments, or future reactions to the same
 novel  or low-probability stimulus. In this conceptual-
 ization, stimulus  presentation elicits a readjustment
 of  templates or  relative probabilities, reevaluation
 of relative costs  of subsequent decisions, etc.  For
 example, Donchin suggested that the  sensitivity of
 P300  to probability could reflect the greater need for
 readjustment  following the  presentation  of a rare
 event.
     In a  guessing  task, response  selection  occurs
 before stimulus presentation; therefore, the fact that
 a  large and reliable  P300 is elicited implicates P300
 involvement in feedback. Feedback can be inferred
 here in terms  of its information properties, i.e., the
 subject  finds  out  whether his guess was right or
 wrong. If he made a high-risk guess, the feedback may
 be interpreted  differently than if he had made a low-
risk guess (Tueting and Sutton 1973).


     In a guessing situation,  P300 is larger to sound
 after light (crossmodal) than to sound after sound
(ipsimodal), and the same is true for light (Levit et al.
 1973,  Zubin and Sutton  1970).  These results corre-
spond  to RT data; crossmodal sequences  yield longer
 RTs than ipsimodal sequences (Waldbaumetal.  1975).
Pilot data  also suggested that P300 amplitude in the
current  trial  depends  upon whether the  subject's
guess in the previous trial was right or wrong (Sutton,
unpublished data).


    Tueting (1968) and Tueting et al. (1970) reported
that P300 in a  guessing task was large for a low-prob-
ability event whether  the event was  determined in
terms  of stimulus probability, sequential probability
(probability of alternation or of repetition), or proba-
bility  of the outcome of the subject's guess.  These
effects were strongest for correct guess outcomes, and
there was some indication that probability of outcome
was an additional  relevant factor. The fact that the
probability effect held for probability of repetition as
well as for probability of alternation and the fact that
 the probability effect  differed for right and wrong
 outcomes indicates that sequential effects for P300
 are not a simple function of time separating identical
 stimuli in a sequence,  and  hence  are not a result of
 recovery  or  simple habituation.  These sequential
 effects, however, could be  related  to cognitive varia-
 bles such as the subject's running calculation of absol-
 ute and sequential probabilities of events and of the
 calculated risks involved in making one guess rather
 than another. The right-wrong dimension in the gues-
 sing task can be interpreted within a match/mismatch
 framework, and templates for probability and analysis
 of  risk  can be  inferred to  develop  in  probability
 learning tasks (Leifer et al. 1976).

     K. Squires et al. (this section) have investigated
 sequential P300 effects in a random program of rare
 target stimuli and frequent background stimuli. These
 effects go  back  in  sequence  at least  10 stimuli (or
 11.7 sec) when the probability of the target is 10%
 and the probability  of the background stimulus 90%.
 The results indicate that a stimulus elicits  a  larger
 P300 if preceded by more of the same than if preceded
 by different stimuli. A similar result has been reported
 by Barrett et al. (1975) in a  threshold situation.

     A more complete discussion of the relationship
 of P300 to feedback can be found in  Sutton  et  al.
 (this section).
 Summary

     Conference discussion  confirmed the fact  that
 evoked potentials (EPs) reflect a number of cognitive
 variables in a systematic manner. Discussions revolved
 around three EP components: Nl, P300, and a nega-
 tive component preceding P300. The latter negative
 component is similar to P300 in that it too is endog-
 enous, but unlike P300, it is modality specific.

     Evidence  is accumulating  that Nl  is related to
 stimulus set, an early  stage of selective attention, in
 auditory, visual, and somatosensory modalities. How-
 ever, most data have come from studies using auditory
 stimuli.  The level  of attenuation  or gating  in the
 nervous  system, whether central or peripheral,  and
 the amount of attenuation in terms of the complexity
 of  the  information being processed were debated.
 One aspect emphasized was the importance of precise
 definition of selective  attention based on behavioral
 measures, preferably obtained  on the basis of signal
 detection theory.

    The relationships of P300  to response set, selec-
tive attention, decision, feedback, expectancy, orient-
ing, uncertainty, etc., were discussed. The  focus was
on  the  proper methods  for relating two sources of
data-on one hand,  behavioral data on cognition, and

-------
ERPs and Information Processing
                                            169
on  the  other hand, EP data. One important behav-
ioral paradigm  for measuring  the time required for
various  mental  operations, reaction time (RT), was
discussed  at length. Certain methodological issues
arise when comparing  RT data to EP data. Conse-
quently, there  is considerable  difficulty in making
sense of the relationships between RT, P300 latency,
and P300 amplitude that have been obtained to date.
Much theorizing in the area of cognition involves the
notion  of a comparison  of stimulus input  against
stored memory representations. In this context, the
usefulness of the match/mismatch logic for interpret-
ing P300 was assessed,  and the difficulty of defining
the template was pointed out.

    The negative component preceding P300 received
a great  deal of attention. At the present time, data
related to this new component are limited, but dis-
covering where this component fits in human cogni-
tion is an important challenge for future study.


    There is considerable data and  theory in cogni-
tive psychology that EP researchers could use  to
advantage in designing experiments  and interpreting
results. EP data may well prove fruitful for an increas-
ed understanding of the area of cognition as well. EPs
may serve to validate inferences made on the basis of
behavioral   measures  concerning  what   intervenes
between  stimulus input and response output. In any
case, it seems likely that these two areas of investiga-
tion-cognition and evoked potentials-should even-
tually mesh. For the time being, any contradictions
are likely to be particularly fruitful areas of research.

-------
 HOW MANY LATE POSITIVE WAVES ARE THERE?'
 W. T. ROTH

 Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine,
 Stanford, and Veterans Administration Hospital, Palo Alto, CA, U.S.A.
    Three variables define the  relatively  faceless phe-
 nomena  to which our research efforts  are devoted:
 voltage, time, and location. The interactions of volt-
 age and time that create peaks with various latencies
 and polarities give a name to the topic of this discus-
 sion late  positive waves (LPWs).  This family of waves
 has also been called collectively P3 or P300jThe lat-
 ter term, which  implies a mean latency of 300 msec,
 is increasingly less descriptive as later and later LPWs
 are reported.  If the term LPW is reserved for waves
 after P2, a rather flexible minimum latency must be
 accepted for LPWs. The P2 component in  the auditory
 mode has mean  latencies that vary at least from 170
 to 205 msec, depending in  part on  task parameters
 (Roth et al. 1976a). Visual P2s are later, extending to
 around 280 msec. Thus, the earliest LPW, P3a, is some-
 times hard to distinguish from P2. The P3 a component
 may merge with P2 or it may be as late  as 300 msec
 (Roth 1973; N,  Squires et al. 1975; Snyder and  Hill-
 yard  1976; Ford et al. 1976).  It is seen most clearly
 when stimuli  are rare and  task-irrelevant. The  P3b
 component usually  occurs in the 300- to  400-msec
 range for rare task-relevant  stimuli. P3b  is the stan-
 dard P300 described in many earlier papers by Sutton,
 Donchin, Tueting, Hillyard,  and  others.  A "positive
 missing stimulus potential" appears  when  stimulus
 omissions from  a regular ?3imulus  train are rare and
 when these  omissions are targets (e.g., Simson et al.
 1976). In the Simson et al. experiment, this wave had
 a mean  latency of 465 msec  for auditory stimuli and
 565 msec for visual stimuli. A peak termed "P4" with
 a latency of 650 msec has been reported by Picton et
 al. (this volume). It was elicited by  an auditory feed-
 back stimulus that gave information as to  the correct-
 ness of a choice in a visual concept learning task.
 Finally,  Courchesne (1976) found that rare  task-rele-
 vant visual targets that evoked a positive wave at 417
msec in adults, evoked a similar wave with a  mean
latency of 702 msec in 5- to 8-year olds. Equally rare
 but nontarget stimuli  evoked a positive wave with a
mean latency of  448 msec in adults and 982 msec in
 the children.
    The components listed above can be called "peaks"
 since they rise and fall in a few hundred milliseconds.
 A late slow positive wave at Pz has been observed to
 follow rare task-relevant stimuli. It begins some time
 after 300  msec and lasts for more  than  1 sec (N.
 Squires et  al.  1975, K. Squires et al. 1977, N. Squires
 et al. 1977). Either CNV resolution or skin potential
 artifacts at the reference electrodes can  produce pro-
 longed shifts,  but these are unlikely explanations for
 the origin  of  this slow wave because  of its scalp dis-
 tribution.

   Of the  three variables that characterize the phe-
 nomena, location  has  become the most  favored  in
 arguments  for the distinctness or unitary nature  of
 the various LPWs. At  least  four  different anterior-
 posterior scalp distributions have been described for
 these waves: (1) A predominantly parietal-central dis-
 tribution (Fz  < Cz <  Pz > Oz) is characteristic  of
 most peaks in  the P3b latency range and missing stim-
 ulus potentials. (2) A  predominantly frontal-central
 distribution (Fz = Cz > PZ > Oz) is said to be charac-
 teristic of the P3a component, or P3s to  task-relevant
 stimuli to which no motor response is required (Hill-
 yard et  al.  1976, Tueting and Sutton 1976), and  of
 unfamiliar rare task-relevant nontarget visual displays
 in adults (Courchesne et al. 1975), or of all rare task-
 relevant, nontarget visual displays in  children (Cour-
 chesne  1976). (3) A  parietal-occipital distribution
 (Fz < Cz < Pz = Oz) was found for "P4." (4) A distri-
 bution in which the polarity differs by lead was found
 for the slow wave.  This wave is positive at Pz, almost
 absent at Cz, and negative at Fz.
   If  there  are  four topographic distributions  of
LPWs, does it follow that there are four different pro-
cesses? If "process" is considered to be an explanatory
framework that brings order to our observations, the
distinctness  of phenomena observed in a particular
experiment cannot be considered sufficient evidence
for the distinctness of the processes underlying them.
     preparation of this paper was supported by NIMH Grant DA 00854 and the Veterans Administration.

-------
How Many Late Positive Waves?
                                             171
Although  there is an inherent ambiguity in grouping
things as "like" and "unlike," an ambiguity similar to
that of defining figure and ground for a pattern recog-
nizer, certain  guidelines are implictly  followed in
classify ing evoked potential phenomena. First, distinct
processes  should be  represented by phenomena that
can  be varied independently.  Statistical procedures
such as principal component analysis are of value here,
but there are many  pitfalls  in determining the inde-
pendence  of phenomena. Some of these pitfalls are
discussed  by Weiskrantz (1968) in  his critique of the
concept  of "double dissociation." Suffice it to say
that processes related to each other in a hierarchical
manner may appear to be dependent  until just the
right experimental manipulation is discovered. Second,
processes  are often thought of as distinct only if they
are discontinuous and intermediate processes are never
encountered. Whatever combination of voltage, time,
and location define the two processes, phenomena
should not  fall into a continuous spectrum between
the two definitions.

   The criterion of discontinuity for distinctness can
be  made  clearer by  reference  to a few examples. If
four different auditory stimulus intensities resulted in
four significantly  different  Nl amplitudes, no  one
would be  misled into thinking that this was evidence
that four  different processes were operating, since Nl
is known  to be a continuous variable related to sensa-
tion level by a reasonably simple monotonic function.
Thus, the four distinct amplitudes only  represented
the experimental parameters chosen. Or if LPWs of
four distinct latencies were created,  principal com-
ponent  analysis might yield four orthogonal compon-
ents. It might even be possible to vary the amplitudes
independently from the latencies. Yet, since available
data suggest that latency is a continuous function of
speed of  stimulus  evaluation, latency  differences are
not considered sure-fire evidence for distinct proces-
ses. For example,  Kutas and Donchin (this section)
found that LPW latency varied  more than 200 msec,
depending on how  visually  displayed  words were to
be  evaluated. More  complex  evaluations that  took
more time, such as deciding whether one word was
synonymous with another, resulted  in  later LPWs
than simpler evaluations, such as identifying a fixed
target word. Similarly,  target  stimuli  that are more
easily discriminable  give shorter latency LPWs  than
less  discriminable  targets (N.  Squires  et  al.  1977).
Children probably take longer than adults to process
stimuli, which could explain some of the longer laten-
cy  LPWs  for children  found by Courchesne (1976).
Even if age has an independent effect on LPW latency,
it is likely to act continuously in that  older children
would have latencies intermediate between  those of
younger  children  and  adults.  The realization  that
LPW latency is a continuous variable makes it easy to
accept the conclusion of N. Squires et al. (this section)
that P3bs elicited by  rare target stimuli of various
intensities and by the  omission of stimuli in a train
are part  of the same process or "functionally equiv-
alent."  These two LPWs have the same scalp distribu-
tions, are affected in the same way by manipulations
of probability, and show  discrepancies in amplitude
and latency that could be explained on the basis of a
continuous variable such as amount of time jitter or
latency of recognition.

   P3a presents more difficult detection and classifica-
tion problems.  It would be convenient if P3a were
simply a very early LPW consistently elicited when
stimulus-processing demands are  minimal. Unfortu-
nately, LPW latency is  not always shorter when sub-
jects are  told to ignore stimuli or given a reading task
(Roth et al. 1976b; and this section). K. Squires et al.
(in press) had difficulty finding P3a in a replication of
N. Squires et al.  (1975).  K. Squires et al. (in press)
did find  a factor that  had peak loadings in the 250
msec  range, but  it accounted for only  2.6% of the
variance  and  the  factor scores were not consistently
related to experimental variables. However, in spite of
ah1  these difficulties  in specifying the exact  circum-
stances  under which positive waves  with latencies
between  200 and 300 msec can be produced, it is
impossible to disregard the evidence that such peaks
exist.

   Evidence for the distinctness of P3a and P3b de-
rives  primarily  from the different topographic distri-
butions of the two components. This raises a question
of the validity of topographical differences as criteria
for establishing the distinctness of a process.2 There
is no logical priority for topographic distribution over
amplitude and latency as a criterion  of distinctness.
The power of distribution as an argument for  essential
differences comes from the concreteness of being able
to imagine generators at spatially distinct locations in
the brain. In fact, different amplitudes and different
latencies, even without different .scalp distributions,
may also be produced by different physical generators
in the sense of different pathways and different selec-
tions of single units. However, these mechanisms  are
so poorly understood  that they are not even a topic
of speculation in the LPW literature. Just as  differen-
ces in amplitude or latency do not necessarily imply
different processes, neither do different distributions
and  different  generators.  Take,  for example,  the
experiment of Kutas and Donchin (1974) in which
the readiness potentials of right-handed subjects were
larger over the hemisphere contralateral to the respon-
ding hand. Two  distinct distributions occurred, each
undoubtedly produced by spatially separated neural
masses.  The authors, however,  never claimed  that
 they had fractionated the readiness potential into two
 *See the Scalp Distribution section of this volume for an extended discussion of related issues.

-------
 172
                                           Roth
 distinct processes, RPr and RP1, because the mapping
 of motor ar. J sensory processes on the cortex relates
 asymmetry  and  task parameters in a fairly simple
 manner. The relationship is not completely continuous
 given the neuroanatomy of the cortex, although inter-
 mediate distributions might be  obtained if the sub-
jects pressed foot pedals. That is, the unity  of a proc-
 ess may  depend  on structural and  developmental
 considerations rather than the  continuity of mathe-
 matical descriptors.

   Similarly, a  unitary  LPW  process that changes
 smoothly in latency and distribution depending on
 stimulus at processing time might be hypothesized.
 For instance, P3a and P3b could simply represent two
 samples of  this  function at different points in time.
 At early latencies, the process would be more frontal,
 and at later latencies, more parietal. Thus,  latency
 and distribution would be dependent. Unfortunately,
 there is insufficient evidence to claim continuity.

   Ford et al. (1976) did find distributions intermedi-
 ate between P3a and P3b in an experiment where an
 attention variable had three levels. These results would
 have  been more  convincing if  the amplitude of P3a
 had been larger. The flat anterior-posterior distribution
 obtained when attention was not directed to the stim-
 uli may have been a floor effect related to the effect
 of noise on their measurement method. They located
 peaks as the maximum or minimum voltage  in a speci-
 fied latency range and, as the signal-to-noise ratio de-
 clined, noise peaks rather than signal peaks may have
 been  detected. Thus, the peaks located would never
 be smaller than the "floor" of fluctuations produced
 by the noise.

   Also damaging to the hypothesis of a single genera-
 tor for P3a and  P3b is the fact that, even within  a
 single  experiment, latency and distribution may be
 independent. Courchesne et al. (1975) reported  the
 same  latencies for LPWs with  a frontal  distribution
 elicited by complex visual displays and for LPWs with
 a parietal distribution elicited by simple visual displays
 (digits). However, these results do not justify postal-
 lating two  distinct processes.  Instead, these results
 suggest that experiments could be devised to test the
distributional continuity of LPWs by varying stimu-
lus familiarity. If a parameter were discovered that
controlled  LPW  distribution continuously, a good
cue could be made for  considering LPWs  part of a
 single process.

   Another type of implicit evidence for the distinct-
ness of processes is the presence of visually distinct
peaks in an evoked potential tracing. However, P3a
and P3b are so close together, and P3a is so close to
P2,  that visual separation is  not  convincing. Even if
 two peaks were discontinuous in latency (i.e., never
 overlapping), multiple peaks could represent a single
process that requires the repetition of a mental oper-
 ation or  a single  process that yields multiple peaks
because  of its specific neuroanatomy (e.g.,  multiple
parallel pathways). In  these cases, principal compon-
ent analysis would be helpful.
   Even peaks that differ in both latency and distri-
 bution  could represent the same process. This may be
 the case for theP4 wave of Picton et al. (this volume),
 which appears in  their  Fig. 5 to occur in the same
 tracing  as an earlier P3. Both peaks are affected iden-
 tically  by the experimental variables. The  authors
 conclude  that "this  distinction" (i.e.,  latency  and
 scalp distribution  differences) "makes it possible to
 hypothesize that the  two waves reflect separate pay-
 chophysiological processes, possibly the  appreciation
 of feedback information  (P3)  and  its utilization in
 conceptual learning (P4)." Based  on the available evi-
 dence,  the authors' conclusions  seem to  be unwar-
 ranted. The peaks are highly correlated in appearance.
 The  most logical argument would be for the authors
 to contrast  their  findings with those of the  many
 other studies that  have reported P3s without P4s. At
 best, this would open  the  possibility that  the two
 peaks represent separate processes.
   In general, the case for multiple LPW processes is
unproven. More data are needed, particularly on the
experimental parameters that affect late wave distri-
bution. If  other ERP processes play an interactive
role, such  processes may not be positive onw-i.e.,
interactions of slow negative processes  with LPWs
may change  LPW  distributions  as well as latencies.
For example,  N. Squires et al. (in press) observed a
0.7S correlation between distributional differences in
P300 amplitude and slow wave size, and a correlation
of 0.86 between latency differences and slow wave
size. Other negative  phenomena have been reported
by  Loveless  and Sanford (1974), Rohrbaugh et al.
(1976), and  Roth et al. (1976a). Rohrbaugh et al.
described a frontal negativity that increased over time,
a  finding  that  could  explain  the  observation by
Courchesne et al. (1975) of LPWs that initially show-
ed a  frontal distribution, but shifted in  a posterior
direction across trials. The nature of these  negative
processes and their interactions with  LPWs remain to
be clarified.

-------
LATENCY OF EVENT-RELATED  POTENTIALS
AND REACTION TIME
W. RITTER

Lehman College, City University of New York, New York, NY, U.S.A.
    Of  the  various  event-related  potentials (ERPs)
elicited by target stimuli in reaction time (RT) experi-
ments, the P3 component (hereafter assumed to have
a parietal maximum) is the only one previously known
to have  a range  in latency great  enough to be related
in time  to RT.  Depending on the experimental con-
dition, the peak  latency of P3 based on averaged ERPs
has been found to vary anywhere from 250 to 600
msec. The range in latency  for single-trial P3s for an
individual subject within a single condition can be as
great as  300 msec (Ritter et al. 1972). The large varia-
bility in P3 latency both within and across conditions
is consistent  with most theories  concerning the func-
tional significance of physiological  activity underlying
P3, such as information  delivery, target selection,
decision-making, sensory discrimination, and match-
mismatch processes.  (For review, see the summary of
preconference   correspondence,  this   section.)   In
addition, these theories tacitly or explicitly presume a
correlation between  P3 latency  and RT. As Hillyard
pointed  out  in  preconference correspondence, some
studies  have found  a correlation  (e.g., Ritter et al.
1972, Picton et  al. 1974), while  others have not(e.g.,
Karlin et al. 1970, Karlin and  Martz  1973).  Kutas
and  Donchin (this section) have shown the correla-
tion between P3 latency and RT  is affected by the
trade-off between speed and accuracy of response,
a finding which may account  for some of the  dis-
crepant  results.

   There is a more  serious problem for most of the
theories  of P3  function than  whether or not  the
latency of P3 correlates  with RT. Ritter et al. (1972)
concluded that P3 onset could occur early enough to
be causally related to RT. That  conclusion was based
on estimated average delays for each subject between
P3 onset and the initiation of motor activity in motor
cortex. For one subject (WR), the estimated average
delay was 20 msec; for other subjects, it was longer.
It therefore appeared possible  forP3 onset  to precede
motor activity,  though it was suggested that the N2
component, which peaks about  100msec earlier than
P3, may reflect a significant  factor in the timing of
RT. In retrospect, it appears likely that a large pro-
portion  of motor  responses  must have  begun  be-
fore the onset of P3. Otherwise, the distribution of
delays between P3 onset and the initiation of motor
activity on single trials would have been surprisingly
small in variability  (an unlikely possibility since  the
correlations  between  P3  and RT  latencies were
generally in the 0.70s  or were quite skewed). There
is certainly no justification in hypothesizing that P3
reflects such processes if subjects can make  discrim-
inative motor responses prior to P3.

    Several investigators have  recognized that  P3
occurs too late to reflect the processes usually attrib-
uted to it  and have accordingly  suggested  that P3
reflects sequalae to  task-related decisions and respon-
ses, such as registration of pertinent  information in
memory, resetting of perceptual analyzers, or other
processes associated with preparation  for future trials
(Donchin et al.  1973, Picton and Hillyard,  1974).
Since P3 apparently occurs too late to reflect target
selection  or information  delivery, we turned our at-
tention to N2, a component that occurs early enough
in  time  to be causally related to motor responses,
appears to be elicited in the same  kinds of situations
that elicit P3, and, like P3, is endogenous in nature.
    There seem to be two reasons for the late arrival
of N2 on the ERP scene. First, N2 is smaller in ampli-
tude than P3 and is often obscured by P2 because of
overlapping latencies of the two components. Second,
since N2 is smaller in amplitude than P3, less variabil-
ity in latency from  trial to trial is required for N2 to
be observed in averaged ERPs. Klinke et  al. (1968)
resolved both of these difficulties with one stroke by
randomly omitting stimuli in a train of stimuli deliver-
ed at a steady, fast  rate. The omitted stimuli elicited
clear N2 and P3  components. The use of a steady,
fast rate of stimulation presumably permitted subjects
to develop internal rhythms that produced excellent
time-locking of N2 to stimulus omissions. The circum-
stance that P2 (an exogenous potential) is not elicited
by omitted stimuli prevented P2 from obscuring  N2.

    Two reports  support the  time-locking explana-
tion. Picton et al. (1974) found that increasing  the

-------
 174
                                                                                                   Ritter
 interstimulus interval in a train of stimuli which con-
 tains randomly omitted stimuli  resulted in smaller
 and eventually unobservable N2 components associat-
 ed with stimulus omissions. (A similar result occurred
 for P3, but it could  still  be observed at the longest
 interstimulus  interval used  because of its greater
 amplitude.) Ruchkin and Sutton (in press) have con-
 cluded that the failure to observe  N2  for omitted
 stimuli that elicited P3 in Sutton et al. (1967) was
 due, at least in part, to the degree of trial-to-trial varia-
 bility in the latency of N2.


      Although  the  Klinke  et  al.  report  appeared in
 1968, P3 continued to hold center stage in theories
 concerning target selection and related processes, pre-
 sumably  because it was not clear that P3 latency was
 too long to reflect  those processes. When the impor-
 tance of N2 became clear  because of RT data, it was
 hypothesized that  N2 and  P3  were both components
 of a  complex  waveform emanating from the  same
 brain tissue. Topographic analyses of ERP associated
 with omitted stimuli (Simpson et al. 1976), however,
 showed  that N2 and  P3   had different intracranial
 sources.  Furthermore, N2 was modality-specific in
 its distribution, whereas P3  was not.

     Similar results were obtained in a vigilance experi-
 ment in which the targets consisted of random changes
 in  physical  parameters (Simson  et al.  1977). The
 modality specificity of N2 suggested that different
 neural ensembles accomplish auditory and visual dis-
 criminations. The circumstance that  N2 preceded P3
 in  latency  for ERPs elicited  by  stimulus omission
 indicated that  N2 reflected the detection  of omitted
 stimuli and P3 some  other process.  Results of the vigil-
 ance study suggested that N2 reflected target selection
 whether  the target was an  omitted stimulus or a
 change in parameter of a physically present stimulus.
 These results  are   detailed elsewhere  (Ritter,  this
 volume).

     As pointed out  earlier P3 is the only component
with  a range in latency  hitherto known to  be great
enough to be related to RT.  If N2 reflects target selec-
tion or related processes, then N2 latency should vary
as a  function of the difficulty of target selection and
correlate with RT.  Recently completed data analysis
supports that conclusion (Ritter et al,, in preparation).
Reanalysis of the data of Ritter et al. (1972) indicates
that, in single-trial measurements of N2  elicited by
the targets of the vigilance task, the mean latency of
N2 increased across conditions as a function of the
difficulty  of the discrimination of the targets within
conditions. Furthermore, the single-trial analysis yield-
ed  somewhat greater product-moment  correlations
within conditions between  N2  and  RT than was pre-
viously found between P3 and RT.
     On  the  basis of these considerations,  it seems
 appropriate to propose that the earlier hypotheses
 relating P3  to information delivery, target selection,
 and related processes be shifted  from P3 to N2. In
 this case, the only relevant hypotheses concerning the
 functional  significance  of  the  physiological activity
 underlying P3 would be those associated with prepar-
 ation for future events. Since there are no published
 studies  of these alternative  possibilities, there  is a
 clear need  for experiments that examine new hypo-
 theses concerning the functional significance  of P3.

     Hillyard (personal communication) has raised the
 possibility  that,  although P3 is too late  to directly
 reflect target selection, it might nevertheless indirectly
 reflect selection by virtue of a  delay between critical
 neural activity and the manifestation of P3  at the
 scalp. Somjen (this volume)  indeed mentions the
 possibility  that some late components could be due
 to glial potentials (with a  delay  of as much as 100
 msec between neural activity and the detection of a
 glial potential  at the scalp), Thus, Hillyard (personal
 communication) points out:  "Suppose, for  instance,
 that the P3 was  generated by  the  depolarization of
 glial cells consequent upon the  release of K + ions by
 the active neurons which do  the selecting. The delay
 between the neural activity and its direct glial 'reflec-
 tion' could account for the delay of P3 re the selective
 response."

     Although  this possibility must be considered, it
 opens a Pandora's box. At  present neural  and glial
 potentials at the scalp cannot be distinguished, except
 for  early potentials with a  latency presumably too
 short to reflect  glial potentials.  The sequence with
 which the neural activity associated with various late
 potentials  would thus be   unknown. Imagine,  for
 example, that either P2 or N2 (or both) reflect neural
 activity directly and P3 reflects glial potentials gener-
 ated by neural activity  that precedes the neural activi-
 ty of either P2 or N2 (or both). Some glial potentials
 might be associated with  a  delay  of 50 msec and
 others with 100 msec or values in between. Thus, the
 neural activity of P2, N2, and P3 could occur in any
 sequence imaginable, including the possibility of their
 simultaneous occurrence. In  view o f the  havoc these
 possibilities entail, and the variety of post hoc specula-
 tions open  to the whim of a given investigator's theo-
 retical inclinations, the following is  suggested: until it
 is  established  empirically  that scalp-recorded ERP
 components can reflect glial  potentials, all ERP com-
 ponents (with the possible exception of dc potentials)
should be considered to be directly related to neural
 activity and, therefore, the latency of a component
should indicate the time of occurrence of the neural
 events that produce that component. These assump-
 tions appear to provide the most parsimonious frame-
work currently available  for  the  interpretation of
ERPs.

-------
EQUIVOCATION AND P300 AMPLITUDE


D. S. RUCHKIN

Department of Physiology, School of Medicine, University of Maryland, Baltimore, MD, U.S.A.

S. SUTTON
Department of Psychophysiology, New York State Psychiatric Institute, New York, NY, U.S.A.
    The  purpose of this paper is to briefly review
evidence for the concept that the information received
by a subject from the occurrence of an event deter-
mines, in part, the amplitude of the P300 component
elicited by the event. We do not mean to imply, how-
ever, that other variables are not also relevant in deter-
mining P300  amplitude or that the  P300 wave is a
unitary phenomenon.

Review
    In the framework of classical information theory
(Shannon and Weaver 1949), the amount of informa-
tion provided by an event is related to the subject's
a priori uncertainty of the  event's  occurrence; the
lower the a priori probability of occurrence, the great-
er the information provided.  However, the amount of
information received by the subject equals the infor-
mation provided  by the event minus an information
loss related to the subject's a posteriori uncertainty of
having correctly perceived the event; the greater the
a posteriori uncertainty, the greater  the information
loss. In information theory terms, this loss is referred
to as equivocation.

    The role of a priori event probability was explicit-
ly recognized  in the initial investigations of P300
(Sutton  et  al.  1965). Sutton et al., Tueting et al.
(1970),  Friedman et  al. 1973), and Ruchkin et al.
(1975) have used a paradigm  in which the subject
guessed what the ensuing stimulus would be. When
stimulus probability was manipulated, P300 amplitude
varied inversely with the joint  probabilities of the
stimulus and guess (outcome probability). Donchin et
al. (1973) modified this procedure  by varying the
complexity of the  sequence in which stimuli were
presented. P300s were smallest when  simple, easily
predictable sequences were used and  became progres-
sively  larger as the sequences became more complex
and thereby more  difficult to predict. Karlin and
 Martz (1973) investigated the  P300 component in
 choice-reaction  time  and delayed-choice paradigms.
 They reported that stimulus and response probability
 jointly  influenced  P300 amplitude,  with  the least
 probable combination eliciting the largest P300. Paul
 and Sutton (1972) and Squires  et al. (1975b) varied
 the probability of occurrence of the signal in auditory
 signal detection  experiments. The  P300 elicited by
 detected signals  was  largest when signal probability
 was lowest.

     Squires et al. (1973) used  a paradigm in which
 the subject performed an auditory intensity discrimi-
 nation.  In each trial,  the subject indicated the degree
 of confidence in his decision. This was followed by a
 visual stimulus that gave feedback on accuracy. It was
 found that when the stimulus confirmed a high-confi-
 dence decision, P300  was smallest. When the stimulus
 discontinued  a  high-confidence decision,  P300 was
 largest.

     A common element of the experiments described
 above is that  the stimulus provided information that
 resolved the subject's uncertainty with respect to a
 prior expectation. As the subject's  estimate of the a
 priori probability of  the occurrence of an event  de-
 creased, the amplitude of the P300 elicited by that
 event increased. To the extent  that the information
 content  of an event may be viewed as being inversely
 proportional to its a priori probability, the amplitude
 of P300 can be interpreted as reflecting, in part, the
 amount  of information provided  by the event.

    There also have been several reports of the effect
 upon P300 amplitude of the subject's a  posteriori
uncertainty of perception. Mast and Watson (1968),
 Hillyard et al. (1971), Paul  and Sutton (1972), and
 Squires  et al.  (1975b) have investigated the behavior
 of P300s elicited by  auditory  stimuli near sensory

-------
  176
                            Ruchkin and Sutton
 threshold. Mast and Watson utilized a signal-counting
 procedure. Their results suggested that P300  ampli-
 tude was related to the subject's criterion for detect-
 ing presence of  a signal.  Hillyard et al, Paul and
 Sutton, and Squires et al. utilized standard signal-de-
 tection  paradigms (Green and Swets 1966). Hillyard
 et al. (1971) investigated the relationship  between
 discriminability (d;)  and P300 amplitude. They re-
 ported  that P300 increased as  discriminability in-
 creased, up to a level of about 90% correct responses.
 Paul and Sutton (1972) used a fixed stimulus inten-
 sity and varied the subject's criterion by manipulating
 pay-off  contingencies. P300 amplitude elicited by
 correct  detections increased  as the criterion became
 more  strict.  Squires et al. (1975b) also used a fixed
 stimulus intensity. Their subjects responded with a
 numeric  confidence   rating.  P300 amplitude  was
 largest when the subject was most confident  of his
 decision.

     The common  finding of these  investigations is
 that P300 amplitude increased as the subjects were
 more certain of their perceptions and hence the infor-
 mation loss due to equivocation was less. Hillyard et
 al. demonstrated this by direct variation of discrimina-
 bility of the stimulus, while Paul and Sutton in effect
 varied the degree of certainty required of the subjects,
 and Squires et al. segregated the data into different
 levels of certainty. Donchin  (1968) reported similar
 findings  in a visual perception experiment involving
 threshold flashes. P300s were large when the subjects
 were certain  of their  judgments and small when the
 subjects were uncertain.

     Further evidence for the role of equivocation in
 determination of P300 amplitudes is provided by dis-
 crimination experiments reported  by Ritter  et al.
 (1972), Adams and Benson (1973), Lang et al. (1975),
 and Ford et  al. (1976). Ritter et al. investigated the
 relationship between P300 latency, reaction time, and
 auditory pitch discrimination in a go-no go paradigm.
 Two conditions were  used. In one, the discrimination
 was relatively easy;  in the  other, it was  difficult.
 While the authors did not specifically report the effect
 of discrimination  difficulty  upon  P300 amplitude,
 inspection of their published waveforms indicates that
 P300 amplitude was  lower when the discrimination
 was more difficult. Langet al. (1975) instructed sub-
jects to  rank five  different equi-intensity tones. The
 tones, presented individually, were 700, 1000, 1100,
 1200,  and  1500 Hz. There were relatively few errors
 for the  700-  and 1500-Hz tones, while errors  were
 relatively numerous for the 1000-, 1100-, and 1200-Hz
 tones. Peak-to-peak amplitude from  NlOO to P300
 was significantly  larger for easily ranked tones than
 for tones that were difficult to rank.

     Adams and Benson (1973) used a relatively high-
 intensity (30 dB SL)  auditory stimulus to signal cor-
 rect performance  of  a difficult psychophysical task.
 Incorrect performance was indicated by presentation
 of a lower intensity stimulus. Objective intensity con-
 trast between the two alternative stimuli was manip-
 ulated by varying the intensity of the stimulus that
 indicated incorrect performance (from 0 to 24 dB).
 As the contrast was reduced, amplitude of the P300
 elicited by  the fixed-intensity stimulus was reduced,
 although the stimulus provided the same a priori infor-
 mation. Increased equivocation as stimulus contrast
 decreased is a parsimonious explanation of this find-
 ing. However, Adams and Benson did comment that
 subjects were able to readily distinguish between the
 two stimuli, although the investigators did not deter-
 mine whether there were any objective differences in
 the  ease with which the distinction could be made.
 The fact  that the subjects apparently readily distin-
 guished between  two stimuli  does not necessarily
 vitiate the equivocation argument.

     Thurmond and Alluisi (1963) demonstrated, in a
 choice-reaction time experiment, that as dissimilarity
 between two signals was reduced, reaction time in-
creased, even though discriminability remained high.
 This indicates that although a relatively small  degree
 of equivocation may not significantly interfere with
 the ultimate accuracy of a decision, it will delay it,
 due to increased difficulty in processing the signal.

     Ford  et al. (1976) have provided further support
 for this interpretation. They presented subjects with a
 sequence  of tone pips at a fixed loudness level and
 frequency. Occasional mismatch tone pips were pre-
 sented on a random basis, there being three levels of
 frequency mismatch:  (1) a 5%  frequency shift (2) a
 25% shift, (3) an  octave shift. Subjects were required
 to respond to the  mismatch tones with a button press.
P300s  elicited by the mismatch tones increased in
 amplitude and reaction time decreased as the  degree
 of mismatch increased, despite the fact that subjects
 responded  correctly to 89% of the 5%-frequency-
shift trials and to 96% of both the 25%-shift and
 octave-shift  trials.  Ford et al.'s  results provide a fur-
ther demonstration  of  the effect  of  equivocation
upon  P300 amplitude.  Their  results  also directly
demonstrate that  P300 amplitude  may reflect the
degree  of difficulty in reaching a decision rather than
the final accuracy of the  decision-making process, as
indicated by the accuracy (which did not vary), reac-
tion time (which increased), and P300 amplitude
(which  decreased) for the octave and 25%-shift con-
ditions.

    Reduction in  average P300 amplitude with in-
creased equivocation may be due to a direct decrease
in amplitudes in individual trials and/or to increased
latency variation of P300 in individual trials. The lat-
ter possibility seems particularly likely if one assumes
that equivocation  leads to variability in decision time
and that P300 latency is linked to the time at which a
decision is made.

-------
 Equivocation and P300 Amplitude
                                                                                                  177
    An opportunity to examine these alternatives
was afforded by investigations of the emitted poten-
tial.  An emitted P300 may  be elicited by the non-
occurrence of a relevant stimulus that may or may
not occur  at a given point  in time. Emitted P300s
have been  observed in guessing experiments (Sutton
et al.  1967,  Ruchkin  and Sutton  1973) and some
signal detection  experiments (Squires et al, 197Sb),
Average  emitted  P300s  in  guessing experiments
are smaller than  corresponding evoked P300s elicited
by stimulus  occurrences.  Ruchkin  and Sutton  (in
press) demonstrated that the lower average amplitude
was due partly to increased latency variation of emit-
ted PSOOs and partly to a direct amplitude difference.
The direct amplitude  decrement as well as the  in-
creased latency  variability may  be  due to temporal
uncertainty and hence  increased equivocation asso-
ciated  with identification  of the nonoccurrence of a
stimulus, There have been several  reports of the occur-
rence of clear evoked P300s in correct detection trials
of signal-detection experiments, but the evidence for
the occurrence of emitted P300s in signal-absent trials
is more limited. Hillyard  et al. (1971) and Paul and
Sutton (1972) were unable to observe emitted P300s.
Squires et al. (1975b) reduced temporal uncertainty
through the use of a cue light and  sorted their data
into high  and low decision-confidence trials. They
observed  emitted  PSOOs in high-confidence correct
rejection and false-alarm trials. They further demon-
strated that emitted P300 amplitude in correct rejec-
tion trials increased as stimulus intensity increased.
Squires et al.'s results suggest that latency variations
may in part contribute to  the  lower amplitudes of
average emitted P300s. However, the effect of de-
cision  confidence  and stimulus intensity upon emit-
ted  potentials  in signal-absent  trials suggests  that
equivocation also has a direct effect upon P300
amplitude.

    The results of the investigations cited in this paper
can  be most parsimoniously described as follows:
P300 amplitude is in part  determined by the total
amount of information received, i.e., the greater the
reduction in uncertainty, the greater the P300. P300
can vary as a function of information received in two
ways: (1) for a fixed level of prior uncertainty, P300
amplitude will  be determined by equivocation-de-
creasing as equivocation increases; (2) for a fixed level
of stimulus equivocation, P300  amplitude  will  be
determined by prior uncertainty-increasing as uncer-
tainty increases. The latter relationship is subject to
the provision that equivocation  is not  so large that
a priori  probability effects are  "swamped"  by the
high degree of a posteriori uncertainty (Squires et al.,
1975 b). What  seems to be at issue in the former
relationship is not necessarily outright discriminability
or correctness, but  the  immediacy  and ease with
which  a decision  can  be made.  This  suggests that
P300 amplitude may reflect processes involved in the
early stages of decision making rather than the final,
conscious step.
Summary

    Evidence  for the concept  that the amount of
information received by a subject from the occurrence
of an  event  determines, in part, the amplitude of
P300 is reviewed. The amount of information received
depends upon the a priori uncertainty of the event's
occurrence minus an information loss, refe&ed to as
equivocation, due to the a posteriori uncertainty of
having correctly perceived the event. Support for this
concept is provided by the results of experiments in
which  either or both a priori probability and equivo-
cation  were variables.

Acknowledgments

    This work was supported in part by U.S. Public
Health Service Grant NS11199 to D. S. Ruchkin and
U.S. Public  Health  Service Grant  MH14580  to S.
Sutton. We thank Dr. E. M. Glaser of the University
of Maryland for his criticism and advice.

-------
  THE  LATE  POSITIVE COMPONENT AND
  ORIENTING BEHAVIOR
  D. FRIEDMAN

  New York Psychiatric Institute, Department of Medical Genetics, New York, NY, U.S.A.
  Review

     The concept of orienting is an old one, dating
  back to the work of Pavlov, who noted (at first, in
  annoyance)  that his  experimental  animals directed
  their attention to any novel or unusual stimulus in
  the experimental chamber.  He thus termed it  the
  "investigatory" or "what is it?" response. The orient-
 ing response consists of a series of physiological chang-
 es  including heart-rate  deceleration  (Graham  and
 Clifton 1966), pupillary dilation (Shakhnovich 1965),
 increments in galvanic skin response (GSR) (Lovibond
 1969) and EEC desynchronization  (Lynn  1966). It
 has been shown to be elicited by such stimulus charac-
 teristics as novelty and uncertainty (see Lynn 1966),
 some of which have also been shown to influence the
 ubiquitious P300 wave.  The orienting response has
 received much experimental attention with the use of
 electrodermal variables, heart rate, and pupillary dila-
 tion, but it is only within the last decade that the late
 positive  component has been added  to the list  of
 "components of the orienting response."

     Ritter et al. (1968) were the first to use the classi-
 cal  orienting response  paradigm  to elicit averaged
 evoked-potential components of this response. They
 presented a series of 1000-Hz tones at 2-sec intervals.
 There were 30 stimuli per run and 24 runs. In order
 to look at habituation within the stimulus train, Ritter
 et al. averaged by stimulus position within  the train,
 across the 24 runs. A substantial decrement in P200
 was observed as a function of stimulus position, but
 in a second experiment where stimulus presentation
 was changed to one every 10 sec, no decrement in
P200 was seen, so that  refractoriness (Davis et  al.
 1966) and not habituation seemed to be the cause of
 the  decrement.  In  a  third experiment,  1000-Hz
 tones, one every 2 sec, were delivered with a pitch
 change  to  2000 Hz at the 21st  stimulus.  Stimulus
 position averaging revealed a short-term decrement in
 P200, but the pitch change did not restore the ampli-
 tude of P200. The change in stimulus quality produced
 a change  in the evoked-potential waveform,  with a
 parietal-maximum, late positive component at 350
 msec.  In a  fourth experiment,  the  paradigm  was
 changed to include the unpredictable onset of a run
 (temporal uncertainty)  and an unpredictable pitch
 change within  a run. Both of these unpredictable
 changes  produced  parietal-maximum  late positive
 components at 350 msec.  Thus, Ritter et al. (1968)
 demonstrated  that P300s were elicited in  situations
 identical to  those used to  elicit autonomic compon-
 ents of the orienting response.

    Support for P300 as a component of the orienting
 response  could be  strengthened by concomitant
 recording of autonomic measures. Evidence from a
 variety of laboratories had indicated that  heart  rate
 (Higgins 1969), GSR (Lovibond 1969), and pupillary
 dilation  (Levine and Hakerem 1969) were similarly
 affected by  manipulations  of stimulus uncertainty,
 one of the early postulated psychological  correlates
 of P300. Tueting et al. (1970) had shown that P300
 varied  markedly in amplitude with changes in stimulus
 probability in an uncertainty paradigm. We (Friedman
 et  al.  1973) followed  these  lines  of evidence by
 recording both evoked potentials and pupillary dila-
 tions in a modification of the  Tueting et al. (1970)
 probability paradigm. Pupillary dilation is easily time-
 locked, with  an approximate  1-sec latency to peak
 and, when averaged, can  be distinguished from noise
 with approximately the  same N as P300 (Hakerem
 1967).  We found that pupillary dilation followed the
 same lawful relationship to  probability of occurrence
 as did P300:  the lower the  probability, the larger the
 amplitude of dilation. In addition, a component of
 the dilation response, prestimulus slope, followed a
 similar  time course and was closely  related to CNV
 amplitude, tending  toward  dilation when subjects
 were guessing  upcoming events (CNVs were  more
 negative) and toward constriction (CNVs were less
 negative and  some were positive) when they were
 told which stimulus  would occur next (Friedman
 1972).  This paradigm was  not the traditional "ori-
enting" paradigm. In informational terms,  however,
both P300 and  pupillary  dilation were larger in am-
plitude  when an infrequent event reduced uncer-
tainty.   The  concept of uncertainty  reduction has

-------
Late Positive Components and Orienting
                                            179
been  used before  within  an  orienting framework
(Pribram 1967,Sokolow 1960).
    Roth et al. (this section) recorded GSR and evok-
ed potentials in an orienting response paradigm where
chords unpredictably  and continuously (unlike the
Ritter et al.  1968 paradigm) replaced background
noise  bursts. They used GSR as a criterion measure
for sorting PSOOs to the chords into high- and low-
amplitude averages. During attend conditions (respond
to the change), GSR and P300 showed similar ampli-
tude  changes.  During read (i.e., ignore) conditions,
there  was a dissociation of the two measure, with
small-amplitude PSOOs associated with large-amplitude
GSRs. Roth et al. (1976) speculated that direction of
attention affected P300 and GSR differently, with a
larger GSR  orienting response when attention is not
directed towards the stimulus train. This result holds
only for the "read" condition.
    In both the Tueting et al. (1970) and Friedman
et al.  (1973) "certain" conditions (subjects told what
stimulus to expect), P300 amplitude was still a mono-
tonic  function  of stimulus probability even though
stimuli were "task-irrelevant." Roth (1973) and Roth
and Kopell (1973) also demonstrated that infrequent,
task-irrelevant stimuli produced reliable P300s, but in
the case of Roth's investigation (1973), P300s were
of much earlier latency (mean of 210 msec) than pre-
viously reported. In further support  of this "early
P300" as  an orienting potential,  these investigators
found a decrement in  amplitude  across quarters of
the experimental run. Response decrement is to be
expected as the novelty  of a repeated change in stim-
ulation  diminishes.  Habituation   of  the orienting
response with repetition is one of the key  postulates
of orienting response theorists (Lynn 1966).


    Theoretically, an orienting response is expected
whenever  background stimulation  (to which subjects
are assumed to  habituate) is changed in any manner-
hence, the use of prolonged runs of standard stimuli
with infrequent changes. The finding that P300s were
present for task-irrelevant, nonattended stimuli (Ritter
et al.  1968, Roth 1973) as well as for attended, task-
relevant stimuli  (Ritter et al. 1972) poses a problem
for any unified theoretical interpretation of the P300
wave. On  the other hand, if there are two P300s, one
elicited when  variables  that produce orienting  are
used  and  the  other when  task-relevance is the key
feature of the experimental design, then interpretation
is easier.

    Squires etal. (1975)distinguished the "orienting"
from  the  "attend" P300 by both latency  and scalp
topography. They utilized a paradigm in which sub-
jects either attended or ignored infrequent background
stimuli  and infrequent  intensity  or pitch  changes.
During the attend condition, subjects counted changes
and  during the ignore condition, they read. The ignore
condition produced P300s only to the lowest proba-
bility stimuli (p = 0.10). This P300 varied in latency
from 220 to 280 msec (similar to that of Roth 1973),
and  had  a frontocentral  topographic distribution.
When subjects  attended, this "P3a" wave was present
but was overridden by a later (310-380 msec), more
posteriorly distributed "P3b" wave.  Squires et al.
(1975)  concluded  that the  presence  of the "P3a"
wave reflected a "mismatch to an ongoing stimulus
train, whether or not it is being attended" (p. 399).
This conclusion  is in accord with Sokolov's (1960)
theoretical model, which postulates that any changes
in background stimulus conditions leads to a mismatch
with the neuronal  model and  results in an orienting
response.

     Courchesne et al. (1975) studied the influence of
novelty on the late positive component of the visual
evoked  potential, using easily  recognizable and com-
pletely  unrecognizable novel  stimuli. Novel stimuli
were interspersed (each with 5% probability of occur-
rence) among frequent nontarget stimuli (the number
2)  and  infrequent task-relevant target stimuli  (the
number 4). The subject had to count the target stim-
uli,  which occurred with 10% probability. Evoked
potentials for the counted 4s and  the task-irrelevant,
completely  unrecognizable  novel stimuli  contained
late positive components that differed in  scalp top-
ography, but not in latency. Novel stimuli produced
a  more anteriorly  oriented  distribution,  while  the
counted 4s produced a more  posterior scalp topog-
raphy.  When evoked potentials elicited by the first,
second, and third presentations of these novel stimuli
were averaged separately, there was a large decrement
(50% for the  second stimulus) in  P300  amplitude,
thus demonstrating habituation of the response and
supporting an "orienting" interpretation of this front-
ally distributed wave.

     The more frontal distribution of the potential for
novel visual stimuli reported  by  Courchesne et al.
(1975)  supports the data of Squires et al. (1975) for
the  auditory modality, but differences in latency and
task mitigate against an association between the audi-
tory "P3a" wave and the late P3 to novel visual stim-
uli.  It  is possible, however, that  the  two  waves are
part of the same  orienting potential, but differ in
latency and amplitude, depending upon stimuli and
task conditions. This latter possibility may well be in
accord  with another of Sokolov's  postulates that the
amplitude of the response varies  with the  degree of
mismatch  from  the neuronal model.  For example,
Sokolovian theory would  predict a larger amplitude
response to a more novel visual stimulus (Courchesne
et al. 1975) than to a less novel pitch change (Squires

-------
180
                                      Friedman
et al. 1975). Comparision of these two studies, in fact,
confirms this  prediction:  PSOOs to unrecognizable
novel stimuli (Courchesne et al.) were larger in ampli-
tude than PSOOs to easily recognizable novel stimuli
(Squires et al.).

    This latter postulate was directly tested by Ford
et al. (1976), who parametrically varied the degree of
disparity of infrequent pitch changes from standard,
frequent background tones. Mismatch tones were 5%
(one-half musical step), 25% (major third), and 100%
(an octave)  discrepant  and were interspersed  during
read (ignore) and respond (button press to mismatch)
conditions. Under both read and respond conditions,
the amplitude  of P300 varied with the degree  of mis-
match,  although this relationship was much  clearer
and stronger under respond conditions. P300 latency
was longer during respond (mean of 336 msec) than
during read (mean of 287 msec), and the topographies
were different as well. The read P300 was more uni-
formly distributed, while the respond P300 was largest
at Pz. These data support an interpretation of P300 in
terms of orienting dependent upon the degree  of mis-
match and support the hypothesis that more than one
P300 generator exists, producing PSOOs of different
latency and amplitude.

Summary and conclusions

    Enough evidence has accrued  to suggest  that
P300 does reflect orienting,  since it is elicited in  the
same situations and by the same variables that elicit
the classical autonomic components of the orienting
response. The  fact that late positive components of
different latencies and  scalp topographies have been
elicited under similar orienting paradigms presents a
problem in  attempting to draw general conclusions
about situations that produce an "orienting P300." In
general, early and more frontally distributed late posi-
tive components have been reported to occur when
subjects were ignoring the stimuli, while longer latency
late positive components have been found when they
were attending the train of stimuli, even though the
stimuli  were task-irrelevant  (i.e.  Courchesne et al.
1975). Fitter et al. (1968) found a late (350 msec)
P300 that had a posterior focus, while  Squires et al.
(1975) reported an early  (220-280 msec), frontally
distributed "P3a" and Ford et al. (1976) observed an
early  (287  msec), but more uniformly distributed
P300 to auditory stimuli during "ignore" conditions.
It has been suggested by Roth (1973) and Squires et
al. (1975) that the late P300s reported by Ritter et al.
(1968)  were "target-detect" P300s,   since  their
subjects were practiced and were expecting the pitch
change, thus accounting for the  longer latency and
posterior topography.

    In any  event, it is likely that one of the various
P300 waves will occur when experimental operations
designed to induce orienting behavior are employed.
The  conditions, stimuli, and  tasks under which the
production  of these waves can be wholly attributed
to orienting remain to be parametrically investigated.
The  addition of autonomic measures to these  same
paradigms might aid in the interpretation of orienting,
and use of single-trial  analysis (Ritter et  al. 1972,
Courchesne et al. 1975), especially in the case of large
amplitude autonomic signals (e.g., GSR), might pro-
vide information that is "washed out" when the tech-
nique of averaging is used.

-------
 DOES  P300  REFLECT  TEMPLATE
 MATCH/ MISMATCH?
 J.M.FORD

 Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine,
 Stanford, and Veterans Administration Hospital, Palo Alto, CA, U.S.A.
    The  mentalistic concept of template matching
has been used by  psychologists to explain pattern
recognition.  When  used  by psychologists,  template
means a standard or a prototype. A stimulus is iden-
tified by  noting its congruence with the  template
(Neisser 1967). Although the concept has some appeal
at first blush, it loses strength when it becomes appar-
ent that in order to explain much of perception, it is
necessary to make the definition much more flexible,
ultimately making  template-matching theories indis-
tinguishable   from   feature-analysis  theories.  For
example, if the task is to detect any letter A, then the
template will be a composite of block A's, script A's,
tilted A's, etc. Regardless of these  definitional prob-
lems, evoked potential researchers often  think in
terms of templates  to explain the P300 phenomenon.
To do this, a flexible notion of how a match or mis-
match is made is adopted. The template is some kind
of memory of stimulus events. New stimuli are com-
pared with  this template, and a match or mismatch
occurs. Depending on task or environmental demands,
the template will be more or less fuzzy.

    A number of variables affect P300. Two of them,
task  relevance  and  probability,  are discussed in this
paper. Since  these  variables interact in a  complex
manner, studies in which only one is manipulated, the
other being held constant, are considered. The success
or failure of template  matching and mismatching in
explaining existing P300 data is examined, after first
defining template matching and mismatching in terms
of task relevance and probability.

Definitions

    Task-relevant stimuli  are stimuli that a subject
needs to make a decision. They are either stimuli
requiring a response (a  target) or stimuli having some
bearing on the decision regarding the target, perhaps
affecting performance.  In experiments where the
subject  is instructed to  detect a particular signal,
some researchers assume  the subject leams what the
target is and develops a template (neural representa-
tion)  of it in  memory.  When  a signal occurs that
matches the template, the subject decides that the
target has occurred. This template will be referred to
as the target template. It has been suggested that it is
a match of the sensory input with the target template
that elicits the P300.

    Probability has also  been shown to affect P300
amplitude. P300 amplitude has been manipulated by
varying  the probability of ongoing events, either the
global or the sequential probability, or by varying the
outcome probability.  These events are psychological
events  manipulated by  the contingencies between
physical events. Templates of these events may be the
environmental templates ofSokolov (1963). To deter-
mine the effect of any type of probability on P300, it
is important to establish that both the probable and
improbable events are task relevant and equally dis-
criminable. In  experiments where the subject grows
accustomed to hearing or seeing a certain background
signal or experiencing a certain event, some researchers
assume  that the subject  develops  a template for the
probable event. When another event of a different
pitch, configuration,  or  psychological value occurs,
the input mismatches his template. This template will
be referred to as a probability template; and when the
observed event mismatches the  probable event, a
P300  results. The probable  stimulus could be con-
sidered a probability template match; and unless it is
also task  relevant, it does not elicit a large P300
(Squires et al. 1975). Expectancy is a term often used
interchangeably with probability. Since the subject is
rarely asked what he really expects in these experi-
ments, it is better to use the more operational term of
probability.

Target  template matching
    The most direct attack on  the issue  of target
matching was made  by Posner  et al.  (1973). The
experiment was a match-to-target design where sub-
jects were shown  a target letter  followed 1 sec later
by another letter. The second letter either matched or
mismatched  the target. A match elicited an earlier

-------
 182
                                            Ford
 and  larger P300 than did a mismatch, but the mis-
 match elicited a larger N2 than did the match. Match-
 es often have faster (Posner and Boies 1971) and less
 variable (Nickerson 1969) reaction times (RTs) than
 do mismatches. The psychological process underlying
 this  phenomenon  is unclear. Nevertheless, inferring
 from RT data, it seems reasonable that P300s to mis-
 matches would be later and the onset of individual
 P300s would be more variable. These factors  could
 explain the late, shallow  appearance of P300 to mis-
 matches, and had they been averaged with respect to
 RT (keeping RT and RT variability equal for both
 matches and mismatches), the PSOOs might have look-
 ed quite similar. However,since the difference between
 standard deviations in RT is usually less than 25 msec
 (Sternberg 1969), it is unlikely that this could explain
 the difference in broadness of P300 seen in Posner's
 data.

     Another example  of target  matching and mis-
 matching, when matches and mismatches were equi-
 probable, is an experiment by Courchesne et al.  1976.
 Seven subjects saw 80% background slides (2's), 10%
 targets (4's), and 10% novels (colorful, unrecognizable
 pictures). Subjects were required to count the targets.
 For the moment, assume that the subject has a tem-
 plate  for  the target so that its delivery results in a
 match and the delivery of a novel results in a mismatch.
 The match evoked  a large parietal P300 and the mis-
 match evoked a large frontal-central P300 (and N2).
 In another condition,  where  subjects counted the
 novel slides, the novel stimuli evoked a large  parietal
 P300, as  did the target 4's in the other condition.
 Since no novel slide was ever presented more than
 once, it is  difficult  to imagine a template of a partic-
 ular novel slide. In order to make these data fit the
 notion of template matching,  we must think of a
 general template for any complex stimulus (Tueting,
 this volume).

    The relationship between target  template-match-
 ing and P300 has  been studied  extensively  by the
 Hillyard/Squires group (Hillyard et al. 1971; Squires
 et  al. 1973a, 1973b,  1975). On the basis of signal
 detection experiments, these authors suggest that sub-
jects establish two templates in memory-one for sig-
 nal presence (a template of the signal itself) and one
 for signal  absence.  P300 is  triggered by  a match to
either template, and  the  closeness of the match is
measured   by decision  confidence.  The  closer the
match, the larger and earlier the P300 (Squires et al.
 1973b). One interesting result in the 1975 study was
that detections of signal presence (hits) are associated
with larger PSOOs than are detections of signal absence
(correct rejections)  made  at the same level of confi-
dence. To explain this result, a second variable, related
to  stimulus probability,  was suggested and  tested.
This variable was the outcome probability of a  trial.
The more  improbable the outcome,  the  larger the
 P300. That is,  when signal  absence is improbable,
 correct rejections also become improbable and evoke
 larger P300s than do correct detections. Squires et al.
 1975 pointed out that, when a signal  is well above
 threshold,  when subjects are confident that a signal
 has  been delivered,  and when task relevance  is  not
 varied, improbability of outcome seems to play a
 more dominant role than target template-matching.

 Probability template mismatch

    Throughout the  literature,  P300s  evoked  by
 improbable events  are reported. In  two  separate
 experiments, Tueting et al.  (1970) varied the sequen-
 tial  and global probabilities of stimulus  occurrence.
 When  subjects were  asked to guess which of two
 stimuli would occur next,  Tueting et al. found that
 P300 amplitude was predicted by the interaction of
 the stimulus probabilities and the guessing behavior
 of the subject. The interaction was called the outcome
 probability; the more improbable the  outcome,  the
 larger the P300.

     K. Squires  et  al. (1976) also  investigated  the
 effect  of both sequential and global probability on
 the ERPwaveform. Instead of probability, they  called
 the variable expectancy. Expectancy was defined as a
 mathematical function of decaying memory for events
 within the prior sequence. For target stimuli, P300
 (as well  as N200 and slow  wave components) varied
 inversely with stimulus expectancy.


     An improbable event can also be response-related.
 Karlin and  Martz (1973) have shown  that, when a
 response is rare, the stimulus  associated  with  the
 response elicits a larger P300 than when the response
 is frequent regardless of stimulus probabilities.

   When improbable feedback is delivered, a large
 P300 is elicited (Squires et al. 1973a, 1975; Tueting
 et al.  1970); when  an infrequent pitch or intensity
 change occurs, large P300s are elicited (Ritter et al.
 1968,  Ford et al. 1976, Squires et al. 1975); when an
 infrequent  line drawing occurs in a series of letters, a
 P300 is elicited (Courchesne et al. 1975); and so on.
 Thus,  it appears that improbable events associated
 with stimulus,  response, or outcome evoked  larger
 amplitude P300s. Many reports can be reinterpreted
 in terms of a mismatch with the probability template
eliciting  a  P300,  although  few investigators have
 used these terms.

    In summary, the largest P300s are elicited in two
 fairly disparate situations-target  template matching
 and  probability  template mismatching. To  reconcile
 this  disparity may not be possible within the frame-
 work of template matching and mismatching, but
some of the possible ways are suggested below.

-------
P300 Latency and Template Match/Mismatch
                                            183
    Tueling  (this section) suggests that target  tem-
plate-matching and probability template-mismatching
might be pulled  together under a unitary concept by
assuming, in the probability case, that the subject
establishes a template for improbable or novel events.
Another speculation is that the two processes can be
considered component parts of a serial process where
probability mismatches are registered  first, followed
by  an evaluation of whether the event matches the
template. When  an event mismatches the expectancy,
an early P300 should result; when it matches a target,
a later P300 should result. This concept fits the data
of N. Squires et al. (1975) and Roth et al. (1976), but
is inconsistent with the data of E. Courchesne et al.
(1978)   showing that probability mismatches elicit
later P300s than do targets in the visual system. Rotli
cl al. (in preparation) have obtained similar inconsis-
tent data in the auditory system.  Furthermore, this
concept  docs not deal with the occurrence of target
mismatches  as in Posner's work (1973), or with how
P300 amplitude and latency might change with the
degree of match or mismatch.
    Squires et al. (1975) formulated a way in which
P300 could be affected by the degree of target match-
ing. They suggested that template matching  may in-
volve  a series  of comparisons of the signal against a
list of templates  that vary  from the most target-like
to the least target-like. P300 would be elicited when a
match is satisfactorily  encountered,  causing  P300
latency variations. The  confidence in the  decision
that a match actually  occurred would be reflected in
P300  amplitude. The  same logic can presumably be
extended to  the probability situation,  in which the
list of templates is ordered from the most expected to
the least expected. N. Squires et al. (1975), however,
adopted a different strategy. Instead of trying  to unite
target  matching  and probability mismatching under
one  rubric, they suggested that the P300s  elicited
under  these two situations are  separate components
(P3b and P3a) that have  different latencies and scalp
distributions  and  reflect different  processes.  P300
data at the present time do not permit the unequiv-
ocal evaluation of the different template match-mis-
match hypotheses discussed in this paper.

-------
  EVOKED POTENTIALS  AND  FEEDBACK1


  S. SUTTON, P. TUETING2, M. HAMMER, AND G. HAKEREM
  Department of Psychophysiology, New York State Psychiatric Institute, New York, NY, U.S.A.
     More  than  10 years  ago, Sutton et al. (1965)
 reported that subjects had larger P300s on discovering
 that a guess was wrong than on discovering it was right.
 However, in subsequent experiments, this pattern was
 noted to be inconsistent. In some experiments, no
 statistical difference was found.  In one experiment,
 most subjects had larger PSOOs associated with feed-
 back that informed them their guess was right (Levit
 et al. 1973).

     On noting how sensitive P300 amplitude was to
 the  interaction of stimulus and guessing probabilities
 (guessing probabilities were not examined in the earli-
 er experiments),  the assertion that being wrong gave
 greater amplitude than being right was  temporarily
 withdrawn, and it  was suggested that  the  issue  of
 right-wrong differences could not be resolved until it
 was disentangled  from the  effects  of probability
 (Tueting et al. 1970). At about this time, Squires et
 al. (1973) looked at feedback that told a subject that
 his discrimination was correct or incorrect, and found
 that P300 was larger when feedback indicated an in-
 correct discrimination. These  differences remained
 even after probability effects were statistically partial-
 led out. Poon et al. (1974) used a guessing procedure
 in which subjects could learn the correct sequence of
 stimuli to  a fairly  high degree of accuracy. Again,
 they  found that P300 was larger for incorrect than
 correct feedback.

     In contrast to studies reporting larger P300s for
 incorrect feedback, a recent study by Leifer et al.
 (1976), also a guessing situation, found larger PSOOs
 associated with correct feedback.  But an  inkling of
 the complexity of the problem is given by their analy-
 sis of changes over successive blocks. P300s associated
 with  feedback  following predictions that the high
 probability  event would occur became smaller in am-
 plitude over successive blocks. However, P300s asso-
 ciated  with  feedback following predictions that the
low probability event would occur did not alter over
successive blocks. On the other hand, in a simpler 50-
50  probability situation  not involving probability
learning, Tueting and Levit (unpublished data) found
that incorrect guesses had larger P300s at the begin-
ning of the experiment, but this  reversed as the ex-
periment  progressed  so that correct guesses had larger
P300s in later trials.

    More recently, we attempted  a reanalysis of the
feedback  concept and have completed some prelimi-
nary experiments based on  that  reanalysis. We devel-
oped the  following formulation:  Whether stimuli pro-
vide feedback with respect to a guess or with respect
to a discrimination, they have the following cross-cut-
ting properties:
  1. Confirmation-disconfirmation — The feedback
     stimulus tells the subject that a guess or discrim-
     ination was correct or incorrect.
  2.  Degree of value-This  can  be illustrated by
     experimental designs in which the experimenter
     places different degrees of monetary reward or
     loss on being correct or incorrect.
  3.  Direction of value-We are used to thinking of
     being right as a good thing and being wrong as a
     bad thing. But  it is not always  so, as illustrated
     in the statement, "Damn it! I knew it was going
     to rain!"or when one is sure that some symp-
    tom means cancer and  then discovers  that it
    indicates only a minor ailment.
  4.  Information-A feedback stimulus delivers infor-
    mation relevant to future guesses or judgments
    (e.g., it may indicate that the last judgment was
    wrong, and that the subject should try to  find
    some other  basis for  the  next response).  We
    have, however,  generally  only assumed  that
    feedback  influences the  subject's guessing and
    discrimination  strategies on  future trials  and
    that these changes  are  reflected  in P300. A
    somewhat different approach  is to  construct
'This investigation received support from Grant No. N1E-G-74-0042. The assistance of Stuart Steinhauer, Marion
 Hartung, and Robert Youdin is gratefully acknowledged.

2Now at  the Maryland Psychiatric Research Center/Department of Psychiatry, University of Maryland Medical
 School, Baltimore, Maryland.

-------
EPs and Feedback
                                               185
      experimental designs in which one controls the
      information provided by the  feedback stimulus
      in order to study more directly the degree to
      which it affects the subject's behavior and evok-
      ed potentials on  future trials. An interesting
      experimental design is one  arranged so that the
      subject's being correct or incorrect on a given
      trial is used by the experimenter to determine
      the subsequent probability with which different
      stimuli will be presented.
    In a recent set of experiments, we  attempted to
disentangle two of these properties of feedback, con-
firmation-disconfirmation  and  direction  of value.
The direction of value concept is more familiarly
known as the win-lose dimension.
    In one experiment, 12 subjects guessed in each
trial whether the next stimulus would be a high-pitch-
ed click or a low-pitched click. Thus, in each trial, as
the click occurred, the subject discovered whether the
guess  was correct or incorrect. This single outcome,
however, did not determine whether the subject won
or lost. The subject had been instructed that winning
or losing was based on  performance in each pair of
trials, according to the following  rules: the subject
won a quarter if both members of the pair were right
or if both members of the pair were wrong. The sub-
ject lost a quarter if one member of the pair was right
and the other was wrong. Note several things about
this design. Being right on the first  member of the
   pair did not in itself determine the payoff. If the sec-
   ond was also right, the subject won; if the second was
   wrong, the  subject lost. Likewise, being wrong on the
   first member of the pair did not determine payoff. If
   the second  was also wrong, the subject won;  if the
   second was right  the subject lost. Furthermore, with
   respect to the second member of the pair, the subject
   could win on  right or on wrong guesses or could lose
   on right or  on wrong guesses. Thus we experimentally
   separated confirmation-disconfirmation from winning
   and  losing. (With some subjects,  we played  the
   game with opposite rules.)

       The overwhelming finding was that P300  to the
   second member of the pah- was much larger than to
   the first member  of the pair, as shown in Rg. 1. This
   difference  in  amplitude might  be  attributed  to the
   fact that the response to the second stimulus involved
   completion of the pair; however, related data suggest
   that this was not the basis for the  finding. Another
   interpretation  is  that the larger P300 for win-lose
   reflects a value dimension that is often forgotten in
   formulations with respect to P300; winning or losing
   a  quarter is more important than  simply  having  a
   guess confirmed or disconfirmed.


       Other findings were less dramatic. For the second
   member of the pair, P300 was usually larger and later
   for lose than win. A more general finding was greater
             S1
             400     800
               RIGHT

               WRONG
                 S2
                                                RIGHT
                              WRONG
400      800
400
800
                       WIN
         	LOSE
  Fig. 1. Vertex auditory evoked potentials for one subject in a guesting paradigm. Negative at vertix is up with respect
  to a linked earlobe reference.  The flrst stimulus of the pair (SI) confirmed or disconflmed the subject's guess. The
  second stimulus of the pair (SI) also confirmed
-------
  186
                                    Sutton et al.
  negativity in the P2-N2 region for the lose waveform
  for the  overwhelming majority  of subjects. On the
  whole, these differences between lose and win wave-
  forms held  whether they occurred in experimental
  formats where the subject lost by being wrong or lost
  by being right and correspondingly won by being right
  or won by being wrong. This is the direction of value
  concept  mentioned  earlier.  Identification  of more
  subtle interactions must wait until all data have been
  retrieved from the computer. Data  for all scalp loci
  also have not been fully examined; only vertex has
  been analyzed at this time.

     In  the confirmation-disconfirmation waveforms-
  responses to the first member  of  the pair—consist-
  ent  differences  were  again in  the  P2-N2  region.
  The discon firm a tion wave form, for almost all subjects,
  was more negative in  the P2-N2 region. Unlike the
  win-lose  comparison, there were  no differences in the
  P300 region.

     Summarizing, then, in feedback designs the P300
  is very  sensitive  to the importance or value of the
  stimulus to  the subject, as reflected by the fact that
  win-lose  P300s were much larger than confirmation-
  disconfirmation P300s. On the other hand, differences
  between  winning and losing, as well as between confir-
  mation and disconfirmation, were reflected most con-
  sistently  in the P2-N2 region.

     In  the correspondence group discussion, preced-
 ing this symposium, Squires reexamined earlier data
  from Squires et al. (1973), which compared confirm-
 ing and disconfirming  feedback  in a discrimination
 situation. The earlier data appear remarkably similar
 to ours. The statement that there  was more negativity
 in the P2-N2 region summarizes what was most con-
 sistent over all of our subjects. We also saw evidence
 for: (1)  a latency  shift, particularly  in  the  N1-P2
 arm-later for lose; (2)  truncation of P2 for lose; (3)
 sometimes a development of a clear  N2 for lose; and
 (4) sometimes a  later,  more  peaked, and larger am-
 plitude P300 for lose than for win.

    In another set of guessing experiments involving
eight subjects by Steinhauer  of our laboratory, sub-
jects decided  before each trial whether they would
bet 504,  254, or nothing, for that trial. Again, only
vertex data have been analyzed. P300 was found to
be larger the  greater the value of the bet, as shown in
 fig. 2. Even more interestingly, in otherwise identical
sessions with  the  same  subjects,  a random-number
program selected  the value of the subject's bet and
the subject was informed of that  value prior to each
trial. The  whole level of P300 amplitude was shifted.
When the  computer selected the value of the bet, all
P300s were half the size of those  obtained when the
subject selected the value of the bet. In the computer-
bet condition, order  tended  to be preserved-larger
  stakes yielded larger PSOOs than smaller stakes-but
  less consistently than in the subject-bet condition.

     The concept that the feedback stimulus may pro-
  vide different degrees  of information needs further
  elaboration. We have not limited the term feedback
  to  those cases where it can be shown that the feed-
  back provides information that alters data obtained in
  subsequent  trials.  Partly, this  is because it seems
  reasonable to infer that systematic feedback almost
  always affects future evoked potentials and behavior,
  and when the question is examined directly, one finds
  that indeed it does. For example, in a guessing situa-
  tion without telling subjects about the relative prob-
  abilities of two events,  but simply letting the stimuli
  indicate whether guesses are right or wrong, most sub-
 jects will match the event probabilities. In discrimina-
  tion situations, the story  is more subtle, but evidence
  exists that feedback results in improved discrimination
  (Jenness, 1972a).

     Tueting et al. (1970) and Friedman et al. (1973)
 have shown that in situations in which subjects match
 event  probabilities, P300 is  larger the  smaller the
 obtained outcome  probability. Outcome probability
 is a term that takes into account both stimulus proba-
 bilities and  the subjects'  guessing probabilities.  The
 relationship was clearest for confirming feedback, but
 seemed  to hold  for disconfirming feedback as well.
 More  recently,  Campbell and  Picton (1976), using
 feedback stimuli in  a time estimation task, reported
 that P300 amplitude was more systematically related
 to  the meaningful information content  of the feed-
 back stimulus than to the information content defined
 classically  in terms  of surprisal value (probability) of
 the stimulus.

     Subtle sequential effects, particularly  in P300,
have been  noted and related to the inferred sequen-
tial expectancy of the subject. In some studies, data
have been averaged as a  function of type of sequence
(Levit et al.  1973; Sutton et al.  1978; Tueting et al.
1970). In other studies, single  trials have been con-
sidered as a function of type of sequence (Squires et
al.,  this volume).

    Levit (1972) compared normal subjects, schizo-
phrenic patients, and depressive patients in a guessing
situation in which clicks and light flashes had a 50:50
probability. Normal  subjects and depressive patients
developed what we  called  ipsimodal expectations. In
other words, when  they received a light  on a given
trial, they  more often guessed that on the next trial
they would  also receive a  light; having received  a
sound, they more often  guessed that on the next trial
they would receive  a sound. Schizophrenic patients
were about 50:50 in their expectation. When P300
was examined as a function of modality of the stim-
ulus in the previous trial, normals and depressives had

-------
 EPs and Feedback

 SUBJECT BET
COMPUTER BET
                1000


              WINO
              WIN 25 c
              WIN 50 c
0
                1000
  SUBJECT BET
   COMPUTER BET
                  1000


                  10
                1000

              LOSEO
              LOSE  25 c
              LOSE  50 c
Fig.  2.   Vertex auditory evoked potentials for one
subject in a guessing paradigm.  Negative at vertex is
up with respect to a linked earlobe reference.  Win
and  lose waveforms are compared as a function of
the monetary  value of the bet.   Waveforms are al-
so compared in a situation where the subject de-
cided on the monetary value of the bet and in a situ-
ation where the computer decided on  the monetary
value of the bet.


larger P300s when  the stimulus in the previous trial
was  in  the  other  modality-just  what one  would
expect on the basis of the notion that P300 is larger
when the stimulus is relatively unexpected. For schiz-
ophrenics, no such difference was found. Admittedly,
such evidence is indirect, but it is in the right direc-
tion,  and it was inferred by  examining  both  the
behavioral and  evoked potential data (see Sutton et
al. 1978).  Unfortunately, few studies have system-
                                                                   187

                        atically related feedback to changes in both behavior
                        and P300 (with the exception of Jeness 1972b and
                        Leifer et al. 1976). More study of the effect of feed-
                        back on both behavior and P300 on future trials is
                        greatly needed.
    We  would like to conclude  by focussing on the
concept of value  or  importance as  an  explanatory
dimension in interpreting P300 findings.  While  we
have no intention of suggesting that this variable pro-
vides an explanation of all P300 findings, the concept
of the importance of the information provided by the
stimulus appears to cross-cut a number of experimen-
tal designs that affect P300 amplitude. Perhaps target
stimuli,  in a sense by definition, are  more  important
than nontarget stimuli. Perhaps, because of the way
organisms are biologically  constructed, rare events,
novel events, or unexpected events may be  inherently
more important than n on orienting events. In the data
presented here, winning 50tf is more important to the
subject than winning nothing. Placing a bet oneself is
more involving, and  in  that sense more important,
than observing the computer's luck, even if the subject
collects the  winnings or pays the losses. Perhaps  the
reason why P300 is sometimes larger for confirm and
sometimes larger for disconfirm  is that we have inad-
vertently made it more important for the subject to
focus on either right or wrong outcomes in determin-
ing guessing strategy. An interacting variable may be
variation  among  subjects. For  personality reasons,
some individuals may focus  on  winnings  and others
on losses. Experiments would need to be directed at
these issues. Such an approach  to the experimental
analysis of P300 in the win-lose paradigm may be use-
ful for other P300 paradigms  as well.

Summary


    The literature on whether P300 is larger when a
feedback stimulus informs a subject that performance
on  a task is correct  as opposed to incorrect was
reviewed. We concluded that the relationship between
P300 amplitude and direction of feedback continues
to be inconsistent. Based on a re-analysis of the prob-
lem, a new  experiment  that yields more  consistent
findings was  performed. Losing money yielded greater
negativity in  the P2-N2 region than winning money,
no matter whether the rules were such that a correct
guess resulted in losing or an incorrect guess resulted
in losing. However,  when the  experimental design
dissociated being correct or incorrect from winning or
losing, in  the sense that correctness or incorrectness
could not influence the monetary outcome  directly,
being incorrect showed greater negativity in the P2-N2
region than  being correct. With  respect to P300,  the
effects were relatively weak; for more subjects P300
was larger for losing than for winning. But there was

-------
188
                                    Sutton et al.
little difference in P300 for correctness and incorrect-
ness when the experimental design isolated this varia-
ble from winning or losing.

    In another experiment, P300 amplitude was larger
the greater the size of the monetary bet placed on the
guess.  However, in  a design in which the computer
selected the monetary value of the bet to be placed,
P300  amplitude was half the size  obtained when the
subject selected the monetary value of the bet to be
placed. It should be noted that even when the com-
puter selected the value of the bet, the subject pock-
eted the winnings or paid the losses.

    It was suggested that the importance of informa-
tion provided by the stimulus may be a key variable
influencing P300 amplitude. A possible source of
inconsistent findings with respect to P300 amplitude
and  right-wrong  feedback might be  that different
experiments have inadvertently steered subjects into
focussing on either being wrong or being right in guid-
ing their strategy.

-------
POTENTIALS ASSOCIATED WITH THE
DETECTION OF  INFREQUENT  EVENTS IN A
VISUAL DISPLAY
R. COOPER, P. V. POCOCK, W. C. McCALLUM, AND D. PAPAKOSTOPOULOS

Burden Neurological Institute, Bristol, England
    As part of a project to investigate changes in vigi-
lance over extended periods of time.electrophysiolog-
ical events that accompany  detection of infrequent
events  occurring in a continuously observed visual
display were studied.

Method

    The observer was seated 2 meters away from a
25-inch television monitor (16-deg angle subtended in
the horizontal plane), on which  a picture of a land-
scape  was displayed from video tape  (Cooper et  al.
1977). At infrequent, irregular intervals (average 4
min) a vehicle-car, van, or lorry-crossed the display
along any of four roads starting from left or right or
from behind bushes in the centre. The observer was
instructed to press a switch with his left thumb when-
ever he saw a vehicle and press another switch with
his left index finger using a prearranged code to indi-
cate the type of vehicle. The angle subtended by the
vehicles was between  0.5 and 0.25 deg. The contrast
of the picture was about 20%, and the brightness was
set for comfortable viewing in the darkened room. A
total of 24 events occurred during a 1.5-hr watch
period.

    In six subjects, recordings were taken from Fpz,
Cz (compensated for eye movement), and Oz referred
to commoned electrodes on the two mastoid processes
and from bipolar montages  01-P3 and 02-P4. EEC
and horizontal and vertical oculograms were recorded
using amplifiers with  an 8-sec time constant. In three
additional subjects, electrodes were placed at Fpz, Fz,
Cz, Pz, Oz, F3, F4, C3 C4, P3, P4.01.02, and 3 cm
posterior to Oz. In these subjects, recordings of the
oculograms were obtained. The upper frequency limit
was 70Hz (-3dB). Respiration, EKG,  galvanic skin
response switch presses, and EMG of operant muscles
were also recorded. Data channels were sampled  at
100 points/sec and stored on digital tape of aPDP-12
computer.  The sampling started  shortly before the
vehicle entered the display  and ended 16 sec later.

Results
    Detection time-i.e., the time between the entry
of the vehicle onto the display and the switch press
indicating detection—varied greatly across subjects,
vehicles, and routes, with an average of 4.3 sec and a
range of 0.4  to 25 sec.  A further 2.4 ± 2 sec was
required to recognize the vehicle.

    Recordings of eye movements showed that up to
1 or 2 sec before detection (indicated by switch press)
subjects were  scanning the display  as shown in a
two-dimensional plot of eye position in Fig. 1. In this
trial the vehicle was moving on a right to left track.
At the start of this trial, the eyes were scanning near
the vehicle,  which had already entered the display,
but the eyes then moved to the left at AB without
seeing it. About 1 sec later (D) the eyes returned to
the left and scanned the area of the display just below
the vehicle. The eyes then  moved upward, overshot
the track at EF, and then  locked onto the vehicle.
The vertex EEC shows a large positivity at time EF,
followed by  the switch  press indicating detection.

    Multichannel  recordings of the  same  event are
shown in Rg. 2. Large positive potentials occur at the
vertex and occipital midline electrodes. EMG begins
to increase immediately after the positivity  culminat-
ing in the button press 5.4 sec after the vehicle has
entered the display.

    All nine  subjects  developed this large positive
potential between the time when an eye movement
brought the gaze to the region of the vehicle and the
switch press indicated detection. In 90% of the trials
of all subjects, the potentials were identified in single
trials stored on digital tape. The average amplitude at
the vertex of this detection potential across nine sub-
jects was 38 juV; mean amplitudes for individual sub-
jects ranged from 20 to 63 /iV. The field distribution
is such  that Cz and Pz are more or less equipotential
and the field falls to about  60% at electrodes Fz, Oz
P3,P4, C3 andC4 (Cooper et al. 1977).

    The occurrence of this potential was  not time-
locked  to the  eye movement taking the eye into the
target area, nor was it at a fixed time before the switch
press that indicated detection. The frequency distri-
bution of the times of occurrence of eye movements
and switch press  with respect to the positivity are

-------
 190
                                   Cooper et al.
                                                                           I       I \
-------
Visual Detection Potentials
                                                                                             191
           Fpz-2M
                                           m	L JwiiJiSii.^k
             BIPOLAR
                                                                                    I
             EMG
SECONDS    1
      I I 111! LI I I I J
                                           4
                                           ll
 8     9     1O
n ll il i ii.i.u li ' j i 11 ml i mi
              Fig. 2. Single-trial multichannel recording showing horizontal eye scan pattern fch 1)
              with fixation and tracking (detection) occurring about 4 sec after the vehicle entered
              the display. Same trial as Fig. 1. A slow potential change started in the right posterior
              region (O2 • P4S ch 6) when the eye moved into the target area. A large positive poten-
              tial occurred at about the time the eye fixated on the target and was largest at the
              vertex fch 3), slightly smaller at the midline occipital (Oz)and hardly seen in the
              frontal region (Fpz). EMG (ch 8) and switch press (ch 9) occurred soon after the
              detection potential.  Tl  corresponds to AB in Fig.  1. Time T2 is the end of the sac--
              cade that  takes the eye  from the left of the display to the target area. Tracking starts
              shortly after this eye movement. Calibration  of EOG, 25QuV,  EEC and EMG.
              25 uV. Negative up.

-------
 192
                                                                                         Cooper et al.
                                   PEAK POSITIVITY FREQUENCY
                                                                     SWITCH PRESS
              EYE MOVEMENT
                                               TIME, sec

Fig. 3.  Frequency distribution of time relationships of eye movements and switch press to detection positivity
for 24 trials in six subjects.  The positive peak occurs at  time zero.  The switch press histogram shows that most
presses (34) occurred between 200 and 300 msec after the positive peak.  Although the distribution of eye move-
ments  was bimodal, the largest number (30)  occurred between 200 and 300 msec before detection positivity.
Due to the difficulty in deciding which was the  last eye movement preceding detection, there are only 110 eye
movement measurements.
 be  expected if the vehicle  appeared in  peripheral
 vision. Analysis of scan patterns revealed no obvious
 disruption  or abrupt termination of the normal scan
 immediately prior to detection, which might confirm
 triggering (interval Tl  T2 in Fig. 2). However, since
 slow  potential  shifts  started at this  point  in time,
 this eye  movement toward the  target area seems to
 have special significance.

    The  voltage distribution  of slow potential shifts
 observed in  these  subjects  was unusual  in  that it
 appeared  in  the occipital-parietal bipolar channel.
 TTiere are two possible interpretations: (1) that a slow
 positive shift occurred  at the parietal electrode or (2)
 that  a negative shift  occurred  at the occipital elec-
 trode. Further information about the field distribution
 from  occipital and vertex referential recordings is not
 easy  to obtain  since these recordings are confounded
 by  the appearance of  the larger positive  detection
potentials  The occipital negativity interpretation is
consistent with other data showing changes of steady
potential in the occipital region during visual  tasks
(Kurtzberg and Vaughan  1977). This bipolar  nega-
tivity lasted  only a few seconds before changing
polarity,  a  change  that may be explained by an in-
creasing  negativity at the vertex (cf. ch 2  Fig. 4)
spreading into the parietal area.

    The large positive potential that occurred at about
the time when tracking started has been called the
detection positivity (Cooper et al. 1976). This poten-
tial showed very little reduction of amplitude during
the 24 events and could usually be seen in the original
data between the eye movement and switch press.  It
was maximum at the vertex and midline parietal area
and spread into the occiput and  frontal regions, but
not to Fpz. It  might be generated when the observer
changed  from  a scanning to a tracking mode of eye

-------
Visual Detection Potentials
                                                                                                  193
 Fpz-2M  A
 Cz-2M
 Oz-2M
                START OF
    SECONDS
                                          AVERAGE ALL SUBJECTS
                                                                      I
                                     t
 AVERAGE PRESS
EYE MOVEMENT
 Fig. 4. EEC averages time-locked to the last large movement (time T2) before detection. EEC data were averaged across
 six subjects (about 20 trials each).  The prefrontal channel shows a remnant of eye movement. The positivity at
 the vertex and occipital electrodes shortly after eye movement (vertical line) is  the average of detection potentials
 imprecisely synchronized to eye movement. Note increasing negativity starting at  eye movement in right bipolar
 recording and long-lasting negativity at vertex.  Time constant. 8 sec.
 movement,  or  it  may  represent the confirmation
 of a visual event being matched against  a particular
 set  or  expectation. If the latter were the case, the
 potential would clearly have much in common with
 the P300,  especially since the last eye  movement
 often preceded the detection potential by 200 to 300
 msec (Fig. 3). P300latency is known to vary consider-
 ably according to circumstances. Similar factors might
 account for the difficulty in establishing a clear tem-
 poral  relationship  between detection positivity and
 environmental or other physiological events such as
 eye movements.  Detection potentials occurred only
 to the  detection of a vehicle and not to  the identifi-
 cation  of the type of vehicle,  which occurred later.
 This observation is consistent with the concept of a
                         match  between the input signal and  a neural tern-
                         plate-in this case, of a vehicle.
                         Summary

                             The first electrophysiological sign of detection of
                         an infrequent event occurring in a visual display is a
                         slow rise of negativity at vertex, parietal, and occipital
                         electrodes. This response usually starts about the time
                         the gaze transfers to the area of the display contain-
                         ing the event. A large  positive potential, which seems
                         to appear  when the presence of the target is confirm-
                         ed, follows.  The switch press  indicating detection
                         follows about 300 msec later.

-------
  ANALYSIS OF NONSIGNAL EVOKED CORTICAL
  POTENTIALS IN  TWO  KINDS OF  VIGILANCE
  TASKS
 D. FRIEDMAN, W. RITTER1, AND R. SIMSON

  Department of Medical Genetics, New York State Psychiatric Institute, New York, and
 Albert Einstein College of Medicine, Bronx, NY, U.S.A.
     A  number of investigators have reported that
 when the ratio of signals to nonsignalsis 1:10, signals
 elicit P300s but nonsignals do not (Hillyard et al.
 1973, Ritter and Vaughan 1969, Squires et al. 1975).
 Using language stimuli, Friedman et al. (1975) found
 that nonsignals did elicit P300s, possibly due to the
 engagement of the P300 system by verbal stimuli (cf.
 Galambos et  al.  1975). The  amplitude of P300 to
 nonsignals could be related to the degree of processing
 required by linguistic analysis, as well as to the relative
 frequency of nonsignals (Friedman et al. 1973, Tuet-
 ing et al. 1970).

    In  order  to  shed light on the  factors affecting
 nonsignal  P300  amplitude,  two experiments were
 devised; in one, the degree of information processing
 required by nonsignals was manipulated, and in the
 other, the relative frequency of signals to nonsignals
 was manipulated. Two kinds of vigilance tasks were
 used in the first experiment. In one condition, subjects
 were instructed to respond to a specific target, and in
 the other  condition, to any stimulus that occurred
 twice  in succession. The processing demands of the
 latter  task were greater because each nonsignal had to
 be remembered for comparison with the next stimulus
 to see whether it was a signal. In the second  experi-
 ment  using only  the single-target task,  the relative
 frequency of nonsignals was manipulated in two ways
 ways—by varying the percentage of nonsignals and
signals and by  varying whether the  nonsignals were
comprised  of many different  stimuli or only one
stimulus.

General procedures

    Subjects ranged  in age from 27 to  45; all had
previous experience in evoked potential experiments.

    Electrical activity was recorded from nonpolariz-
able Ag/AgCl electrodes at Fz, Cz, Pz, Oz, and above
the right eye, referred to the right earlobe. The ampli-
 fier time-constant was 1 sec. For two subjects, EOG
 was recorded  from left and right outer canthi  to
 monitor possible lateral eye movements.

     Visual stimuli, 50 msec in duration, were moni-
 tored from a DEC VR-14 slave scope by videocamera
 and presented  at an interstimulus interval of 1.5 sec
 on a video monitor with a central fixation point. The
 monitor was continuously illuminated, and the stim-
 uli were of moderate intensity and subtended a visual
 angle of 2 deg 20 mhv. Stimulus presentation and data
 acquisition were controlled by a POP 11/10 computer.
 EEC was digitized every 4 msec to obtain a 100-msec
 prestimulus baseline  and  1000-msec poststimulus
 epoch;  data were stored on digital  tape  for off-line
 analysis.

 Results

 Experiment 1

     Stimuli in each task were the numbers 02 to 19,
 presented in Task A in blocks of 60 and in Task B in
 blocks of 64. In Task A, the signal number 08 occur-
 red 15 times per block, and 15 of the numbers from
 02 to 19 occurred 3 times each per block for a total
 of 45 nonsignals. In Task B, signals were the repetition
 of any immediately preceding number. Signals occur-
 red 16  times per block; and nonsignals, which were
 12 of the  numbers  from 02 to 19, occurred 4 times
 each per block, for a total of 48 nonsignals. The ratio
 of signals to nonsignals was 1:4 in both tasks. Tasks
 were alternated for a total of eight blocks of each
 with the order  of tasks counterbalanced  across sub-
jects. Subjects were instructed to respond to signals
with a brisk wrist extension and rectified EMG was
 used to  record  reaction time. Averages for each sub-
ject were computed across blocks for 120 signals and
 360 nonsignals (Task A) and for 128 signals and 384
 nonsignals (Task B).

    Averaged evoked potentials for two subjects  to
each stimulus class in the two tasks are presented in
 Fig.  1.  The mean topographic distribution of P300
1 Al«o at Herbert H. Lehman College, City University of New York, Department of Psychology.

-------
Nonsignal EPs in Vigilance Tasks

     NONSIGNAL
                                                                          195
  Pz
 ,-.P300   —TASK A

/•A\      .....TASK Bfl  B \\
                      DF     "X
  Pz
                       BJ
         A     ss~>~

  Oz     Vtf
      •\ I I  I  I I  I I  T
         1100 mMC
                    1*
5 /iV                 15 MV
      • \  i i  , i  i i  i i  I-
         1100
Fig  1 Averaged evoked potentials from parietal and
occipital electrodes for two subjects. A complete ex-
perimental protocol is shown. Stimulus onset at 100
msec. Time lines every 110 msec.
across the four subjects measured from the prestim-
ulus baseline is presented in Fig 2. P300 to both the
signals and nonsignals had a parietal  focus. T-tests for
correlated means  were used  to  assess significance.
Because  multiple comparisons were performed, the
Bonferroni criterion (Hays 1963)  was used to correct
for the number of tests.  A corrected alpha  level of
0 05  was used and only  those t-values that had an
associated probability of .01 or less were accepted as
significant. P300 amplitudes to signals in both condi-
tions did not differ significantly and were always
larger than the respective nonsignal P300. P300 to
nonsignals in Task  B was of larger amplitude than
P300 to  nonsignals  in Task A at Fz (p < .001), Cz
(p<.005),andPz(p<.001),butnotatOz(p<.05).

     The  findings indicate a large task-specific effect
of processing complexity  on P300 amplitude to non-
signals, an effect that is independent of probability.
 Reaction time analyses revealed longer reaction times
 to Task B signals (mean=480 msec) than to Task A
 signals (mean=451 msec), showing an effect of proces-
 sing complexity on behavioral responses as well.
                                                         10 -
                                                          8 -
                                                     a

                                                     §    6
                                                     a.
                                                     1
                                                          4 •
                                                     a.
                                                     6
                                            	K  TASK  A NS

                                            ---•  TASK  B NS

                                            ---«  TASK  A SIG

                                            	•  TASK  B SIG
                                                Fz
                                                                    I
                                                                    Cz
Pz
Oz
                                           Fig 2. Mean baseline to peak P300 amplitude for four
                                           subjects. Data for signals (SIG) and nonsignals (NS)
                                           from each task are shown.

                                                To check the intrasubject reliability of these find-
                                           ings, the data of subjects BJ and DF were divided into
                                           first-half and second-half averages. In both cases, the
                                           effect of task demands on nonsignal P300 amplitude
                                           was replicated in  each half of the experiment. In fact,
                                           the experiment was repeated with subject BJ a week
                                           later, and the results were the same. EOG recording
                                           revealed no appreciable lateral  or vertical eye move-
                                           ments.  Vertical EOG was  extremely low amplitude
                                           and did not show any systematic changes between or
                                           within conditions.
                                Experiment 2

                                     In Task A of Experiment 1, clear P300 compon-
                                ents were observed in nonsignal waveforms where the
                                probability of the class of nonsignals was 0.75 and
                                where each nonsignal occurred  with a probability of
                                0.05. To test the hypothesis that P300 to nonsignals
                                in this situation  was  affected by the probability of a
                                given nonsignal,  additional experiments were  run in
                                which blocks of trials with many different nonsignals
                                (15, three times each, yielding  a probability of 0.05
                                for each were alternated with blocks where only one
                                nonsignal (the  number  02 with  a probability of
                                0.75) occurred. The signal (08), and other procedures
                                were the same as Task A of Experiment 1. Two sub-
                                jects served  for eight; blocks of each condition. As
                                shown on the left side of Fig. 3,*P300 to nonsignals
                                was present  and of  equal amplitude in  both condi-
                                tions. Signal P300s were also equivalent in these two
                                conditions.

-------
  196
                                                                                        Friedman et al.
             1 IN 4
             1 IN 10

	 ALL DIFFERENT


— ALL SAME
                                                                   1 IN 4
                                                              ALL DIFFERENT
    1 IN 10
ALL DIFFERENT
     t.MM...
          1100 mue
                       BjUV
          1100 mMo
 F/f 3. Left-hand  panel:  Nomignal stimuli from a
 condition In  which they occurred 75 percent of the
 time taken together and 5 percent for a given non-
 signal fall different/ or 75 percent for one nonsignal
 (all same); N is 360 per averaged response, Right-hand
 panel: Nonsignal stimuli where the probability taken
 together \vat  0.90 and approximately 0.06 for a given
 nonsignal fall different) and 0,90 for one  nonsignal
 fall same); N is 108 per averaged response.

     Since the usual vigilance situation Is one In which
 the ratio of signals to nonsignali It lower (e.g.,  1 in
 10), an  additional experiment was conducted uilng
 the lame  procedure described  above, but  with  a
 signal/nonsignal ratio of 1:10. Two alternating blocks
 with all  different nonslgnals (probability " 0.06 for
 each and a claw probability of 0.90), or a single non-
 signal (probability  •  0.90) were run with one subject
(WR).  Waveforms on the right side in Fig. 3 disclose
 that for the  1:10 ratios, P300s are not discernible
 regardless of the probability of individual nonsignals,
 Furthermore, for this same subject (Fig. 4), compari-
 son of nonsignal waveforms in the 1:4 conditions
(class probability « 0.7S)  with nonsignal waveforms
in the  1:10 condition (class probability » 0.90) shows
 that P300 is  unobservable when the probability is
 increased to 0.90.

     Although preliminary, these results suggest that
 P300 amplitude in a simple vigilance situation is an
 inverse function  of the  probability of occurrence of
 nonsignals taken as a class rather than the probability
 of a given nonsignal, and that at greater nonsignal
 densitiea  P300 may  not be observable  in nonsignal
 waveforms at all.

Discussion

     The  results indicate that  P300i are elicited by
 nonsignal stimuli, and their production is dependent
                                        1lOOmMO
   1100mm
                               Fig, 4. Comparison of waveforms to nonsignal stimuli
                               from one subject (WR) In conditions where taken to-
                               gether they occurred with a probability ofO. 75 fl;4)
                               or 0.90 (1:10).

                               upon  two factors: (1) processing demands  of non-
                               signals and (2) probability of a class of nonsignals and
                               not probability of a given nonsignal within that class
                               (see Friedman et al. 1975). These findings suggest the
                               hypothesis that these two factors Interact in  produc-
                               ing P300  amplitude changes  to  nonsignal  stimuli.
                               When probabilities are equivalent (as in Tasks A and
                               B) but processing demands differ, P300  amplitude
                               will be  larger to the nonsignal stimuli that require
                               greater processing. When probability is high (e.g., 0.9)
                               and task demands are not great (as in Task A), nonsig-
                               nal  stimuli will not  elicit observable P300s at  the
                               scalp. The type of cognitive processing required in
                               Task A was not  sufficient  to produce P300s at a
                               nonsignal probability of 0.90, whereas subject WR
                               had discernible P300s to the same nonsignal stimuli at
                               a probability of 0.75.

                                  Task B experiments in which the  ratio of signals
                               to nonsignals  is 1:10 are now being  carried  out.  If
                               P300s are  present to these nonsignal stimuli when
                               processing  demands are greater, this will confirm the
                              hypothesis   that both  probability and  processing
                               demands interact to alter nonsignal P300 amplitude.

                                  It is possible  that differential attention favoring
                               the nonsignals of Task B contributed to the observed
                              effect. In one subject, longer analysis times revealed
                              no  consistent  negative  slow  wave  development
                               between successive nonsignals In either task, ruling
                              out CNV return as a factor in  this result. In Task A,
                              the standard vigilance situation, the subject  simply
                              has to determine the presence or absence of the signal

-------
Nonsignal EPs in Vigilance Tasks

on  any  given trial (long-term  template  match).  In
Task  B, the subject  must  store each nonsignal  in
memory and wait  for the next to  see if it recurs
(short-term template match). Longer reaction times
to Task B signals are consistent with the suggestion of
increased processing load in that condition.

    The finding that class probability and not the
probability of a given  nonsignal affects P300 ampli-
tude can be generalized to signals in these tasks. In
experiment 1, signals of Tasks A and  B occurred with
a class probability of 0.25, but each different signal in
Task  B had a much lower probability of occurrence
(approximately 0.015). However, signal-evoked P300
amplitude did not differ between tasks,

    The finding that probability affects  nonsignal
P300  amplitude  is consistent with  the  results  of
Squires et  al. (1975) who found PSOOs to nonsignals
when probability was 0.50. The fact  that P300s were
not seen to nonsignals when they occurred 90 percent
of the time does not necessarily mean that there was
no  P300 brain response to these stimuli.  Rather, the
probability findings suggest that, when  nonsignals
occur  90 percent of the time, the brain's response is
too small to be seen at scalp electrodes.

    In summary, these experiments demonstrate that
nonsignal P300 amplitude is positively related to the
degree of cognitive processing required by nonsignal
stimuli, as demonstrated for signal PSOOs in a wide
variety of cognitive tasks (e.g., Donchin et al. 1973,
Poon et al. 1976, Friedman et al, 1975), is negatively
related to the frequency of the class  comprising non-
signals, and is  unrelated  to  the probability of any
particular  nonsignal within  such a  class. The data
implicate memory storage  and/or retrieval as a possi-
ble  functional correlate of P300, and suggest that one
should  be  able  to manipulate P300 amplitude  by
changing the complexity of memory storage or search.

Summary

    Visual  evoked  potentials were  recorded from
four subjects in  response  to numeric stimuli during
                                             197

 two kinds of vigilance  tasks in  which the signal-to-
 ri onsignal ratio was 1:4.  hi  Task A, subjects  were
 required to respond to the same stimulus (08) through-
 out a block of trials; and in Task B, subjects  were
 required to respond to any stimulus that occurred
 twice in succession. Signal stimuli from the two  tasks
 produced similar-amplitude PSOOs, while nonsignals
 of Task B produced significantly greater P300 ampli-
 tudes  than  nonsignals  of Task A, This P300 ampli-
 tude  difference   reflected   the  greater  processing
 demands of nonsignals in Task  B, since in that task
 subjects were required  to  remember  each nonsignal
 and then wait to see if the number recurred.
     A second experiment assessed the effect of non-
 signal probability on P300 amplitude by requiring the
 same response  as in Task A with the same signal-to-
 nonsignal ratio (1:4), but with a change in the relative
 probability  of each  nonsignal. In one condition, the
 same nonsignal (02) was used throughout, and in a
 second condition several nonsignals (each with a rela-
 tive probability of 0.05) were used. No difference in
 nonsignal P300 amplitude was found, which indicates
 that  the  probability  of a  class  of nonsignals  rather
 than the probability of individual nonsijgnals affected
 P300 amplitude. In part 2 of this experiment, a signal-
 to-nonsignal ratio  of 1:10 was used, with either the
 same or different nonsignals. No P300s to nonsignals
 in either condition were observed. Collectively, these
 findings (1) demonstrate that P300 amplitude to non-
 signal stimuli is dependent upon both the probability
 of a class of nonsignal stimuli and on the processing
 demands  elicited by these  nonsignal stimuli, and (2)
 suggest that these  two factors  interact in affecting
 nonsignal P300 amplitude.
 Acknowledgments

    The authors  wish to thank Dr. L. Erlenmeyer-
Kimling, in whose laboratory this  research was per-
formed, and  Mr.  Jim Hollenberg for computer pro-
grams used in this study.

-------
 VARIATIONS  IN THE  LATENCY OF P300 AS A
 FUNCTION  OF VARIATIONS IN SEMANTIC '
 CATEGORIZATIONS

 M. KUTAS2 AND E.  DONCHIN
 Cognitive  Psychophysiology Laboratory, University of Illinois, Champaign, IL, U.S.A.
     Several recent  studies  of P300  (Ritter  and
 Vaughan  1969, Tueting et al. 1970, Squires et al.
 1977) utilized the following experimental paradigm:
 the subject is presented with a stream of stimuli, each
 of which  may belong to one  of two categories. The
 subject counts and reports the number of stimuli that
 belong to one of the two categories. The probability
 is low that a stimulus will belong to one of the cate-
 gories and correspondingly high that it will belong to
 the other category. Stimuli in  the low-probability
 category  were  found to  elicit  an  enhanced P300
 component. In  most  experiments reported to date,
 categorizations  were based on physical  features of
 stimuli, such as the frequency of tones, hue of light
 flashes, or specific pattern of the visual stimulus. It
 seemed necessary  to determine  if the same  results
 would be  observed if the categorization required of
 the  subject  was based on semantic features  of the
 stimuli (cf. Friedman et al. 1975).

     For this purpose, subjects were presented with
 sequences  of words, each of which could be categoriz-
 ed, on the basis of a semantic rule, into one of two
 categories.  The two categories  appeared with  the
 probability of either 0.20 or 0.80. The intent  was to
 determine  the  extent to which the  appearance of
 stimuli belonging to the rare category would enhance
 the P300 component. As the complexity and latency
 of the categorization response varied with semantic
 categories, the relationship between the duration of
 cognitive operations  and the  latency of the P300
 component could also be examined. If P300 reflects
 specific cognitive processing activity (Donchin 1975,
 Donchin et  al.  1973),  then  the latency of P300
 relative to  the physical stimulus would depend upon
 the latency and duration  of the cognitive process and
 would  vary  as a function of its complexity. This
 proposition was tested in  the present study.
Methods

    The experiment  utilized PLATO, a computer-
assisted instruction system developed  at the Univer-
sity of Dlinois. The PLATO terminal  uses a plasma
panel  for display (Smith and Sherwood 1976). The
display is achieved by illuminating any of 512 x 512
luminous dots. In this experiment, the PLATO sys-
tem was programmed to present a sequence of words
on the terminal, one at a time every 2000 msec. Each
word was preceded by an external trigger, which was
led to a PDP-8/E computer. The trigger activated the
digitizing process so that EEC data would be acquired
in relation to the presentation of the stimuli.

    Data  from three  studies are reported. In each
study, subjects were  presented with four different
sequences of words. Each sequence consisted of about
200 words selected randomly on  each  trial with the
appropriate  probability. The  following series  were
used:
      fixed names. The words were either "Nancy"
      or "David." "Nancy" appeared 20% of the
      time.
      Variable names. Words were selected from a
      list of 20 female or 20 male names. Each name
      was a two-syllable word consisting of five let-
      ters, Twenty percent of the names were selec-
      ted from the female name list and 80% from
      the male name list.
      Rhymes, Words were selected either from a list
      of different  words rhyming with "cake" or
      from a list of 20 four- or five-letter nonrhym-
      ing one-syllable words. Rhymes were presented
      20% of the time.3
      Synonyms. Words were selected  from a list of
      20 arbitrarily chosen words and 9 synonyms
      of the word "prod." Synonyms  were present-
      ed 20% of the time.4
 1.
2.
4.
* An extended report of these experiments has been published elsewhere (Kutas et al. 1977).
2 Present address: Department of Neurosciences, University of California Medical School, La Jolla; CA, U.S.A.
3The nine words rhyming were: bake, rake, stake, lake, make, fake, wake, flake, and take. In the second experi-
 ment, the words ache and steak were included.
4The synonyms presented were goad, poke, shove, nudge, urge, push, prompt, spur, and press.

-------
 P300 Latency and Semantic Categorizations
                                            199
     Brain  potentials were  monitored with  Burden
electrodes  placed at  Fz, Cz, Pz, C3, and C4 referred
to linked mastoids in the first experiment and to the
chin in others. EOG was recorded between supraorbi-
tal and canthal positions. The subject was grounded
on the forehead. EEC was recorded on analog tape
with a 2-sec time constant and 30-Hz high-frequency
cutoff  amplitude. Data were digitized off-line by an
IBM 1800  computer and stored on digital magnetic
tape. Digitizing started 220 msec prior to the stimulus
and ended  780 msec after the stimulus. The sampling
rate was 10 msec per point.  A PDP-8/E  computer
determined, on line, whether eye movement artifacts
were present during each trial by  comparing EOG
variance to a criterion value. 5 Contaminated trials
were not included in the average (the synchronizing
pulse on the analog tape was inhibited).

     Subjects sat  in a comfortable chair in a semi-dark-
ened, shielded room and completed the four condi-
tions in the following order:  fixed names, variable
names, rhymes, and synonyms. Prior to each block of
trials,  the subject was instructed to watch the words
and count  stimuli  from the rare category. At the end
of each run, the subject was asked to report his count.
Each condition consisted of approximately 200 trials
(40 rares).

Results and discussion

Experiment 1

     Six subjects participated in the  first experiment,
in which the task was to count the number of stimuli
belonging  to the  rare  category.  Fig. 1 presents the
data from  one subject and  superaverages computed
over the entire subject group.

     Evoked responses elicited by stimuli belonging to
the frequent category lack a P300 component, while
ERPs elicited by stimuli belonging to the rare category
show  a marked  P300 component. Latency  of the
P300 component varied widely-shortest for the fixed
name  categorization and the longest for synonyms.
Variable names and rhyming words showed intermedi-
ate latencies. The same order of latencies characterized
data from all subjects. P300 amplitude elicited during
the fixed-name condition was larger than in the other
three conditions. There were no systematic differen-
ces among the amplitudes of the average P300s.

    These  data demonstrated that the P300 response
was associated with the categorization of stimuli even
when categorization  depended upon semantic rather
than  physical characteristics  of  stimuli.  Although
averaging was over a diverse array of physical stimuli,
        FREQUENTS
                                    RAKES
                      CENTRAL,
                    SINGLE SUBJECT
                      FRONTALS
                      CENTRALS
                      PARIETALS
           TIME, mite
     - FIXED NAMtS
     -VARIABLE NAMES
     -RHYMES
     - SYNONYMS
Fig. 1.  Superimpositions of average ERPs obtained
during  experimental conditions requiring different
semantic categorizations. At the top are sample Cz
waveforms to the rare and requent stimuli from an
individual  subject.  The remaining  waveforms are
superaverages across six subjects for Fz, Cz, and Pz
positions. Only the rare stimuli (prob. = 0.20) were
counted. Each rare  waveform  consists  of approxi-
mately 18 to 30 single trials.
such as different female names, a clear P300 response
was elicited. The data were consistent with the sugges-
tion that P300 latency varies systematically with the
complexity of information processing required.

Experiment 2

    To validate the extent  to which  differences
between series were related to different subject deci-
sion times, the experiment was repeated with  five
other subjects, who were asked to respond rapidly by
pressing one of two buttons upon the appearance of
any  stimulus  from  the two  categories. The  results,
shown in Fig. 2, differ from those obtained in Experi-
ment 1 in two important respects. First,  it appeared
that the execution of a motor response changed the
appearance of the "frequent" evoked response in that
5 Trials were rejected if the sum of squared digitized values (220 msec prestimulus and 780 msec poststimulus)
 exceeded a criterion value determined by visual inspection of a large sample of EOG traces and their correspond-
 ing digital values.

-------
 200
                                             Kutas and Donchin
     FREQUENTS
                                    RARES
                     FRONTALS
                     CENTRALS
                    PARIETALS

                 I  I  I
         TIME, msec
  — FIXED NAMES
     •VARIABLE NAMES
  	RHYMES
  — SYNONYMS
TIME,
Fig.  2. Superimpositions of ERPs averaged across five
subjects from four different experimental conditions.
Subjects were required to perform a choice reaction
time response,  responding to frequent stimuli with
one hand and rare stimuli with the other.

a positive component, presumably P2 of the motor
potential, could be  observed. This positivity could
not  fully account for the marked enhancement of
positivity with a latency of about 400 msec associated
with rare stimuli.  Latencies  obtained in this second
experiment, however, were somewhat less differenti-
ated than those observed in the first experiment. The
fixed-name latency was still considerably shorter than
that associated with  the other three conditions; how-
ever, these three conditions were no longer as differen-
tiated as they were during the count condition. Table
1 presents means and standard deviations of reaction
times (RT) averaged  over the five subjects for each of
the  experimental conditions. These means are based
on RT scores obtained  from all trials on which sub-
jects responded correctly.  The large variance of RT is
noteworthy. The  fixed-name RT  was  substantially
shorter than the RT  of the other conditions, which
were essentially equal. The variance of P300 latency
(Fig. 2)  was equally large, suggesting a substantial
degree of trial-by-trial variation  in P300 latency and
RT.  An  analysis of  the relationship between P300
latency and RT is presented by Kutas et al. (1977).


     Failure  to  observe  differences in P300 latency
with three  of four experimental conditions could be
attributed to the fact that in the second experiment
speed of response  was emphasized  without requiring
accuracy in  categorization. Thus, subjects tended to
                 execute erroneous categorizations as they attempted
                 to maximize  response speed.  The error rate varied
                 across  experimental conditions, with  fewest errors
                 occurring during the fixed-name condition and most
                 occurring  during the  synonym condition.  Clearly,
                 subjects could have  traded accuracy  for speed. A
                 third experiment was therefore  run to assess this
                 possibility.
Experiment 3

    In the final experiment, the rhyme series was not
used.  Five subjects participated under three experi-
mental conditions with each of the remaining series.
The "count" condition replicated Experiment  1 and
the "RT-accuracy" condition subjects made a choice
reaction to  the stimulus, but were instructed  to be
very accurate. The results are shown in Fig. 3.  When
accuracy was emphasized, P300  latencies varied  in
the same manner as in the count condition.  When
speed was a prime consideration, subjects seemed to
maximize speed by reducing processing time invested
                       FREQUENTS
                                                  RARES
                  	FIXED NAMES
                  	VARIABLE NAMES
                  	SYNONYMS                    TIME, mwc
                 Fig. 3. Superimposition of central (Cz) ERPs averaged
                 across five subjects for three semantic categorizations
                 obtained during three different response regimes.

-------
P300 Latency and Semantic Categorizations
                                        201
        Table 1. Means and Standard Deviations of Reaction Times, Experiment 2
                                  (RT-speed, five subjects)
Condition
Fixed names
Variable names
Rhymes
Synonyms
Rare stimuli
Mean RT
514.41
613.25
633.48
666.55
Standard
deviation
83.31
89.82
123.79
109.43
Frequent stimuli
Mean RT
419.43
522.22
524.22
513.48
Standard
deviation
96.88
107.24
125.39
105.78
         Table  2.  Grand  Means  and Standard Deviations of  Reaction Times,
               Experiment 3 (Count, RT-accuracy, RT-speed, five subjects)


Condition
Fixed names-
accuracy
Variable names-
accuracy
Synonyms-
accuracy
Fixed names-
speed
Variable names-
speed
Synonyms-
speed
Rare stimuli

Mean RT

543.56

573.03

619.73

455.47

506.92

531.43
Standard
deviation

111.19

164.67

100.60

77.23

100.65

94.70
Frequent stimuli

Mean RT

453.27

498.79

520.25

353.15

409.29

413.75
Standard
deviation

110.09

99.38

119.56

102.89

92.33

96.99
in the categorization, thereby reducing the variabil-
ity in P300 latency. Grand means and standard de-
viations of RT for the six conditions an presented in
Table 2.

Conclusion

    The  latency of the  late positive  component
(P300) associated with rare occurrences of relevant
stimuli varies with  stimulus  evaluation  time. The
differences in P300 latency cannot be attributed to
the effects of the motor response on  P2. Latency
differences observed during the count condition were
quite similar to those observed in the  RT-accuracy
condition, yet  no manual responses were required
during  count conditions. The data  are  consistent
with the view that the variable-latency parietal-maxi-
mum, late positive waves are manifestations of the
activity of the same intracranial processor. Alternate
views that tend to differentiate between late positive
components  by  their latency (e.g., Thatcher 1977)
seem less parsimonious.

-------
 TOPOGRAPHICAL  STUDY OF THE EMITTED
 POTENTIAL OBTAINED AFTER THE OMISSION
 OF AN  EXPECTED VISUAL STIMULUS
 B. RENAULT1 AND N. LESEVRE2

 Laboratory of Electroencephalography and Applied Neurophysiology, Salpetriere Hospital,
 Paris, France
    Several  laboratories  have  reported a cerebral
 potential, the 'emitted potential,' following an ex-
 pected but missing stimulus (Barlow et al. 1965;
 Sutton et al.  1967; Klinke  et al. 1968; Ritter and
 Vaughan  1969; Barlow 1969; Weinberg et al. 1970,
 1974;  Picton  and Hillyard 1974). The emitted po-
 tential  is considered to  be  similar to the positive
 component  (P300)  that  peaks  about 250 to  500
 msec after a stimulus that provides information to
 the subject, especially under conditions of uncert-
 ainty (see review by Tueting, this section). Simson et
 al.  (1976) and Renault  and Lesevre (1976)  have
 shown  that the  emitted positive  potential is pre-
 ceded by a negative  component that  usually peaks
 in the  posterior region when the missing stimulus
 is visual.  Moreover, Simson et  al. (1976) demon-
 strated  that this negative emitted  component changed
 topography  according to stimulus modality, where-
 as the positive wave did not.
    Renault and Lesevre (1976) also showed that
both the negative and the  positive components of
the emitted potential were made up of successive
or  overlapping peaks with various  topographies.
Latency variability in single  trials due to the absence
of a precisely  timed external stimulus (Ruchkin and
Sutton 1978)  could explain these average successive
peaks if  they  all  had the same topography. Since
this is not the case, emitted components were topo-
graphically analysed  trial by  trial in  the present
study, utilizing the spatiotemporal mapping method
of Remond (1961).  To test the possible topograph-
ical specificity of this phenomenon, a comparison
was made between the missing visual stimulus potent-
ial and the early part of the visual evoked potential
obtained with pattern stimulation.
   Two experiments were  conducted to evaluate
successive or overlapping components of the emit-
ted potential as related to different brain generators
and psychophysiological mechanisms.  In  particular,
the variation of these components with different
tasks involving or not involving motor response was
investigated.  Topographical  specificity of emitted
potential components was then examined by compar-
ing such potentials obtained under half-field stimul-
ation  conditions with  potentials  obtained  under
whole-field conditions and with the early part of the
pattern evoked potential.

Methods

   Five normal adults served as subjects. In the first
experiment, three  runs of 450 visual stimuli, each
of the same  type (whole-field  stimulation), were
recorded for each subject. The stimulus was a check-
erboard projected  at  1-sec intervals for  a period of
20 msec on a screen, which had a fixation point in
the center. During each run,  10% of the stimuli were
randomly omitted.  The  response  required  of the
subjects was different in each run:

   1. Go  condition-The subject was  instructed  to
     respond as quickly as  possible  to the missing
     stimulus  with  a  right-finger  displacement
     toward the right.

   2. No-go condition-The subject was instructed to
     respond to each stimulus, but to withhold re-
     sponse to the missing stimulus.

   3. Counting condition-The subject was instruc-
     ed  to  count  the missing stimuli mentally
    without motor  response to  the presentation
     or omission of the stimulus.
1 Attache de Recherche, CNRS.
2 Maitre de Recherche, INSERM.

-------
 EP after Omitted Stimulus
                                             203
         vs
                                     OMISSION
Fig. 1. Five chronograms (above) obtained from the longitudinal part of the montage and the corresponding spa-
tiotemporal map (below), representing the average emitted response to 45 missing stimuli occurring in a series of
whole-field checkerboard stimulations during a counting task.  The emitted potential is preceded and followed by
average visual evoked potentials.  On the time scale, 0 indicates the moment the stimulus should have appeared
(omissions} or the occurrence of a visual stimulus (vs). On the map, the ordinate represents space, i.e., the loca-
tion on the scalp of the five electrodes of the longitudinal arm of the montage. The values between two succes-
sive electrodes are obtained by mathematical interpolation. Amplitude is represented in the form of isopotential
lines;  in this figure,  the potential increases in 2-n V steps from one isopotential line to the next.  Thick lines indi-
cate potential 0; thin lines indicate negative potentials; broken lines indicate positive potentials. Peaks are indi-
cated  by + or -.  Pattern evoked potentials include N100 and P200 components peaking in the preoccipital
region.  The emitted potential is characterized by a negative wave that peaks (4 i*V) at 125 msec in the parietal
region and later in the preoccipital region, and a positive  wave (8 [iV) that peaks at the vertex (380 msec) and
later in the parieto-occipital area (400 to 700 msec). Data from the same subject are shown in all figures.
   The second experiment also involved three runs
of 450 visual stimuli. A different type of stimulation
was  employed in  each run:  (1)  whole-field, (2)
right half field, (3) left half-field.  Subjects  counted
omissions, which  occurred  with 10% random pro-
bability.

   Recordings were made with a montage of nine
equally spaced electrodes 4 cm apart forming a cross.
The  longitudinal part of the cross was on the midline
extending from 4 cm anterior to 12 cm posterior to
the vertex (Cz). The  transverse branch crossed mid-
line  4  cm posterior to the vertex (approximately  on
the central sulcus)  and extended 8 cm on the right
and  left. Each electrode was referred to linked ears.
The  time constant  was 0.7 sec, with an upper band-
pass limit of 220 Hz. Horizontal and vertical electro-
oculograms were recorded,  and every response oc-
curring during or after an eye movement was elimin-
ated from the  analysis.

   On-line analog-to-digital conversion was done by
computer at a 2-msec sampling rate and displayed in
the form of chronograms and spatiotemporal maps
(Remond  1961).   Spatiotemporal  maps were con-
structed from averaged  data (45  responses to omis-
sions for each run), as well as  from single-trial data
when possible. (Spatial interpolation acheived in the
mapping process is  itself a way of improving signal-
to-noise ratio.)

Results
Average emitted potentials to missing stimuli

    An  average potential time-locked to the moment
the  visual  stimulus should have appeared was ob-
served in all subjects in  go and no-go reaction time
(RT) situations as well  as the counting task. This
potential consisted of a  negative  wave (N) followed
by a positive wave (P) (Fig. 1 and 2 and Tables 1
and 2). Each of these averaged components  usually
consisted  of at least   two  different  components
peaking  at  different times  and locations on the
longitudinal part of the montage. These components
were designated Na, Nb, Pa, and Pb.

     Whole-field stimulation:  Table 1  shows mean
latencies  and  amplitudes of  emitted  components

-------
  204
                            Renault and Lesevre
                      NO STIMULUS
REF:
                                                         V. STIMULUS (WHOLE FIELD)
 Ffc 2.  The upper map ft an enlargement of the part of Fig. 1 corresponding to the emit ted potential. Note the
 apparent flowing of the positive component from the frontocentral region toward the ocipital region.  The second
 map represents the same event obtained on the transverse arm of the montage.  The two bottom maps show the
 average pattern evoked potentials to  the stimuli following omissions as  obtained on the longitudinal and the
 transverse part of the montage. Isopotential steps are 2 nV.
  (measured on the longtitudinal part of the montage)
  for the counting, go, and no-go conditions. Latencies
  and  amplitudes  varied widely  among subjects  in
  each  experimental situation. Only the negative wave
  amplitude differed  consistently  according  to the
  type  of  task; the  negative  component was  larger
  in the go and no-go conditions than in the counting
  condition.

     Table 2 indicates the topography of peaks on the
  longitudinal  and  transverse  parts  of  the  montage.
 The exact position of the longitudinal peaks (referred
  to Cz)  varied among subjects, but the general spatial
 organization  of each component  was  very  similar
 for the  five  subjects.  In  the  counting condition,
 topography of  the first negative wave for four sub-
 jects was more posterior than that of the first positive
 wave.  Spatiotemporal  organization  of  the  positive
 wave  was quite similar in all subjects in that it always
 began to peak next to the vertex (Pa located at Cz
 ±  4  cm) and  culminated later in  the parietal or
 parieto-occipital region  (Pb located 6 to 12 cm pos-
 terior  to  Cz).  This longitudinal organization  was
 seen under all three  task conditions. For all subjects
 (12 situations out of 13) and all tasks, the negative
 component of the averaged  transverse emitted po-
 tential was larger  on  the  right hemisphere (3.7 juV)
 than left hemisphere (1.7
    Reaction time was longer for all subjects when
 a motor response  was required to omissions (x=555
 msec,  S.D.=141 msec) than to non-omitted visual
 stimuli (x=314 msec, S.D=100 msec).
Half-field stimulation:  in  the  case  of omissions
occurring after half-field stimulation, three subjects
(RR,  BR, and NL) showed an asymmetrical nega-
tive peak (Na) located contralaterally in both situa-
tions  (right-field and left-field). This contralaterality
was never observed for the positive component. In
the longitudinal plane, the NA component was always
located in posterior  regions.  One of the remaining
two  subjects  showed an  ipsilateral  peak  during
right-field stimulation, while the other had a midline
peak  in  the  same  situation;  thus, no comparison
could be made.

-------
EP after Omitted Stimulus
205
Table 1. Mean Latency and Amplitude of Negative (N) and Positive (P) Components of the
Emitted Potential Measured on the Longitudinal Montage during Whole-field Stimulation


Latency, msec
(S.D.)
Amplitude, jttV
(S.D.)
Counting
Na Nb Pa Pb
178 321 469 565
(33) (94) (97) (97)
2.8 2.6 4.4 4.7
(0.7) (1.2) (1.5) (2.7)
No/go
Na Nb Pa Pb
150 280 451 594
(32) (62) (83) (100)
3.9 5 4.6 3.9
(2) (1) (3) (1.2)
Go
Na Nb Pa Pb
132 272 371 535
(34) (30) (44) (64)
3.07 4.7 5.1 4.8
(0.6) (3) (4.2) (2.4)
      Table 2. Topography of Emitted Potentials during Whole-field Stimulation9

Subject
RR
JF
JPJ
BR
NL
Montage
Long.
Tr.
Long.
Tr.
Long.
Tr.
Long.
Tr.
Long.
Tr.
Counting
Na Nb
+12 Cz
8R 8R
+8 +12
8R.L 8R.L
+4 Cz
8R 8R,L
+10 +12
8R 8R
+12 +12
4R 4R
Pa
-2
Ml
Cz
Ml
+8
Ml
+4
Ml
Cz
Ml
Pb
+6
Ml
+8
4R
+10
Ml
+8
Ml
+8
Ml
No/go
Na
Cz
4R
+6
8R
+4
Ml


Nb

+8
8R
Cz
Ml

+12
4R
Pa
+4
Ml
Cz
Ml
Cz
4L
-4
Ml
Cz
4L
Pb
+10
Ml
+4
Ml
+8
4L
+8
Ml
+8
Ml
Go
Na
+4
2R
+10
8R

+4
Ml

Nb
+4
Ml
+8
8R


+12
8R
Pa Pb
Cz +8
Ml Me
-2 +6
4R 4R


+4 +10
Mt Ml
• Location of peaks on the longitudinal and transverse part of the montage for three tasks. The location of longitudinal peaks
Is expressed in cm with respect to Cz, minus is anterior and plus is posteriorto Cz. The location of transverse peaks is noted
with respect to mldllne (Ml), on the right (R), on the left (L), or bilateral (R.L).

-------
  206
Renault and Lesevre
 REF:  EARS
 Fig. 3. Vertex patterns of single-trial emitted potentials following the omission of an expected visual stimulus in
 the no-go task run. Longitudinal part of the montage. Upper map: the negative wave peaks on the parieto-occipi-
 tal region; the positive one at the vertex.  Lower map: the negative and positive waves both peak at the vertex.
 Isopotential steps are 4V-V.
 REF:  EARS
 Fig. 4. Parietal pattern of single-trial emitted potential following the omission of an expected visual stimulus
 (cross montage).  The first 100 msec following the omission are not represented. The long positive wave peaks
 first at 450 msec in the parietal region; and a second peak occurs 650 msec after the omission in the preoccipital
 region. It is preceded by at least three successive waves of short duration, also located in the parietal area. Note,
 especially in the transverse part of the montage, the similarity of this emitted pattern and the evoked potential
 of Fig. 2. Isopotential steps are 2 pV.
REF:  EARS
    4/A/
 Fig.  5.  Mixed pattern of a single-trial emitted potential following the omission of unexpected visual stimulus
 (cross montage). The first  100 msec following the omission are not represented. This emitted potential looks
 like the spatiotemporal addition of the maps of Fig. 3 and 4. Isopotential steps are 4 uV.

-------
EP after Omitted Stimulus
                                             207
Single-trial analysis of emitted potentials

    Single-trial spatiotemporal maps were constructed
for the five subjects (45 trials x 3 tasks=135 maps
per subject). With the exception of subject JF who
emitted very high-amplitude potentials, only approx-
imately 60% of single-trial waveforms could be de-
tected due to the low signal-to-noise ratio.

   Figs. 3-5 depict  different topographical patterns
of emitted  potentials  observed in these  single-trial
maps.  Positive components  peaked  either  at  the
vertex (Fig. 3.),  in  the  parietal region (Fig. 4),, or
successively  at the  vertex  and  then postcentrally
(mixed  pattern, Fig. 5).   Positive  components were
preceded by a negative  component that  peaked at
the vertex (Fig. 3 lower)  or postcentrally (Fig. 3 up-
per and Fig.  5 upper). Parietal positive peaks were
sometimes preceded by three successive parietal wave i
(negative-positive-negative) of short duration (Fig. 4
upper). Topographical patterns of emitted potentials
varied  in  the same  subject  from  trial  to trial.  The
relationship  of emitted potentials  to  the task could
not be clearly established, although the vertex  neg-
ative- vertex positive  pattern tended to occur within
the initial five or six omission trials of a run.

     Preliminary measurement of amplitude and laten-
 cy parameters in the single-trial maps of two subjects
 (166  maps) indicated that:  (1) the  latency of the
 vertex  positive component  (x =  374 msec, SD.  =
 87 msec) was shorter than the latency of the parietal
 positive component (x = 467 msec, S.D. = 92 msec);
 (2)  the latency  of negative  and positive  emitted
 components  increased  across  runs, regardless  of
 topography; (3)  reaction time also increased across
 runs in the go condition; (4) the first  positive  peak
 always occurred   before  the  motor  act  (x interval
 = 128  msec, SD. = 78 msec) and was highly cor-
 related with RT  (0.77); and (5) the positive compo-
 nent  of the vertex negative-vertex positive pattern
 exhibited higher  amplitude  (X =  17.6 juV, SD. =
 5.4 juV) than the positive parietal component  (X =
 Discussion

     Averaged as well  as  single-trial spatiotemporal
 maps  suggest  that  the  emitted  potential following
 an omitted stimulus is  not a unitary phenomenon,
 but consists of at least two components of different
 polarity, which seem to  reflect different brain activi-
 ties.   Two different patterns  of emitted  potentials
 may  be distinguished in terms of chronotopograph-
 ical organization. One  pattern is characterized by
 a positive vertex component  (sometimes preceded
 by a vertex negative component). The other pattern
 is similar  to  the association cortex potential (ACP)
 described  by Ritter et  al. (1972) and  is made up
 of two parietal components, one  negative and  the
other positive. Both  patterns of emitted potentials
can be observed during the same task and both can
occur together in a single  trial, yielding  a complex,
mixed pattern.

    The averaged negative component of the emitted
potential  to  omitted  visual stimuli was observed in
the parieto-occipital region posterior  to the positive
component. Simson et al.  (1976) showed that the
negative component of the emitted potential changed
topography when omissions occurred within a series
of visual or auditory stimuli,  a finding that suggested
modality  specificity. In the present study, a contra-
lateral negative component was observed  in the aver-
age maps of  three  subjects whenever omissions oc-
curred  within  a  series of  half-field stimulations,
whereas the positive component remained symmetri-
cal.  This  finding  is  consistent with  the modality-
specific character of the negative component reported
by Simson et al. However, more information concern-
ing the topographical organization of this component
in single-trial data  for missing half-field stimuli is
needed before drawing conclusions.

    During  whole-field stimulation,  the  transverse
montage  of  four subjects  showed  an asymmetrical
negative  component larger in the right hemisphere,
whereas  the  positive  component was symmetrical-
ly distributed, suggesting that the negative and posi-
stive components are  independent. This right asym-
metry is difficult to explain because, with the excep-
tion of one subject, the visual evoked potential in full-
field stimulation always appeared symmetrical when
recorded   with the  same montage. Since  the  right
hemisphere is  known to be  active during spatio-visual
processing, this  asymmetry  could be  related to the
spatial visualisation (visual memory)  demands  of a
task that requires counting, motor response, or with-
drawal of motor response  in the  absence of a visual
pattern. Experiments  employing  verbal rather  than
visual stimuli are  needed  for better understanding
of the negative component of the emitted potential.
   Trial-by-trial  analysis did permit  differentiation
(on  the  basis of  topographical criteria)  of several
patterns  of emitted  response during the same task.
These topographical patterns seem to reflect different
psychophysiological mechanisms that may play a role
dependent  on the  task (cf. Renault et al. in press).
Amplitude  of the  negative component was  larger
during both  go and no-go tasks compared to  the
counting task. This  result  may be due to  a  larger
slow  negative wave, reflecting a  more  imp  tant
preparatory process when a motor response is  quir-
ed.

   The pattern consisting of two vertex components
(negative and positive)  may not be  a  true emitted

-------
 208

 potential as described by Weinberg (1970) orSimson
 et al. (197') since it is not preceded by a negative
 modality-specific wave in the parieto-occipital region.
 This pattern resembles the classical vertex potential
 described  long  ago by P. Davis (1939), Bancaud et
 al. (1953), and Gastaut (1953) and may constitute
 an orienting response since it decreases with the pas-
 sage  of time,  an effect which  may reflect  habit-
 uation. The positive  component of this vertex  po-
 tential seems similar  to  the P300 evoked by novel
 stimuli (Courchesne  et  al. 1975).  Consistent with
 Courchesne's  findings, the vertex positive peak in
 this study was of shorter latency and higher  ampli-
 tude than  the parietal positive peak. The vertex posi-
 tive  component  preceded by  a  parieto-occipital
 negativity,  however,  seems to  constitute a proper
 emitted response. This  pattern  does not  seem  to
 habituate  with  time,  is lower in amplitude,  and is
 longer in  latency than  the  pure  vertex pattern,
 suggesting that a more complex mechanism, possibly
 related to time estimation, is involved.
   According  to  Squires  et al. (1975) and to Ford
et al. (1976), the positive component of the emitted
response has a centroparietal distribution when at-
tention is involved, whereas it peaks in centrofrontal
regions when  attention is  not required. Attention
was  not manipulated in  these experiments, so this
hypothesis cannot be directly evaluated. Nevertheless,
the topographical similarity between the pure parietal
pattern and the visual evoked  potential suggests that
the  negative component of  the  emitted potential
may reflect a sensory information retrieval mecha-
nism (Weinberg   et  al.  1970, 1974; John  1972),
                            Renault and Lesevre

thus involving attention processes and acting as  a
feedback  control  on perception. When a  sequence
of  frontocentral  and parietal positive waves  occur
in the same trial, the two components may reflect
two distinct stages in a complex decision  process,
separated  in  time  by  approximately  100  msec.


Summary
   The  chronotopographical  organization  of  the
emitted  potential to expected, but missing,  visual
stimuli was studied  in  five subjects. Three runs of
450 whole-field  visual stimuli  were  presented  in
which  a random 10% of the stimuli were omitted.
Three  task conditions were used: (1) counting omis-
sions, (2) motor  response to omissions (go), and (3)
motor response to visual stimuli but not to omissions
(no-go).   Left half-field and right half-field stimuli
were  also  used  in  a second study. Data were  dis-
played in both averaged and single-trial  spatiotempo-
ral maps.

   Results indicate that the emitted potential consists
of at least two components, a negative  modality-spe-
cific wave  that peaks in the  parieto-occipital region
and a  positive  wave  that peaks either centrally  or
postcentrally.  A complex  pattern  including  two
positive waves peaking in  different regions was also
observed. When  negative  and  positive components
both  peaked at  the  vertex,  the  pattern was  inter-
preted as an orienting response rather  than emitted
potential.  Emitted potentials were discussed  as in-
dices of complex  decision  mechanisms involving time
and sensory information retrieval.

-------
AUDITORY EVOKED POTENTIALS,  SKIN
CONDUCTANCE  RESPONSE, EYE MOVEMENT,
AND REACTION TIME  IN AN  ORIENTING
RESPONSE  PARADIGM '

W. T. ROTH, J. M. FORD, P. L. KRAINZ, AND B. S. KOPELL
Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine,
Stanford, and Veterans Administration Hospital, Palo Alto, CA, U.S.A.
    Unpredictable stimulus events sometimes elicit a
complex of reactions  called the orienting  response
(OR). Ritter et al. (1968) were the first to suggest
that a late positive component (P300 or P3) of the
evoked potential (EP) was also a component of the
OR, and we have pursued this theme in a  series of
papers (Roth 1973, Roth and Kopell 1973, Ford et
al. 1976). Both the amplitude  of P3 and the ampli-
tudes of  other OR components are inverse func-
tions of stimulus probability (Tueting et al. 1970,
Roth  1973, Sokolov 1963). Both P3 and other OR
components can occur in the absence of instructions
that  make the  eliciting  stimuli task-relevant (Roth
1973, N.  K. Squires et al. 1975), but both increase
in mean amplitude when such instructions  are given
(Picton and Hillyard 1974, Sokolov 1963, Bernstein
et al. 1975). Task-relevance is usually assumed to
influence some part of the attention process.

    The primary purpose of this study was  to deter-
mine the  relationship between the skin conductance
response (SCR) and the  auditory  evoked  potential
(AEP) in  an OR paradigm. SCR has often been used
as an indicator of the OR (Raskin 1973). We hypoth-
esized that, since both P3s and SCRs may be compon-
ents of the OR, larger SCRs would be associated with
larger P3s. In addition, we wanted to examine critical-
ly the relationship between eye movement and the
AEP. If eye movements or blinks are sometimes com-
ponents of the OR (as they are of the closely related
startle or  defensive response), one would also expect
more eye  movement to be associated with larger P3s.
Finally,  we  wanted  to  examine the relationship
between reaction time (RT) and AEPs.

    To test these relationships, we sorted individual
trials as to whether they were above or below a criteri-
on level for SCR, eye movement, and RT.  A special
feature  of our sorting procedure was that it would
not accept  consecutive trials  above (high trials) or
below (lew trials) the criterion level. Instead, high
and low trials were selected alternately, a procedure
that yielded high and low AEP averages from compar-
able  points in time. Without this restriction, high
trials for a given variable might be selected from the
beginning of a run and low trials from the end. Roth
(1973)  and Raskin et al. (1969) have previously
shown that P3 amplitude and  SCR amplitude decline
over time if the eliciting stimuli are task-relevant. Our
sorting procedure allowed us to control for the effect
of time and thus to look for relationships between
variables that are independent  of time.

    Task-irrelevant stimuli of  moderate or low inten-
sity elicit SCRs that habituate to an asymptote very
close to zero after 3 to 20 repetitions. In  order to
obtain an adequate number of trials with stimulus-
specific SCRs for EP averaging, we used a 95-dB SPL
musical chord as a stimulus. Stimuli of this intensity
persist in eliciting SCRs at least intermittently (Ras-
kin et al. 1969).

Method

    Eight male and eight female subjects with a mean
age of 25 years participated in  a single 3-hr session.

    The electroencephalogram (EEC) was  recorded
from subdermal pin electrodes at Fz, Cz, and Pz refer-
enced to linked disc electrodes attached to the ear-
lobes. The electrooculogram (EOG)  was  recorded
from disc electrodes applied 3.5 cm above and 2.0 cm
below the pupa of the right eye. EEC and EOG were
amplified and recorded with a system set to a band-
pass of 0.03 to 100 Hz (3dB points of 6 dB/octave
rolloff  curves).   The  amplifiers have an  input
 'This research was supported by NIMH Grant DA 00854 and the Veterans Administration

-------
 210
                                       Roth et al.
  impedance of 100 megohms at these settings.  Skin
  conductance level (SCL) and skin conductance re-
  sponse  (SCR) were  recorded  from a single pair of
  7.5-mm-diameter disc electrodes applied  to  the pal-
  mar surface  of  digits 2 and 3 of the nondominant
  hand.   Skin conductance was transduced by  a de-
  vice that applies a constant 1.0 volt across the elec-
  trodes and  has  as its output a voltage proportional
  to skin conductance.

     Before  stimulation,  each  subject  opened  and
 closed his eyes to allow  quantification of eye artifact
 at each EEC lead.

     Two types of auditory  stimuli were presented
 binaurally through earphones. Each type was control-
 led by electronic switches set to rise and fall times of
 2.5 msec and to  a duration of 50 msec. One stimulus
 was a 95-dB SPL chord consisting  of 550-, 900-, and
 1080-Hz tones. Interstimulus intervals (ISIs) between
 chords were multiples of a 1-sec interval and followed
 a fixed pseudorandom sequence based on a Bernoulli
 distribution. The probability of a chord occurring
 after any 1-sec interval was 0.1. In two of the four
 stimulation  periods, chords were  the only  type of
 stimulus presented (no-background conditions). In
 the other two stimulation periods, a 65-dB noise burst
 was  presented at 1-sec intervals (background condi-
 tions). At the same ISIs as in the no-background con-
 ditions,  the 65-dB noise  burst was replaced by  a
 chord. During both conditions, a  50-dB white noise
 background was used to conceal extraneous sounds.

     Instruction conditions were also varied. In press
 conditions,  subjects  were told to  press a button as
 quickly  as possible whenever they heard a chord. In
 read conditions, subjects were told to read a popular
 mystery novel and to ignore all sounds. Each subject
 received the four possible combinations of stimulus
 and  instruction conditions in four runs counterbal-
 anced in order across  subjects. Each run lasted about
 20 minutes, during which time 65  chords were pre-
 sented. After each run, there  was a brief rest period.

    The value of the SCR for each trial was calculated
 as the sum of absolute deviations between a prestim-
 ulus  baseline and  the skin conductance over an inter-
val from 2.4  to  6.4  sec after  the  chord. Sampling
 occurred every 5  msec. Such a sum is closely propor-
 tional to the integral  of  the absolute deviation  over
 the interval. EOG deviation was calculated similarly
over  the  interval from 2 to 420 msec. Trials with RTs
more than 770 msec were excluded.

    The  criterion  levels were  the mean RT across all
trials of a run for RT sorting, the mean EOG deviation
across all trials of a run for EOG sorting, and 70% of
the mean SCR across all trials of a  run for SCR sort-
ing. Criterion  levels were established separately  for
each run of each subject.  Since the  sorting algorithm
 required high  and low trials to alternate-i.e., after
 selection of a high  or low trial, subsequent high or
 low trials were bypassed until a trial of the opposite
 type was found—the number of averages for a given
 AEP depended on  the time course of the criterion
 variable. Pilot studies indicated that the criterion level
 of mean EOG and  mean  RT provided an adequate
 number of both high and low trials for AEPs. Mean
 SCR, however, did not provide enough high and low
 trials, so that 70% of the  mean SCR was used  as the
 criterion level.  Data from each EEG and EOG channel
 were  sorted into six different categories (high/low
 SCR, high/low EOG, and high/low  RT) and averaged
 over a 420-msec epoch. Whether the initial trial in the
 average  was high or low was balanced across condi-
 tions.

     EP  averages  were corrected for eye movement
 artifact by subtracting a predetermined percentage of
 the EOG average from each EEG average. These per-
 centages varied across load and subject.

    Peak amplitudes and latencies were measured by
 finding the minimum or maximum amplitude in fixed
 latency ranges  and subtracting a prestimulus baseline
 from these points. On the basis of visual inspection of
 individual AEPs, peak latency criteria were defined
 that would be valid for the greatest number of curves.
 These criteria were as follows: Nl, 70-140 msec; P2,
 150-230 msec; N2, 200-300 msec;  and P3, 250-380
 msec. Fixed latency ranges allow replicability between
 laboratories, but are not entirely  satisfactory.  For
 example, one subject's first  prominent negative peak
 occurred at  180  msec under certain conditions. In
 certain  AEPs from  several  subjects, it  was unclear
 whether the prominent positive peak should be con-
 sidered a late P2 or an early P3. N2s were particularly
 inconsistent and  appeared  to depend on the ampli-
 tudes and latencies of the surrounding positive peaks.

    Trials picked by  the three criterion variables were
 largely independent. The  mean number of averages
 for all sorts was 11;  the mean overlap between trials
 picked  for both  SCR and  RT sorts was  1.7 trials,
 between SCR and EOG, it was 2.0, and between EOG
 and RT, it was 2.2.

    A  repeated measures of variance was used for sta-
 tistical evaluation of results. Results reported below
 are significant at p <  .05.

 Results

    Fig.  1  shows grand averages across all subjects of
vertex  AEPs sorted  by SCR and RT. Nl peaks  are
more  prominent  in  the  no-background conditions,
while P3 peaks are more prominent in the background
conditions, especially when a button press was requir-
ed. Fast RTs  are associated with  larger P3s in  the

-------
 210
                                       Roth et al.
  impedance of 100 megohms at these settings.  Skin
  conductance level (SCL) and skin conductance re-
  sponse  (SCR) were  recorded  from a single pair of
  7.5-mm-diameter disc electrodes applied  to  the pal-
  mar surface  of  digits 2 and 3 of the nondominant
  hand.   Skin conductance was transduced by  a de-
  vice that applies a constant 1.0 volt across the elec-
  trodes and  has  as its output a voltage proportional
  to skin conductance.

     Before  stimulation,  each  subject  opened  and
 closed his eyes to allow  quantification of eye artifact
 at each EEC lead.

     Two types of auditory  stimuli were presented
 binaurally through earphones. Each type was control-
 led by electronic switches set to rise and fall times of
 2.5 msec and to  a duration of 50 msec. One stimulus
 was a 95-dB SPL chord consisting  of 550-, 900-, and
 1080-Hz tones. Interstimulus intervals (ISIs) between
 chords were multiples of a 1-sec interval and followed
 a fixed pseudorandom sequence based on a Bernoulli
 distribution. The probability of a chord occurring
 after any 1-sec interval was 0.1. In two of the four
 stimulation  periods, chords were  the only  type of
 stimulus presented (no-background conditions). In
 the other two stimulation periods, a 65-dB noise burst
 was  presented at 1-sec intervals (background condi-
 tions). At the same ISIs as in the no-background con-
 ditions,  the 65-dB noise  burst was replaced by  a
 chord. During both conditions, a  50-dB white noise
 background was used to conceal extraneous sounds.

     Instruction conditions were also varied. In press
 conditions,  subjects  were told to  press a button as
 quickly  as possible whenever they heard a chord. In
 read conditions, subjects were told to read a popular
 mystery novel and to ignore all sounds. Each subject
 received the four possible combinations of stimulus
 and  instruction conditions in four runs counterbal-
 anced in order across  subjects. Each run lasted about
 20 minutes, during which time 65  chords were pre-
 sented. After each run, there  was a brief rest period.

    The value of the SCR for each trial was calculated
 as the sum of absolute deviations between a prestim-
 ulus  baseline and  the skin conductance over an inter-
val from 2.4  to  6.4  sec after  the  chord. Sampling
 occurred every 5  msec. Such a sum is closely propor-
 tional to the integral  of  the absolute deviation  over
 the interval. EOG deviation was calculated similarly
over  the  interval from 2 to 420 msec. Trials with RTs
more than 770 msec were excluded.

    The  criterion  levels were  the mean RT across all
trials of a run for RT sorting, the mean EOG deviation
across all trials of a run for EOG sorting, and 70% of
the mean SCR across all trials of a  run for SCR sort-
ing. Criterion  levels were established separately  for
each run of each subject.  Since the  sorting algorithm
 required high  and low trials to alternate-i.e., after
 selection of a high  or low trial, subsequent high or
 low trials were bypassed until a trial of the opposite
 type was found—the number of averages for a given
 AEP depended on  the time course of the criterion
 variable. Pilot studies indicated that the criterion level
 of mean EOG and  mean  RT provided an adequate
 number of both high and low trials for AEPs. Mean
 SCR, however, did not provide enough high and low
 trials, so that 70% of the  mean SCR was used  as the
 criterion level.  Data from each EEG and EOG channel
 were  sorted into six different categories (high/low
 SCR, high/low EOG, and high/low  RT) and averaged
 over a 420-msec epoch. Whether the initial trial in the
 average  was high or low was balanced across condi-
 tions.

     EP  averages  were corrected for eye movement
 artifact by subtracting a predetermined percentage of
 the EOG average from each EEG average. These per-
 centages varied across load and subject.

    Peak amplitudes and latencies were measured by
 finding the minimum or maximum amplitude in fixed
 latency ranges  and subtracting a prestimulus baseline
 from these points. On the basis of visual inspection of
 individual AEPs, peak latency criteria were defined
 that would be valid for the greatest number of curves.
 These criteria were as follows: Nl, 70-140 msec; P2,
 150-230 msec; N2, 200-300 msec;  and P3, 250-380
 msec. Fixed latency ranges allow replicability between
 laboratories, but are not entirely  satisfactory.  For
 example, one subject's first  prominent negative peak
 occurred at  180  msec under certain conditions. In
 certain  AEPs from  several  subjects, it  was unclear
 whether the prominent positive peak should be con-
 sidered a late P2 or an early P3. N2s were particularly
 inconsistent and  appeared  to depend on the ampli-
 tudes and latencies of the surrounding positive peaks.

    Trials picked by  the three criterion variables were
 largely independent. The  mean number of averages
 for all sorts was 11;  the mean overlap between trials
 picked  for both  SCR and  RT sorts was  1.7 trials,
 between SCR and EOG, it was 2.0, and between EOG
 and RT, it was 2.2.

    A  repeated measures of variance was used for sta-
 tistical evaluation of results. Results reported below
 are significant at p <  .05.

 Results

    Fig.  1  shows grand averages across all subjects of
vertex  AEPs sorted  by SCR and RT. Nl peaks  are
more  prominent  in  the  no-background conditions,
while P3 peaks are more prominent in the background
conditions, especially when a button press was requir-
ed. Fast RTs  are associated with  larger P3s in  the

-------
AEP, SCR, RT and Orienting
                                                211
                           SCR SORT
                   RT SORT
      PRESS
      READ
                                                     B
            0    100  200  300  400
                    msec
                     0    100   200  300  400
                             msec
           Fig. 1. AEPs sorted by skin conductance
           ed over 16 subjects. Only Cz is illustrated.
background  condition.  In  read conditions, AEPs
beginning from about 100-200 msec are more posi-
tive in the low than in the high SCR trials.

    Rg. 2D presents the mean SCR and RT criterion
variables  for trials selected by the sorting procedure
in various conditions. SCRs were smaller in press than
in read conditions for high SCR trials. SCLs (not illus-
trated) preceding  high SCR trials were not different
from SCLs preceding low SCR trials. Table 1 provides
data for comparing means of sorted trials with means
and within-subject standard deviations (SDs) over all
trials.

    Fig. 2A, 2B,  and  2C present mean Nl and P3
amplitudes for each of the sorts. Background, lead,
and attention effects were generally the same regard-
less of the method of sorting.  These effects are sum-
marized below.

    Background. Nl was larger for all sorts in no-
background conditions. Although there was a consis-
tent interaction between  background and lead, this
background effect was significant at each  lead. For
the EOG and RT sorts, Nl was later in no-background
conditions. Neither P2 nor N2  were affected by back-
response (SCR) and reaction time (RT) combin-
B = background; NB = no background.
    ground. P3  was later in background conditions for
    SCR and EOG sorts.

        Lead. Nl, P2, N2, andP3 amplitude consistently
    showed main lead effects. Nl was largest at Cz, P2
    and P3 were largest at Pz, and N2 was largest at Fz.
    Nl,  P2, and P3 latencies also  consistently showed
    main lead effects.  Nl was earliest at Pz,  and P2 and
    P3 were earliest at Fz.

        Attention. Nl amplitude was smaller during read
    than  during press in the SCR sort, and it was earlier
    during read than during press in the SCR and EOG
    sorts.  P2 amplitude showed no  attention effect, but
    P2 latency was consistently earlier during read than
    during press.  N2 amplitude was also unaffected by
    attention, but N2 latency was  earlier during press
    than during read. P3 amplitude was consistently larger
    during press than  during  read.  An attention x lead
    interaction appeared to be due to the attention effect,
    being  strongest at  Pz. P3 latency was unaffected by
    attention.

        High/low SCR. The only effect of this factor was
    on P3 amplitude, which showed a significant atten-
    'tion x high/low interaction. Low  SCRs were associated

-------
212
                                 Roth et al.
                          SCR SORT
                                                             RT SORT
                                              P3
                                              N1
 K25-





 + 15-


. -10-


 -15-


 •20-
                                                                 •-• LOW PRESS
                                                                 o-o HIGH PRESS
                                                                 *-• LOW READ
                                                                    HIGH READ
                           EOG SORT
            B         NB

        CRITERION VARIABLE
             P3+15-:
                + 10-
                -10-
             N1 -15-


                -20-j
  10-

  6-|

  2™





250-
                                                       SCR
                                                       RT
                          B
                                     NB
            B
                      NB
             fig. 2. A-C; Mean P3 and Nl amplitude across all leads and subjects for each
             condition sorted by skin conductance response (SCR), electrooculogram (EOG)
             and reaction time (RT). D: Mean values for SCR and RT criterion variables when
             sorted by SCR and RT, respectively. In A-D, low - low SCR, low EOG voltage,
             or fast RT.
     Table 1. Mean Over All Trials, Mean Within-Subject Standard Deviation (SD)
      Over All Trials, Mean of High Trials, and Mean of Low Trials for the Three
                                   Criterion Variables

SCR,
jumhos/cm^ x 10"*
EOG,
arbitrary units
RT,
msec
Mean
Mean SD
Mean high
Mean low
Mean
Mean SD
Mean high
Mean low
Mean
Mean SD
Mean high
Mean low
Backc
Press
6.269
6.832
9.553
1.646
90
63
214
34
287
72
342
248
round
No press
7.393
7.932
11.711
1.762
36
19
127
17

No background
Press
6.911
6.705
8.785
1.874
93
59
224
31
261
59
354
225
No Press
8.050
7.880
12.525
2.250
38
17
120
16


-------
AEP, SCR, RT and Orienting
                                            213
with large P3s only during read conditions. This effect
is small, with mean differences about 2 pV.

    High/low EOG. No amplitude or latency effects
were present.

    High/low RT. For both P2 and P3 amplitudes,
large peaks were associated with fast RTs, but only in
the background condition.  N2 latency was  earlier
when RTs were fast.

Discussion

    Contrary to the initial hypotheses, no evidence
of a positive association between AEP peaks and SCR
or EOG was found. The AEP was affected, however,
by background and attention parameters. The effects
of these parameters on Nl, P2, and P3 were similar to
effects  observed  previously (Hillyard  et al.  1973,
Picton  and Hillyard 1974, Roth et al. 1976, Ford et
al.  1976b). There was  some evidence that either P2
and P3 were not completely separated by the latency
criteria  or  a single positive process was present at
both latency ranges. Both P2 and P3 were largest at
Pz, whereas P2 would be expected to be largest at Cz
(cf. Roth etal. 1976).

    The effects of RT sorting on P3 amplitude in the
background condition may also be an attention  effect,
since fluctuation  in RT may occur with fluctuation
in attention. Similar findings were reported by Karlin
et al. (1971). It is unlikely that the 7 /iV P3 difference
between fast and slow  RT averages is explainable on
the basis of EPs to the motor response itself. The  rele-
vant components of the finger movement potential
have a mean peak-to-peak amplitude of only  3  juV,
even when averaging is response-synchronized (Deecke
et al. 1976). A readiness potential would not be pre-
sent to a significant degree  since the subject  is per-
forming  an unwarned  RT task. If  the  effect  of RT
sorting is based on attention, the absence of this effect
in the no-background condition is puzzling.  It is also
possible that P3 latency jitter plays a role in the RT
sorting effect.

    There are several possibilities why our hypotheses
were not confirmed. Because of the intensity  of the
stimuli, a defense response may have  been elicited
instead  of  an OR. The OR and defense response are
both  accompanied by  SCRs  and  eye  movements;
however, the OR is associated  with increased sensi-
tivity to stimuli  and the defense  response may be
associated with decreased sensitivity (Sokolov 1963;
Loveless, this  volume). Defense responses, however,
are supposed to be resistant to habitation, whereas
in  this study  SCRs or  eye movements tended to
decrease after  the first  few occurrences of the chord.
This leads to a second  possibility, namely, that inter-
mittant SCRs  and eye movements were habituated
ORs that have different correlates  than  the  initial
ORs.  Evidence from other experiments indicates that
this can  be the case for the SCR (Siddle  1974).
In addition, some SCRs or eye movements might have
occurred  spontaneously, following the chords only by
coincidence. A third  possibility is  that eye  move-
ments, SCRs, and P3s are indeed components of the
OR, but that they have only low positive correlations,
which disappear when the effect of time is removed.

    The  grand averages in  Fig.  1 suggest that the in-
verse association between P3 and SCR amplitudes in
the read conditions may be the result of a slow poten-
tial shift. Several  such  shifts  have been described
(Loveless and Sanford 1974, Naatanen  1975,  N. K.
Squires et  al.  1975, Roth et  al.  1976). One could
hypothesize that there is a negative slow potential OR
component, although the lack of OR-related changes
in Nl, P2, or  N2 amplitude  or distribution or in P3
distribution makes that possibility tenuous.  It is un-
clear why SCR sorting had an effect only in the read
conditions, unless the motor response somehow alter-
ed  the properties of P3 or the SCR. Although  one
would  expect  motor  response to increase SCRs in
general, high SCR trials  had larger SCRs in the read
rather than press condition.

    The  negative  result  for EOG sorting serves  as a
rationale   for continuing  to  eliminate eyeblink-con-
taminated trials in order to control for the spread of
potentials generated by the eyes. There was no associ-
ation  between potentials in  the EOG and the EEG
besides this artifactual one, which was minimized by
our correction procedure. Of course, this rationale is
only valid  when experimental  parameters similar to
ours are used.

    This  experiment is not the final word on the rela-
tionship between P3 and non-EEG orienting response
components. Our suggestions as to why our hypothesis
was not confirmed should be taken as suggestions for
new experiments. We think that an artificial gap has
arisen  between the older autonomic psychophysiol-
ogy and  the  newer EP  psychophysiology, and  that
the simultaneous  use  of autonomic and  EEG vari-
ables can lead to  a fruitful integration of  the  two
areas in the future.

Summary

    Auditory evoked potentials (AEPs) to a pseudo-
random series  of 50 msec, 95-dB chords were studied
in 16 subjects. The electro-oculogram (EOG) and  skin
conductance  response (SCR)  were  also  recorded.
There  were four  runs with either chords alone (no
background) or chords inserted in a train of 50-msec,
65-dB noise bursts (background). Subjects were re-
quired either to read or to press a reaction time (RT)
button to  the chords. In  all conditions,  the chords

-------
214

had a 0.1 probability of occurrence in any 1-sec inter-
val.

    AEPs were obtained by sorting individual trials
by high/low SCR, high/low EOG, and high/low RT.
The sorting was based on the mean of the sorting vari-
able  for individual runs, and required  high and low
trials to alternate.
                                    Roth et al.

due to changes in amplitude of the classical P3 wave,
but is a result of changes in positive  or negative slow
potential processes.  Fast  RTs were associated with
large P2s and P3s when background tones were present.
Many previously observed findings concerning Nl, P2,
and P3 were replicated.

Acknowledgments
    Large SCRs were associated with small P3s dur-
ing the read conditions. This effect is apparently not
    We  thank  S. J. Lewis, M. J. Rosenbloom, and
M. T, Cnargin who assisted in this study.

-------
 SEQUENTIAL DEPENDENCIES OF THE
 WAVEFORM  OF THE  EVENT-RELATED
POTENTIAL:  A  PRELIMINARY REPORT1
K. C. SQUIRES, D. WICKENS, N. K. SQUIRES, AND E. DONCHIN

Department of Psychology, University of Illinois, Champaign, IL, U.S.A.
    One basic assumption underlying research involv-
ing the average event-related potential (ERP) is that
when stimulus and task variables are  held constant,
ERPs elicited by all occurrences of a particular event
are identical. As the number of variables shown  to
influence the ERP continues to increase, investigators
have resorted  to increasingly sophisticated methods
for creating averaged ERPs from presumably  homo.-
genous populations of single-trial waveforms.  One
popular technique has been to randomize the order of
stimulus presentation to eliminate differential prepar-
edness (Naatanen 1970) and to segregate trials for
averaging according to which stimulus was presented
on each  trial. Studies using this paradigm have contri-
buted significantly to our understanding of the effects
of attentional variables, task requirements, stimulus
relevance, and stimulus probability (e.g., Donchin and
Cohen 1967, Ford et al. 1973, Harter  and Salmon
1972, Picton and Hillyard 1974, Ritter and Vaughan
1969, Rohrbaugh et al.  1974, Ruchkin and Sutton
1973, N. Squires et  al.  1975, Sutton et al. 1965,
Tueting et al. 1970, and Wilkinson and Lee 1972).

   A recent study by Squires and Donchin (1976),
however, suggests the need to reassess this assumption.
The standard method of segregating trials for averaging
was  supplemented by categorization  of ERP wave-
forms according to waveform statistics generated  on
each  trial. Wide variations in averaged ERP  were
found for events often considered identical, and close
inspection of waveform statistics suggested that varia-
tions were  at  least  partly due to effects of the
sequence of stimulus presentations.  This explanation
seemed plausible in view of reports  by Remington
(1969) and others (Falmagne  et al.  1975) showing
that choice reaction time, which has often been linked
to the P300, is exquisitely sensitive  to the specific
sequence of preceding trials. Since  the  existence of
such sequential  dependencies would violate the fun-
damental assumption that ERPs under the conditions
described can be considered identical, an investigation
was designed to parallel the reaction time study of
Remington (1969).

Method

    Seven  subjects listened to series of tone bursts
(60 msec, 60 dB SPL) presented at a rate of one every
1300 msec. On each presentation the tone was either
high  pitched (1500 Hz)  or  low pitched (1000 Hz)
with equal  probability (P = 0.5). Subjects counted the
occurrences of the high-pitch tone. In a second experi-
mental condition, everything  remained the same
except that stimulus probabilities were changed to
0.3 (high pitch) and 0.7 (Low pitch).

   EEC was recorded  from electrodes at Fz, Cz, and
Pz referred to linked mastoids with a wrist ground.
The bandpass of the amplifying system  was set for a
time constant of 0.8 sec with an upper half-amplitude
frequency  of 35 Hz. A768-msec epoch of EEC, begin-
ning 100 msec prior to stimulus onset,  was digitized
on each trial and stored for later analysis. Additional
electrodes were  situated to record eye-movement and
blink potentials. EEC epochs with eye-related artifacts
were excluded  from the waveform analysis, but all
trials entered into the tabulation of stimulus sequence.

Results and discussion

   One subject's  ERPs (Cz) to the counted high-
pitch tones (P = 0.5) are shown in Fig. 1. The termin-
ology of Remington (1969) has been used.  An "A"
represents  the stimulus trial n  (here the high-pitch
tone).  Thus the  ERP for the  first-order sequence
(A) represents the average from all presentations of
the high-pitch tone. There were two possible patterns
for second-order sequences, "AA" and "B." The ERP
for" AA" was formed from trials on which a high-pitch
tone was preceded by a high-pitch tone, and the "BA"
      research was supported by the Advanced Research Projects Agency of the Department of Defense under
  Contract No. DAHC 15 73 C 0318 and Contract No. ONR US NAVY N-000-14-76-C-0002, both to E. Donchin.

-------
 216
                                  Squires et al.
                                                                         20JUV
Fig. 1. Average ERP waveforms of one subject for different stimulus sequences. "A" and "B" represent occur-
rences of the high- and low-pitched tones, respectively. ERPs in each case are elicited by the final high-pitched
tone in the designated sequence.
ERP from high-tone trials preceded by low-tone trials.
Similarly, there were  four third-order patterns,  8
fourth-order patterns, and 16 fifth-order patterns.

    In Fig. 1, waveforms are presented at nodes of a
tree structure. A similar tree was shown by Remington
to be useful in organizing reaction time data. Reming-
ton derived the tree from a repetition hypothesis,
according to which the addition of a like stimulus (A)
in front of a sequence leads to a shorter reaction time
to stimulus A on trial n than the addition of an unlike
stimulus (B).

    There is a systematic variation in the ERP wave-
form elicited by high-pitch  tones as a function of the
sequence of preceding  stimuli. Examination of the
upper  and lower  limbs  of the tree structure shows
that the  amplitude of the P300 component (along
with the associated N200 and slow wave components)
(K. Squires et al., (1977); N. Squires et al.  1975) in-
creases with increasing numbers of unlike stimuli pre-
ceding the eliciting stimulus and  decreases with  in-
creasingly long  runs of like stimuli.  No systematic
shifts in latencies of waveform peaks were  noted.
Remington's  reaction time data,  however, suggest
that latency  shifts are likely to be rather small (a
maximum of about 20 msec between the  "A" and
"BBBBA"  sequences); peak-latency  shifts of this
magnitude  are  beyond the resolution  of  the  tech-
niques used in this study.

    Within each order (i.e., within sequences of equal
numbers of stimuli), there were systematic variations
in waveform  that were independent of the absolute
number of like and unlike stimuli within  the short
sequence   (cf.   waveforms   for   "ABBBA"  and
"BBBAA"),  but  were  critically  dependent  upon
where in the sequence the As and Bs appeared and

-------
 Sequential Dependencies of ERPs
                                             217
  OS
  O.4
  0.3-
                                                                  600 m»c
    I
                                                  TRIALS
Fig. 2. Expectancy for the stimulus on each trial in a sequence of trials along with the average ERP waveforms for
the corresponding sequence (same subject as Fig. 1).  High-pitched and low-pitched tones represented by short-
and long-stimulus markers, respectively.
whether or not patterns (e.g., alterations) emerged. A
detailed analysis of waveform variations is presented
elsewhere (K. Squires et al. 1976).

    To summarize, a model for predicting waveform
statistics  based upon the concept of expectancy has
been developed. The model assumes that the expect-
ancy for each stimulus is determined by three factors:
(1) the memory for event frequency within the prior
stimulus  sequence (an  exponentially  decaying func-
tion of position within the sequence), (2) the specific
structure of the sequence (whether patterns of alter-
nations are set up), and (3) the global probability of
the stimulus. According to the model, the  amplitude
of the complex of late components (N200, P300, and
slow wave) increases as the expectancy  for the pre-
sented stimulus decreases.  The  regression equation
based upon the mean data for seven subjects at stim-
ulus probabilities of 0.3, 0.5, and 0.7  accounted for
78% of the variance in the statistical data.

    Fig. 2 illustrates the dynamic nature of changes in
expectancy from trial-to-trial and the corresponding
effect on the ERP waveform. These results show that
the ERP waveform is remarkably sensitive  to trial-to-
 trial variations in the  sequence  of events preceding
 the eliciting event, and provide new evidence for the
'association  of late  ERP  components  with subtle
 cognitive processes.  These results also  suggest that
 before we can assume  homogeneity of events associ-
 ated with the average ERP, more sophisticated group-
 ings of  single-trial waveforms than have  previously
 been utilized must be developed.
 Summary
     Subjects counted one  of two auditory stimuli
 that occurred in a Bernoulli sequence. Event-related
 potentials (ERPs) were averaged for each stimulus
 according to the sequence of preceding stimuli. Late
 ERP components decreased in amplitude with increas-
 ing numbers of like stimuli and increased in amplitude
 with increasing numbers of unlike stimuli within the
 preceding sequence. This effect was found to extend
 as far back  as trial N-5. It was concluded that these
 variations in amplitude of late components were relat-
 ed to variations in the subject's expectancy for a stim-
 ulus.

-------
 FUNCTIONAL EQUIVALENCE OF  SIGNAL-
 PRESENT,  SIGNAL-ABSENT, AND  THRESHOLD-
 DETECT PSs1
 N. K. SQUIRES AND K. C. SQUIRES
 University of Illinois, Champaign,  IL, U.S.A.
 S. A. HILLYARD

 University of California, San Diego,  CA, U.S.A.

    The omission of a highly expected stimulus has
 been shown to elicit a positive wave recordable from
 the human scalp, with a latency of over 300 msec.
 This late-positive component can be obtained by omis-
 sion of stimuli in several modalities, including auditory
 (Sutton et al. 1967; Picton et al. 1974; Ruchkin et al.
 1975; Ruchkin and Sutton,  in press, a,b),  visual
 (Barlow 1969;  Ruchkin  and  Sutton 1973, in press),
 and tactile (Klinke  et al. 1968). These results have
 frequently  been interpreted in terms of the delivery
 of information; the omission of a stimulus elicits a
 late-positive component  when that omission carries
 task-relevant  information,  for  example, when  the
 omission is a target to be counted (Picton et  al.
 1974, Barlow 1969) or confirms or disconfirms a
 prior guess (Sutton et al. 1967; Ruchkin  and Sutton
 1973, in press a,b). Within this framework, the late-
 positive component to  omitted stimuli  may  be a
manifestation of the same process that underlies the
 P3  component  that follows the reception of target
 stimuli in  a  number  of paradigms (cf.  K.  Squires
et al.  1973a  and Hillyard and Picton, in press, for
 reviews). Ruchkin et al. (1975) have provided support
 for this conjecture by demonstrating that the prob-
ability of event occurrence has similar effects on the
 amplitudes of the late-positive components associated
with stimulus presence and stimulus absence. More-
over, Ritter et al. (in press) have found that the scalp
 topography of the P3  elicited by omissions from an
 ongoing  train (either  auditory or visual)  is indis-
 tinguishable from   that  elicited  by intramodality
 targets embedded in the same train.

    Under slightly different circumstances, i.e., when
 subjects are required to judge  the presence or absence
 of a near-threshold tone pip, there is considerable evi-
 dence showing that signal-absence  does  not readily
                                                 elicit a late-positive component (Hillyard et al. 1971,
                                                 Paul and Sutton 1972, K. Squires et al. 1973b, K.'
                                                 Squires et al. 1975a), despite the fact that stimulus
                                                 absence certainly carries task-relevant information in
                                                 this paradigm. The following experiments were design-
                                                 ed to reconcile these results by systematically varying
                                                 three important variables  that differentiate the pre-
                                                 viously described  paradigms  and that alone, or in
                                                 some combination, might  account for the discrepan-
                                                 cies noted: the probability of signal presentation, the
                                                 intensity of the signal, and the manner of signal pre-
                                                 sentation, whether on a trial-by-trial  basis or in an
                                                 ongoing train. Measurement of the scalp  distribution
                                                 of the  late components was also included  in this
                                                 study to provide a further basis for comparison.

                                                 Methods

                                                     Five adults with normal hearing, including authors
                                                 KS and NS, served as subjects in a series of 15 to 20
                                                 two-hour sessions.

                                                     Evoked potentials to all events were recorded
                                                 from frontal, vertex, and parietal  midline sites (25%,
                                                 50%, and 75% of the inion-nasion distance,  respec-
                                                 tively), using Ag/AgCl disc electrodes referred to right
                                                 mastoid. Vertical EOG was recorded and averaged in
                                                 all subjects to rule  out contamination  from ocular
                                                 artifacts.  The half-amplitude frequency bandpass of
                                                 amplifiers was set at 0.15 to 500 Hz.

                                                     During the experiment, the subject sat in a reclin-
                                                 ing chair in an acoustically shielded chamber with a
                                                 panel of lights and response buttons in front  of him
                                                 or her. In the first experiment, the subjects reported
                                                 on a trial-by-trial basis whether or not a signal had
                                                 occurred. Each trial began with a  warning tone (SI),

.._ research was supported by NASA Grant NCR 05-009-198 to Robert Galambos and NIH Grants NH 25594
to S. Hillyard and NS 07454 to Donald Norman, who provided the experimental facilities.

-------
Functional Equivalence of P3s
                                           219
which was a binaural 1-kHz, 50-msec tone burst, at an
intensity of 65 dB SPL, presented against  a continu-
ous background of white noise (65 dB SPL). After an
interval of 600 msec, from  the onset of the warning
tone, a second 50-msec, 1 kHz tone (S2)occurred with
a certain  probability (0.9, 0.5,  or 0.1  on different
blocks of trials). The intensity of S2 was also  varied
across blocks-90 dB (SPL), 65 dB (SPL), or threshold
level. Threshold intensity was determined individually
for each subject by the method of constant stimuli so
that  each  subject's detection performance  was 90 to
95% correct.  This intensity was 44 dB SPL for most
subjects. Thus, there were  nine  stimulus  configura-
tions, three S2 intensities at three a priori probabili-
ties.  Following each trial, a small neon bulb was light-
ed  1600 msec after the onset  of SI to  request a
response of the subject.  Two response buttons were
available, one to indicate that S2 had been presented
and  the other  that  S2 had not been presented. The
time from the response-light presentation to the begin-
ning of the  next trial varied randomly from 1500
msec to 3500 msec.  There were  100 trials  per block,
all undei  one condition of stimulus intensity  and
probability. Experimental conditions were presented
first in order of decreasing  probability of  a signal
occurrence,  with  decreasing intensity  within each
probability, and then the entire sequence  was repeated
in the reverse order. Completion of an entire sequence
required 3  to 4 hours.

     Evoked response recording began 100 msec  before
the warning tone and continued until 750 msec after
the signal, for a total of 1450 msec. Separate averages
were made for  signal-present  and signal-absent trials.
Randomization  of trials was under the control of a
PDP-9 computer, as were evoked-response averaging,
signal presentation, and collection of the behavioral
response.

Results

    A  complete  set of  averaged evoked responses
(from the vertex) for subject HC is shown in Fig. 1.
Evoked responses from the signal-present trials (solid
lines) and from the signal-absent trials (dotted lines)
are superimposed. Decreasing S2 probability is from
left to right, and decreasing S2 intensity is from top
to bottom.  The late-positive  component evoked by
signal omission can be  seen at all intensity levels for
the S2 probability of 0.9. At each signal intensity the
amplitude  of this component decreased  as  the S2
probability  decreased.  Simultaneously,  the  P3  to
signal  presentation increased in  amplitude  as that
signal was made less probable.

      The averaged amplitude (re a baseline over the
100-msec interval prior to S2 occurrence  or omission)
and latency data from all subjects are shown in Fig. 2.
Although the amplitude of the signal-absent late-posi-
tive component is less  than that  of the signal-present
P3 (Fig.  2A; F(l, 4) =  9.41, p < .05), and latency is
longer (Fig. 2C; F(l, 4) =  45.46, p < .01), the scalp
distributions of the two components are similar (Fig.
2A), both being maximal in amplitude at the parietal
                   P(S2) » 0.*
           90 dB
          6BrfB
       TRESHOLD.A
                    1460 IDMC
                                                         S2PRESENTED
                                                         S2 OMITTED
                        5JUV
c
           Fig. 1, Evoked potentials to stimulus presentation and stimulus omission for subject HC
           at each of three signal intensities and three signal probabilities.

-------
220
                                                                                     N. Squires et al.
           P  c.   P
          BLICTMOM
    .400
   i
   2 tte
   c
»0.t
              0.*  0.1  0.1
              PROBABILITY
STIMULUS MEBENTED

    •  00 dB

    •  68 dB

    «  THHEBHOUO

STIMULUS OMITTED

    o  ao da

    o  aa da

    *  THRESHOLD
     —   BO  aa   T
            INTENSITY

   Fig. 2. A. P3 amplitude as a function of electrode
   location for signal-present and stgnafabnnt tvok-
   td potentials at each signal intensity. B. P3 ampli-
   tude  (parietal electrode) for signal-present and
   tignal-abitnt at each level of probability of the
   eliciting event.  C. P3 peak latency as a function
   of signal intensity.
 electrode (F(2,8) - 6.42, p < .05). The electrode by
 condition (present-absent) Interaction wai nonslgnifl-
 cant. Furthermore, both component!  increased In
 amplitude as  the probability of the eliciting event
 decreased (Fig. 2B; F(2, 8) - 8.78, p < .01) and both
 increased in latency with decreasing signal intensity
 (Flg.2C;F(2,8)-10.71,p<.01).
      In  a second experiment  in which stimuli were
  occasionally  omitted from  an ongoing train,  a late-
  positive component wai also evoked by the stimulus
  omission, with a parietal-maximum scalp distribution.
  •The vertex waveforms  for one subject are shown in
  Fig 3.  As In the previous experiment, the latency of
  this component Increased as  the signal intensity de-
  creased. Furthermore, the scalp topography of P3 to
  the omitted stimulus was virtually identical to that
  obtained under similar conditions in the first  experi-
  ment;  the mean amplitudes  at the parietal,  vertex,
  and frontal  electrodes were  10.2, 7.5 and 3.9 ^V,
  respectively, for the signal  level  of 90 dB (compare
  with Fig. 2A, open circles).
Discussion

    These data demonstrate a close resemblance be-
tween the signal-present and signal-absent late-positive
components:  scalp distributions are the same, ampli-
tudes of both increase with decreasing probability of
the eliciting event, and latencies of both increase with
decreasing signal intensity. This latter finding suggests
that the timing of the decision as to whether or not a
signal has been presented is determined by the loud-
ness of that signal.

     While the functional similarity of the signal-absent
and signal-present P3s is apparent, the signal-present
P3 is consistently earlier and of greater amplitude.
This asymmetry between  the  effectiveness  of signal-
presence and signal-absence in eliciting P3s has pre-
viously been  noted by investigators using signal detec-
tion paradigms (Hillyard et al. 1971; K. Squires et al.
 1973b, 1975a) and guessing paradigms (Sutton et al.
 1967; Ruchkin and Sutton, in press a). Ruchkln and
Sutton (In press a) have suggested that stimulus omis-
sion results In both poorer time-locking of the P3 and
in greater equivocation Cess a posteriori information),
and that both of these factors reduce the amplitude
of the stimulus-absent P3.  In support of this analysis,
                                                    STIMULUS! PRESENTED
                                                    STIMULUS OMITTED
                                                                         SO dB
                                                                          68  dB
                                                                          44  dB
                                                     7BO meac

                                     Fig  3  Evoked potentials to stimulus  presentation
                                     and stimulus omission for subject KS in the counting
                                     procedure where stimulus omission was the infrequent
                                     event (P - 0.10) and the target to be counted.

-------
Functional Equivalence of P3s
                                            221
Ruchkin and Sutton showed that correcting for P3
latency variation  on a  trial-by-trial basis reduces the
amplitude  differential  between stimulus-absent and
stimulus-present P3s, but does not eliminate  it, sug-
gesting that the individual stimulus-absent P3s are of
lower amplitude as well  as more  variable in time of
occurrence. These  conclusions are  also consistent
with the recent findings of K. Squires et al. (1975b)
who found that by obtaining better time-locking of
the signal-absent P3 with the help of a visual cue dur-
ing the observation interval, much, but not all, of the
asymmetry  in the effects of the two types of events
is eliminated.

    Finally, the similarity of the P3 latency changes
and the P3 scalp distributions obtained in the trial-by-
trial (signal detection) procedure  and the  counting
procedure  suggests that the late-positive components
obtained under the two sets of circumstances reflect
the same brain  process.  A similar correspondence
between  the distributions of the threshold-detection
P3 and the omitted-stimulus P3 has been found by
Hillyard et al. (in press).
Summary

    Comparisons were made among the late-positive
waves (P3s) to presented and omitted auditory stimuli
in signal detection and counting procedures. Three
probabilities  of signal presentation were  used (0.9,
0.5, and 0.1) at three intensity levels (90 dB, 65 dB,
and threshold). The similarities of amplitude, latency,
and scalp-distribution variations with intensity  and
probability support the functional equivalence of the
stimulus-present and stimulus-absent  P3s,  as well as
the equivalence of P3s found in signal-detection  and
counting paradigms.

-------
 A CNV REBOUND  FUNCTION: PRELIMINARY
 REPORT1


 J. J. TECCE2
 Laboratory  of  Neuropsychology,  Boston  State  Hospital, Department of  Psychiatry, Tufts
 University School of Medicine, Boston, MA, U.S.A.
     Contingent negative variation (CNV) is an event-
 related brain potential that shows optimal  develop-
 ment  in a constant-foreperiod simple reaction-time
 task where a first stimulus (SI) serves as a ready sig-
 nal  for  a  second stimulus (S2)  to which  a  motor
 response (MR) is required (Walter et al. 1964).  In this
 situation, there is a unitary attention set to prepare
 for response  to S2. When attentiveness to S2 is frac-
 tionated by requiring that subjects time-share between
 the  reaction  time task and another overlapping task,
 CNV development is  disrupted. For example, CNV
 amplitude is reduced when a short-term memory task
 for letters is interpolated between SI and S2 on each
 trial (Tecce et al.  1976,  1978). The divided attention
 set produced in these studies (press quickly to S2 and
 remember  the letters) produced  reduction  in  CNV
 amplitude and slowing of reaction time. The associa-
 tion of CNV reduction and slow reaction time was in-
 terpreted as a CNV distraction effect. The question
 remains as to whether this effect can be demonstrated
 by intermixing  trials having the  short-term  memory
 task for letters  and trials  not having this memory
 load.  The present study was designed to assess this
 possibility.

 Method

    Thirty-six males served as_paid volunteers.  Their
 ages ranged from  18 to 26 (X = 21.06; SD = 2.66).
 All subjects except two  were right-handed. Subjects
 were screened medically and psychiatrically.

    The  basic procedure  was  a constant-foreperiod
 simple reaction-time paradigm.  The preparatory stim-
 ulus (SI) consisted of  a brief (0.15 sec) flash of a
 black "X" (2 cm  in height) appearing on a circular
 patch of dim  light (2.5 cm dia). The projector, which
 stood 44 cm from the  floor, was located  1 meter
 from the subject's eyes at an  approximate angle  of
 25° from the  horizontal.  S2 was a 1000-Hz tone of
 approximately 70 dB sound pressure level presented
 through earphones 1.5 sec after the "X" and terminat-
 ed by a telegraph key press. The preferred hand was
 used for keypressing.

    There were two experimental conditions:  "con-
 trol" and "50% letters." A control run consisted  of
 31  trials of Sl-S2-motor response,  lasting approxi-
 mately  7  min. Intertrial intervals  varied randomly
 from 8 to 14 sec (X =  11) within a rectangular distri-
 bution of values 1  sec apart. The 50% letters condi-
 tion was a  modification of the Tecce  and  Scheff
 (1969) short-term memory paradigm. This condition
 involved 31 trials randomly presented in two ways-
 with or without letters. The 16 no-letters trials were
 identical  to  those in the control run and involved
 only the "X" flash, tone,and key press.  The 15 letters
 trials were similar to no-letters trials except that four
 letters (A, E, I, and 0) were spoken within the X-tone
 interval through an intercom and subjects were requir-
 ed to repeat the letters upon hearing "OK" spoken by
 the experimenter  1 to  4 sec after the subject's key
 press. The letters were given in a different randomized
 sequence on each trial.  The first (no-letters) trial was
 omitted from data analysis. The sequence of test runs
 was (1) control condition, (2) 50% letters condition,
 and (3) control condition. This  ABA design is suitable
 where (A) a baseline condition (first phase) is estab-
 lished; (B) a treatment is given; and (A) a recovery
 test is made (Johnson and Lubin 1972). There  was a
 6-min rest between test runs.

    EEC was recorded from Fz, Cz, and Pz with link-
ed earlobes as reference. EOG was recorded from 3 cm
above and 2 cm below the right eye as measured from
center to center of pupil and electrode. Electrodes
'Supported by USPH Research Grants AG-00515 from the National Institute on Aging and MH-19211 from the
 National Institute of Mental Health.

2Research Scientist Development Awardee 1-K02-00016 from the National Institute of Mental Health.

-------
CNV Rebound
                                           223
were Beckman Ag/AgCl standard (EEC) and miniature
(reference and  EOG) types. The time constant for
EEC and EOG was 8 sec. High-frequency cutoff was
75  Hz (50% amplitude reduction),  with 12 dB per
octave roll-off. CNV and EOG data were recorded on
magnetic tape. Trials with eye movements, premature
key presses within the S1-S2 interval, and extraneous
baseline shifts such as those produced by body move-
ments, were omitted  in off-line averaging with  a
CAT  1000.  Average  CNVs were  based  on  6 to 12
trials per run. »h« number being constant for a given
individual. CNV amplitude  was measured as the dif-
ference in average voltage (sampled every 16 msec)
between   the 256   msec  epoch  pre-S2  and  the
512 msec epoch pre-Sl; this  difference was refer-
enced to an on-line 25-juV calibration pulse.

    EKG for cardiotachometric analysis was  recorded
from sternum to lower left chest. Overall heart  rate
(beats per minute) for each condition was determined
by  obtaining the mean of a random sample of 20% of
individual momentary heart rate in a 7-min run (20%
and 100% samples yield comparable results). Eye-
blinks were  defined as an EOG excursion of at least
50  juV and of less than 900 msec in duration (usual
duration: 150 to 300 msec). Eyeblink rate (blinks per
minute) was based on the number of blinks occurring
during an entire 7-min run. For the letters-recall trials,
accuracy of recall (percent correct trials) was deter-
mined by dividing the number of trials in which recall
of the  four letters was correct by 15 (total number of
trials scored). A trial was correct if letters were repeat-
ed in the same sequence given.

    For comparisons  among control, letters, and no-
letters trials, mean differences were evaluated by corre-
lated t-tests with 35 df (n = 36). Reported differences
are significant at the .05 level or less.

Results

    Preliminary evaluation  of CNV amplitude, reac-
tion time, heart rate, andeyeblink frequency for base-
line and recovery control runs indicated no significant
differences on each response measure. Consequently,
a pooled  control  value (mean of the two  control
values) was used for statistical analysis.

    Table  1 indicates means and  standard deviations
of CNV amplitude for pooled controls, trials with let-
ters, and trials with no letters.  For simplicity of pre-
sentation, CNV amplitude values appear in the table
as algebraically positive. Differences in CNV amplitude
(letters trials minus  pooled controls and  no-letters
trials minus pooled controls) are shown in  Rg. 1  for
the three recording sites. Rg. 1 and Table 1 indicate
that CNV amplitude is reduced in letters trials for the
Pz  recording site; in  no-letters  trials, CNV amplitude
is elevated  at CZ and Pz.  Table 1   also indicates
that CNV  amplitude is lower  for letters trials than
             Table 1. Means (and Standard Deviations) of CNV Amplitude and
                           Reaction Time for Three Types of Trials
Trial
type
Letters-
recall
No
letters
Pooled
control
CNV amplitude, MV
Fz
6.14
(5.81)
7.28
(6.15)
5.84 c
(3.67)
Cz
11.81'
(7.07)
14.62 *'b
(7-29)
12.19°
(5.80)
Pz
5.58 *'b
(5.63)
10.38§'b
(5.04)
8.31°
(4.70)
Reaction
time,
msec
276.03 b
(79.85)
260.33 "
(63.85)
232.86
(37.76)
           • Letters-recall and no letters significantly different from each other.
           b Significantly different from pooled controls •
           c Significant differences: Cz >  Pz> Fz-

-------
 224
                                         Tecce
                                 LETTERS
                                 NO-LETTERS
             Fz          Cz         Pz
                RECORDING SITES

Fig.  1. CNV amplitude difference scores (letters trials
minus control trials and no-letters trials minus control
trials) for Fz, Cz, and Pz recording sites in a group of
young normal  volunteers (n =  36). The downward
direction of solid  bars for Cz and Pz indicates CNV
reduction in letters trials compared to controls (CNV
distraction effect). The upward direction of the diag-
onal bars at each site indicates CNV enhancement in
no-letters trials compared to controls  (CNV rebound
effect).

no-letters trials at  both Cz and  Pz. Fig. 2 shows ex-
amples of CNV traces for one individual. A compari-
son of control values of CNV amplitude for the three
recording sites shows CZ larger than Pz and Pz larger
than Fz (Table  1).

   Reaction  times were slower in both letters and
no-letters trials compared to pooled controls (Table
1).  Heart  rate  levels were  significantly  elevated  in
the 50%  letters condition (X. = 72.99 beats permin.,
SD =  10.08) compared to pooled controls (X = 70.19,
SD - 9.79). There was no difference in eyeblink fre-
quency between the 50% letters (X = 14.08 blinks per
min, SD  =  6.73) and the  pooled  controls (X. = 13.14,
SD= 12.18) conditions. Mean percent of trials having
correct recall of letters was 96.67 (SD = 4.65).

Discussion

    Compared to control trials, the  short-term mem-
ory task produce da selective reduction of CNV ampli-
tude at Pz  and a slowing of reaction time to S2 (CNV
distraction effect). This finding suggests that centre-
posterior brain regions may mediate distraction effects
involving intermittent lexical stimulation and that dis-
ruption in  CNV development can be produced selec-
tively within a  block of trials. Similarly, the pattern
of CNV reduction in letters trials, where two overlap-
ping tasks  were performed (listening  to letters and
preparing for response  to tone)compared to no-letters
trials, where the subject only prepared for response to
tone, indicates that CNV development can be disrupt-
ed by the presence of a  high information load that
fractionates attentiveness to S2  (tetters trials), but
can be enhanced by the  unexpected removal o f this
information (no-letters  trials). That  these opposite
effects occurred in  the same block of trials demon-
strates fine-grained experimental  control  over one
type of event-related slow potential (CNV) through
the systematic  alteration of information processing.

    Perhaps the most  important finding in this study
is the unexpected elevation of CNV amplitude in the
no-letters  trials beyond control  values,  despite the
fact  that control and  no-letters trials  are physically
identical; i.e., each type  is made  up of a flash-tone-
key  press. These supranorrnal CNV amplitudes may
represent a disinhibitory effect produced by the shift
from a divided attention set in letters trials (listening
for letters and preparing  for response  to tone)  to an
undivided  (unified) attention set in no-letters  trials
(simply preparing for response to tone). This interpre-
tation is supported  by subjective  reports that, once it
was  clear early in the  flash-tone interval that no


                               -CONTROLS

                               -LETTERS

                               - NO LETTERS
    S1
                                      0.5 sec
Fig. 2. Examples of CNV traces (Pz) of one individual
for letters, no-letters, and control trials. Compared to
the control trace,  averaged CNV is reduced in letter
trials  (distraction effect) and is enhanced in no-letters
trials  (rebound effect).  CNV resolution (return  to
baseline after S2)  is incomplete for the letters ("dis-
traction")  trials, and the latency of the late positive
component of the evoked response to SI is delayed
for letters trials compared to control trials. The S1-S2
(flash-tone) interval is 1.5  sec. The subject's motor
response to tone is not depicted. Relative negativity
at Pz referred to linked earlobes is upward.

-------
CNV Rebound
                                           225
 letters would be heard, subjects quickly shifted their
 entire concentration to the tone task. The possibility
 that  the  CNV  rebound phenomenon  represents  a
 switching of attention  processes is in agreement with
 recent findings of diminution of this CNV effect in
 elderly individuals, who characteristically  have diffi-
 culty in changing attention sets. Whatever the mech-
 anism of  action, the CNV rebound effect  appears to
 be a reliable characteristic of the normal intact human
 brain and may provide a useful measure of plasticity
 in brain  functioning.  One  group of  patients with
 bimedial  prefrontal leucotomies showed a complete
 absence of the CNV rebound effect (Tecce et al., this
 volume).

     The finding of slower reaction times in no-letters
 trials as well as letters trials suggests that this measure
 may  reflect a  tonic distraction process in  the 50%
 letters condition produced by the uncertainty of not
 knowing  whether  letters will be present or absent,
 whereas CNV amplitude reduction may reflect a more
 selective type of phasic distraction process intrinsic to
 letters trials. The fact that heart rate was elevated In
 the  50%  letters condition suggests that this measure
 of cardiovascular  function may be a more  sensitive
indicator of experimental effort  and/or  distraction
than is oculomotor function as measured by the blink
response.
Summary

    Thirty-six  normal men were  tested  in  a simple
reaction time task consisting of a flash-tone-key press
(control trials) and in a similar task where a short-term
memory task for auditory letters was either presented
within the flash-tone interval or not presented (no-
letters trials).  CNV  amplitude showed a pattern of
reduction  in letters trials accompanied by a slowing
of reaction time (CNV distraction effect). CNV ampli-
tude  showed an unexpected pattern of elevation in
no-letters trials beyond control values (CNV rebound
effect). The rebound function of CNV may be a use-
ful  indicator  of plasticity  of functioning in the
normal intact human brain.

Acknowledgments

    The research assistance  of Mary Beth  Boehncr,
June  Savignano-Bowman, Connie Dessonvillc, Debbie
Melnbresse, James Kahle, and  Baiba  Liepins is grate-
fully  acknowledged.

-------
  AVERAGE  EVOKED POTENTIALS AND TIME
  PERCEPTION1
 P. TUETING

 Department of Psychophysiology, New York State Psychiatric Institute, New York, NY, U.S.A.
     Cognitive models of time perception relate time
 perception data to expectancy, information proces-
 sing, coding  and  memory.  Sensory  physiological
 models of time perception, on the other hand, propose
 an internal time keeper or clock whose rate is influ-
 enced by general  physiological activation (Rutsch-
 mann 1973, Triesman 1963, Vroon and van Boxtel,
 1972).

     Inferred physiological activation in time estima-
 tion has been related to independent physiological
 measures, particularly the EEC. Faster EEC frequen-
 cy composition has been taken as an index of greater
 physiological  activation, and thus  faster clock rate,
 than  slower EEC  frequency composition. Surwillo
 (1968) proposed that  the alpha rhythm is the visual
 representation of the internal clock, but data support-
 ing this hypothesis are not definitive (Anliker 1963,
 Coffin 1974, Surwillo 1964, Vroon and van Bostel
 1972, Woodruff 1975).

    Average evoked potentials (EPs) would appear to
 hold considerable promise in relating time perception
 to both physiological and cognitive models. Certain
 EP components have  been related  to physiological
 activation; others have been implicated in cognitive
 processes involving expectancy, uncertainty, and the
 matching of sensory input against memory.

    Recent studies illustrate how EPs might be indic-
 ative of the substrate  of behavioral  time estimation.
 In a study  in  which EPs and time estimation  data
were obtained  on different trials, cortisol was found
to decrease the effects of attention on the visual EP
(occipital) and to increase behavioral productions of
elapsed time (Kopell et al.  1970).  Weinberg et al.
 (1974) reported emitted potentials occurring at the
 point in time near the  end of an estimated duration.
 Me Adam (1966) reported that the contingent negative
variation  (CNV) preceding keypress increased as sub-
                                                   jects learned to produce a temporal interval. However,
                                                   CNV  decreased as  subjects became  more accurate.

                                                       Some  EP studies have used methods other than
                                                   production or reproduction to measure time estima-
                                                   tion. Simernitskaya (1973) asked subjects to discrim-
                                                   inate flashes of light according to duration when, in
                                                   reality, flash duration remained constant. A negative
                                                   component emerged in the parieto-occipital region at
                                                   a latency of 1 SO msec, whereas EPs from the centro-
                                                   frontal region were depressed (except for a 240-msec
                                                   component recorded from the  right  hemisphere).
                                                   Divenyi (1973) reported a relationship between over-
                                                   estimation  of short time intervals between two tones
                                                   and an increase in time between the peaks of N100
                                                   and P200.  Warhonowiez (1974), on  the other hand,
                                                   found  no  rlationship between  EPs  and  time error
                                                   when  subjects were required to judge the similarity
                                                   or dissimilarity in duration of sound pairs.

                                                       In the  following study, average evoked potentials
                                                   were simultaneously recorded  while subjects made
                                                   judgments  of a constant time interval  (1 sec) between
                                                   two clicks.  The rationale was that differences found in
                                                   EPs  to the  same  interval could be attributed to
                                                   differential   time perception since the actual time
                                                   interval remained constant.

                                                   Method

                                                      Eight males aged 18 to 27 years were paid  for
                                                   participating. Ag/AgCl electrodes were placed at  Cz,
                                                   Oz, and under the left eye with a left earlobe refer-
                                                   ence. EEC  and EOG amplifiers were  set to pass  fre-
                                                   quencies between 0.02 and 1000 Hz (half-amplitude
                                                   points).

                                                      Each trial began with the onset of a fixation light
                                                   that  remained on throughout the trial. One second
                                                   later, two clicks (1000-to 3000-Hz bandpass) separated

  is investigation received support from HEW Grant No.  MH14580 and NIE Grant No. C-74-0042 awarded to
Dr.  Samuel Sutton. The assistance of Dr. Sutton, Ms. Marion Hartung, and Mr. Dennis Murphy is gratefully ac-
knowledged.   I am indebted to Dr. Daniel Ruchkin and Dr. Jacques  Rutschmann for a critical  reading of the
manuscript.

-------
EPs and Time Perception
                                            227
by a 1-sec interval were presented. The subject's task
was to judge the  silent  interval  between the first
and the second click. One second after the  second
click,  a  flash cued the subject to indicate his judg-
ment by pressing either a left or  right  key with his
right index finger. The intertrial interval between the
flash  and presentation  of the fixation light  for the
next trial was 5.9 to 6.9 sec.

    Subjects were instructed that the task would in-
volve a difficult discrimination of two different silent
time intervals bounded by clicks. Before each block
of 20  trials, they were given an example of the short
interval (e.g., 990 msec) and of the long interval (e.g.,
 1010  msec),  ostensibly for  reference and practice.
During experimental trials,however, only a 1000-msec
interval between clicks were presented, although sub-
jects were told that approximately 50% of the inter-
vals would be short and 50% of the intervals would be
long. A random program  was used to determine the
subject's  "accuracy"  and payoff (a dime added for
each "correct" discrimination, or subtracted for each
 "incorrect" discrimination), and subjects were inform-
ed at the end of each block how much they had won
or lost in that block. There were 400 trials, 20 blocks
of 20  trials each, and some subjects were retested on
a second day.

     In addition, one subject was run on a variation of
 the above design.  Experimental blocks were compos-
 ed of a random sequence of trials with either a 990-
 or 1010-msec interval between clicks. EPs were aver-
 aged separately for short and long intervals, and for
 correct and incorrect discriminations. Approximately
 1 to 3 sec following the subject's response, a feedback
 click was delivered following a correct discrimination
 only (first session) or following an incorrect discrimi-
 nation only (second session).

     Trials with eye movement were discarded. Com-
 ponents were measured from peak-to-peak and from
 baseline-to-peak with baseline determined by visual
 estimation  of voltage over a 100-msec period before
 presentation of either the  first  or second click.  In
 addition, the CNV was measured  as the most negative
 point within the 1000-msec interclick interval relative
 to the baseline preceding the first click. Differences
 were assessed by t-tests for correlated data.

 Results

    The N94, PI87,  and N274  vertex components
 were consistent across subjects and were systematical-
 ly measured, however, occipital records were not
 measured because of inconsistencies across  subjects
 and alpha contamination in two subjects. For some
 subjects, a late occipital  positive component follow-
 ing the  second click  could be  identified (P3007),
 particularly for judgments that  duration was short.
    No significant  differences  between vertex EPs
associated with judgments of short versus judgments
of long were found for the first click or for the height
of the CNV.

    Vertex EPs for the second click averaged separate-
ly for trials judged short and for trials judged long are
shown in Fig. 1 for all eight subjects. None  of the
measurements of N94, P187, or N274 from baseline
was significantly different between conditions. How-
ever,  the peak-to-peak  measure of P187-N274 was
larger  for long judgments (5.5 juV), than  for short
judgments (3.7 MY) measurements indicated that the
difference was largely due to  variation in the N274
component.
                	JUDGED LONG
                	JUDGED SHORT
                 SECOND CLICK
              1
                           ^
                       ?.5 JIV
              0  0.4       0   0.4
                    TIME.MC

 Fig. 1.  Vertex auditory EPs (N « 200) elicited by
 the second click delimiting the end of the estima-
 tion interval for eight subjects.  Responses associated
 with intervals judged short and intervals judged long
 are aligned on P200.  Vertical lines delimit the laten-
 cy window within which  area difference measures
 were made.

      EPs of long and short trials differed in the latency
  range between 213  and  387 msec following the
  presentation of the second click. This region is bound-
  ed by the two vertical lines in Fig. 1. Since the differ-
  ence appeared consistently within a given EP latency
  region, the following measurement was devised: EPs
  for  long and  for short judgments were aligned on
  P187 (as in Fig. 1), and the area of the difference in
  the 213-to387-msec region was measured with respect
  to whether the waveform was more negative or posi-
  tive  for long than short judgments. Results indicated
  that the waveform was more negative  for long  than
  short judgments within this time window (p < .01).

      Fig.2 shows data obtained from the single subject
  run  on  the  random  sequence of short  and  long

-------
 228
                                       Tueting
                                                       ——  JUDGED LONG
                                                       .......  JUDGED SHORT
                             SHORT INTERVAL
                               (990 muc)
      LONG INTERVAL
      (1010 muc)
        SECOND CLICK
         CLICK
                              8HORT RESPONSE
       LONG RESPONSE
                                          0.8
                  0.8
                                                                                I
                                                TIME, sec
        Fig, 2. Vertex auditory EPs (N'* J00,/ elicited by the second click delimiting the end of the
        estimation Interval for one subject. Waveforms at the top (overlapped at baseline) show differ-
        ences associated  with short and long judgments for the 990-msec and 1010-msec intervals
        separately. The same waveforms are overlapped at the bottom to indicate that the waveshapes
        are similar for the same response category regardless of whether the actual interval was 990 or
        1010 msec.
intervals. Again, the  waveforms for intervals judged
long is more negative in the N250 region whether or
not the Interval was actually short or long. (The sub-
ject judged 53% of the trials correctly). EPs obtain-
ed for the feedback clicks showed a larger early P300
that was larger for feedback  indicating an incorrect
than correct judgment.


    The behavioral data were inspected for evidence
of a "lengthening" of the perceived interval over time
(indicative of a slowing down of clock rate with time).
"Lengthening" in this case would be indicated by an
increasing proportion of short judgments. No consis-
tent increase in the proportion of short judgments
either across blocks or  within blocks of 20 trials was
found.  (Four subjects showed  lengthening across
blocks, one showed shortening, two showed no change
and one subject's behavioral data were not recorded.)
However, subjects produced  more  short (x « 66%)
than long (x » 34%) judgments during the experiment
(p<.02).
Discussion

    Auditory EP differences related to perception of
elapsed time were found. Differences  between long
and short judgments were apparently subjective and
unrelated to actual elapsed time, which was constant
or was varied by such a small amount that chance
performance resulted. This aspect of the design is sim-
ilar to that of Begleiter and Porjesz (1975). In their
study, visual EPs to the same medium-intensity flash
differed depending upon whether the flash was judged
as dim or bright.

    Although  it is  important to  demonstrate that
average EPs reflect subjective time  judgments,  the
differences can be interpreted further. A relationship
between increased amplitude of the N2SO component
and lowered arousal has been suggested (Wilkinson et
al. 1966), and the P300 EP component has been relat-
ed to variables of a more cognitive nature (see  Tuet-
ing, this volume). The P300 component may be partial-
ly overlapped in time with the N250 component at

-------
EPs and Time Perception
                                           229
vertex under certain experimental conditions (Fried-
man et al. 1973, Prichep et al. 1976, Tueting 1968;
Tueting, this volume). In fact, it is known that several
components can occur in this latency region, and it is
not always possible to delineate them.

    EP differences to the second click appear to be
opposite to differences expected on the basis of clock
theory. In terms of a physiological model of time per-
ception, judgments that the interval was long should
be associated with a faster clock rate and thus greater
physiological activation (smaller N2SO). Instead, the
results indicated that long judgments were associated
with a  larger N250 component. In  addition, no  EP
differences were found to the first click that initiated
the beginning of the  estimation interval. Presumably,
differences in a tonic state of physiological activation,
if present, would have affected the click initiating the
estimation period as  well  as the click ending the esti-
mation period.

    The findings appear to relate better to a cognitive
model than to a physiological model of time percep-
tion. As mentioned above, a larger positive component
associated with cognition may overlap N250 in vertex
recordings. The presence  of a P300 component  for
short judgments can be seen in Fig. 2, and the com-
ponent was also seen in occipital recordings for some
subjects.

    There is no direct evidence from this study con-
cerning what cognitive aspects may have resulted in
larger P300 for judgments of "short." P300 amplitude
has been related to lower response probability (Karlin
and Martz  1975), but in the present study 66% of
the judgments were "short"—the category with larger
positivity in the P300 region. Information from other
studies  indicates that larger P300  components  are
associated with stimuli  possessing greater salience
(Jenness  1972)  and greater  decision  confidence
(Squires et al.  1975). It is not immediately obvious
which  of these aspects  might  have been involved.
Greater salience for  short is suggested  by the larger
percentage  of short judgements,  However, greater
decision  confidence  for short judgements is  also a
viable  explanation  because the  accuracy  of time
judgment (in absolute units) generally  decreases the
longer the intervals to be judged, and  greater tem-
poral uncertainty is related to smaller P300 amplitude
(McCarthy and Donchin 1975).

    In order to assess alternative explanations pro-
posed for the data in terms of P300, more systematic
study is  obviously required. A direction for further
study would propose the use of a wider range of time
intervals  to be judged and a wider range  of verbal esti-
mates by the subject. The relationship of EPs to time
perception accuracy could then  be more precisely
studied, perhaps within a signal detection framework.
Since the  P300 component  can be  isolated from
N2SO by the modality-specific distribution of N2SO,
information on topographical distribution would be
desirable. In  any case, the findings of the  present
study suggest that investigation of EP correlates of the
judgment of elapsed time may reveal previously un-
available information concerning the state of the sub-
ject during a behavioral time estimation  task.

Summary

    Average evoked  potentials associated with short
versus long judgments of elapsed time were recorded.
On each  trial, two clicks separated by a silent interval
to  be estimated  were presented. This interval was
always 1 sec in duration. EP amplitude to the second
click delimiting the end of the estimation interval was
more negative in the N250 region (213 to 387 msec)
for judgments that  the interval was long than short.
No differences were found for EPs elicited by the first
click  or  for  the  CNV. These findings indicate that
average  EP differences can reflect time judgments
that are entirely subjective. Further, the results do
not seem to correspond to a sensory physiological
model relating activation to the rate of a hypothetical
internal  clock.  A cognitive explanation involving the
WOO component (which may be partially overlapped
in time with N250) was considered likely, e.g., greater
salience  or more decision confidence for  "short"
judgments.

-------
ATTENTION AT THE  COCKTAIL  PARTY:
BRAINSTEM EVOKED  RESPONSES REVEAL NO
PERIPHERAL  GATING

D. L. WOODS AND S. A. HILLYARD
Department of Neurosciences, University of California, San Diego, CA, U.S.A.
    When a subject is presented simultaneously with
two different spoken messages, one fed into each ear,
he is usually able to focus his attention effectively on
one message to the exclusion of the other (Cherry
1953). In fact, when instructed to attend to a normal
speech passage in one ear, subjects typically can recall
only rudimentary information about the passage in
the unattended channel (such as the sex of the speak-
er), as if somehow that message had been "shut out"
before processing was complete. The problem of how
one can attend to one channel of auditory input to
the exclusion of other competing channels has come
to be known as the "cocktail party problem," since
similar mechanisms are presumably involved in follow-
ing a conversation  in  a noisy room. The subjective
experience of shutting out, or gating, the irrelevant
inputs has found theoretical expression in theories of
attention that propose a filtering (Broadbent 1958)
or attenuation (Treisman 1964) of unattended inputs.

    In its most radical physiological form, the gating
hypothesis suggests that sensory inputs  are attenuat-
ed near the auditory periphery through descending ef-
ferent inhibitory systems. Several years after Cherry's
demonstrations of dichotic attention, Hernandez-Peon
(1956) claimed to find support for such a radical posi-
tion; click-evoked responses in cats were reduced in
amplitude at the cochlear nucleus when the cat was
distracted by visual or  olfactory stimuli. Subsequent-
ly, however, this influential experiment was criticized
on methodological 'grounds:  head and  pinna move-
ments of the cats could have reduced effective stim-
ulus intensity and diminished cochlear nucleus evoked
responses regardless  of efferent attenuation (Worden
1966).  When care was taken to stabilize acoustic in-
put,  evoked responses  at the  cochlear nucleus were
found to remain stable despite attentional manipula-
tions (Wickelgren 1968).

    In  several recent  experiments that maintained
careflil control of the acoustic input, however, modi-
fications in  brainstem  evoked activity were reported
with changes in behavioral state  (Oatman 197i,Suga
and Shirnozawa 1974). Similarly, changes have been
reported in  auditory  evoked unit  responses in the
cochlear nuculeus (Buchwald and Humphrey 1972),
Olesen et al. 1975), medial geniculate (Gabriel 1975),
and brainstem reticular formation (Olds et al.  1972)
as  a function of habituation or conditioning proced-
ures. These studies reaffirm the possibility that effer-
ent modulation of input plays a role in some forms of
auditory attention.

    Techniques have  recently become available for
measuring  click-evoked,  far-field electrical activity
generated in the human brainstem (Jewett and Willis-
ton 1971).  Such brainstem evoked responses (BERs)
permit a reexamination of the peripheral gating hypo-
thesis in paradigms known to produce powerful selec-
tive-attention effects in man. In the first such study,
Picton and Hillyard (1974) examined auditory (evok-
ed potentials  during  click  intensity discrimination
and in a control condition when clicks were ignored.
They found no significant change in any brainstem
evoked  component, but substantial enhancements in
the late  waves (Nl-P2)when attention was focussed
on the  train of clicks. Similar effects  of attention
upon the late cortical evoked potentials have been
reported in a variety of selective-attention paradigms
where the subject was required to focus attention on
one channel to the exclusion of others (see Hillyard
and Picton  1978 for a review). In a recent study,
Hink and Hillyard (1976) found late wave correlates
of selective attention to dichotic speech  passages.
Both Nl (80 to 120 msec) and P2 (160 to 200 msec)
components were enhanced to neutral vowel probe
stimuli when attention was directed to a superimposed
prose passage in that ear.

    In the  present  study, the dichotic speech para-
digm was used to investigate possible attention-relat-
ed modifications of probe-evoked BERs. Some evi-
dence suggests that the efferent auditory projections,
which are anatomically well-established in primates,
might be functionally engaged by such ecologically
appropriate  and  acoustically  complex  stimuli. For

-------
BERs and Gating
                                            231
example,  Dewson  (1968) reported that macaques
were  impaired  in their ability to make difficult pho-
neme discriminations presented in masking noise fol-
lowing section  of the efferent olivo-cochlear bundle.
Presumably, such efferent modulation, if it also exists
in man in comparable situations, should be reflected
in attention-related changes in the BER.

Methods

    Six paid student volunteers served as subjects.
They  reclined in  a  dimly lit, sound-treated chamber
and  listened through  headphones while  different
prose  passages were presented independently in each
ear. A female  voice read  from a  novel in one ear,
while  a male read a separate story in the other; super-
imposed on each  passage were bursts of click stimuli
delivered independently in the two ears. Click triggers
and spoken prose passages were recorded  on a four-
channel audio tape  system  in such  a way that during
active conditions, the click probes and the speech it-
self always occurred concurrently in a given  ear; i.e.,
clicks never occurred  without speech in the same
channel.  During  control  conditions,  speech inputs
were disconnected and  clicks  were presented in iso-
lation.

    Click stimuli were 100-^sec pulses of gated white
noise, presented with ISIs of 40 msec, in bursts of six.
The minimum interburst interval was 100 msec. Click
intensities in each ear were adjusted to 61 dB sound
pressure level (SPL) and  maximum speech intensities
for both male  and  female  voices were set at 60 dB
SPL. Subjects attended to one channel of prose or the
other in a  counterbalanced design, and answered a
short questionnaire  on its contents after each 5. to 10-
minute  segment.  Male and female  voices were also
counterbalanced across ear of delivery.

    BERs  were recorded from vertex-right  mastoid
and left mastoid-right mastoid configurations,  ampli-
fied (bandpass set at 10 to 3000 Hz), and stored along
with appropriate  trigger  pulses on  audio tape (band-
pass 30 to 16,000 Hz) for subsequent off-line averag-
ing.  Mid-latency  and  late evoked potentials  were
recorded from  Cz,  T3,  and T4 derivations and from
the upper orbit, each referenced to the right mastoid
process. They were  amplified (bandpass set at 0.15 to
500 Hz) and stored  on an FM tape recorder (bandpass
0 to ISO Hz) for subsequent off-line analysis on a sig-
nal averager.  All  inter-electrode  impedances  were
maintained  below  4  KI2.  Each  experiment last-
ed 2  hours, with BERs recorded  to approximately
120,000 probe  stimuli.

Results

    Subjects attended  effectively  to the  relevant
channel and reported typical  cocktail party effects;
i.e., the unattended messsage faded away, was shut out,
or was blocked from awareness, Despite this powerful
subjective impression of  input filtering, none of
the BER components changed significantly in either
latency or  amplitude  as  a function of  attention.

   In  vertex-to-mastoid  derivations,  six positive-go-
ing components  (waves I-VI   of  Jewett) and  the
prominent negative deflections after waves I (Label-
led I'), HI (III'), and VI (VI') were identified in all
subjects  (Fig. 1  shows  representative tracings from
two  subjects).  Mastoid-to-mastoid  records  showed
prominent  early  waves, particularly wave III, with
much  attenuated  late components (IV-VI); again,
there  were no significant attention-related changes in
amplitude or latency of any component.

    Tables 1 and 2 show the stability of amplitudes
and latencies of different components of the vertex-
to-mastoid BER as a function of attention:  analysis
of variance (subjects x ear x sex of speaker x direction
of attention) did not reveal any significant  attention-
related changes in amplitude or latency  of any com-
ponent. The large number of  click-evoked responses
(96,000 per subject) provided  high signal/noise ratios
and low variances for each component, as shown in
Table  2. For example,  the  95% confidence  interval
established around the  peak-to-peak V-VI'  measure
showed that any attention-related changes would be
less than ±5.27% of the overall mean amplitudes.

    In most  subjects, midlatency  (10 to 30 msec)
responses were also recorded; they were highly varia-
ble in amplitude both within  and  between  subjects
and  were not systematically  related to attentional
manipulations. This variability is presumably due to
the large myogenic components in this latency range
(Picton et al. 1974b). The Nl and P2 components eli-
cited  by each burst were too small to be reliably quan-
tified, consistent with  trie well-established refrac-
toriness of these vertex potentials, both with repetitive
monaural stimuli (Davis et al.  1972) and between the
ears (Butler  1972).
 Discussion

     The stability of the BER suggests that peripheral
 gating of the  rejected  input  does not occur when
 attention is focused  on  one channel of dichotic
 speech. Such observations are consistent with theories
 of attention that suggest that all sensory  inputs are
 fully analyzed in parallel regardless of attentional
 focus (Deutsch and Deutsch  1963, Norman 1968).
 This result is also in accord with the Broadbent/Treis-
 man models of attention, which claim that selection
 between stimulus channels can occur on the basis of
 simple cues, such as spatial  position or frequency,
 provided such selection occurs beyond the brainstem

-------
232
                                            0.5 juV
                                             ATTEND LEFT

                                             ATTEND RIGHT
      Woods and HiUyaid

RIGHT EAR
                                                                C.C.
                                                                 L.C.  i
                                                                 C.C.
                                                                 L.C.  \
                                                                           PROBE
                                                                            PROBE
                                                                         ,  I .1.1.1.1
                                              TIME, ms«c

 Fig. 1.  Click-evoked BERs from two subjects (sisters C.C. and L.CJ during attention to left and right ear mes-
 sages.  Click probes were superimposed on female and male voices in left and right ears, respectively, in the top
 pairs of tracings; voices were reversed in lower tracings.  N= 12,000 responses/average.

-------
BERs and Gating
                                        233
Table 1. Mean Latency and Batellne-Peak Amplitude of Each Component In the Bralnatem
                          EP  to Attended and Nonattended Clicks

Latency, msec
Attended
Nonattended
Standard deviation
Amplitudes, /uV
Attended
Nonattended
Standard deviation
Component
I

1.58
1.56
± 0.036

0.08
0.06
± 0.009
II

2.71
2.80
±0.065

-0.03
0.01
±0.033
III

3.67
3.65
±0.015

0.11
0.15
±0.030
IV

5.11
5.06
±0.024

0.25
0.26
±0.024
V

5.86
5.77
±0.050

0.39
0.41
±0.019
VI

7.31
7.34
± 0.020

-0.22
-0.21
± 0.025
VII

8.38
8.34
±0.033

-0.42
-0.39
±0.027
Table 2. Mean Peak-Peak Amplitudes (/;V) of Bralnstem EP Components and the Percent
     Difference In Amplitude (±se) between Attended and Nonattended Conditions
                                                    Components

Attended
Nonattended
Standard deviation
Difference
Standard error
l-l'
0.213
0.193
±0.016
+9.8
±7.8
I'-lll
0.244
0.257
±0.012
-5.18
±5.9
Ill-Ill'
0.165
0.167
± 0.016
-1.20
±8.9
III'-V
0.444
0.455
±0.022
-2.45
±4.7
V-V
0.600
0.601
±0.016
-0.17
±2.7
v-vr
0.808
0.808
±0.015
0.00
±2.3
 level. In fact, it has been suggested that such a telen-
 cephalic selective mechanism is reflected in attention-
 related  changes  of  the  auditory-evoked Nl  wave
 (Hillyard and Picton 1978).

    While the present study seems to rule  out a crude
 attenuation  of all input  to. the unattended ear, the
 possiblity of fine-tuned peripheral attention cannot
 be excluded. Recent work by Hecox (1974)' estab-
 lished that the BER is generated almost exclusively
 by high-frequency-sensitive portions  of  the cochlea
 and auditory pathways.  Since the frequencies that
 contain most of the linguistic information in human
 speech are below those that contribute to the BER,
 attenuation  restricted  to speech frequencies might
 still occur in unattended channels with little accom-
 panying change in the BER itself.


     These results are concordant with studies demon-
 strating the stimulus-bound character of the BER
 (Picton et al. 1974, Hecox and Galambos 1974), and
 the stability of the BER during changes in arousal or
 attention  (Amadeo and  Shagass 1973,  Picton and
 Hillyard 1974, Jewett and Williston 1971).
   The absence of  any attention-related  change  at
Nl  latencies in this study is sdmewhat surprising,
since such'effects have been shown to occur in similar
experiments where thft probes mdre closely approxi-
mated the acoustic properties of the speakers' voices
(Hink and Hillyard  19'76). Perhaps the evocation  of
Nl is more finely tuned during attention to  normal
speech than it  is in less natural  tasks,  so that high-
frequency probes were excluded  from  the attended
channel by spectral  considerations alone,  regardless
of the ear of presentation. A second possibility is that
the  attention-related  increases  in  80-to-120 msec
negativity of Nl (Schwent and Hillyard 1975) do not
occur when ISIs are so short as to eliminate the Nl
altogether.
Acknowledgments


   Special  thanks go to Steve Vanvoorhis for tech-
nical assistance. This work was  supported by N1H
Grant MH-22594.01, NASA Grant NCR 05-009-198,
and NSF  Grant BNS-77-149231. D.  Woods held a
USPHS Predoctoral Fellowship.

-------
 INFORMATION PROCESSING REFERENCES
 Adams, J, C. and  Benson,  D, A, Task-contingent
   enhancement  of the auditory evoked  response.
   Blectroenceph. clin. Neurophysiol., 1973,35:249-
   257.

 Amadeo, M, and Shagass, C. Brief latency click evoked
   potentials during waking and sleep in man. Psycho-
   physiology, 1973, 10:244-250.

 Anliker, J. Variations in alpha voltage of the electro-
   encephalogram. Science, 1963, 140:1307-1309.

 Arezzo, J., Pickoff, A. and Vaughan, H. G. The sour-
   ces and intracerebral distribution of auditory evok-
   ed  potentials in the alert rhesus monkey. Brain
   Res., 1975,90:57-73.

 Bancaud,  J., Block, V.  and Paillard, J.  Contribution
  . EEC a  1'etude des potentials evoques chez I'homme
   au niveau du vertex. Rev. Neurol., 1953, 89:399-
   418.

 Barlow, J. Some observations on the electrophysiology
   of timing in the central  nervous system. Electro-
   enceph. clin. Neurophysiol., 1969, 27:545.

 Barlow, J. S., Morrell, L. and Morrell, F. An evoked
   response in relation to temporal conditioning to
   paired  stimuli  in man. MIT Res. Lab. Electron.,
   QPR 1965, 78:263-272.

 Barrett, G., Halliday,  A. M. and Halliday, E. P300
   amplitude and sequential stimulus  interaction in a
   somatosensory detection task. Electroenceph. clin.
   Neurophysiol.,  1975,39:223-224.

 Begleiter,  H. and Porjesz, B. Evoked  brain potentials
   as indicators of  decision-making.  Science, 1975,
   187:754-755.

Bernstein,  A. S., Taylor, K. W. and  Weinstein, E. The
   phasic  electrodermal response as a differentiated
   complex  reflecting stimulus significance. Psycho-
   physiology, 1975, 12:158-169.

Bostock, H.  and Jams, M. J. Changes in the form of
   the  cerebral evoked response related to the speed
   of simple reaction time. Electroenceph. clin. Neuro-
   physiol., 1970,29:137-145.

Broadbent, D. Perception and communication. Perga-
   mon Press, London,  1958.

Buchwald, J. S. and Humphrey, G.  L, Response plas-
   ticity in cochlear nucleus of decerebrate cats during
   acoustic habituation procedures. J. Neurophysiol.,
   1972,35:864-879.

Butler, R. A. The influence of spatial separation of
   sound sources on the auditory evoked response.
   Neuropsychologia, 1972, 10:219-225,

Campbell, K. B. and Picton, T. W. Stimulus informa-
   tion content and the  late positive component of
   the human  evoked  potential. Neurosci. Absts.,
   1976, 11:237.

Cherry, E. C. Some experiments on the recognition
   of speech with one and two ears. J. Acoust. Soc.
   Am., 1953,25:975-979.

Coffin, S. C. Behavioral correlates of variability in the
   duration of the cortical excitability cycle. Doctoral
   dissertation, Stanford  Univ. University Microfilms,
   1974,  No. 74-26,007.

Cooper, R., McCallum, W. C., Newton, P., Papakostop-
   oulos, D., Pocock, P.  V. and Warren,  W. J. Corti-
   cal potentials associated with the detection of vis-
   ual events. Science, 1977, 196:74-77.

Courchesne, E.  Stimulus novelty, recognition,  task-
   relevance and the visual evoked potential in  man.
   Unpublished doctoral dissertation, Univ. of  Cal.,
   San Diego, 1975.

Courchesne, E. Event-related brain potentials:  A  com-
   parison  between children  and adults. Science,
   1977,  197:589-592.

Courchesne, E., Courchesne, R. and Hillyard, S.  A.
   The effect of stimulus deviation on P3 waves to
   easily recognized stimuli. Neuropsychologia, 1978,
   16:189-200.

Courchesne, E.,  Hillyard, S. A. and Galambos,  R.
   Stimulus  novelty, task relevance and the visual
   evoked  potential  in  man.  Electroenceph.  clin.
   Neurophysiol., 1975,39:131-143.

Crow, H. J.  Intracerebral polarisation and multifocal
   leucocoagulation  in   some psychiatric illnesses.
   Psychiat.   Neurol.  Neurochir.  (Amst.),  1973,
   76:365-381.

Davis, P.  A.  Effects of acoustic stimuli on the  waking
   human brain. J. Neurophysiol., 1939, 2:494-499.

Davis, H., Mast, T., Yoshie, N. and Zerlin, S. The slow
   response of the human cortex to auditory stimuli.

-------
Information Processing References
                                          235
   Electroenceph. clin. Neurophysiol., 1966,21:105-
   113.

Debecker, J. and  Desmedt, J, E. Cerebral  evoked
   potentials correlated with  speed overload  in  a
   sequential acoustic detection.  In J.  E. Desmedt
   (Ed.), Cerebral evoked potentials in man. Brussels
   (in press).

Debecker, J. and  Desmedt, J. E. Cerebral  evoked
   potential  correlates in forced-paced tasks. Nature
   NewBiol., 1971,234:118-120.

Deecke, L.,  Grozinger, B. and Kornhuber, H.  H. Vol-
   untary finger movement in  man: Cerebral poten-
   tials and  theory. Biol. Cybernetics, 1976, 23:99-
   119.

Desmedt, J.  E. and Debecker, J. Differential enhance-
   ments of early and late components of event-related
   evoked potentials in crossmodality and intramodal-
   ity  cognitive  processing;  their  dissociation  from
   CNV  and slow potential  shift. In J.  E. Desmedt
   (Ed.), Prog. clin. Neurophysiol. (in press).

Deutsch, J. A. and Deutsch, D. Attention: Some theo-
   retical considerations. Psychol.  Rev., 1963,70:80-
   90.

Dewson, J.H.,  III.   Efferent  olivocochlear  bundle:
   Some relationships to  stimulus discrimination in
   noise. J. Neurophysiol.,  1968, 31:122-130.

Dincheva, D., Atsev, E. and  Popivanov, D. Investiga-
   tions on  mental work capacity by modern electro-
   physiological  methods.  Comptes rendus de t'Aca-
   demie Bulgare des Sciences,  1972, 24:829-832.

Donald, M. W. Topography of evoked potential ampli-
   tude fluctuations. In W.  C. McCallum and  J. R.
   Knott (Eds.),  The responsive brain. Wright, Bristol,
   1976:10-14.

Donchin, E.  Average evoked potentials and uncertain-
   ty resolution.  Psychonom. Sci.,  1968, 12:103.

Donchin,  E. Brain electrical  correlates of  pattern
   recognition. In G. F. Inbar (Ed.), Signal  analysis
   and pattern recognition in biomedical  engineering.
   Wiley, New York, 1975:199-218.

Donchin, E. CNV-P300 relationships: A correspon-
   dence. In W. C. McCallum and  J. R. Knott (Eds.),
   The responsive brain. Wright,  Bristol, 1976:222-
   234.

Donchin, E.  and Cohen, L. Averaged evoked potentials
   and intramodality  selective  attention.   Electro-
   enceph. clin. Neurophysiol., 1967, 22:537-546.

Donchin, E., Kubovy, M.,  Kutas, M., Johnson, R., Jr.
   and  Herning,  R. I.  Graded changes in  evoked
   response  (P300) amplitude as a function of cogni-
   tive activity. Percept. Psychophys., 1973,  14:319-
   324.
Donchin, E., Tueting, P., Ritter, W., Kutas, M. and
   Heffley, E. On the independence of the CNV and
   the P300 components of the human averaged evok-
   ed potential.  Electroenceph.  clin. Neurophysiol.,
   1975,38:449-461.

Eason, R.  G., Barter, M. R. and White, C. T. Effects
   of attention and arousal on visually evoked cortical
   potentials  and  reaction time  in  man. Physiol.
   Behav., 1969,4:283-289.

Falmagne, J. C., Cohen, S. P. and Dwivedi, A. Two-
   choice  reactions as an ordered memory scanning
   process. In P.M.A.  Rabbitt and S. Dornic (Eds.),
   Attention and Performance,  V. Academic  Press,
   New York, 1975.

Ford, J. M. Effects of attention on the scalp-recorded
   auditory  evoked potential of man.  Unpublished
   doctoral dissertation, Stanford Univ.,  1975.

Ford, J. M., Roth, W. T., Dirks, S. J.  and Koppell,
   B. S. Evoked potential correlates of signal recogni-
   tion between and within modalities. Science, 1973,
   181:465-466.

Ford, J. M., Roth, W. T. and Koppell, B. S. Auditory
   evoked potentials to unpredictable shifts in pitch.
   Psychophysiology, 1976a, 13:32-39.

Ford, J. M. Roth, W. T. and Koppell, B. S. Attention
   effects on auditory evoked potentials to infrequent
   events. Biol. Psychol., 1976b, 7:65-77.

Friedman,  D. Effect  of stimulus uncertainty on the
   pupillary dilation  and pupillary slope responses.
   Paper  delivered  to the American  Psychological
   Association, Symposium on the Pupil, 1972.

Friedman,  D., Hakerem, G.,  Sutton, S. and  Fleiss,
   J. L. Effect of stimulus uncertainty on the pupillary
   dilation response and the vertex evoked potential.
   Electroenceph. clin. Neurophysiol., 1973,34:475-
   484.

Friedman,  D., Simson, R.,  Ritter, W.  and  Rapin, I.
   Cortical evoked potentials elicited by real speech
   words  and human  sounds.  Electroenceph. clin.
   Neurophysiol., 1975,38:13-19.

Friedman, D., Simson, R., Ritter, W.  and Rapin, I.
   The late positive component (P300) and informa-
   tion processing in sentences. Electroenceph. clin.
   Neurophysiol., 1975,38:255-262.

Gabriel, M., Saltwick, J. and Miller, D. Conditioning
   and reversal  of short-latency  multiple unit respon-
   ses in the rabbit medial geniculate nucleus. Science,
   1975,189:1108-1109.


Galambos, R.,  Benson,  P., Smith, T.  S., Schulman-
   Galambos, C. and Osier, H. On hemispheric differ-
   ences in evoked potentials to speech stimuli. Eleo-
   troenceph. clin. NeurophysioL, 1975,39:279-283.

-------
236
           Information Processing References
Gastaut, Y. Lea pointes negatives evoquees sur le ver-
   tex:  Leur  signification psychophysiologlque et
   pathologique. Rev. Neurol., 1953, 89:382-398.

Gerbrandt, L. K., Goff, W. R. and Smith, D. B. Distri-
   bution  of the human average movement potential.
   Electroenceph. clin.  Neurophysiol.,  1973,34:461-
   474.

Graham, F. K. and Clifton, R. K. Heart-rate change as
   a component of the orienting response. Psychol.
   Bull., 1966,65:305-320.

Green, D.  M. and Swets, J. A. Signal detection theory
   and psychophysics, Wiley, New York, 1966.

Hakerem,  G. Pupillography. In  P. Venables (Ed.), A
   manual of psychophysiological methods. North-
   Holland, Amsterdam, 1967:337-349.

Harter, M. R. and Salmon, L. E. Intra-modality selec-
   tive attention and evoked potentials to randomly
   presented patterns. Electroenceph. clin. Neurophy-
   siol., 1972,32:605-613.

Hays, W. Statistics for Psychologists. Holt, Rhinehart
   and Winston, New York, 1963,

Hecox, K.  Frequency specificity of brainstem evoked
   responses in man. J.  Acoust. Soc. Am., 1974a,
   56:S63.

Hecox, K. and  Galambbs, R.  Latency  changes in
   human  brainstem evoked potentials as a function
   of signal in tensity and age of the subject. J. Acoust.
   Soc. Am., 1974,55:488,

Hernandez-Peon,  R.,  Scherrer,  H.  and  Jouvet, M.
   Modification of electrical activity in cochlear nucle-
   us during attention in unanesthetized cats. Science,
   1956,123:331-332,

Higgins, J. D. Set and uncertainty as factors influenc-
   ing anticipatory cardiovascular responding in
   humans. J. comp. physiol. Psychol., 1971, 74:272-
   283.

Hillyard,  S, A., Courchesne, E.,  Krausz, H.  I. and
   Picton, T.  W. Scalp topography of the "P3" wave
   in different  auditory  decision  tasks. In  W. C.
   McCallum and J. R. Knott (Eds.), The responsive
   brain. Wright, Bristol, 1976:81-87.

Hillyard, S. A., Hink, R. F., Schwent, V. L. and Pic-
   ton, T. W. Electrical signs of selective attention in
   the human brain. Science, 1973, 182:177-180.

Hillyard, S, A., Squires, K. C., Bauer, J. W. and Lind-
   say, P.  H.  Evoked potential correlates  of auditory
   signal  detection. Science,  1971, 172:1357-1360.

Hillyard,  S. A. and Picton, T. W. Event-related brain
   potentials and selective information processing in
   man. In J. Desmedt (Ed.), Prog.  clin. Neurophysi-
   ol., Vol. 6, Cognitive components in cerebral event-
  related  potentials and selective  attention. Karger,
  Basel, 1978:1-50.

Hink, R.  F.  and Hillyard, S. A. Auditory evoked
  potentials  during  selective  listening to dichotic
  speech  messages.  Percept. Psychophya.,   1976,
  20:236-242.

Hyman, R, Stimulus information as a determinant of
  reaction time. J. exp. Psychol.,  1953,45:188-196.

Jenness, D. Auditory evoked-response differentiation
  with  discrimination learning to humans. J. comp.
   Physiol. Psychol., 1972a, 80:75-90.

Jenness, D. Stimulus role  and gross differences in the
  cortical evoked  response.  Physiol. Behav.,  1972b,
  9:141-146.

Jewett, D. and Williston, J. Auditory-evoked far fields
  averaged from the scalp of humans. Brain, 1971,
  94:681-696.

John, E. R. Switchboard  versus statistical theories of
   learning and memory. Science, 1972, 177:850-864.

 Johnson, L.  C.  and  Lubin,  A. On planning psycho-
   physiological experiments. In N. S. Greenfield and
   R. A. Stembach (Eds.), Handbook of psychophys-
   iology. Holt, Rinehart and Winston,  New York,
   1972:125-158.

 Kagan, J. Do infants think? Sci. Amer., 1972, 226:74-
   83.
 Karlin, L. Cognition, preparation  and sensory-evoked
   potentials. Psychol. Bull., 1970, 73:122-136.

 Karlin, L. and Martz, M. J., Jr. Response probability
   and sensory-evoked  potentials.  In S. Kornblum
   (Ed.), Attention  and performance IV. Academic
   Press, New York,  1973:175-184.

 Karlin, L, Martz,  M. J.,  Brauth, S. E. andMordkoff,
   A. M.  Auditory evoked potentials, motor potentials
   and reaction time. Electroenceph. clin. Neurophys-
   iol., 1971,31:129-136.

 Karlin, L., Martz, M. J.  and Mordkoff, A. M. Motor
   performance and sensory evoked potentials.  Elec-
    troenceph. clin. Neurophysiol., 1970, 28:307-313.

 Klinke, R., Fruhstorfer, H. and Finkenzeller, P. Evok-
   ed responses as a function of external and stored
   information. Electroenceph. clin. Neurophysiol.,
    1968,25:119-122.

 KopeU, B. S., Wittner, W. K., Lunde, D., Warrick, G.
    and Edwards, D. Cortisol effects on averaged  evok-
    ed potential, alpha-rhythm, time estimation, and
    two-flash  fusion  threshold.  Psychosom.  Med.,
    1970,32:39-49.

 Kurtzberg, D. And Vaughan, H. G, Electrophysiologic-
    al  observations on visuomotor system and visual

-------
 Information Processing References
                                          237
   neurosensorium. In  J. E. Desmedt (Ed.),Visual
   evoked potentials in  man:  New  developments.
   Clarendon Press, Oxford, 1977:314-331.

 Kutas,  M. and  Donchin, E. Studies of squeezing:
   Handedness, responding hand, response force, and
   asymmetry of readiness potential. Science, 1974,
   186:545-548.

 Kutas, M., McCarthy, G. and Donchin, E. Augmenting
   mental chronometry: The P300 as an index of
   stimulus evaluation time. Science,  1977, 197:792-
   795.

 Lang, P. J., Gatchel, R.  J. and Simons, R. F. Electro-
   cortical and cardiac rate correlates of psychophysi-
   cal  judgment. Psychophysiol.,1975, 12:649-655.

 Leifer, L. J., Otto, D. A.,  Hart, S. G. and Huff, E. M.
   Slow  potential correlates  of  predictive behavior
   during a complex learning task. In W. C. McCallum
   and  J.  R.  Knott  (Eds.),  The  responsive brain.
   Wright, Bristol, 1976:65-70.

 Levit, R.  A. Averaged evoked potential correlates of
   information processing in schizophrenics, psychotic
   depressives  and  normals.  Doctoral dissertation,
   Columbia   Univ.,  1972.  Dissertation Abstracts
   International, 1972,33:4468.

 Levit, R.  A., Sutton,  S. and Zubin, J. Evoked pote-
   tial correlates of information processing in psychi-
   atric patients. Psychol. Med., 1973, 3:487-494.

 Levine,  S. and Hakerem,  G.  Pupillary dilation as a
   function of stimulus  uncertainty. Paper presented
   to the  Amer.  Psychol. Assoc., Symposium  on
   Attention (Human), 1969.

 Loveless, N. E. and  Sanford, A. J. Slow potential cor-
   relates of  preparatory  set. Biol. Psychol., 1974,
   1:303-314.

 Lovibond, S. H. Habituation of the orienting response
   to multiple stimulus sequences. Psychophysiology,
   1969, 5:435-439.

 Luna, A. R. Human brain and psychological processes.
   Harper and  Row, New York, 1966.

 Lynn, R. Attention, arousal and the orientation reac-
   tion. Pergamon, New York, 1966.
Mast, T. E. and Watson, C. S. Attention and auditory
   evoked responses to low detectability signals. Per-
   cept. Psychophys., 1968,4:237-240.

McAdam,  D. W.  Slow potential  changes recorded
   from human brain  during learning of a temporal
   interval. Psychonom. Sci., 1966, 6:435-436.

McCallum, W. C. Brain slow potential changes and
   motor responses  in  a vigilance situation. In W. C.
   McCallum and J. R. Knott (Eds.), The responsive
   brain. Wright, Bristol, 1976:46-50.
McCarthy, G. and Donchin, E. Temporal and event
   uncertainty and auditory event related potentials.
   Paper presented to the Society for Psychophysio-
   logical Research, Toronto, October 1975,

McGuigan, F. J. and Schoonover, R. A.  (Eds.). The
   psychophysiology  of thinking.  Academic Press,
   New York, 1973.

Naatanen, R. Selective attention and evoked poten-
   tials. Annals of the Finnish Academy of Science,
   1967,151:1-226.

Naatanen, R. Selective attention and evoked poten-
   tials  in humans-a critical review. Biol.  Psychol.,
   1975,2:237-307.

Naatanen,  R.  Evoked potential,  EEC and  slow
   potential correlates of selective attention.  Acta.
   Psychol. Suppl (Amst.), 1970,33:178-192.

Neisser, U. Cognitive Psychology. Appleton-Cenrury-
   Crofts, New York, 1967.

Nickerson,  R. S.  Same-different response  times: A
   model  and a  preliminary  test. Acta  Psychol.,
   1969,30:257-275.

Norman, D. A. Toward  a theory of memory and
   attention. Psychol. Rev., 1968, 75:522-536.

Oatman, L. C. Role of visual attention on auditory
   evoked potentials in unanesthetized cats.  Exp.
   Neurol., 1971,32:341-356.

Olds, J., Segal., M., Hirsh,  R.,  Disterhoft, J. and Korn-
   blith, C.  Learning centers of the rat brain mapped
   by measuring latencies of conditioned unit responses.
   J. Neurophysiol.,  1972,35:202-219.

Oleson, T.D., Ashe, J. H. and Weinberger, N. Modifica-
   tion  of auditory and somatosensory system activity
   during pupillary conditioning in the paralyzed cat.
   J. Neurophysiol.,  1975, 38:1114-1139.

Parasuraman, R. and Davies, D. R. Response and evok-
   ed potential  latencies associated with  commission
   errors in visual monitoring. Percept. Psychophys.,
   1975, 17:465-468.

Paul, D. D. and Sutton, S. Evoked potential correlates
   of response criterion in auditory signal detection.
   Science, 1972,  177:362-364.


Picton, T. W., Hillyard, S. A., Krausz, H. and Galambos,
   R. Human auditory evoked potentials. I: Evaluation
   of components. Electroenceph.  Clin. Neurophysiol.,
   1974,36:179-191.

Picton, T. W. and Hillyard, S. A. Human auditory evok-
   ed potentials. II: Effects of attention. Electroenceph.
   clin.  Neurophysiol., 1974, 36:191-200.

Picton, T. W,, Hillyard, S. A. and Galambos, R. Cortical
   evoked responses  to  omitted   stimuli. In M. N.

-------
 238
             Information Processing References
   Livanov (Ed.), Major problems of brain electrophys-
   ology. Nauka, Moscow, 1974:302-311.

Poon, L. W., Thompson, L. W. and Marsh, G. R. Average
   evoked potential changes as a function of processing
   complexity. Psychophysiology, 1976, 13:43-49.

Poon, L. W., Thompson, L. W., Williams, R. B., Jr. and
   Marsh, G. R. Changes of antero-posterior distribution
   of CNV and late positive component as a function
   of information processing demands. Psychophysiol-
   ogy, 1974, 11:660-673.

Posner, M. I., Klein, R., Summers, J. and Buggie, S. On
   the selection of signals. Memory  and Cognition,
   1973,1:2-12.

Posner, M. K. and Boies, S. J. Components of attention.
   Psychol. Rev., 1971, 78:391408.

Pribram,  K. H.  The new neurology and the biology of
   emotion:   A structural  approach.  Am. Psychol.,
   1967,22:830-838.

Prichep,  L.  S., Sutton, S.  and  Hakerem, G. Evoked
   potentials in hyperkenetic and normal children under
   certainty and uncertainty: A placebo and methyl-
   phenidate study. Psychophysiology, 1976, 13:419-
   427.

Raskin, D. C. Attention and arousal. In W. F. Prokasy
   and D. C. Raskin (Eds.), Electrodermal Activity in
   Psychological Research. Academic Press, New York,
   1973:125-155.

Raskin, D. C, Kotses, H. and Bever, J. Autonomic indi-
   cators  of orienting and  defensive reflexes. J. exp.
   Psychol.,  1969,80:423-433.

Remington, R. J. Analysis of sequential effects in choice
   reaction times. J. exp. Psychol., 1969,  2:250-257.

Remond, A.  Integrated and topological analysis of the
   EEC.  Electroenceph. clin.  Neurophysiol.,  1961b,
   20:64-67 (supplement).

Renault,  B.  and Lesevre, N, Variation des potentiels
   evoques par des patterns (P.E.P.) chez 1'homme dans
   une situation de temps de reaction:  Influence de la
   vigilance et delamotricite. Rev. EEC Neurophysiol.,
   1975,4:360-366.

Renault,  B.,  Ragot, R., Furet, J. and   Lesevre,  N.
   Etude  des  relations entre le "potential emis" et
   les mecanismes de preparation perceptivo-motrice.
   In J.  Requin  (Ed.) Fonctions  Anticipatiees  du
   Systeme  Nerveux   et   Processes   Psychologique.
   CNRS, Paris (in press).

Ritter, W., Simson, R. and Vaughan, H.G., Jr. Associa-
   tion cortex potentials and reaction time in auditory
   discrimination. Electroenceph. clin. Neurophysiol.,
   1972,33:547-555.

Ritter, W., Simson, R. and Vaughan, H. G. Topographic
   analysis of task-related cerebral  potentials.  In J. E.
   Desmedt (Ed.),  Prog,  clin Neurophysiol.,   1978,
   Vol. 6 Karger, Basel (in press).

 Ritter,  W. and Vaughan,  H.  G., Jr. Averaged evoked
   responses in vigilance and discrimination: A reassess-
   ment. Science, 1969, 164:326-328.

 Ritter, W., Vaughan, H. G., Jr. and Costa, L. D. Orient-
   ing and habituation to  auditory stimuli: A study of
   short-term  changes in  average  evoked  response.
   Electroenceph. clin. Neurophysiol,,  1968, 25:550-
   556.

 Ritter, W., Vaughan, H. G., Jr. and Friedman, D. A brain
   event realted to the making of a sensory discrimina-
   tion, (in preparation).

 Rohrbaugh, J. Cortical potentials associated with deci-
   sion  making.  Unpublished  doctoral dissertation
   Univ. of Illinois, 1973.

 Rohrbaugh, J. W., Donchin, E. and Eriksen,  C. W. Deci-
   sion making and theP300 component of  the cortical
   evoked  response.  Percept.  Psychophysiol   1974
   15:368-374.

 Rohrbaugh, J. W., Snydulko, K. and Lindsley,  D. B.
   Brain wave components of the contingent negative
   variation in humans. Science, 1976, 191:1055-1057.


 Roth, W. T. Auditory evoked response to unpredictable
   stimuli. Psychophysiology, 1973, 10:125-137.

 Roth, W. T., Ford, J. M. and Kopell, B. S. Event-related
   potentials and reaction  time fluctuation. (In prepa-
   ration.)

 Roth, W. T., Ford, J.  M., Lewis, S. J. and Kopell, B. S.
   Effects of stimulus probability  and task-relevance
   on event-related potentials.  Psychophysiology,
   1976b, 13:311-317.

 Roth, W. T. and Kopell, B. S. P300-an orienting reac-
   tion in the human auditory evoked response. Percept.
   Motor Skills, 1973, 36:219-225.

Roth, W. T., Kopell, B. S., Tinklenberg, J.  R., Darley,
   C.  F., Sikora, R.  and Vesecky, T. B. The contingent
   negative variation during a memory retrieval task.
   Electroenceph. clin. Neurophysiol.,  1975, 38:171-
   174.

Roth, W. T., Krainz,  P. L., Ford, J.  M., Tinklenberg,
   J. R., Rothbart, R. M. and KopeD, B. S. Parameters
   of temporal recovery of the human auditory evoked
   potential. Electroenceph. clin. Neurophysiol., 1976a
   40:623-632.

Ruchkin, D., McCalley, M. G. and Glaser, E.  M. Event
   related   potentials and time estimation. Psycho-
   physiology, 1977, 14:451-455.

Ruchkin, D. and Sutton, S. CNV and P300 relationships
   for emitted and  evoked potentials. In J.  Desmedt

-------
Information Processing References
                                            239
   (Ed.), Prog. clin. Neurophysiol.,  vol. 6, Karger,
   Basel, 1978 (in press).

Ruchkin, D. S. and Sutton, S. Visual evoked and emit-
   ted potentials and stimulus significance.  Bull. Psy-
   chonom. Soc.,  1973, 2:144-146.

Ruchkin, D. S., Sutton, S. and Tueting, P. Emitted and
   evoked P300 potentials  and variations in stimulus
   probability. Psychophysiology, 1975, 12:591-595.

Ruchkin, D. S.  And Sutton, S. Latency characteristics
   and trial by trial variation of emitted potentials. In
   J. E.  Desmedt(Ed.), Prog. clin. Neurophysiol., 1978,
   vol.  6 (in press).

Rutschmann, J. Time judgments by magnitude estima-
   tion and magnitude production and anxiety: A prob-
   lem  of  comparison  between normals and certain
   schizophrenic patients.  J. Psychol.,  1973, 85:187-
   223.

Schwent, V. andHillyard, S. A. Auditory evoked poten-
   tials  and multi-channel selective attention. Electro-
   enceph.  clin. Neurophysiol., 1975, 38:131-138.

Schwent, V. L., Snyder, E. andHillyard, S. A. Auditory
   evoked potentials during multichannel selective list-
   ening: Role of pitch and localization cues.  J. exp.
   Psychol.:  Human Percept. Perform., 1976,  2:313-
   325.

Shaknovich, A. R. On the pupillary components of the
   orienting reflex during action of stimuli specific for
   vision and non-specific (extraneous) stimuli. In L. G.
   Voronin et al. (Eds.), Orienting reflex and explora-
   tory  behavior. Am. Inst.  Bio.  Sci., Washington,
   1965:249-258.

Shannon, C. E. and Weaver, W. The mathematical theory
   of communication.  Univ.  Illinois  Press, Urbana,
   1949.

Siddle,  D. A. T. Overextinction of the evoked skin con-
   ductance response: An EEC study. Psychophysiol-
   ogy, 1974, 11:630-638.
Simernitskaya,  E.  G. Application  of the  method of
   evoked potentials to the analysis of activation pro-
   cesses in  patients with lesions of the frontal lobes.
   In K. H.  Pribram and A. R. Luria (Eds.), Psycho-
   physiology  of the  frontal  lobes. Academic Press,
   New York, 1973:29-51.

Simson, R., Vaughan,  H. G., Jr. and Ritter, W. The
   scalp topography of potentials associated with mis-
   sing visual or auditory stimuli. Electroenceph. clin.
   Neurophysiol., 1976,40:33-42.

Simson, R., Vaughan,  H. G., Jr. and Ritter, W. The
   scalp topography of potentials in auditory and vis-
   ual discrimination tasks. Electroenceph. clin. Neur-
   ophysiol., 1977,42:528-535.
Smith, S. G. and Sherwood, B. A. Educational uses of
   the PLATO computer system. Science, 1976, 192:
   344-352.

Snyder, E.  and Hillyard,  S. A. Long-latency evoked
   potentials to irrelevant, deviant stimuli. Behav.
   Biol., 1976, 16:319-331.

Sokolov, E. N.  Inner speech and thought. Plenum
   Press, New York, 1972.

Sokolov, E. N. Neuronal models and  the  orienting
   reflex. In M. A. B. Brazier (Ed.), The central ner-
   vous system and behavior.  Josiah Macy, Jr. Foun-
   dation, New York, 1960:187-286.

Sokolov,  E.N.   Perception  and  the Conditioned
   Reflex.  Pergamon Press, New York, 1963.

Squires, K. C. and Donchin,  E.  Beyond averaging:
   The  use of discriminant  functions to recognize
   event related potentials elicited by single auditory
   stimuli.  Electroenceph. clin. Neurophysiol.,  1976,
   41:449-459.

Squires, K. C., Donchin, E., Herning, R. I. and McCar-
   thy, G. On the influence of task relevance  and stim-
   ulus probability on event-related potential compon-
   ents.  Electroenceph.  clin. Neurophysiol.,  1977,
   42:1-14.

Squires, K. C., Hillyard, S. A. and Lindsay, P. H. Cor-
   tical potentials evoked by confirming and discon-
   firming  feedback following an auditory discrimina-
   tion. Percept. Psychophysiol.,  1973a, 13:25-31.

Squires, K. C., Hillyard, S. A. and Lindsay, P. H. Ver-
   tex potentials  evoked  during auditory signal detec-
   tion: Relation to decision criteria. Percept. Psycho-
   phys., 1973b,  14:265-272.

Squires, K. C., Squires,  N. and Hillyard, S. Vertex
   evoked  potentials in a rating-scale detection task:
   Relation to signal probability. Behav. Biol., 1975a,
    13:21-34.

Squires, K. C., Squires,  N. K.  and Hillyard, S. A.
   Decision-related cortical potentials during an audi-
   tory signal  detection task with cued observation
   intervals. J. exp. Psychol. Human Percept. Perf-
   form.,  1975b, 1:268-279.
 Squires, K.  C., Wickens, C. D., Squires, N. K. and
    Donchin, E. The effect of stimulus sequence on
    the waveform of the cortical event-related poten-
    tial. Science, 1976, 193:1142-1145.

 Squires, N. K., Donchin, E. and Squires, K. C. Bisen-
    sory stimulation: Inferring decision-related proces-
    ses from the P300 component.  J. exp. Psychol.
    (Human Percept.), 1977,3:299-315.

 Squires, N. K., Squires, K. C. and Hillyard, S. A. Two
    varieties of long-latency positive waves evoked by

-------
 240
            Information Processing References
    unpredictable  auditory  stimuli  in  man. Electro-
    enceph. ;lin. Neurophysiol., 1975,38:387-401.

 Suga, N. and Shimozawa, T.  Site of neural attenua-
   tion  of responses to  self-vocalized sounds  in
   echolocating bats. Science, 1974,  183:1211-1213.

 Surwillo, W. W. Some observations  on the relation of
    response speed to frequency of photic stimulation
    under conditions of EEC synchronization. Electro-
    enceph. clin. Neurophysiol., 1964, 17:194-198.

 Surwillo, W. W. The relation of response time variabil-
    ity to age and the influence of brain wave frequen-
    cy. In G. A. Talland (Ed.), Human Aging. Academ-
   ic Press, New York, 1968.

 Sutton, S. The specification of psychological variables
   in an average  evoked  potential experiment. In E.
   Donchin and D. B. Lindsley(Eds.), Average evoked
   potentials:  Methods,   results,  and  evaluations.
   NASA SP-191, Washington, D. C, 1969:237-298.

 Sutton,  S.,  Braren, M.,  Zubin, J. and  John,  E. R.
   Evoked potential correlates of stimulus uncertain-
   ty. Science, 1965, 150:1187-1188.

 Sutton, S.,  Spring, B. and Tueting, P. Modality shift
   at the crossroads.  In L. Wynne, R. Cromwell and
   S. Matthysse (Eds.), The  Nature of Schizophrenia:
   New Findings and Future  Strategies. Wiley, New
   York, 1978.

Sutton, S.,  Tueting, P.,  Zubin,  J.  and  John,  E. R.
   Information  delivery  and  the  sensory evoked
   potentials. Science, 1967, 155:1436-1439.

Tecce, J. J., Savignano-Bowman, J. and Meinbresse, D.
   Contingent negative variation and the distraction-
   arousal  hypothesis. Electroenceph.  clin. Neuro-
   physiol.,  1976,41:277-286.

Tecce, J. J. And  Scheff,  N. M. Attention reduction
   and  suppressed  direct-current potentials in the
   human brain. Science,  1969, 164:331-333.

Thatcher, R. W. Evoked-potential correlates of hemis-
   pheric lateralization during semantic information-
   processing. In S. Hamad, R. W. Doty, L. Goldstein,
   J. Jaynes and G. Krauthamer (Eds.), Lateralization
   in the nervous system.  Academic Press, New York,
   1977:429-448.

Thurmond,  J. B.  and Alluisi, E. A. Choice time as a
   function of stimulus dissimilarity and discriminabil-
   ity. Can. J.Psychol., 1963, 17:326-337.


Triesman,  A.  M.  Temporal discrimination  and the
   indifference interval:  Implications for a model of
   the internal clock.  Psychol., Monogr., 1963, Mon.
   77, No. 576.

Treisman, A.M. Verbal cues, language and meaning in
   selective attention. Am. J. Psychol., 1964, 77:206-
   219.
Tueting, P. Uncertainty and averaged evoked response
   in a guessing situation. Unpublished doctoral disser-
   tation,  Columbia Univ., Ann Arbor, 1968, No. 69-
   3097.

Tueting, P. and Sutton, S. The relationship between
   pre'-stimulus negative shifts and post-stimulus com-
   ponents of the averaged  evoked potential. In S.
   Kornblum (Ed.), Attention  and performance IV.
   Academic Press, New York, 1973:185-207,

Tueting, P. and Sutton, S. Auditory evoked potential
   and lift/no lift  reaction time in relation to uncer-
   tainty.  In W. C. McCallum and J. R. Knott (Eds.),
   The responsive brain. Wright, Bristol, 1976:71-75.

Tueting,  P.,  Sutton, S. and Zubin,  J. Quantitative
   evoked potential correlates of the  probability of
   events.  Psychophysiology, 1970, 7:385-394.

Vaughan,  H.  G.,  Jr.,  Costa, L.  D., Gilden,  L. and
   Schimmel, H. Identification of sensory and motor
   components of cerebral activity  in simple reaction
   time tasks. Proc. 73rd. Conv. Am. Psychol.  Assoc.,
   1965,  1:179-180.

Vaughan,  H. G., Jr., Costa, L. D. and Ritter, W. Top-
   ography of the human motor potential. Electro-
   enceph. clin. Neurophysiol., 1968, 25:1-10.

Vroon, P.  A.  and van Boxtel, A. Testing some impli-
   cations of the sensory physiological model  of the
   time sense. Psychol.  Forsch., 1972, 35:81-92.

Waldbaum, J. K., Sutton, S. and Kerr, J. Shift  of sen-
   sory modality and reaction time in schizophrenia.
   In M. L. Kietzman,  S. Sutton  and J. Zubin  (Eds.),
   Experimental  approaches to   psychopathology.
   Academic Press, New York, 1975:167-176.
 Walter, W. G., Cooper, R., Aldridge, V. J., McCallum,
   W. C. and Winter, A. L. Contingent negative varia-
   tion:  An electric sign of sensorimotor association
   and expectancy in the human brain. Nature, 1964,
   203:380-384.

 Warhonowiez, A. Evoked potentials and acoustic time
   error.  Electroenceph. clin. Neurophysiol.,  1974,
   36:442p.

 Weinberg, H., Walter, W. G., Cooper, R. and Aldridge,
   V. J.  Emitted cerebral events. Electroenceph. clin.
   Neurophysiol., 1974,36:449-456.

 Weinberg, H., Walter, W. G. and Crow, H. J. Intracere-
   bral events in human related to real and imaginary
   stimuli. Electroenceph. clin. Neurophysiol., 1970,
   29:1-9.
Weiskrantz, L.  Some  traps and pontifications. In L.
   Weiskrantz (Ed.),  Analysis of behavioral change.
   Harper and Row, New York, 1968, 415-429.

-------
Information Processing References
                                             241
Wickelgren, W.  O. Effects of walking and flash stim-
   ulation on click-evoked response in cats. J. Neuro-
   physiol., 1968,31:769-777.

Wilkinson, R, T. and  Lee, M.  V. Auditory evoked
   potentials  and selective attention. Electroenceph.
   clin. Neurophysiol.,  1972, 33:411-418.

Wilkinson, R. T., Morlock, H. C. and Williams, H. L.
   Evoked cortical response during vigilance. Psycho-
   nom. Sci., 1966,4:221-222.

Woody, C. D. Characterization  of an adaptive filter
   for the analysis  of  variable latency  neuroelectric
   signals. Mod. biol. Eng., 1967, 5:539-553.
Woodruff,  D.  S. Relationship among EEC alpha fre-
   quency, reaction time,  and age: A  biofeedback
   study. Psychophysiology, 1975, 12:673-681.


Worden, F. A. Attention and auditory electrophysi-
   ology. In E. Stellar and J. M. Sprague (Eds.), Pro-
   gress in Physiological Psychology, Vol. I. Academic
   Press, New York, 1966.
Zubin, J. and Sutton, S. Assessment of physiological,
   sensory, perceptual, psychomotoi, and conceptual
   functioning  in schizophrenic patients.  Acta Psy-
   chiat. Scand., 1970, 46:247-263 (Supp. 219).

-------
IV.  LANGUAGE
     Section Editor:

     Robert M. Chapman
     Center for Visual Science
     University of Rochester
     Rochester, New York, U.S.A.

-------
LANGUAGE  AND  EVOKED  POTENTIALS1
R.M. CHAPMAN
University of Rochester, Rochester, NY, U.S.A.
Introduction

   Although  the  study  of language and  evoked poten-
tials  (EPs)  is  relatively  young,  with  few  published
studies,  the  field  appears  to be gaining momentum
and  sophistication.  This  area is  particularly  difficult
to  evaluate  since  nearly  all  the  problems  of other
areas of EP  research must be considered,  along with
the linguistic ones.  Distinguishing language effects per
se  from both  lower-order  effects,  such as  sensory,
motor,  and artifactual, and higher-order effects, such
as general  states and cognitive processes, is difficult.
Thus, key  issues often  revolve around the question of
the specificity of EP differences that might  be related
to language.

   In  the  preconference  correspondence among the
panel  members  (listed  later), some  21  issues associ-
ated  with  relating language  functions to evoked po-
tentials were identified. Discussion in this section has
been organized  around six topics,  each of which  is
treated in  a short  essay  following this brief overview.
The topics  are (1)  Issues in Neurolinguistics: Evoked
Potential Analysis  of  Cognition  and  Language, (2)
Distinguishing  Linguistic  and  Stimulus Effects, (3)
Lateral  Asymmetry  of  Evoked  Potentials  and  Lin-
guistic Processing, (4) Individual Differences and Simi-
larities in Language Effects on EPs,  (5) Contributions
of  Linguistics and  Other Data Bases, and (6)  Methods
of EP Analysis in Linguistic Research.

   Following  these essays, three papers on specific
research  related  to language and evoked potentials are
presented:  (1)  a study  of semantic versus lower  order
effects  of linguistic  stimuli by  A.L. Megela  and T.J.
Teyler, (2) a study of response to language stimuli  in
children  with  reading  disabilities by S.A.  Shelburne,
and (3)  a  study of the  effect of electrode placement
and recording  montage on CNV amplitude  preceding
speech  by  S.H.  Curry, J.F.  Peters,  and H. Weinberg.
Panel members and mini-reviews of their
research

   Members  of  the  panel on language  and evoked
potentials were:

   1.  W.S. Brown, University of California, Los Angeles,
      California.
   2.  R.M.   Chapman   (chairman),   University   of
      Rochester, Rochester, New York.
   3.  D. Friedman,  New York State Psychiatric Insti-
      tute, New York, New York.
   4.  D. Molfese, Southern Illinois University, Carbon-
      dale, Illinois.
   5.  S.A. Shelburne,  Jr.,  University  of Cincinnati,
      Cincinnati, Ohio.
   6.  T.J.  Teyler,  Harvard   University,  Cambridge,
      Massachusetts.
   7.  R.W. Thatcher,  New York  University School of
      Medicine, New York, New York.

   At the beginning  of the language  session,  each
panelist  briefly  reviewed his research  to  provide  a
common basis  for discussion  and  to focus attention on
specific issues.2 Mini-reviews  are useful here also, since
the essays focus on the same issues and refer to relevant
aspects of research only as examples.

   Shelburne has been studying reading disabilities with
the  aim  of developing  diagnostic neurophysiological
techniques. He visually presents letter stimuli to subjects
whose  task is  to indicate whether  the stimuli make  a
word or  nonsense syllable. The stimuli  are formed of
three letters presented at 1-sec intervals: blank, letter,
letter, letter, blank.  For each word there is a match-
ing   nonsense  syllable  that  has  the  same  first  two
letters. Thus,  the subjects cannot distinguish between
word and nonsense  syllables until the  third stimulus
is flashed.
 'Supported in part by the National Institutes of Health Research
  Grant 5 RO1 EY01593 to Robert M. Chapman.
 7 A transcript of discussion that took place at the meeting has
  been prepared and is available from the Chairman at cost.

-------
246

    Each trial run included a randomized presentation of
 50  words and  50  matched  nonsense syllables. When
 subjects achieved a score  of 95  percent  or more, a
 clearcut difference  emerged between  EPs to the  third
 letter and  EPs to the first or second letter and a marked
 difference was evident between third and first position
 responses. The  latter was probably  a  P300 effect,
 perhaps related to  resolution  of uncertainty. Children
 with reading disabilities scored near chance level of 50
 percent and  did not  show  the   third position-first
 position EP effect.  Normal children too young to read
 did not exhibit  the EP  effect. Children with reading
 disabilities have  the following profile: (1) poor task
 performance in older child (9 to 12 years), (2) no visual
 EP  difference between the final and  first two stimuli
 within  a trial, and (3)  low  level of  eye  movements
 (EOG). Normal children  show considerable eye move-
 ment during this task.

   Molfese  has   been  especially  concerned  with EP
 aspects related to acoustic and speech sounds. Speech is
 composed  of acoustic stimuli organized with particular
 characteristics. Molfese has studied parts of the brain
 response  that  reflect  changes  in  a  consonant-vowel
 syllable versus a vowel syllable, with non-speech controls
 of narrow bandwidth that do not sound like speech. He
 has  investigated a variety  of subjects along the develop-
 mental  continuum. The data from neonates (only a few
 days of age) are particularly important. Using principal
 components analysis, EP components  related to these
 speech stimuli differences were identified. For example,
 one component seems to differentiate  between steady-
state stimuli  and   stimuli  with transitions.  Another
component appears  to be related to the bandwidth  of
 the auditory stimuli.

   Brown has studied EP correlates  of contextual mean-
ing of words  using  auditory presentation. He employs
homophone stimuli, i.e., words that sound alike but have
different meanings  depending on context. Brown pre-
sents these words in a  phrase or sentence context that
gives them specific meaning and then separately averages
the  EPs  to   the  same  sounding  word in  different
contextual  meaning phrases.  For  example,  "rose" is
studied  in such phrases as:  "a pretty rose,"  "chairs in
rows," "his temperature rose," and "a boatman rows."
For each electrode locus, he  compares the EP waveforms
derived  from the four meanings. Brown has used three
statistical methods:  (1) coefficient of correlation be-
tween EP waveforms, (2)  discrimination index based on
correlation coefficients, and (3) stepwise discriminant
function analyses. He  consistently  found  that over the
left  hemisphere, particularly  from the  left  anterior
electrode  close to  Broca's  area, different average EP
waveforms were  associated  with different word mean-
ings.  Over  the right  hemisphere the EP waveforms for
different word meanings were more similar. In these
experiments, the  subjects do not speak; therefore, this
 technique  is  not comparable to   phonation. The EP
                                          Chapman

differences  were related to whether  the stimuli were
ambiguous  in  meaning,  which  was  manipulated  by
order of  phrase presentation and  instructions  to  the
subject.

   Teyler has explored semantic  meaning effects using
verb/noun  homophones. In  one  phase, subjects were
instructed to think of the word when  a  click probe was
presented approximately a  second after the subject
heard  the   word.  Waveshapes of EPs  were distinctly
different for verb and  noun meanings  associated with
physically identical stimuli, but were  quite similar over
both hemispheres for a given  word meaning. Different
subjects  appeared  to have relatively  unique EP wave-
forms for a  given word, and these were  relatively stable
across experimental conditions. A positive peak at about
180 msec tended to  have a shorter latency for the verb
meaning. Subsequently, Teyler explored an habituation/
generalization paradigm in which a word was repeated
seven  times  and then  either  the same  or a different
stimulus  word was  presented.  Later  parts  of the  EP,
predominantly P300, were very sensitive to the general-
ization  task.  For  semantically  similar  stimuli,  e.g.,
synonyms  like  "couch"  and "sofa,"  EPs  from  the
Wernicke's  area derivation did  not  dishabituate.  By
contrast, EP dishabituation was found when the stimulus
word  was changed to semantically different words that
were physically  rather  similar, e.g., "cut" and  "cup."
Frontal derivations were insensitive to manipulation of
linguistic stimuli.

   Thatcher  employed  a  delayed  semantic matching
paradigm using synonyms, antonyms, and neutral words
presented visually. The control stimuli were random dot
patterns. For example,  a  trial might contain control
stimuli and the words "small" and "little." The subject's
task was to  indicate the semantic category of the  second
word  in  relation to  the first. The second word is  a
synonym a third of the time, an antonym a third of the
time, and neutral a third of the time. This experiment is
designed  to extend  the  match-mismatch  paradigm be-
yond  stimulus features to a semantic  level. EP asym-
metries were found to second words, particularly in the
400- to 500-msec range (P400). EP asymmetries to  the
random  dot  stimuli were  not  obtained.  Significant
differences  in P400  amplitude   were  found between
antonyms and neutrals, between synonyms and neutrals,
and between second and first words.  The effects were
maximal in posterior  rather than anterior regions.

   Friedman reported work with two kinds of paradigms
using word stimuli, a vigilance paradigm and sentences
with one word  incomplete. In the  vigilance paradigm,
the subject  was asked  to detect a designated  target
stimulus among a group of auditory stimuli. The stimuli
were  words, e.g., "pint,"  "bowl," and "kick,"  and
human-generated, nonlanguage sounds, e.g., whistle,
cough, and "psst." Another condition was a nontask in
which the subject  listened passively.  P300 amplitudes

-------
Language EPs
                                                247
from  smallest  to largest  were obtained  for no-task
stimuli,  nontarget  stimuli,  and  target stimuli.  This
occurred equally for words and human-generated sounds.
Although this paradigm used words, it  did not require
engagement of linguistic processes, since the task could
be  successfully  performed on  the basis  of acoustic
characteristics.

   Friedman and coworkers then used  a paradigm in
which subjects had to use meaning in order to complete
the task. Sentences in which the first grapheme of one
word  was  missing  were presented visually.  The basic
sentences were: "The heel  is on the shoe." "The peel is
on the orange." and "The wheel is on the axle." In one
condition,  the  first grapheme of the second word  was
omitted, and so the  subject saw  "eel" in  all  cases.
Following  the  last  word  of the sentence, the subject
indicated what the second  word was on the basis of the
context  provided by  the  last word.  In the  second
condition,  the second word was presented in full so  that
the last word provided no  further task-relevant informa-
tion. Whenever a word delivered information, i.e., last
word  in the first condition and  the second word  in the
second condition, the peak latency of P300 was longer.
Another finding was that P300 was always largest for the
last word  in  the  sentence regardless  of  whether  it
delivered critical information. This effect  was  called
"syntactic closure."

   Chapman  has studied EP effects related to a particu-
lar  area of semantic  meaning, i.e., Osgood's (1964)
analysis  of connotative  semantic meaning.  From Os-
good's analyses, specification of connotative meaning of
words  has  emerged on  three  orthogonal  dimensions:
evaluative, potency, and activity (E, P, and A). Words
that belong to six semantic classes (E+, E-,  P+, P-, A+,
A-) were selected as the  extremes  on these semantic
dimensions. For example, three words that belong to the
E+ class are "fresh," "pleasant," and "quality." At the
opposite end of the evaluative dimension, examples of
the E- class are "enemy," "tragedy,"  and "devil." For an
experimental list, 20 words in each of these six parts of
connotative semantic space were selected. On each trial,
the sequence was a fixation stimulus, followed 0.5 sec
later by a  stimulus word. The task at the simplest level
was  to  repeat the  word  late  in the  2.5-sec interval
following the presentation  of each word. Stimulus words
were randomly intermixed so that neither semantic class
nor  particular word could be anticipated.  EPs were
averaged across the 20 words in each  semantic class.

   Multivariate  analyses  revealed  reliable  EP  effects
related to semantic classes.  These EP  analyses were based
on standardizing the data within each subject's data set,
computing a principal components analysis on the entire
set of data,  and  using the component scores as the EP
measures entering  the multiple discriminant analyses.
The effects were  consistent  enough  to  classify  EPs
belonging to the various semantic classes by discriminant
analyses, which used  the  same classification functions
for all subjects. When classifications to opposite ends of
one semantic dimension at a time were made, the overall
success  rate was 97%. The jackknifed cross-validation,
which leaves each EP out of the development set and
then classifies  it, has an overall success rate of 90%. Two
lists of words  were used so that the generality of the EP
effects could  also be assessed by applying classification
functions based  on data from one  word list to data
obtained with the other word list. The overall other-list
cross-validation success rate was 73%. Additional multi-
ple discriminant analyses were computed to classify EPs
into all six semantic classes at once, in which case chance
is   16.7%. For  these multidimensional  analyses, the
overall success rate was 55%, jackknifed cross-validation
was 42%, and other-list cross-validation was 40%. These
success rates were more than twice chance level.

Concluding remarks

   Because  there  is  no  universal agreement on the
definition of linguistics or linguistic theory, seeking EP
relationships may appear to be  a severe problem. The
status, however, is similar to that of other fields, such as
learning, for which agreement on definition and theory
is  lacking. Nevertheless, many instances  of learning are
agreed upon,  and detailed effects of learning  variables
have been studied. Psycholinguistics is a relatively young
field, but research has been sufficient  to promote a body
of linguistic data that could form the basis of related EP
research. It is not necessary to wait for linguistics to get
its house in order before embarking on  EP research; in
fact,  the  results of EP research may  help delineate and
constrain linguistic theory. Many on the panel believe
that applying  some aspects of these linguistic data would
be helpful  in  EP research rather  than simply using
common-sense notions of linguistics.

   Two  major  categories of  language  processing  are
production  and  reception.  Both categories present  a
similar  problem of  distinguishing language processing
from the more  peripheral effects of motor and sensory
processes. If language production were invariantly linked
with  particular  motor processes and  language reception
with  particular  sensory  processes, language-specific ef-
fects in  EPs would  be  very  difficult to  distinguish. At
some levels of linguistic  analysis there is evidence that
such  linkages are not invariant. Thus, disassociation of
linguistic effects from sensory and motor effects may be
possible in EP research. Many linguistic distinctions have
alternative representations with which to vary sensory/
motor variables  independently of linguistic variables.
    One  of the strongest  supports from non-EP  re-
search  for differentiating language-specific  effects  in
 EPs  comes  from  findings  of cerebral  localization  in
linguistic studies of brain damage and  dichotic studies
 in normals.  As  the result of this type  of research,
 emphasis has been  placed on  lateralization  and local-
 ization  in  EP-language  research. This  heavy  emphasis
 on place may have detracted from interest in studying

-------
248
functional  relationships  between  EPs  and  linguistic
variables themselves.

   Linguistic  variables  that might  be studied include
phonemes,  graphemes,  morphemes,  syllables,  words,
phrases,  syntax, and semantics. Beginnings have been
made in many  areas. In some studies a linguistic  and
nonlinguistic  variable  are  compared,  and  in  others
variations of  linguistic  variables themselves are investi-
gated. To recommend  a  single type  of experimental
strategy  or  linguistic  variable  would  be  premature.
Rather, the  recommendation is to design studies that are
based on information  available in  linguistics, psycho-
linguistics, and psychology with careful consideration of
various kinds of controls. These include consideration of
sensory and motor controls, as well  as general cognitive
and state controls.

   EP effects  have been related to  various aspects of
linguistics. Research varies considerably in sophistication
and how convincingly the EP effects may be uniquely
tied  to the linguistic variables  being  manipulated. In
many cases, alternative  explanations may  be  found in
terms of sensory differences in stimuli, different states
of the subject, or different cognitive functioning. This is
a common situation  in science; evidence supports hy-
potheses while alternative  hypotheses may account for
the same  data.  A great deal of careful research may be
needed  to assure that a particular effect  is reasonably
interpreted as  a linguistic effect rather than the result of
other variables. Amassing  data of essentially  the same
type can be used to assess the reliability of particular EP
effects,  but it is of little use in assessing the validity of
the interpretation.

   In order  to  ascribe an EP effect  to a  linguistic
variable, different strategies are needed. One strategy is
to systematically relate EP  effects to an established
conceptual  framework,  i.e.,   intralinguistic  variation
within the framework provided by one of the well-tilled
subfields  in  linguistics. This  strategy  suggests a  sys-
tematic approach to  EP measurement in which specific
parts  of the EP can be identified, measured, and related
to linguistic variation. The more detailed the meshing of
EP data with linguistic data is, the more convincing the
linguistic interpretation.

   Another related strategy is based on converging lines
of evidence. If a particular linguistic  variable is known to
affect communication  or  behavioral  performance in
predictable ways in several different situations, one can
study  the generality of  related  EP  effects in these
different  situations. One example would be to find an
EP effect in normal adults that appears to  be  related to
function words such as  articles and prepositions. This EP
effect could  then be  tested  in patients with Broca's
aphasia, a  feature of  which is the omission of these
function  words.  A further line of evidence  could be
sought in children's data, since these function words are
                                          Chapman

omitted  during certain stages  in  language  acquisition.
An  interesting  aspect of  this  example  is  that  the
developmental  order  of appearance of  these words in
a vocabulary is different from the difficulty ordering
in Broca's  aphasia adults.

   A  third strategy  to foster  convincing linguistic in-
terpretations  of particular EP effects  is to  use sys-
tematic control procedures. One  could design experi-
ments  to hold  linguistic effects constant while varying
the confounded variable.  If EP changes are  observed
in this design,  the  question  is whether  the  part  of
the  EP  that   changes  is  different   from  the  part
ascribed  to the  linguistic variable. For example,  the
same  semantic  meaning may  be  carried  by  different
words  so  that  the  confounded  variable  of  different
physical  stimuli could  be  varied  to test for  the  speci-
ficity  of  particular  semantic  meaning effects.  The
converse procedure  of holding the confounded  vari-
able  constant  while  varying  the   linguistic  variable is
also  possible. For example, some physically  identical
words  have quite distinct  semantic meanings as in the
research  on  homophones.  Another  example,  illus-
trating a different kind of confounded variable, is to
hold   the  subject's   task  constant,   e.g.,  a match-
mismatch  task, while  varying  the linguistic  elements
of paired presentations.

   The specificity of  language  effects depends in part
on the dimensionality of EP  measures. For  example,
if only a  single  measure,  say amplitude of apparent
P300,  were  extracted  from EPs, then demonstrating
specificity  may  be  hopeless. Let us pretend  that
nouns  and verbs elicit different amplitude P300s. One
might  argue  that this is not a specific language  effect
because  P300  amplitude  is also  modulated  by  other
variables, such  as stimulus uncertainty. This  situation
would be  very  similar to  finding  a difference  in pupil
dilation  to nouns and verbs;  one would  hardly argue
that the pupil  was  directly involved (a necessary link
in the causal   chain)  in  the   linguistic processing of
nouns and verbs. One might  still find it useful, how-
ever, to utilize the pupil measure (or P300  measure)
in  studying linguistic processing  since  a  reasonable
interpretation  is  that the linguistic  processes per se
systematically   influence  these general processes.  If
there  were particular  parts of EPs that  related more
specifically to  noun-verb  processing,  they  would not
be found  by simply measuring P300 amplitude in the
manifest waveform. It would  be  necessary to use EP
measures that  focus on purely linguistic parts of EPs.
Possible  solutions include:  (1) to use the  difference
between  EPs   with  and  without the particular lin-
guistic processing  and  (2)   to  use   multivariate  sta-
tistical analyses,  which potentially take  into account
all  of the  time points within  EPs  as well  as  their
relationships.  To separate  a   noun-verb  part from  a
stimulus-uncertainty   part  of  the EP, at  least  two
measures of the  EP that respond differently to  these

-------
Language EPs                                                                                     249

two  kinds  of variables must be made. These measures    interpretations of specificity are limited to the dlmen-
must vary  independently or at least  be manipulated    sionality  of the EP measures and the dimensionality of
to different  degrees.  Thus,  the dimensionality of the    the experimental design.

-------
 ISSUES  IN  NEUROLINGUISTICS: EVOKED-POTENTIAL
 ANALYSIS OF COGNITION  AND  LANGUAGE
 R.W. THATCHER

 Department of Psychiatry, New York University School of Medicine, New York, NY, U.S.A.
 Introduction

   Currently, there is controversy and confusion about
 the presence  or absence, as well as the exact nature of
 the distinctions between cognition and language. Some
 authors argue that there  are no  distinctions (Menyuk
 1975);  others argue that  cognition is  dependent on
 language development (Blank  1975). Still others  con-
 sider language secondary to the development of cogni-
 tion (Furth  1975). Much  of the present controversy is
 due to differing definitions of  cognition and language.
 No clear agreement on  exactly what the word "cogni-
 tion" means has  been  decided  upon. This lack of
 agreement seems to be due, in part, to the fact  that
 different investigators have different perspectives. For
 instance, developmental psychologists generally  define
 cognition  in  terms of the development from birth to
 adulthood of skills such as assimilation, object perma-
 nence, conservation, and development of logical opera-
 tions (Piaget  1971a, Langer 1974). Speech pathologists
 and linguists tend to define cognition in terms of specific
 mental abilities with less emphasis on age dependence,
 such as adult thought patterns, symbolic representations,
 and mathematical function  (Rieber 1975). On the other
 hand, cognitive psychologists  tend  to use a broader
 definition  such as  that offered by Neisser (1967) as
 "processes  by which sensory  input is transformed,
 reduced,   elaborated, stored,   recovered,  and  used."
 Finally, neurophysiologists have yet to enter into this
 area in a meaningful way and thus have not attempted to
 formulate a clear definition.

   This section of the EPIC IV Conference is concerned
with evoked-potential correlates of language and cogni-
 tion; however, the  field of neurolinguistics, which  is in
its infancy, has not adequately come to grips with the
problem of defining cognition. For  want of a  better
 procedure,  therefore, the  broader definition  used by
cognitive psychologists will be adopted for purposes of
 the present discussion. This strategy is justified on the
 basis that  sensory, memory, and motor transformations
hold  a  fundamental position  in all of  the current
 definitions of cognition  and  yet  do  not preclude
 elaborations performed on these elementary functions.
 Furthermore, emphasis such as Neisser's (1967) on the
 concept  of transformations helps  facilitate  building a
 workable bridge  between  cognitive psychology  and
 neurophysiology (Thatcher and John 1976).

   Given  this definition  of  cognition, it seems that
 language is  a subordinate class of cognition in that it
 involves  the initial  transformation of sounds and pat-
 terned  visual  information   (phonemes,  morphemes,
 words, sentences) into elementary internal representa-
 tional systems. Subsequent manipulations of these repre-
 sentations (e.g., comparisons to memory, semantic cate-
 gorization, and the  use of syntactical  structures) are
 performed to create more complex levels of information
 representation.

 Neural representational systems

   It is generally agreed that environmental information is
 first transduced at the peripheral sensory level into neural
 space-time  activity,  which is subsequently transformed
 into internal representational systems (Perkel and Bullock
 1968, Szentagothai  and Arbib 1974). That is, there are
 general mapping processes in advanced organisms by which
 elements in the external world are represented or mapped
 onto elements  in the internal milieu.  In terms  of
 information theory,  a representational system is defined
 as any structure for which the elements are mapped onto
 symbolize, or correspond to  the elements of some other
 structure (MacKay 1970). The physiological mechanisms
 underlying the formation of neural representations are not
 completely known; however, recent symposia published
 in the Neuroscience Research Programs have  reviewed
 this  area and have  attempted to establish conceptual
 guidelines (Szentagothai and  Arbib 1974). For instance,
 Szentagothai and  Arbib  proposed  an action-oriented'
 hierarchical organization of perceptual and  memory
 processes in  which an animal  actively synthesizes new
 models and updates old models of the external world. In
 this  regard, the  updating process most likely involves
matching and mismatching of representational models
with sensory input (Thatcher  1976).

-------
Issues in Neurolinguistics
                                                251
   Changes in the early components of the human-scalp
 evoked  potential often correspond to changes in the
 physical attributes of sensory stimuli such as intensity,
 contour, density, color, and spatial  frequency (Regan
 1972). The longer latency components of the human-
 scalp EP are often insensitive to the specific content of
 evoking  stimuli  provided  subjects are  attentive  and
 processing information. For this reason the EP is often
 partitioned  into an  early  or  "exogenous" process,
 sensitive to the physical features of sensory stimuli, and
 later or "endogenous" process, sensitive  to higher-level
 functions related  to  attention,  decision  making,  and
 second-order information processing.

    Linguistic information  processing can be understood
 as afferent information transformed by specialized nerve
 endings  into coded  impulses  that are conducted  cen-
 trally.  The  physical  features  of the sensory stimulus
 (phonemes in the case of auditory input and lines, edges,
 angles, etc., in the case of visual input) are mapped onto
 a neural representional system. Various neurophysiologi-
 cal models can be used to explain the mapping process
 (Hubel  and Wiesel 1962,  Pribram  1971,  Barlow 1972).
 Most  of these models rely  on feature extraction. That
 is,  neural  elements  respond optimally only to a par-
 ticular feature of the sensory stimulus. Feature extrac-
 tion models are  usually  hierarchical (Barlow 1972) in
 that there is a hierarchy of levels of greater complexity
 both of the feature extraction and the representational
 systems. There are a  few nonfeature extraction models
 (see Gibson 1969, Schwartz 1977) in which the anatomy
 of  the brain  uniquely  determines  mappings of  the
 external world onto neural representational systems. For
 instance, Schwartz (1977) elucidated a conformal sen-
 sory map based explicitly  on anatomy. He demonstrated
 that a logarithmic spiral on the periphery (on the retina
 or skin surface) is mapped as a straight line in the cortex.
 According to this model, perceptual phenomena such as
 size invariance and rotational invariance are inherent to
 the  way peripheral  receptors  are  connected to  the
 cortex. In  any case, whether feature extraction models
 or nonfeature extraction  models eventually prevail, the
 concept of transformation of environmental patterns of
 energy  into internal  representations is fundamental to
 the field of neurolinguistics as it  pertains  to cognition
 and language.

 Formal operations versus content of
 operations

    In order to understand evoked potential analyses, it is
 important  from the  outset to distinguish  between the
 neural  representation of information and  the neural
 operations performed on  that information. This distinc-
 tion is emphasized by Piaget (1971a, b) in his discussion
 of adaptation and the development of formal operations.
 In  this context  a formal  operation such as logic can be
 invoked in a variety  of circumstances, all of which may
 involve  different  information inputs (e.g., there are
numerous  possible  syllogistic  statements). Thus,  the
formal aspects of an operation, such as those involved in
logic, are invariant over time, whereas the content being
operated upon  changes. In language, grammatical struc-
ture can  be likened  to a  formal  operation, whereas
meaning  may  be  considered the  content.  Another
example is the act of remembering versus the content of
the memory. Similar distinctions can be made for a wide
range  of cognitive  operations  such as,  imagination,
information comparison, image rotation, linguistic trans-
lation,  and  mathematical  operations.  At any  given
moment the operation and the content of the operation
are inseparable; but these two  features can be distin-
guished experimentally by the fact that the  content
varies over time, whereas the operation is invariant. This
distinction  is particularly relevant to the study of the
electrophysiology of language  since  operations involving
information manipulations  and comparisons  can con-
tribute  to recorded electrical  activity. A  primary ques-
tion in  evoked-potential studies is: does an electrophysi-
ological event reflect the specific content of an  opera-
tion,  or  does  it reflect a general  operation  that  is
invariant while content changes?

Primary versus secondary representational
systems

    Chomsky's (1965) elegant formulation of transformal
grammar  led to  a hypothesis of universal  linguistic
capacities, i.e., cognitive capacities  that are specific and
unique  to  language.  Alternative  views, however,  are
concerned with universals advocating universal cognitive
capacities  that are fundamental to all perceptual and
cognitive experience but are not specific to language per
se  (Thatcher  1976,  Greenfield 1976).  An exhaustive
analysis of these contrasting views is beyond  the scope
of this summary; however, a  useful approach  to under-
standing  this issue is to first consider  the distinction
between primary representational systems and secondary
representational systems.

    Primary representational systems for  both audition
and vision  are  defined as the  internal neural representa-
tion  depicting  the elementary features of a  sensory
stimulus (Thatcher 1976, Thatcher and John 1976). For
example, natural speech stimuli are made up of vowel
and consonant  sounds, which form phonemes.  In this
case  ti'e primary  representational  systems reflect the
processes  by  which  the elementary features  of speech
stimuli (phonemes) are received.

    Secondary neural representational systems map the
larger and  more meaningful units  of languages such as
 morphemes, words, phrases and sentences. In this regard
 the secondary  representational systems reflect the more
holistic units of language. Another distinction  is that the
 primary  systems involve more phylogenetic histories
 (i.e., anatomical and inherited), whereas the  secondary
 systems involve more experiential history or memory.

-------
 252
                                          Thatcher
    Are there primary representational processes specific
 or unique to language? This question is  answered, in
 part,  by  studies using artificial synthesis  of speech
 sounds (Liberman et al. 1967, Liberman 1970). Work by
 Liberman  and his colleagues at the Haskins Laboratory
 revealed two Important aspects of  speech perception.
 First, speech sounds are  perceived  categorically.  In
 nonspeech conditions a human observer can discriminate
 among many more stimuli than he can label or identify,
 but in speech perception this does not occur. That is,
 speech discrimination is  no better  than  identification
 (see Lliker and Abramson  1964b). A second discovery
 by the Haskins group (Liberman et al. 1952, 1967) was
 that speech sounds violate the assumptions of perceptual
 linearity  or invarlance. That  is, an invariant sound
 segment can represent different phonemes in different
 contexts  or, conversely,  a  single  phoneme can  be
 represented by  different  sounds in  different contexts.
 This work shows that the basic units of language involve
 representations with unique characteristics.

    Recent  EP  analyses  have  provided  correlates  of
 categorical perception in neonates (Molfese et al. 1976).
 The latter  studies demonstrate that  neonates exhibit
 different average EPs to different speech sounds as well
 as marked hemispheric asymmetries. These studies, when
 considered as a whole, support the notion that language
 perception  involves  unique  primary representational
 systems and that these systems are innate and functional
 at birth.

   To what extent does language involve unique secon-
 dary  representational systems? There is  no  definitive
 evidence for unique secondary representations as yet. It
 is  clear that the left  cerebral hemisphere is  more
 important  in  language comprehension and production
 than  the  right  cerebral  hemisphere.  It  is  not  clear,
 however, whether this reflects lateralization of universal
 cognitive  operations, unspecific to  language, or  the
 lateralization  of language operations  themselves.  For
 example, damage  to  the  left temporal or left parietal
 lobes in adults  results in  a  wide number  of cognitive
 deficits and  not simply  linguistic  ones (Luria 1966,
 1973; Konorski  1967).  As noted by Luria (1973) and
 Konorski  (1967), left parietal-temporal  damage inter-
 feres  with general analytical functions  including  the
 ability to abstract, to perceive complex patterns, and to
 maintain auditory  sequences. In line  with these  con-
 siderations, Thatcher (1976) and Thatcher  and  April
 (1976) suggested  that  a  universal  aspect of analytic
 function such as logic, language and  mathematics in-
 volves memory match and  mismatch. Some support for
 this was provided in an EP study wherein hemispheric
 asymmetries  appeared  to  be  related to  the general
 operation of memory match-mismatch and  not language
content  itself  (Thatcher  1977b).  Another  universal
cognitive  function,  suggested  by Greenfield  (1976),
involves the structural principles of hierarchical organiza-
 tion.  A hierarchical organization is an organization of
levels whereby lower level units or subordinates combine
to  form  higher  level units or superordinates. Many
aspects  of emotion,  perception,  and memory  are
hierarchically organized.  Language also exhibits  a liier-
archical organization whereby phonemes combine  to
form morphemes and  morphemes combine  to form
sentences;  or in  language  development, the  earliest
phonological units precede the first morphological units,
which, in turn, precede the syntactic one.

   Thus, the relationship between language  and cogni-
tion it  multileveled. Unique  language  representations
may operate at the primary  sensory level. On the other
hand, it is likely that semantic and syntactic language
functions are not unique capacities, but involve universal
cognitive operations.
Evoked-potential correlates of language
information processing
Lateralization and localization

   Evidence  for  anatomical  localization  of linguistic
information  processing  is  provided  in several  studies.
Brown et al. (1973,  1976) showed  that EPs recorded
from F3 and T3 (left  frontal  and left temporal) exhibit
different waveshapes in response to physically identical
words  when the words are  used in different speech
contexts.  Evoked-potential waveshape differences were
significantly greater when recorded from the left hemi-
sphere than  from the right hemisphere  (Brown et al.
1976).  The  waveshape  differences  were not  due to
increased variance and were stronger from F3 than from
T3 derivations.  It is not clear in these studies, however
whether  hemispheric  asymmetries reflect lateralization
of unique language operations or a  universal cognitive
capacity  unspecific to language. Teyler et al. (this
volume) showed maximal EP differences in T3 and not
F3 to  similar  sounding  words  with  very different
meanings. These differences were  reduced to words that
conveyed the same meaning but  had different features
such as words formed by upper- and lower-case letters
(e.g.,  DOG  versus  dog). These  findings suggest  that
invariant EP waveforms from T3 reflect invariant se-
mantle Information. Thatcher (1976) demonstrated EP
amplitude  and  waveshape asymmetries  to physically
identical words representing synonyms or antonyms in a
delayed semantic matching task. The interhemispheric
asymmetries  were  maximal  in  posterior T5/6, P3/4
O1/2)  regions but absent in anterior regions. A similar
anterior-posterior localization of lateralized EP processes
was noted in a paradigm requiring subjects to extract
identical meanings from words presented in  two differ.
ent languages (Thatcher 1977b). As mentioned previ-
ously,  it is  unclear whether  these asymmetries reflect
unique   language  capacities  or  universal  cognitive
operations.

-------
Issues in Neurolinguistics
                                                253
Correlates of syntax

   Syntax refers to  the way in which  words are put
together  to  form phrases  and  sentences. Brown et al.
(1973) examined EP waveforms elicited by identical
words  that terminated different phrases. For example,
they examined EPs elicited  by the  word "fire"  in the
phrases "sit by the fire" and "ready, aim, fire." The verb
produced a different EP than the same  word used as a
noun. Teyler et al.  (1973) used a similar paradigm but
examined EPs elicited by click probes that followed the
word presentation. Teyler  et al. (1973) reported that EP
amplitude and waveform differences were dependent on
the  meaning  and  syntactic context  of the  preceding
word.  Recently, Brown  et al.  (1976) used  an  EP
paradigm in which both the  trigger word and the  phrase
were the same, e.g., they used the phrase "it was/led/."
In this case  the subjects  were instructed prior  to the
phrase presentation  to interpret the  last word  of the
phrase as a verb ("the horse was led") or as a noun ("the
metal was lead"). Clear EP waveform differences elicited
by the last word of the phrase were noted in F3 when
the  word was differently interpreted in the two con-
ditions. Although these studies demonstrate clear  EP dif-
ferences  to physically identical words, it is difficult to
determine whether the differences are  due to the dif-
ferent  meanings of the words or the different syntactic
contexts.
   Recently,  Friedman et  al. (1975a) examined EPs
elicited by words comprising a sentence.  They found
clear differences between  EPs elicited by articles, words
representing  the object of  the sentence, and the last
word of a sentence. The most pronounced effect was
observed in EPs elicited by the last word of a sentence,
which resulted in an increased latency and an enhanced
amplitude  of  the  late  positive  component  (P400
process). The  latter experiment had less confounding
between word meaning and syntax than was evident in
the  previous studies since the meaning of the word and
its syntactical function were counterbalanced. Friedman
et al. (I975a) suggested that the EP changes noted to the
last  word of the  sentence  reflected  the process  of
"syntactic closure."

Correlates of semantic information processing

   Semantic  refers  to the meaning of  representational
content  of a word or phrase. Several experimenters have
investigated   this  aspect  of language  using  evoked-
potential analyses. Very  few if any have successfully
overcome the problem noted earlier about separating the
content  of an operation  from the  operation itself. For
example, Teyler et al. (this volume) showed that words
with the same meaning  but  different  physical  figures
(e.g., DOG versus dog) elicit very similar  EPs.  On the
other hand, words that are physically similar but possess
different meanings (e.g.,  cut versus cup) elicit different
EPs. A similar phenomenon was observed by Johnston
and Chesney  (1974) using an ambiguous figure that
could be interpreted as a 13 or a B and by Grinberg and
John  (reported  in  Thatcher and John 1976)  using a
vertical  line that could be interpreted as a number or a
letter. In the latter studies general operations related to
number versus letter operations were not controlled. In
the case of cut versus cup, EP differences may have been
due to verb versus noun operations.

   The distinction between content  and a general opera-
tion where  the  operation is invariant  and the  content
changes  is  a subtle but  important distinction. This
distinction is best investigated in studies that control a
given cognitive  operation while manipulating content.
For example, the contribution of content versus general
operations in the studies such as by Brown et al. (1973,
1976) and Teyler et al. (1973, this volume) may  possibly
be  resolved by  averaging EPs to groups  of verbs  and
groups of nouns and showing different EP waveforms (or
clusters) within the verb or noun categories as a function
of  word  meaning.  If there are no differences within
categories, then the differences between the categories
may  reflect  a general process distinguishing the noun
versus  verb  operation.  It  is still  unclear, however,
whether these operations are uniquely linguistic.

   One of the most  successful experimental distinctions
between an operation and the content of an operation
was provided by Chapman et  al.  (1977  and at  this
conference). These  workers averaged EPs to six different
semantic categories of words, based on Osgood's (1964)
"Evaluative (+,.)," "Potency (+,-)," and "Activity (+,-)"
dimensions,  presented  in  random  order.  Subjects sat
passively and observed the words  without  making an
overt decision. The central idea was to make semantic
meaning  within a  category invariant while randomly
changing the physical characteristics of the word. In this
way variations due  to word features should cancel out
while invariant responses,  as  a function of  semantic
category, should summate. Chapman et al. (1977) found
that  the  average EPs to each of the six semantic word
categories were markedly similar. When  a  Z-transform
was performed on each EP  data point by comparing the
grand mean of all  six categories to the EP means of an
individual semantic category, then very marked differ-
ence waves  were noted to occur. Normalized  Z-differ-
ences for each of the six categories were markedly stable
 and replicable  as demonstrated by discriminant analysis
of an  independent sample wherein EP semantic cate-
 gories were differentiated with 55% accuracy. A chance
 result using six categories would be 16.7%. Results of
experiments by Chapman  et al. (1978) indicate  that
semantic content  is  reflected in slight  variations or
 deviations from average  latency and amplitude of the
 evoked response. Walter Freeman (1972, 1975) has  been
 a strong proponent of precisely such a model where
 information is  represented by phase deviations from an
 invariant reference process. Another interpretation of
 Chapman's findings, suggested  at  the conference, was
 that Z-differences  could be due  to arousal and emotive

-------
 254
                                          Thatcher
inequalities existing between semantic categories. There
were six different categories indicating six different
arousal  or emotive states, however,  which strains the
classical  definition of arousal. If the number of word
categories were  to be expanded and replicable Z-differ-
ences remained in each category, then the state interpre-
tation becomes  largely  synonymous to semantic  in-
formation.  In  any case, Chapman  et  al.'s findings
indicate  that the averaged EP is strongly influenced by
general operations whereas the content of the operations
may be reflected in deviations about the mean.

Summary

   Several studies have  demonstrated EP correlates of
language information  processing. These include evoked
responses to elementary phonemic and syllabic speech
sounds  in  children and adults, differentiated  evoked
responses to  physically similar words  with different
meanings, and similar evoked responses to physically
different words that convey the same meaning. Evoked
potential correlates of syntax and semantics have been
observed  as well  as  some  evidence  of hemispheric
lateralization and localization of cognitive functions.

   Although these studies show that various aspects of
language can be profitably studied with EP procedures,
this  survey  indicates  that  clear distinctions between
general cognitive function and specific language capacities
are difficult to draw.  Language  appears  to be  unique
primarily in the coding of the elementary  aspects of
speech stimuli.  Higher levels of information transactions
such as  those involved in semantics and syntax are most
likely specializations of universal  analytical or cognitive
operations. General cognitive operations such as abstrac-
tion, comparison, symbol sequencing, the extraction of
inter- and intra-modal invariances, and the formation of
hierarchical structure are extensively involved in language.
These operations, however, also occur with nonlinguistic
information.

-------
DISTINGUISHING LINGUISTIC AND  STIMULUS EFFECTS
D.L. MOLFESE
Department of Psychology, Southern Illinois University, Carbondale, IL, U.S.A.
   The  application  of evoked  potential methods to
linguistic research entails the presentation of auditory or
visual  stimuli.  Physical  characteristics  of linguistic
stimuli such as intensity, brightness, duration, length,
and  rise  time must  be carefully controlled as in  other
areas of ERP research. Other  acoustic characteristics
such  as  formant  frequencies or consonant transitions
uniquely identify sounds as speech signals that convey
linguistic meaning.  Stimulus parameters,  both  specific
and nonspecific to speech, which must be considered in
evoked potential studies of language, are reviewed in this
paper.

General stimulus parameters

   Much previous evoked potential  research has focused
on the effects of simple acoustic manipulations on the
amplitude, latency, and waveshape of evoked potentials.
Future  research on the  physical  aspects of language
should be  designed in light of past  findings. For this
reason,  it  is important that stimuli  be carefully de-
scribed. The number of graphemes  (visual presentation)
or phonemes (auditory presentation) should be stated.
All stimuli should be of the same length or duration, or
subtend  the same  visual  angle unless  one  of   these
elements is the  variable under  study  (Taurozzi 1973).
The loudness or brightness of stimuli should be matched
(Picton  et al.  1970).  The rise time of acoustic stimuli
should be  specified since this variable may affect the
auditory evoked  potential  (Lamb and  Graham  1967,
Lindsey 1971, Skinner and Antlnoro 1971).

   All of the  physical factors contribute to waveform
differences between stimuli. If the  independent variable
under study is not  one of these variables, research that
fails  to  control  such factors  may be seriously con-
founded. Given the influence of these stimulus charac-
teristics  on the  evoked potential waveform, a detailed
description  of the  physical parameters of  the stimuli
should  be provided  in the published report.  For ex-
ample,  if  auditory  linguistic  stimuli  are  employed,
sonographic records should be included.

   The   manner  in  which stimuli  are produced is
an important  factor.  Use of a parallel or series speech
synthesizer permits precise control of the speech signal.
On the other hand, present  computer models of speech
lack many features  that characterize  natural  speech,
although such features may  not be necessary for simple
identification. The  role  of these cues has  not  been
investigated.  If natural speech stimuli are employed, a
great deal of control may be lost. For example, while the
onset of one  formant transition in relation to other for-
mants  may be manipulated using computer-generated
speech sounds, such manipulations  are greatly limited
when natural speech  is used. In both  cases, with natural
and synthetic speech, it is important  to control peak or
average intensity levels, rise time, and stimulus duration.

   A question could be raised concerning the legitimacy
of averaging across stimuli differing in these characteris-
tics. If stimuli do differ in some of the factors discussed
above, should they be collapsed together? AEPs elicited
in  response to stimuli differing  in rise time would be
expected to  differ. If these responses were  then  aver-
aged, one would expect that variability would increase.
Experimental designs should be  examined for  possible
confounding  of physical  parameters of the stimuli with
linguistic  experimental conditions to avoid the  possi-
bility of a false  finding.

   Given the literature on  habituation  effects,  order of
presentation  and  the interstimulus interval (IS!) are of
critical  importance.  By  varying  the ISI  and presenting
stimuli  in  a random order,  habituation  effects  can be
reduced (Regan 1972).  ISI duration is another factor
that should be considered. Enough time should elapse
between stimuli to allow the background EEG to return
to  its  prcstimulus baseline  level. This factor appears to
vary with the age of the  subjects tested. Latencies of EP
components  in younger  subjects are longer than those
for older subjects. Consequently,  return to baseline takes
longer in the younger group (Barnet and  Lodge  1967,
Ohlrich and Barnet 1972, Callaway and Halliday 1973).

 Acoustic determinants of speech perception

    Research on the perception of speech signals provides
a direct approach for distinguishing between linguistic
and cognitive  functions (Liberman 1970). Given the

-------
 256
                                                                                                   Molfese
 variety of speech sounds employed in  the world's lan-
 guages (Ladefoged  1971),  speech sounds appear to a
 great  extent to be arbitrary elements  restricted only by
 the dimensions and characteristics of the human vocal
 tract.  These sounds constitute a perceptual, as well as a
 productive code for individual languages (Liberman et al.
 1967). Such  sounds can  be broken down  into basic
 units,  many physical characteristics of which have been
 identified  (Stevens  and House  1972). These include
 acoustic bands of concentrated energy called formants,
 rapid transitional elements,  and voice onset time (VOT).
 Vowel identity  is generally signaled by the frequencies
 of the lower three formants in a speech sound,  and
 consonant information is conveyed by the VOT cue and
 by an  interaction of a rapid frequency shift (transition)
 with the formant frequencies. The precise role, however,
 of these cues in speech perception is  not completely
 understood.

   One issue in speech perception centers around the
 inability of  a human language listener to detect changes
 in  the  acoustic  signal  until  these changes  cross  a
 boundary that differentiates  one  speech  sound from
 another. This  perception  of speech  sounds is quite
 different from the ability to detect small changes in the
 frequency of a pure tone.  Although adults are able to
 discriminate many more tones  than they are able to
 identify  or  name,  the  ability  to  detect changes in
 consonant sounds is  only as good as the ability to assign
 a unique label to these sounds.

   The ability  to discriminate only those speech sounds
 that can be  labelled  uniquely is called  categorical
perception and is thought to be due to VOT (Lisker and
 Abramson  1964b) and/or  onset of the  first formant
 transition (Stevens and Klatt 1974). Voice onset  time
 refers  to  the  relation between laryngeal  pulsing  and
 consonant release. A number of studies (cf. Liberman et
 al. 1967) have demonstrated that adults classify bilabial
 stop consonants with VOT  values of 0 and +20 msec as
/b/, and those with VOT values of +40 and +60 msec are
identified as /p/. Adult English speakers fail to discrim-
inate between sounds with VOT values of 0 and +20
 msec  or between  those  with VOT's of +40 and +60
msec.  Both  sets contain a  20-msec difference between
 pairs. These adults are able to detect a  20-msec change
 when  it crosses the phoneme boundary. Intriguingly,
 native  adult speakers of Spanish are unable to detect this
 difference between +20 and +40 msec  VOT. Although
 Spanish does not use this boundary to distinguish speech
sounds, it does make use of one between -20 and 0 msec
VOT. This boundary cannot be detected by native mono-
lingual English speakers. Thus, some speech cues appear to
be used to discriminate  and identify speech  sounds in
 general (e.g., formants and  transitional  elements), and
 others, such as VOT, are specified by the  language itself.

   Preliminary use of EP procedures to study VOT percep-
 ,™ appears encouraging.  Molfese  (1978)   recorded
 auditory  evoked  potentials from T3  and  T4  scalp
 locations  from  16 adults who  were  involved in  a
 phoneme  identification  task.  Sixteen  series of  four
 randomly  arranged bilabial stop consonants (0, 20, 40,
 60 msec) differing only in VOT were  presented  to
 subjects who were instructed to press one telegraph key
 if they heard the syllable /ba/ and a second if they heard
 the syllable /pa/.  Averaged AEPs were  submitted  to a
 principal  components  analysis (after Chapman 1974a),
 and four  factors,  which accounted  for  91% of  the
 total  variance, were isolated.  Independent analyses of
 variance identified two factors (Factors 1 and 4) that
 were  sensitive  to  VOT changes across  but not within
 phoneme boundaries. Both effects occurred only for the
 T4 lead, as can be seen  in Fig. 1. A similar effect was
 found in studies involving 4-year-old  children (Molfese
 and Hess,  1978)  and  4-month-old  infants, but  not
 newborn  or premature infants (Molfese and Molfese, in
 press).


   While the frequency relations between the first three
 formants provide an invariant cue for vowel  identifica-
 tion,  no such invariance  has been found for consonant
               ADULT FACTORS
             0204060  0204060
            VOICE ONSET TIME, mMC
Fig.  1.  Changes in mean score for Factors 1 and 4 as a
function of VOT and hemisphere. The means for the
dependent variable in the analysis of variance (the weight
for the factors for the different stimulus conditions) are
plotted along the ordinate. The stimulus values are plotted
on  the abscissa. Factor scores are plotted separately for
left hemisphere  (T3) and right hemisphere (T4J record-
ings. The temporal pattern of factor loadings is shown in
the boxes on the right  where time is plotted along the
abscissa.

-------
Linguistic, Stimulus Effects
                                             257
sounds (LIberman et al. 1967). Rather, it appears that
the consonant  transition changes as a function of the
vowel context. For example, the consonant transition
that carries  information  concerning the  /d/  in the
syllable  /di/ Is characterized by a riling second transi-
tion, and the consonant in the syllable /du/ is marked by
a rapidly  falling  transitional element.  In  both cases,
however, the initial consonant transition Is identified as
/d/. This issue is yet to be resolved by speech scientist!,

   The  acoustic cues  used to differentiate  between
speech  sounds may  become more  readily Identifiable
through the application of EP  procedures. Research in
this direction by Molfese et al. (1976) entailed recording
AEPs from sixteen 2-day-old human infants in  response
to a series of acoustic stimuli, which varied in formant
bandwidth and presence of a transition. The stimuli are
presented  in Fig. 2. The two stimuli in the left column
are perceived as: top, the speech syllable /gae/ character-
ized by  initial transitions followed  by  steady-state
formants and, bottom, a vowel  syllable /ae/, which does
not contain a transition. These sounds contain formants
of normal  bandwidth.  The stimuli in the  right column
contain narrow bandwidth formants (sinewave). Again,
the top stimulus contains a transitional element, but the
bottom does not.

   AEPs were  recorded from  T3 and  T4 referred  to
linked earlobes. Averaged AEPs submitted to a principal
components  analysis yielded four components, which
accounted  for 96% of the  total variance. Analyses of
variance identified one component as reflecting changes
in bandwidth and  another as reflecting the presence or
absence of formant transitions. These results suggest that
        2 •-
           PORMANT,
           WITH
           TRANSITION
8INEWAVE
WITH
TRANSITION
                              2"

                              1"
                                           300
                                 WITHOUT
                                 TRANSITION
                    300
                 DURATION, mMO
          300
Fig.  2. Acoustic  structure of syllables  /gae/ and /ae/.
Stimuli with normal width formant structure are on the
left of the figure, and those with narrow band formant
structure  (sinewave) are  on  the right. Stimuli with
transitional elements (consonant-vowel syllable) are pre-
sented at the top of the figure; steady-state syllables
(vowels)  are at  the bottom. (Reproduced  with  per-
mission from Molfese et al. 1976.)

evoked   potential  techniques   may  be  sensitive  to
systematic changes in acoustic parameters  related to
speech   perception.  The   question  of  acoustic  in-
variance  in consonant  recognition  still remains;  per-
haps these  techniques  could  be applied  to  seek  an
answer.

-------
 LATERAL ASYMMETRY  OF  EVOKED POTENTIALS AND
 LINGUISTIC PROCESSING
 D. FRIEDMAN

 Department of Medical Genetics, New York State Psychiatric Institute, New York, NY, U.S.A.
 Introduction

   Numerous behavioral studies using the dichotic tech-
 nique  (reviewed by Kimura 1967) and tachistoscopic
 recognition  procedures  (reviewed  by  White  1969)
 pointed to the inequality of the cerebral hemispheres in
 processing verbal and nonverbal information. With the
 advent  of commissurotomy for relief of intractable
 seizures, studies of the "split-brain" human added to our
 knowledge of hemispheric  specialization  of  function
 (reviewed  by Gazzaniga 1970 and Nebes 1974). Recent
 anatomical evidence has shown that the cortical mass in
 the region of the planum temporale (Wernicke's area) is
 larger on the left than on the right hemisphere in both
 infants  (Witelson and Pallie  1973) and adults  (Gesch-
 wind and  Levitsky  1968, Wada et al.  1975). That the
 brain's electrical activity would mirror these behavioral
 and  anatomic asymmetries was a natural assumption for
 evoked potential researchers to make.

   Early researchers in this area  used repetition of the
 same or different stimuli with no task imposed upon the
 subject  (e.g., Cohn  1971) or used discrimination para-
 digms (e.g., Morrell  and Salamy 1971, Wood et al. 1971)
 where phonemic information was necessary to perform
 the task. With the exception of Molfese's data (Molfese
 1975), results of these  studies  have shown extremely
 small amplitude asymmetries. Furthermore,  many  of
 these studies  could  be  faulted  on  methodological
 grounds (see Friedman et al. 1975a).
   Recent  studies have  attacked the evoked potential-
language problem from  a meaning or "deep-structure"
level. Thatcher (1977a)  employed a semantic  match-
 mismatch  paradigm,  while  Teyler et  al. (1973) and
 Brown  et  al. (1973, 1976) used the  same word  in
 different meaning contexts. Friedman et al. (1975b) and
 Shelburne  (1972) utilized a modification of the informa-
 tion  delivery paradigm where subjects had to know the
 meaning of a linguistic unit in order to solve the
 problem. Although  Friedman et al. (1975a, b), Galam-
 bos  et  al. (1975), and Shelburne  (1972) found no
 consistent amplitude asymmetries, Teyler et al. (1973)
 and Thatcher (1977a) did. Brown et al. (1976), using
 stepwise discriminant analysis, found greater discrimin-
 ation over the  left  hemisphere than over  the right of
 evoked  potentials elicited by the physically  identical
 word in differing contexts.
 Difficulties in assessing the significance of
 reported asymmetries

 Analytic methods

   While amplitude  measures  are straightforward  and
 allow identification  of discrete peak effects, analyses
 that focus on only a few points of the waveforms may
 be inappropriate if the evoked potential reflects changes
 in processing over time. Some  investigators (e.g., Chap-
 man,  this section; Brown  et al. 1976; Thatcher 1977)
 have,  therefore, used whole-wave analyses, such as factor
 analyses, stepwise discriminant functions, and correla-
 tions. These  kinds of analyses have an advantage in that
 they  take the whole wave into  account  (i.e., all  the
 sampled  points), but have some  difficulty in handling
 artifacts, such as latency shifts between waveforms.
 Ideally,  one  should  use both amplitude measures  and
 whole-wave analyses in drawing conclusions. This issue is
 complicated  since  the way one measures often leads to
 as many different results as there are measuring tech-
 niques, as well as difficulties in comparing results across
 studies.

 Existence of lower-order asymmetries

  The investigator demonstrating an asymmetry to a
linguistic stimulus cannot immediately conclude that
linguistic processing accounted  for the asymmetry. For
example, Davis  and  Wada (1974)  have  shown that
asymmetries  exist to lower-order stimuli, such as clicks
and flashes with no  task imposed upon  the subject
Further, in data from my laboratory, we have seen  the

-------
Lateral Asymmetry of EPs
                                               259
same direction standing asymmetries to linguistic as well
as nonlinguistic stimuli, even after switching amplifiers
to control  for  gain effects. Most  of the  more recent
investigations (e.g., Brown  et  al.  1976,  Wood  1975,
Thatcher 1977a, Friedman et al. 1975a) are based upon
an information  processing approach. Thus, it would be
extremely important to use a "noninformational" con-
trol  and not simply  a "nonlinguistic" stimulus as a
baseline  from  which  to measure  asymmetric effects.
Thatcher (1977a) used this kind of control by employ-
ing random dot configurations during intertrial intervals.
Friedman et al. (1975a) used a "no task" control as a
baseline condition, in which the subject simply listened
to the stimuli without a task. Thus, the evoked potential
researcher who wants to demonstrate  an asymmetric
effect due to  linguistic processing must  have  evoked
potentials elicited by noninformational control stimuli
with which to compare evoked  potentials in response to
experimental stimuli with respect to both direction and
magnitude of asymmetry.

Task and subject variables

   Evoked potential studies of hemispheric lateralization
have  not as yet employed tasks that unequivocally
involve  semantic  processing and/or  are  of sufficient
difficulty to  tap  more  than  automatic  aspects of
language function. This, in part, has probably accounted
for  the paucity  of robust evoked-potential  effects.
Approaches toward the goal of "sufficient processing"
have been made (e.g., Thatcher 1977a,  Brown  et al.
1976).  We should  also design  experiments that more
closely  approximate the real world of speech processing
using, for example,  connected discourse, since  this is
natural  for human beings and ensures involvement of the
language processor. Another possibility would be to use
tasks and  stimuli known  to involve  a good  deal  of
competition between the hemispheres, since the tasks we
do use may not engage the hemispheres differentially.

   A related issue is the cognitive style of each subject.
Each person is  unique and engages in different methods
for processing  the same information. Thus, one subject
might process a stimulus (which the investigator specifies
as "verbal") in a nonverbal way, while another  might
process  the same stimulus using verbal  coding. Note
Brown's discussion of individual differences elsewhere in
this section.

   The  influence  of task variables (coding strategies) on
behavioral  asymmetries has been demonstrated by Sea-
mon and Gazzaniga (1973) who asked subjects to code a
visually presented word by either subvocally rehearsing
it or by generating an imagjnal representation. In the
rehearsal  set,  probes  to  the left  hemisphere yielded
significantly  faster reaction  times  than probes  to the
tight, while  the reverse held  for probes to the  right
hemisphere when  subjects  were  in an  imaginal  set.
Additional evidence for the  influence of coding strate-
gies comes from a study by Bever and Chiarillo (1974),
which  showed  that  musically  sophisticated  listeners
recognized simple melodies better  from the right  ear
than  from  the left,  while  musically  naive  listeners
showed  the  well-known  left-ear  effect (see  Kimura
1967). This evidence  indicates that it was  the  kind of
processing applied to  the musical stimulus that deter-
mined which hemisphere was dominant. Thus, musically
sophisticated listeners were able to organize the melodic
sequence  "in terms  of the  internal  relations of  its
components" (an analytic or left-hemisphere function),
while  musically naive subjects  organized  the  melodic
sequence  in  a  gestalt (a right-hemisphere function).
These results suggest that the task assigned to the subject
will  have at least as strong  an effect as  the  type of
stimulus.

Difficulties inherent in averaging

   Since we must repeat stimuli in order to obtain an
average, we may "wash out" effects that are present in
the single trial. For example, the repetition of the same
word,  whether or not in a different context, may lead to
semantic  satiation.  Thatcher has proposed the use of
"light  averaging" (N's of 12  or less), and Friedman has
suggested the scoring  of single trial data (especially in
P300 experiments) as done by Ritter et al. (1972). An
alternative approach is the use of several different words
with no repetition (cf. Chapman  1974b, this volume).
Thatcher  (this  volume)  has  also  pointed  out   that
differences  in  skull  thickness and volume conduction
idiosyncrasies may affect the degree to which evoked
potential  techniques  can  resolve asymmetries at   the
scalp.

   An  additional  problem is the interindividual  vari-
ability that exists in the location  of cortical landmarks
(see Whitaker and Seines 1976), as well as  ambiguity in
the location of the classical speech reception area (Bogen
 1975). Thus, we may not be able to assume that we are
recording from the same anatomical area in  each subject.
This is  further confounded when one  uses only  two
electrodes or averages across subjects.

   The assumption has also been made that  a voltage
difference across the head implies a functional specializa-
tion difference  between  the hemispheres,  i.e.,  that
greater voltage  implies greater output from the cortex
over which the  scalp electrode lies. Vaughan and Ritter
(1970) using a volume conduction model  and Kelly et al.
(1965) have suggested that  this may be  an erroneous
assumption.

Assumption of a unilateral processor

   We  still work under  the assumption  of  a strictly
lateralized  language  processor,  although  Gazzaniga
(1970), Nebes (1974), and Zurif (1974) have shown that
the   right,  so-called  "nondominant"   hemisphere  is

-------
 260
                                         Friedman
capable of  receptive language function. Zurif, in fact,
has shown that some processes necessary for decoding
meaning  are   carried  out  exclusively  in  the  right
hemisphere.  Thus,  we  may  be washing  out  asym-
metric effects  by using stimuli and  tasks that involve
simultaneous,  but differential, processing by the two
hemispheres.

Summary

   Whether  evoked  potentials can  reflect lateralized
functional  processes  is  an  unsettled issue. Contradic-
tory  findings can be partially explained by differences
in stimuli, tasks, and analytic methods. Although asym-
metries do exist, it is difficult to determine whether
observed differences are related to linguistic or lower-
order  processing.  More  stringent  controls  and tasks
directly related to speech  reception are necessary to
reduce "subject option" and to ensure that the hemi-
spheres are engaged differentially.

Acknowledgments

   I  would like to thank my  colleagues  Walter Ritter
and  Richard  Simson  whose  critical  thinking  contri-
buted to some  of the ideas expressed in this paper.

-------
INDIVIDUAL DIFFERENCES AND SIMILARITIES  IN
LANGUAGE  EFFECTS  ON EVOKED  POTENTIALS
W.S. BROWN
University of California at Los Angeles, Los Angeles, CA, U.S.A.
   There  is general  agreement  in  evoked  potential
research  on humans  that  common perceptual experi-
ences,  cognitive processes,  and behavioral responses of
different  individuals  produce  similar  effects on the
waveform, amplitude, or latency of the average  scalp-
recorded  evoked  potential.   Increasingly   finer-grain
cognitive  processes have been demonstrated to have
correlates in the evoked electrical activity of the  brain,
as exemplified in current research on EP correlates of
language processing. Studies of the effects of linguistic
meaning on EPs are beginning to approach  the electro-
physiological study  of  the  mental activity that we
subjectively experience as thought.

   The question  arises,  then,  at  what point will we
begin  to  deal with  EP idiosyncracies due to unique
experiences, cognitive styles and strategies, and thought
processes of each individual? EP morphology may well
be like a fingerprint: the  gross structure is similar for
most individuals, but the fine  grain is idiosyncratic. As
one investigates more complex levels of cognition, both
the  mental  processes   and  their  elec trophy si ological
correlates  should be increasingly  unique for different
individuals. For  example, factors  that might lead to
idiosyncracies in EPs to  language stimuli  would be:
small differences in the  perception of stimulus meaning
due to variations in cognitive speed or processing time,
variability in the anatomical structure or physiological
localization of language functions, and unique memory
storage or retrieval systems.

   Individual differences or similarities in the effects of
language processing on  EPs are essentially  an empirical
 question. The P300, particularly to linguistic stimuli,
 has been shown to  be  very  similar for different sub-
jects (Shelbume  1972,  1973; Friedman et  al. 1975a,b;
 Thatcher,  1977b; Kutas  and Donchin, this volume).
 For example, Friedman et al. demonstrated that most
 subjects produce a larger P300  to  the last word of a
 paced sentence,  despite the fact that the information
 of the  sentence was  delivered  several words  earlier.
 Shelbume (1973), however, demonstrated an important
difference in  P300 to the last letter of CVC trigrams
between  children who performed  well  compared to
those who performed poorly on a reading task.

   Hemispheric   asymmetries  in  the   amplitude  or
latency  of EP  components to language versus non-
language  stimuli have also been shown to be similar for
the majority of right-handed subjects (Matsumiya et al.
1972; Morell and Salamy 1971; Cohn 1971; Wood et
al.  1971; Buchsbaum and  Fedio  1969,  1970;  Neville
1974). An obvious source of subject  effects  in  this
type of research would lie in  differences in the degree
of  left  lateralization of language  function. Subject
differences in  the  asymmetry of EPs  to  language
stimuli thus would  be expected when comparing left-
and right-handers. Molfese et  al. (1975, 1976)  has
demonstrated that  similar amplitude  asymmetries in
evoked potentials to language and nonlanguage  stimuli
can be  seen in  records  from  infants,  children,  and
adults. Molfese  has also reported similar effects on
average  EPs between infants  for the  presence or
absence  of  a   transition  in   a  synthesized  speech
stimulus,  the bandwidth  of  the stimulus, and  the
hemisphere  being  recorded. A  subject difference  was
demonstrated in terms of an EP correlate of the sex of
the infant. Thus, the research on  EP correlates of the
lateralization  of  language  functions   has  generally
demonstrated similar  effects for the  majority of sub-
jects, with  possible  exceptions due to  the effects of
handedness and  a nonlateralized correlate of sex in the
response of infants.

    Individual difference  or  similarity  is  particularly
 relevant  in  studies  of EP correlates of language mean-
 ing. Are the EP effects  of word  meaning similar for
 different subjects?  Teyler et  al.  (1973, Roemer and
 Teyler  1977)  found that responses  to a  click  pro-
 ceeding  the  expression  of the  word "rock," for
 example, were  dissimilar  depending  on whether the
 subject meant "a rock" or "to rock."

    Although across-subject averages showed  noun-verb
 differences, the characteristics of response to the  two-

-------
  262
  word meanings  appeared  to be  idiosyncratic. This  was
  particularly true for the verb meaning of the stimulus
  homophones.  Brown et al.  (1973, 1976; Marsh  and
  Brown, 1977) employed similar homophone stimuli in
  sentence or phrase contexts to demonstrate EP wave-
  form  differences to noun and verb meanings. Brown et
  al.  (1976)  specifically  tested the  question  of inter-
  subject similarity  using  multivariate  statistical  pro-
  cedures. Separate EPs for the two-word meanings for
  each  of 15 subjects were submitted  to stepwise  dis-
 criminant  function analysis (SWDA) in an attempt to
  find common EP time  points  at  which responses to
 noun  and  verb  forms  of the stimulus could be  dis-
 criminated  reliably.  A  discriminant   function   was
 derived   from   left-hemisphere   responses,   which
 identified   EPs  from the  two  conditions with  77%
 accuracy;  but  differences  were  not  apparent from
 right-hemisphere  responses.  Thus,  intersubject  com-
 monality in EP  correlates of contextual meaning was
 demonstrated.  Subsequently,  similar  analyses  were
 done in an attempt to discriminate subaverages of five
 stimuli within single subjects. Although left-hemisphere
 responses  could  be  reliably  sorted by  noun  or  verb
 meaning of the  stimulus,  there  appeared to be little
 similarity  between  different  subjects  in  the  specific
 time  points  chosen  for  the discriminant  functions.
 Thus,   there  was  some  degree  of  idiosyncracy  in
 responses to different word meanings.

   Perhaps  the  strongest case  for  intersubject simi-
 larity in EP correlates of language meaning comes from
 the  work   of  Chapman  et  al.   (1977,  1978).  These
 investigators  have  demonstrated  EP   correlates   of
 general semantic  meaning as it is  measured by Osgood's
 Semantic Differential scale. EP waveforms to groups of
 words occupying opposite polar positions  on one  of
 the meaning scales differed in the presence or absence
 of  a specific component.  Components differentiating
 EPs to words from the two categories were sufficiently
similar  for different subjects  that average responses  to
similar words from different subjects could be reliably
sorted  with a template-matching procedure.

   Intersubject similarity in the  experiment of Chap-
man et al.  might have  been  expected  on  several
grounds. First, Osgood's scale represents perhaps the
most   general  classification   of  semantic  meaning.
                                              Brown
 Second, it is constructed on  the  basis of intersubject
 similarities in the ratings of general semantic meaning.
 Finally, Chapman  et al.  averaged across individual
 words  from a  specific  category,  and  thus may have
 averaged  out  idiosyncracies  in responses to  specific
 lexical  meaning of words.

    Comparison  of the results  of  Teyler, Brown, and
 Chapman  indicates  that  intersubject  difference   or
 similarity  in  EP  correlates  of language  meaning is a
 matter  of  the  level of linguistic-EP analysis.  If one
 could observe the  EP to  a single presentation of a
 word in some  ongoing linguistic context, the responses
 would  have  definite idiosyncracies due to the  unique-
 ness  of thought,  imagery, and association of different
 people, or  of the same person listening  to the  same
 message at different  times. Perhaps this  explains the
 failure  of  Teyler,  who  used averages   of  only  10
 responses,  to  find  similar  meaning-related waveform
 differences between subjects,  or of Brown to  demon-
 strate  similarity  in  EP  discriminators within single
 subjects using averages of only five trials.  As responses
 to  the  same word are averaged across  a large number
 of  contexts (either  stimulus or mental), the response
 for individual subjects might be expected to approxi-
 mate  a  generalized  representation  of  the  brain's
 response to word meaning. Since the function of words
 is to  express similar meaning to people who speak the
 same  language, one  might expect  similarities between
 subjects  to increase above  the single  trial  or small
 N level. Thus, Brown et al. were able to find inter-
 subject  similarities in  the correlates of word meaning
 when comparing averages  of 100 responses to the same
 word. Finally, as one averages  across multiple presenta-
 tions  of different  words  from  a single semantic class,
 demonstrated  to  be  stable   among  individuals and
 cultures, one would expect  intersubject similarities to
 emerge  as  demonstrated by  Chapman and his col-
 leagues.

  This analysis ignores important differences in  experi-
 mental designs used by different investigators.  Never-
 theless,  it provides a reasonable synthesis of the results
 on  the issue of individual  differences and similarities in
 the  effects  of language  meaning on averaged  evoked
potentials. This issue certainly  merits further investiga-
tion.

-------
CONTRIBUTIONS  OF LINGUISTICS AND
OTHER DATA BASES
TJ. TEYLER
 Harvard  University, Cambridge, MA, U.S.A.
   Neurolinguistics, the  study  of brain and  language,
 has  been  spawned  in  an  era  of  interdisciplinary
 research  interest.  The   degree   of  interdisciplinary
 activity, however, may be more  theoretical than real.
 Most   neurolinguists   are   reformed  physiological
 psychologists. Thus, current workers in the field come
 from the physiological as opposed to  the linguistic or
 cognitive tradition. This  is  understandable  since  a
 general surge  of interest in "more complex events" is
 prevalent in the neurosciences. We reflect a portion of
 this scientific  evolution. We  all  agree that  increased
 contact  with  other disciplines  would be beneficial  for
 this fledgling  field, and, presumably, cross-fertilization
 would occur as well. Various data bases are presented,
 and the achieved  or potential  cross-fertilization is dis-
 cussed in this essay.

 Linguistics

   Linguistics  is   the   study  of  the  structure  of
 languages. Attention has been focussed  primarily  on
 formal description  of grammatical  regularities. Whereas
 most  effort  in neurolinguistics emphasizes phonemic,
 word, or  semantic aspects of language, linguistics is
 primarily concerned with more global  aspects of langu-
 age.  The most  comprehensive  theory  of  language
 structure  is  the  generative   grammar  proposed  by
 Chomsky (1965)  and others as  a set of formal rules
 capable  of describing  the relationships found in langu-
 age. It is not and  was not intended to be a model for
 biological aspects of language. Thus, there has been less
 interdisciplinary   communication  than   one  might
 assume.  As  the  neuroscientist  begins  to deal with
 syntactic relations in more  complex verbal contexts,
 the relationship  to   formal linguistic theory  should
 become clearer. Conversely, our activities at elementary
 levels  of language  may  trigger a resurgence of interest
 in these aspects of language on the part of linguists.

 psycholinguistics

    Psycholingulstics can be  defined  as the study  of
 how language  is used and acquired. Neurolinguistics, at
present, is concerned with brain correlates of language
use, either production or reception. The interface here
is  better developed  and has  already proved  fruitful.
Still,  most  psycholinguists,   like  linguists,   are  not
primarily concerned  with simple  elements, but  rather
with  the  relationships between elements  at  relatively
complex syntactic, semantic,  and organizational  levels.
Psycholinguists  deal  extensively  with  the   uses  of
language as  do  neurolinguists, and  the prospects for
continued  interaction  are  bright.  We  can  aid  the
psycholinguist today in studying brain  correlates of
categorical perception  across  species  and in further
clarification  of the localization/lateralization issue.

   Psycholinguistic theory has emphasized the develop-
ment of models more from a psychological  than from
a  biological   point  of   view.  What  would  be  most
valuable for neurolinguists is a psycholinguistic  theory
of language acquisition  and utilization that  accounts
for biological  as well as psychological factors and has
the  formal  precision  of  Chomsky's approach.  Ad-
mittedly, this  is a tall order.

   Many psycholinguists feel that the single best  insight
into language behavior  can be obtained from studying
language acquisition. It  is remarkable that, while there
are individual  idiosyncrasies in language acquisition, the
overall   developmental   pattern  is  markedly  similar.
These  commonalities  hold  across  cultures  and  for
handicapped (deaf) children as well. This invariance in
language  development  may well be attributed  to the
underlying biological maturation and specialization  of
brain tissues.  This  area may  be  extremely fruitful for
the  future  interface between  brain  and language re-
search.

Clinical neurology

   Clinical neurologists  are interested in disturbances of
language  behavior  (aphasias) due to brain pathology.
Correlations between structure and function, although
imperfect, offer a  means  of studying  the  brain  in
relation to  language  behavior.   In  addition,  clinical

-------
  264
 neurology, of all the disciplines interested in brain and
 language, is most versed and concerned with the under-
 lying biology, brain structure and function.

    The lack of good correlation between pathology and
 disturbance  is   not  unexpected   because  language
 behavior  is  undoubtedly  not  the  consequence  of
 activity in any  single brain  area. While it is probably
 true that many brain  areas interact in language be-
 havior,  it  is undeniably true  that the anatomy  and
 physiology  of the brain are  directly  related to  lan-
 guage. Our goal  might well  be the  study of these
 interrelationships between areas of the  brain. The task
 is  complicated by  the inadequacy  of our tools  and
 paradigms and by our not knowing where  to look (or
 not being  able  to look, e.g.,  subcortically). To com-
 pound the problem,  most pathological  symptoms are
 not exclusive to language behavior, but involve other
 cognitive processes as well (see  below).

    In neurolinguistics we have no animal models. For a
 biological  science, this is  a severe handicap  because
 most tools of the trade  are invasive and inappropriate
 for human subjects.  Complicated and unclear as they
 are, human suffering from "natural" lesions may offer
 a  model  for examining biological aspects  of language.

   Clinical neurologists  interested  in aphasia have de-
 veloped theories  and  specific predictions usually related
 to  localization of  function. Although the  theories are
 highly controversial and often rest on inadequate data,
 they are, in many cases, amenable  to  test  by neuro-
 linguistic  techniques.  Conversely, we may  be able  to
 devise clinical tools to aid in the differential diagnoses
 of brain damage.

Neurobiology

   Interactions with neurobiology  (the study of  the
nervous  system)  are  well  established. In  one sense,
neurolinguistics   is  a  subspecialty  of  neurobiology.
There is good evidence that the acquisition of language
and brain maturation are closely linked. Since language
is acquired by means of interactions between organism
and environment, it may be safe to assume  that there
are  concomitant  changes  in  neuronal  relationships
similar,  if  not  identical,  to  neuronal plasticity  as
studied in other contexts. Since  we cannot  directly
study the mechanisms of these  neuronal alterations but
must rely  upon  electrical manifestations, we will have
to depend heavily  on neurobiology  for  the  elucidation
of  the mechanisms themselves. Neurobiology,  being a
 more  molecular  discipline,  will  probably  contribute
 more to neurolinguistics  than the converse. The contri-
 butions can  potentially arise in many areas:  an under-
 standing of  the  neuronal (or nonneuronal) generators
 of   the  electrical  signals we   record, the  underlying
 anatomy, the nature of physiological events in associa-
 tion cortex, and principles of  feature extraction. Pro-
 vided with  these contributions-for  example, in under-
                                             Teyler

 standing  the  neuronal  processes  underlying  feature
 extraction-it will be our job to extend them theoreti-
 cally  and  functionally to  the  neuronal  mechanisms
 involved in language behavior.

 Cognitive psychology

   There  is probably no human  brain pathology that
 disturbs  language  behavior  without  affecting  other
 cognitive  functions. The  brain  is an  integrated organ.
 Psychobiology tells us  that  a good deal of integration
 occurs at  virtually all  brain loci. Language behavior is
 the   complex  intersection   of  sensory,   perceptual,
 memory,   syntactic,  and  semantic  relationships.  In
 language  behavior, the brain is  engaged in a dynamic
 process   presumably   involving   widespread   neuro-
 anatomical loci.  It is not unreasonable to assert that
 language behavior employs brain  processes common to
 many  forms of sensory, perceptual, memory, and other
 cognitive   functions.  It is  a substantial  challenge  to
 neurolinguists  to disentangle (if  possible)  those  inter-
 actions to examine  the degree  of either  overlay  of
 functions or  the specificity of certain brain processes
 in language behavior.

   Many cognitive  psychologists  are using brain record-
 ing  techniques in addressing questions that may  be
 basic  or  at least  related to language behavior.  Thus,
 cognitive  psychology and neurolinguistics  appear to be
 entering an era of fruitful dialogue.

 Other data bases

   At  least three other areas bear close relationships to
 neurolinguistics:  ethology,  developmental  psychology,
 and the applied field  of learning  disabilities.  To  the'
 extent that language  behavior is  an  evolutionary exten-
 sion of general communication schemes in animals, we
 may be able  to learn from and instruct ethologists.
 Language,  as  a formal  system,  is specific  to humans]
 but  communication schemes are  widespread. To what
 extent can  parallels be  constructed  and brain processes
 be delineated?

   The acquisition  of language behavior, as mentioned
 previously,  not  only  falls  within  the   bounds   of
 psycholinguistics, but,  along with the development  of
 other cognitive abilities, is a central theme of develop-
 mental psychology.  We  would  be  well  advised  to
 attend to  the  activities of workers in this area.

   If neurolinguistics can offer a contribution to the
condition  of man in the near future, it will probably
lie in  diagnostic tools  for the clinical  neurologist  in
relation to brain damage and learning disabilities. While
we  have no delusions  of offering  remedies  for  these
devastating human  conditions, we can perhaps assist in
the development of  sensitive tools for the  detection
and  classification of these  conditions. Such  a con-
tribution would, by no  means, be  unappreciated.

-------
METHODS  OF  EVOKED-POTENTIAL  ANALYSIS IN
LINGUISTIC RESEARCH


R. M. CHAPMAN

University of Rochester, Rochester, NY, U.S.A.
   The problems  associated  with  methods of evoked-
potential (EP) analysis do not  appear to be different
for the study of language and EPs than for other areas.
Two  aspects  of relating  EP  effects   to  language
variables, however,  foster a sense of urgency in de-
veloping  EP  measurement  techniques.  Oni  aspect  is
that in addition to focusing on linguistic variables, it  is
necessary to consider many other variables, both lower-
order, such as sensory, motor, and artifact, and higher-
order,  such  as general states and cognitive processes.
Consideration  of  these additional variables is necessary
to  demonstrate   specific relations  between language
variables  and EPs. This tends to promote experimental
designs  with  a large number  of conditions  and  to
promote  EP measurement techniques that can separate
EP effects related  to  the various classes of variables.

   Another aspect that lends urgency to  developing EP
analysis  is  the   relatively  small  size  of EP  effects
associated with linguistic variables. Small EP effects do
not mean they are  not important.  If we were simply
trying  to explain EPs,  then  starting  with large EP
effects would  make sense. If we are seeking EP effects
related to particular variables such  as linguistic  ones,
then we need to use techniques that focus on  these
relations  and  be  satisfied  with  reliable relationships
whether they involve large or small EP effects.

   At an early stage  of research, a reasonable approach
is to use techniques that assess general EP differences
between conditions. At later stages, one wants to know
more about specific  differences. If  an EP peak ampli-
tude  is  reliably  different  between conditions that
reasonably  are related to linguistic  variables, and not
other  variables,  then  useful  information  has  been
obtained. Negative findings are  not  as useful, however,
because of problems  in accepting the null hypothesis.
What was  the power  of the statistical  analysis? Was
there a better EP measure that might have been made,
perhaps another  time  point  not at a peak? Was the
peak measure not  sufficiently  stable,  e.g., would   a
measure  that  incorporated  information  from several
time points be more stable?
   Another technique for more global  assessment of
differences between EPs is computing correlation coef-
ficients. The  common test of correlation significance
based only on the number of paired observations (time
points)  is not appropriate since  the  amplitudes at all
time points are not independent.  Appropriate statistical
tests can be found, however.  For example, the correla-
tion coefficients may be treated Uke any  other measure
for subsequent statistical  test (e.g., t-tests, ANOVA)
provided  several r's are available per  condition  for
assessing variability. Before averaging correlation coef-
ficients  or entering them into  subsequent  tests,  the
Fisher r-to-z  transform  should be  applied. Used with
appropriate  procedures, correlations can  be used to
indicate overall similarities  of EPs and similarity  to a
scoring template (e.g., Chapman 1974b; Chapman et al.
1977).

   The  most common measurement technique in EP
research  is  to measure  the  amplitudes and perhaps
latencies of various positive and negative  peaks In EP
waveforms. These  peaks (e.g., PI, Nl, P2, and P300)
are  often  referred  to as EP components and  often
implicitly assumed to be functionally singular entities.
Such  reiflcation  of manifest  EP peaks may not be
fruitful  for understanding  EP relationships in general,
Including  linguistic  ones. Most would agree that EPs
are composites of a number of underlying components.
Why  assume  that  each underlying component  cor-
responds to a manifest EP peak? The possibilities exist
that several underlying  components  might contribute
to a given peak and that a given underlying component
might contribute to several EP peaks. These are empiri-
cal questions. Evidence  for both of these possibilities
has been  found (Chapman  1974a; Chapman, et al., in
press).

   The  use of multivariate analyses  (e.g., John et. al.
1964; Donchin 1966, 1969; Chapman et al.  in press)
may be of considerable  help  in EP analyses since it is
concerned with multiple measures and with considering
them in combination as a system of measurement. Two
of these techniques, which have been used with EPs,

-------
  266
  are factor analysis and discriminant analysis.  Among
  the several factor analytic techniques, the method of
  choice is often principal components analysis followed
  by   an  objective,  analytic   rotation,  e.g.,  varimax
  method. Principal components analysis can be used to
  uncover and measure  latent EP components, which are
  independent and  may overlap at various  time  points.
 The measures of the latent components, in contrast to
 the original measures at each of the time points, are
 more parsimonious, have increased reliability (measure-
 ment  stability), and  may be  used  to  measure  the
 separate  contributions  to various  time  points. The
 relation of  these  latent  components to the  experi-
 mental  conditions is a separate  question that may  be
 analyzed by submitting  the component scores to other
 procedures, such as ANOVA  or multiple  discriminant
 analysis.

   Discriminant  analysis is another form of multivariate
 analysis, which  can be used  to  relate  EP  measures,
 including latent component  scores,  to  experimental
 conditions.  It  combines EP  measures into  functions
 that best discriminate the designated conditions.

   In research on  the connotative meaning of  words,
 attention was focused on EP effects related  to  the
 semantic classes  in a generalizable way (Chapman et  al.
 1978).   Stable  EP  measures  and   relationships  to
semantic classes  that were generally true for the entire
group of subjects  were  sought. Because the EP wave-
forms of different  subjects are reliably different from
each other,  EP data were standardized for each  subject
separately. Each subject's data at each time point were
transformed to z scores  (mean = 0; standard deviation
» 1). These transforms removed  the  average EP wave-
form  of each subject from further consideration, since
individual differences were not of interest. By analogy,
if we wish to measure the growth of trees, it is helpful
not to include in our  measurements the mountains on
which they grow. Next,  the  standardized  data  for  all
subjects  were combined  in a varimaxed principal com-
ponents  analysis (Dixon 1975). This was computed to:
                                         Chapman

(1)  determine  the  EP  components  found  in  this
semantic experiment  and (2) measure the amount of
each component in each EP. This approach provided
more stable EP measures than using the  raw amplitudes
at each time point.

   The  next step  was evaluating the extent to which
these EP components contained semantic information.
The latent component  scores  for  all  subjects were  t
entered  into  multiple   discriminant analysis  (Dixon
1975)   to  develop classification  functions  that  dis-
criminated  among  the semantic classes.

   A common set of classification  functions was de-
veloped for all subjects, and  their success provided one
test of  the generalizability of the semantic EP  effects.
The results were cross-validated by  a jack knifed pro-
cedure, which assesses the classification success when
each case is left out of the development set and then
classified. The results were  further  cross-validated  by
applying the classification functions developed from EP
data for one list of words to EP data for another list.
The classification  functions used various combinations
of  a  number of  the orthogonal EP components to
distinguish  the six semantic  classes.  Thus,  one  or  two
general  EP components,  such  as  P300  and  CNV-
resolution  (Chapman  et al., in  press) were  not suffi-
cient  to account  for the dimensionality of the data.

   In general,  the answer  to  the  question  of the
specificity  of  language effects depends  on the  dimen-
sionality of EP measures. Multivariate  analysis tech-
niques  can  help  us answer  in  an organized way the
kinds of questions we want to ask about EP data and
their  relation to  experimental  variables.  Of  course,
sophisticated data  analysis cannot  carry  the  burden
alone. Data analysis can only meaningfully  measure EP
effects   produced by  experimental manipulation. It  is
paramount  that  experimental designs control variables
of interest  independently, or at  least manipulate them
to different degrees, to  provide a valid data base for
measurement.

-------
EVENT-RELATED  POTENTIALS ASSOCIATED
WITH  LINGUISTIC STIMULI: SEMANTIC
VS LOWER-ORDER EFFECTS1
A.L. MEGELA

Langmuir Laboratory, Cornell University, Ithaca, NY, U.S.A.

T.J.TEYLER

Neurobiology Program, Northeastern Ohio Universities College of Medicine,
Rootstown, OH, U.S.A.
    Recent studies of event-related potentials (ERPs)
 have stressed the value of these measures as possible
 indicators of cognitive and linguistic functioning (e.g.,
 Teyler et al. 1973, Brown et al. 1973, Johnston and
 Chesney 1974, Megela et al. 1977). This research has
 suggested that linguistic meaning can influence ERP
 waveforms, implying that semantic as well as sensory
 characteristics of stimuli may be reflected in electro-
 physiological  recordings  from the human scalp.

    We designed  an experiment  allowing  a  more
 stringent test of the hypothesis that underlying pro-
 cessing of word meaning may be detected in the ERP.
 In this study, ERPs were recorded in two classes of
 words, synonyms (words nearly identical in meaning
 but different in physical characteristics) and homo-
 nyms  (words identical in physical  characteristics but
 different in meaning).  If ERPs reliably encode lin-
 guistic meaning, then synonyms should elicit poten-
 tials with similar waveforms, and  homonyms should
 elicit potentials with dissimilar waveforms. Since the
 P3 component is  related to semantic and cognitive
 characteristics (Sutton et al. 1965, Friedman et al.
 1975), while the  N1P2 component of the ERP is
 related to  sensory characteristics of stimuli, it was
 predicted that semantic meaning would influence the
 P3.
  * Support for this research was provided  by the
  National Institutes of Mental Health for A.L.M. and
  by the Stimson Fund of Harvard University for T.J.T.
  The work was done at  Harvard University. The assis-
  tance of Paul Clark and R.H. Dworkin is gratefully
  acknowledged.
Method

    Six right-handed females participated in the ex-
periment. They were seated in a comfortable chair
in  a sound-deadening recording chamber and were
instructed to remain alert and to avoid eyeblinks and
body  movements  during presentation of stimuli.
The  experimental  session lasted  approximately 1
hour.

    Scalp electrical activity was recorded by Beck-
man  miniature  Ag/AgCl  biopotential  electrodes
mounted in a flexible lycra cap  worn by  the sub-
jects.  In three subjects, electrodes were placed over
left and right frontal  areas (at locations equidistant
between T3, F3, and C3 for the left lead, and be-
tween T4, F4, and  P4 for the right lead), and in the
other three subjects, over left and  right  temporal
areas  (equidistant between T3, C3, and P3 for the
left lead, and between T4, C4, and P4 for  the right
lead).  These  scalp placements may reflect activity
near presumed language areas of  the brain. Linked
earlobes were  used as reference, and a forearm elec-
trode was used  as ground. Eye movements were
recorded  by disk  electrodes placed lateral to and
underneath the right eye.

     Stimuli were  common nouns, equated for fre-
 quency and familiarity.  Synonyms were rated as to
 degree of similiarity  in meaning, and homonyms as
 to degree of difference in meaning by 30 students
 in an introductory psychology class.  On the basis
 of these ratings, one synonym pair and one hom-
 onym pair were chosen  to serve as stimuli.  Short
 phrases were constructed with these stimulus words
 occurring in the last position of the phrase. Stimulus

-------
268

phrases are presented in Table 1. In addition, a short
phrase  identical to the synonym phase, but ending
with a different stimulus word, was constructed to
ensure  that any ERP waveform  similarities between
synonyms were not due to the fact that the syno-
nyms were preceded by identical phrases. All phrases
were equated for  number of syllables and were  pre-
sented to each subject  at the beginning of the experi-
ment in order to make  their meanings clear.
           Table 1.  Stimulus Phrases
                                         OCK
  Synonym:
                THROW THE SMALL
  Homonym:
              SHUFFLE THE-

              WALKONTHE-

Phrase control: THROW THE SMALL -  BOOK
    Stimulus phrases were tape-recorded in a random
order  with 8-  to  12-second interstimulus  intervals.
The phrases were then played to the subject through
an overhead loudspeaker. Each phrase was presented
30  times. Although stimulus volume in the two ears
can be affected by shifts in head position and  thus
can produce different ERPs from  the two  hemi-
spheres (Regan  1972), the  use of randomized stim-
ulus presentations in this  study makes it unlikely
that the mode  of  stimulus  presentation affected the
results.
                              Megela & Teyler


    EEC was  amplified  through a Grass Model  7
polygraph with a bandpass of 1  to 75  cps  and was
recorded, along with pulses signalling stimulus onset
generated by Massey-Dickinson programming equip-
ment, on FM  tape for  off-line computer  analysis.
ERPs were  obtained for  stimulus words only.  EEC
was averaged at a sampling rate of 512 cps for 1 sec.
In  addition,   product-moment   correlations   were
computed between pairs of waveforms and for peak-
to-peak   and   peak-to-prestimulus baseline ampli-
tudes within the  following  time windows:   100 to
200 msec (N1P2) and 250 to 400 msec (P3).

Results

Correlational analyses

    Correlations  were  obtained  for each subject at
each electrode site between  the following  stimulus
pairs: synonym-synonym, homonym-homonyn, and
synonym-phrase control word for each  time window
(N1P2,  P3, and   entire  ERP).  Correlations  were
also  computed over  the entire waveform  between
right  and left hemisphere leads for each stimulus.
These data  are presented in Table 2 for both frontal
and  temporal  leads. Each entry  represents a  mean
correlation  over  three  subjects. At  all electrode
sites, the  highest correlations  were obtained be-
tween the homonym pair, especially for the 250  to
400-msec time segment (P3). Correlations for  other
pairs show essentially random variation; the several
positive  correlations between  synomyn  I  and the
phrase control word may be attributable to the fact
that  these words shared a  common final sound.
                              Table 2.  Product-Moment Correlations
                               between Stimulus Pairs for All Leads
Left frontal
Stimulus pair
Synonym
Synonym
Synonym
1 -synonym
2
1 -phrase control
2-phrase control
Homonym-homonym
N1P2
-0.
0.
-0.
32
47
37
0.18
P3
-0.09
0.09
0.12
0.61
Entire
0.20
0.35
-0-10
0.37
Left temporal
Synonym
Synonym
Synonym
1 -synonym
2
1 -phrase control
2-phrase control
Homonym-homonym
0.18
0.01
-0.26
0.20
•0.24
0.15
-0.14
0.46
0.02
0.11
0.05
0.34
• i 	 •"-
Right frontal
N1P2
-0.19
0.36
0.09
0.19
P3
-0.10
-0.04
0.04
0.46
Entire
0.15
0.22
0.04
0.30
Right temporal
i
-0.05
0.01
0.04
0.37
_ — —•• — —
-0.01
0.23
0.04
0.43

0.09
0.06
0.07
0.23


-------
ERPs and Linguistic Stimuli
                                            269
Over the entire  waveform,  the  mean  correlations
between hemispheres were very high, ranging from
0.61  to  0.74  for both frontal and temporal leads.

    Correlations  were  transformed  into  z-scores
(eight correlations per time segment per subject) and
then  analyzed  by  a three-way repeated  measures
analysis  of  variance (Dixon  1975) with  electrode
placement  (frontal  versus temporal)  used  as  the
grouping factor and  stimulus pair and  hemisphere
used as repeated  measures.   For each  of the three
analyses  of variance performed, results were similar.
A significant main effect of stimulus pair was found
(early  segment: F(3,12)  = 4.87,  p < .02; late  seg-
ment:  F(3,12) = 9.07, p <  .002; entire waveform:
F(3,12) = 4.44, p <  .03).  Analysis of  the mean z-
transformed correlations  showed  that  these  were
highest in  all  cases  for  the homonym-homonym
comparison.  No significant  group (frontal vs tem-
poral) or hemisphere effects were found.

Amplitude analyses

    Peak-to-peak  amplitudes  of N1P2 and P3 were
computed for  each  stimulus word  at each electrode
location  for all subjects.  These  amplitudes  were
then  analyzed  by  a  three-way  repeated-measures
analysis of variance with electrode placement (fron-
tal vs temporal)  as  the grouping  factor and hemi-
sphere and  phrases as repeated measures.  For the
N1P2  components,  a significant  group  by hemi-
sphere interaction appeared:  over all phrases, ampli-
tudes  at the  left temporal  lead  were  higher than
amplitudes at the right temporal lead, F(l,4) = 8.49,
p <  .05.  Similarly, for  P3  components, left tem-
poral amplitudes were  higher  than  right  temporal
amplitudes,  F(l,4) =  11.28,  p <  .05.  Table 3 pre-
sents means and  standard deviations of amplitudes
at each  electrode site for both early and late com-
ponents.  These results suggest that larger  amplitude
ERPs  to semantic stimuli are evoked  over the left
temporal hemisphere than over the  right  temporal
hemisphere.   Since  a  nonlinguistic control group
was not included in  this study,  however, it is un-
clear whether this finding represents a specific seman-
tic effect or a generalized effect to all stimuli.

         Table 3. Laterally Comparison of N1P2
               and P3 Amplitudes (juV)a

N1P2
Frontal
P3
N1P2
Temporal
P3
Left hemisphere
6.95(2.18)
6.29(1.42)
6.67 (2.97)
8.07(2.6)
Right hemisphere
6.45(1.69)
5.93(1.82)
4.81 (2.59)
4.73(1.57)
    Sample  ERPs  from two  subjects are presented
in Fig. 1.  This figure shows the similarities  between
ERPs to homonyms, and the dissimilarities  between
ERPs to  synonyms.   Moreover, ERPs to  the same
stimulus recorded at left and right hemisphere leads
had very similar waveforms.
LFk^ y\ ^-~^
Lrf N — ^ V-- '
RF-L -^ \ S\s*^- -i
1 ^ /\ _/\
*• v ^
ni>- \s - \ •+ v «
1 HOMONYM 2
1 Tl -r\ 1 >> xx /^ / \
L|r-"V "-1 lt-^__ "v\/J ^-^S V
FT.I . A,, , , ^v J - . ,^/X^.^.
  "Each amphludt raproami lh« nwan ovtr thru lubjactl. Standard dtviitiont
   in partnthttti.
        \7v             ~-            	k
            1       HOMONYM      2   100msec1

 F(y. 7. 5«wp/e £7?ft ro sy/ionym 7, synonym 2, /iom-
 0/j.ym 7, a«d homonym 2 from frontal and tem-
 poral leads. Waveforms were traced from photographs
 of a computer slave screen. Note the similarities in
 waveshapes of ERPs to homonym pairs and the dis-
 similarities in  wave shapes of ERPs to synonym pairs.
 Frontal ERPs are one subject and temporal ERPs are
 from a  different  subject.  Vertical  lines  represent
 stimulus onset.  LF: left frontal; RF: right frontal;
 LT: left temporal;RT: right temporal.
 Discussion

     This study demonstrates that lower order sensory
 effects, related to physical or  acoustic  features of
 stimuli, are  more  important  than semantic effects
 in  influencing ERP waveforms.  ERPs were more
 similar between words of different meanings  but
 similar physical characteristics than between  words
 of similar meanings  but different physical charac-
 teristics.  These results held  for both early (N1P2)
 and  late  (P3)  components   and  for both frontal
 and temporal  ERPs.   Thus, ERPs may  not provide
 a sufficiently sensitive  metric to discriminate  higher

-------
 270

 order  semantic processing in a design where the sub-
 ject is relatively passive and is not required to make
 cognitive or linguistic decisions.  Future research on
 these  questions should incorporate behavioural tasks
 designed  to actively engage the language processing
 centers. For instance, Megela et al. (1977) reported
 a  more reliable semantic effect on ERPs recorded in
 a habituation/generalization  paradigm.
    No significant hemispheric asymmetries at fron-
tal scalp  electrode locations were  found.  On  the
other hand, both  N1P2  and P3 amplitudes over all
phrases were  larger over  the  left  temporal  hemi-
sphere than  over  the  right temporal  hemisphere.
These  differences  were  not of large  magnitude, al-
though  they  were consistent  across  subjects  and
across individual stimuli.
                               Megela & Teyler

     In summary, it appears that consistent semantic
effects  on  ERPs  are difficult  to distinguish.   In
these data,  scalp electrophysiological  activity  was
influenced predominantly  by  sensory rather  than
semantic characteristics of stimuli.

Summary

    An  experiment  designed to  test semantic in-
fluences on  event-related potentials using synonyms
and homonyms was performed.  Correlational analy-
ses  between  pairs of waveforms showed that  con-
sistent patterns were  found  between homonyms,
while ERPs  to  synonyms were  not  significantly
correlated.  For all stimuli, both early and late com-
ponents  were  of  higher  amplitude  over the  left
temporal  hemisphere  than  over the right temporal
hemisphere.  Future research on semantic correlates
of ERPs  should  utilize  paradigms requiring active
linguistic processing by the subject.

-------
VISUAL  EVOKED POTENTIALS TO  LANGUAGE  STIMULI
IN CHILDREN  WITH  READING DISABILITIES1
 S.A. SHELBURNE, JR.
College of Medicine, University of Cincinnati, Cincinnati, OH, U.S.A.2
   Shelbume (1972,  1973)  investigated visual evoked
 potentials (VEPs) to language stimuli in normal adults
 and children. In the 1973 study, normal children were
 presented  letter  stimuli  in  sequence  as  consonant-
 vowel-consonant  (CVC)  trigrams.  The task  was to
 decide whether the CVC formed a word or a nonsense
 syllable. When  the  child performed well on  the task,
 VEPs to the third-position stimulus (last consonant in
 the  CVC sequence) had greater  positive amplitude than
 VEPs to first-position (initial consonant)  and second-
 position (vowel)  stimuli.  If the child could not  dis-
 tinguish between  word and nonsense syllables, no such
 difference  in VEPs was  found. It was suggested  that
 VEP differences were related  to the children's ability
 to make a correct decision in the problem-solving task.

    Results  from previous  studies suggest children with
 reading disabilities (dyslexia) would perform poorly on
 the task  and  show  no differences  in  VEPs from
 third-position  stimuli  compared to  VEPs from first-
 and second-position  stimuli. The  present  experiment
 was undertaken to test these hypotheses.

 Methods

    Nine male children  aged 9  to  14  (mean age, 11
 years)  were  paid  participants  in the  study. Children
 with relatively pure  reading disabilities  (i.e., no  evi-
 dence of neurological, sensory,  or psychiatric handicap)
 were selected on  the basis  of  evaluation  by the  Cin-
 cinnati  Center for  Developmental Disorders.  Screening
 included  complete medical history and  physical  ex-
 amination, extensive  psychological testing, speech  and
 hearing evaluation,  psychiatric and neurological   con-
 sultation, and school  achievement evaluation.

    Testing  procedures  have been  detailed  elsewhere
 (Shelburne  1972, 1973). Visual stimuli  consisting of
 white letters  on a  black background were  presented
 sequentially to form  consonant-vowel-consonant (CVC)

 'Study supported by U.S.  Public  Health  Services Grant No.
 HD05221.
^Present  address: Department of Neurology, Children's Hospital,
 National Medical Center, Washington, D.C. 20009.
trigrams. Each  trial consisted  of the presentation of
blank-C-V-C-blank at fixed 1-sec  intervals with a 3-sec
intertrial interval. CVCs  formed either words or paired
nonsense syllables with the same first  two letters as the
word. Trial runs consisted of the randomized presenta-
tion of  50 words and  50  paired nonsense syllables.
Subjects completed  four runs and were retested in the
same manner  1 year later.

   The  child  was seated in  a comfortable chair in a
dark room.  He  was instructed to observe the letters
during each  trial and  to decide whether or  not the
CVC was a  word.  After the  second blank the  child
pushed  a toggle switch to  the right  (word) or left
(nonsense syllable). Immediate auditory feedback in-
dicated  whether  the  response was correct  (tone) or
incorrect (buzzer).

   EEC was  recorded using  a  1.3-sec time constant at
Cz, P3,  and  P4  referred to linked  ears. During  the
initial session, eye  movements were  recorded between
electrodes at the  right outer can thus and below  the
right  lower  lid. Subjects were  retested  a year  later.
Vertical  and  horizontal eye  movements were recorded
separately during the second session.

   VEPs for  each  position  in the CVC trigram were
averaged separately. 95% confidence intervals were also
calculated  for  each point in the averages.  Point-by-
point comparisons between waveforms were  performed
to  determine   the  latencies  at which  statistically
significant (0.05   level)  differences  occurred.  The
criteria  for scoring  VEP differences for individual sub-
jects was the occurrence of three consecutive points in
averaged  waveforms  where  the 95%  confidence in-
tervals of first-  or  second-position VEPs did  not  over-
lap with the  third-position VEP. See Shelburne (1972,
 1973) for further description of  this  method.

Results

   No  differences  in  VEPs for words compared to
nonsense stimuli were  observed. Subsequent  compari-
sons were pooled across these stimulus categories. Fig.

-------
 272
                                         Shelburne
1 illustrates VEPs to first- and third-position stimuli in
four dyslexic children. No consistent difference in VHP
amplitude  attributable to  CVC position  was observed
in the learning disabled group. Table 1 shows perform-
ance  scores and VHP latency differences observed  in
individual  dyslexic  subjects  during the  first  session
(runs  14) and  second  session (runs 5-8) recorded a
year later. One subject (S.M.) was not  retested. The
results indicate some improvement  in  performance of
six  children during retesting, although the error rate of
the dyslexic group (31.6%) was still high  compared to
that of normal children (12.1%).
          Cz-NONSENSE
          N-200
                            •FIRST POSITION
                            • THIRD POSITION
    J.M.
                                                        O.S.
         48% ERRORS
                                                             31% ERRORS
   M.H.
                                                        D.J.
          51% ERRORS
                                                             45% ERRORS
                                                                               IMC
  Fig. 1. Mean of 200 vertex VEPs to word and nonsense syllable stimuli in four children with reading disabilities.
                         Table 1. Individual Data for Children With Reading Disabilities
Subject
O.J.
L.H.
D.H.
D.S.
O.J.
J.M.
LB.
M.H.
S.M.
Age8
9
12
9
14
11
9
10
11
13
Runi 1-4
% error?
45
25
28
30
48
48
41
51
13
Latency window of
VEP differences6
.
.
.
.
410-550
.
390-680
.
-
Runs 5-8
% errors
36
11
11
22
36
49
36
52

Latency window of
VEP differences
.
-
560-850
-
•
•
450-590
-
Not done
              •Age at initial testing.
              ^Latencies define the beginning and end of the epoch during which significant VEP amplitude
              differences (third position > first or second position) were observed.

-------
VEPs, Reading Disabilities
                                           273
       D.H.
       N-200
                           FIRST POSITION
                           THIRD POSITION
                        RUNS 1-4
                                                                       RUNS 5-8
    Cz
                                                                     900 m»«c
Fig. 2. Mean of200 VEPs to word and nonsense syllable stimuli for subject D.H. Runs 5-8 were performed 1 year after
rum 1-4.
   The data of subject  D.H. merit  comment.  Among
 the  subjects  re tested, D.H. showed the greatest im-
 provement (17%). This performance change was accom-
 panied by  the  development  of a first-third position
 VEP amplitude difference in the latency range 560 to
 850 msec. This difference is clearly shown in Fig. 2.

   Data from previous experiments  with normal chil-
 dren (Shelburne  1973) were combined with  data from
 dyslexic children. Subjects were  then  separated into
 two groups defined by the presence  or absence of VEP
 differences, irrespective  of reading  ability. The  mean
 percent errors was  7.6  for the group with VEP dif-
 ferences and  35.0 for the  group without VEP differ-
 ences (t = 7.52, df = 32, p < .001).

   VEPs of two  children (OJ. and L.B.) in the present
 study exhibited eye movement  contamination (Fig. 3).
 Differences in averaged eye movement occurred at  the
 same  latencies  as VEP  differences. VEP  differences
 listed in  Table  1 for these subjects, therefore, can be
 attributed to eye movement artifact  Other dyslexic
 subjects showed  very little  eye  movement. As a group,
 dyslexic  children had  a  mean  eye movement rate
 (measured grossly on the electro-oculogram) of 16/min
 compared to a rate of 25/min for normal children.

 Discussion

   Word  and nonsense  CVC trigrams  were presented
 visually to children  with reading disabilities. In pre-
 vious experiments (Shelburne  1973), normal children
who  performed  well  on  this  problem-solving  task
showed greater positive amplitude of VEPs from third-
position  stimuli  than  VEPs  from  first-  or  second-
position stimuli. With the exception of subject D.H. in
runs  5-8, the  reading disability  children  showed no
significant VEP differences when eye movement arti-
facts were excluded. In contrast,  17  out of 20 normal
children  tested previously showed significant VEP dif-
ferences. The  absence  of VEP  differences in  both
dyslexic and normal children was associated with poor
task performance. Dyslexic children as a group had far
more difficulty than normal children in  discriminating
between words and  nonsense syllables.

   Eye movement contamination  poses a serious prob-
lem in ERP research, particularly in VEP studies of
children  (Shelburne  1973). Analysis  of eye movement
patterns  indicated that dyslexic  children made fewer
eye movements than normal children made during the
task.  Eye movement patterns may,  in  fact, provide a
useful measure  to  differentiate  children with reading
disabilities from normal  children who perform poorly
on  the task and show no  VEP differences. The total
number of eye  movements  and amount of EOG arti-
fact appears to be  less for  the  dyslexic than for the
normal child.

   Is  there  a  causal relationship between eye  move-
ments and reading ability?  Critchley (1964) and Gold-
berg and Arnott (1970) both concluded that faulty eye
movement  patterns  were the result, rather  than the
cause,  of   reading disabilities.  Reading  disabilities

-------
     274
                                       Shelburne
           L.B.
                                    N-200
                   • FIRST POSITION
                   •THIRD POSITION
                          •SECOND POSITION
                          •THIRD POSITION
     Cz
                                                                                          •--•.BIUV
    EYE
                                                                                           ••-..*•
                                                                         900 mMC
Fig 3 Mean of 200 VEPs to word and nonsense syllables. Each channel is computer-averaged eye movement recordings
from lateral canthus and infra-orbital ridge. Calibration for VEPs only.
 become less apparent with time and remedial therapy.
 One  subject (D.H.) showed  improvement in  perform-
 ance  associated with  development of  normal VHP
 differences after a 1-year interval. A much larger study
 of VEPs in children  with reading disabilities  is cur-
 rently in progress.
 Summary

   Visual  stimuli  in  the  form  of consonant-vowel-
 consonant (CVC) word and nonsense syllable trigrams
were  randomly presented  to  nine  male children with
specific  reading  disabilities  (dyslexia).  In  previous
experiments,   normal  children  who   could   decide
whether or not the CVC formed a word showed greater
amplitude VEPs to third-position stimuli. This  differ-
ence appears  to be related to the subjects' ability to
solve the problem and resolve  the uncertainty. Children
with  reading  disabilities  were  characterized by: (1)
poor  performance  on the word  task, (2) no VEp
differences between third-position  responses and first-
and second-position  responses,  (3) low averaged eye
movements, and (4) low total  eye movement rate.

-------
CHOICE OF  ACTIVE ELECTRODE SITE AND RECORDING
MONTAGE AS VARIABLES AFFECTING  CNV AMPLITUDE
PRECEDING SPEECH1


S.H. CURRY,2 J.F. PETERS,2 AND H. WEINBERG

Simon Eraser University, Burnaby, British Columbia, Canada
   Previous  investigations  concerning  the  electro-
physiological  correlates of  language production have
focused on two  separate  but  related aspects of brain
electrical activity:  the motor  potential (MP) and  the
contingent  negative variation  (CNV).  The MP, which
has  received  the  bulk  of research  attention,  was
examined initially  by  Ertl and  Schafer (1967) and
Schafer (1967).  Although these  authors were able  to
record "nonrandom cortical activity" prior to onset of
vocalization,  they  did  not discuss their  findings  in
relation  to  hemispheric  dominance.  Subsequently,
Me Adam  and   Whitaker  (197 la)  recorded   similar
potentials preceding speech, the amplitudes of which
were  asymmetrical. They  presumed  that  the hemi-
sphere  showing  the largest  negative  shift was  the
dominant hemisphere  for speech and language. Their
data have been  criticized on  methodological  grounds
by  Morrell  and  Huntington  (1971),  Szirtes  and
Vaughan  (1973),  and  Grabow  and  Elliot  (1974).
Grozinger et  al. (1975) have examined  in  detail  a
number  of  these criticisms and  concluded that  with
appropriate  controls, interhemispheric asymmetries of
brain potentials  preceding speech can be demonstrated.

   Assuming that the fixed  foreperiod CNV paradigm
might  minimize artifacts  introduced  by  preparatory
changes  in  tongue  position and throat muscle tone,
Low et al. (1974,  1976)  employed  such a paradigm to
evaluate differential hemispheric involvement in speech
production. These authors conducted a series of studies
involving 51 subjects  (40 normals, 11 epileptics).  Of
particular interest are their  data from the epileptics; in
 10 of them CNV asymmetries accurately predicted the
 result  of carotid amytal testing. Although the grouped
 mean  differences  reported by  Low  et  al.  for the
 normal subjects  showed significantly larger shifts in the
 hemisphere  contralateral  to the  preferred  hand, they
 *Work supported by grants to Weinberg from Medical Research
  Council of Canada.
 * See list of contributors for current address.
were careful to point out that  not all the normals
showed  this  relationship.  Zimmerman  and  Knott
(1974) reported larger CNV's preceding speech in the
left hemisphere of 4 of 5 normal  adults, whereas only
22% of a  population  of stutterers showed  similar
findings. Large left hemisphere CNV amplitudes have
also been reported by Kostandov and Drilling (1973).
Michalewski (1975), on  the other hand, found only
isolated  incidence  of left  greater  than right CNV
asymmetries in 18 subjects.

   In summary there appears to be some evidence both
from MP and CNV studies to support  the  statement
that there is an electrophysiological correlate of speech
production  which  is consistent with clinically obtained
data.  A  review  of  the  literature  reveals,  however,
several methodological differences with regard  to the
placement of both active and reference  electrodes and
the number of replications per subject.  For example,
the  usual  procedure  has  been to  employ  active
electrodes  at  IF3  and  IF4 referenced   to  linked
mastoids  with one recording session  per subject. Low
et al., on the other hand, favor an electrode placement
termed  "inferior-temporal" (Tl, T2)  referenced   to
contralateral ear.  These  procedural  differences may
reflect critical methodological considerations since the
choice of  active   and  reference  electrode  sites and
recording montage have been shown to influence both
magnitude  and  waveform  of  the  averaged  evoked
potential (Goff et al. 1969). Furthermore,  Low (per-
sonal  communication) has found that in most subjects
asymmetries  increased  in  magnitude  over repeated
sessions.  This is a particularly interesting observation
since much of the evidence on test-retest reliability  of
the CNV is conflicting.  For example, Cohen (1969)
reported  a  correlation  of +0.80 between  two test
sessions separated  by 2  to 8 days  using 34  normal
subjects.  Straumanis et al. (1969) reported a  correla-
tion of +0.20,  whereas Roth et al. (1975)  noted a
median correlation of +0.68 between retests separated
by from 5 min to 7 days.

-------
 276

    The present investigation was  therefore designed  to
 evaluate  CNV  amplitudes  recorded  from   inferior
 frontal (IF3, IF4) and inferior temporal (Tl,  T2) sites
 employing   both  contralateral   and  linked  mastoid
 references.  Serial recordings were carried out  to evalu-
 ate  the contention by  Low et al. that the asymmetries
 became   more  prominent  with  repeated recordings.
 Behavioral  data  were also  collected from the subjects
 concerning  dominant handedness (Annett 1967) and
 cerebral  dominance  for language  (dichotic  listening,
 Kimura 1967).

 Methods

   The six  subjects were asymptomatic volunteers with
 no previous experience in  a CNV experiment. Subject
 characteristics are given in  Table  1. All subjects served
 in  four recording  sessions  separated by 1 to 2 weeks.
 Due to technical difficulties in the final data  analysis,
 data  from  only  three  sessions  for  each  subject are
 presented.

   Brain  electrical activity  and  eye  movements  were
 recorded  with a  time  constant  of 5.0 sec and  high-
 frequency cutoff of 30 Hz. Ag/AgCl  disc electrodes
 were affixed at  Cz,  IF3, IF4,  Tl and T2. Electrodes
 IF3 and  IF4 were located 11  cm  lateral and 4 cm
 anterior  to Cz  (McAdam and Whitaker 197la) while
 Tl and T2  were placed 2 cm inferior to the bisection
 of F7-T3  and F8-T4, respectively (Low, personal  com-
 munication). Pairs  of similar electrodes attached to the
 left  (Al,  A3) and right (A2,  A4) mastoid processes
 served as  references.  The reference combination A3  +
 A4 is subsequently referred to as linked (L) while Al
 and  A2  are  the  contralateraf  references.  All  inter-
electrode  impedances were  less than 2.5 k&. Beckman
biopotential electrodes  at the outer can thus and supra-
orbital ridge of the right eye referenced to L were used
to record  EOGs. EEC and  EOG were digitized on-line
and averaged over a 3.496-sec epoch. Eight artifact free
trials were used in each average.

   Subjects  were  positioned comfortably on a  bed in  a
sound-attenuated   chamber.  A   television   monitor
                                        Curry et al.

 located approximately 4  m in front of the subject was
 used for  tachistoscopic  presentation of visual  stimuli.
 In the word speak (WSP) and word button press (WBP)
 conditions, SI was a 3- to 6-letter word presented for
 50 msec  followed 1500 msec later by a question mark
 (?), which served as S2. In the WSP condition, subjects
 were instructed to vocalize the word presented at SI as
 rapidly as possible  after S2. In the WBP condition, the
 subject pressed a button  held in the preferred hand to
 terminate S2. In the standard (STD) condition, SI was
 a 1000-Hz tone  burst presented for 150 msec followed
 1500  msec  later by S2,  a series of  10  c/s  clicks,
 terminated  by  button press.  Both  SI  and S2 were
 presented via loudspeakers placed under the bed. In all
 conditions,  the intertrial  interval  varied  randomly be-
 tween 5  and  30 sec. The three  conditions were run
 twice  on  each experimental day with the sequence of
 conditions randomized across subjects.

   Hand preference was  quantified with  a modification
 of the Annett  questionnaire  and  scoring  procedure
 (Annett   1967).  The  12  test  items concerning hand
 usage for various actions  were  retained and scored as
 +1 or -1 for right and left hands, respectively, while
 mixed usage  was scored  zero. The range  of  possible
 scores was +12 to -12.

   Cerebral  dominance for language was assessed twice
 for each subject  with a dichotic listening task (Kimura
 1967).  Each  test was separated  by at least 2  weeks.
 Fifty  sets  of three pairs of  digits  from 1  to  10
 (omitting  6  and 7) were recorded on tape with  a
 500-msec   interpair   interval  and  a  10-sec  interval
 between sets (Neville 1974). The digits were presented
 through Hosiden  stereo earphones connected to  a Sony
 TC-270 tape  unit. To control for possible auditory
 inequalities, subjects equated the intensity of 30 sec of
 white  noise  in  the  left and  right  earphones.  In
 addition,  the  earphones  were reversed for trials 24 to
48. The  first two sets  of digits  were always practice
 trials. Subjects were instructed to listen  to  each set of
digits and verbally  report as many as possible,  in any
order,  during  the  10-sec  interval.  Responses that
showed a left or right ear superiority were scored as +1
                                       Table 1. Subject Characteristics
Subject
LAT2
LAT3
LAT4
LATB
LAT7
LAT8
Age
22
25
21
21
25
30
Sex
M
M
F
M
F
F
Hand
R
R
R
L
L
R
Annett8
+ 9
+12
+11
- 5
-12
+ 8
Dichotic 1b
+ 1
-17
-12
-11
-24
•13
Dichotic 2*>
+ 6
-10
•12
-10
-17
•10
                       •Annett valu« range from +12 to -12 with + values indicating preferential
                        right hand uug«.
                       ODIchotte icore* have epoiilblerenge from+48 to -48 with -tcore* Indicating
                        right ear preference, I.e. left hemiiphere tuperiority.

-------
CNV, Electrode Site, and Speech

and  — 1, respectively. A zero was assigned  if both ears
were equally accurate. Scores were  summated for each
subject and could range from +48 to -48.

   CNVs were  quantified as the mean  amplitude over
the 200-msec interval prior to S2 relative to the pre-Sl
baseline. These mean values  were converted  to juV by
comparison to  a 50-MV calibration  pulse averaged  in a
manner analogous to the EEC.

Results

   Fig.  1 is an overlay  of the average CNV's obtained
at each electrode in each  condition  across three experi-
mental   days  for  subject LAT  5.  This  subject  was
chosen  for presentation as  he shows a marked differ-
ence between IF3 and  IF4  during the WSP condition.
Hemispheric  differences recorded from  the  remaining
subjects  were not as definitive in either direction.

   Table  2  contains  the grand mean  and  standard
deviation  of  CNV  amplitudes   obtained   at  each
electrode site.  Repeated  measures t tests  and analysis
of variance applied to the data showed  only significant
(p < 0.05) condition effects. As seen in Table 2, these
differences are confined  to  a significant reduction  in
amplitude  in the two-word  conditions (WBP and WSP)
as compared  to  the  standard  (STD). No significant
differences were  found  between the WBP  and WSP
conditions. Additional  analysis showed no significant
day  or electrode effects or interactions.

Discussion

   Since CNV amplitude was not significantly different
for linked and  contralateral  references, there appears at
this  time  to be no reason  for  favoring one  reference
system over the other.

   Furthermore,  the data do not  offer  any uncom-
promising  evidence  for hemispheric differences in the
CNV attributable  to preparation for speech. Of  the
three subjects  showing larger CNVs  in the left hemi-
sphere during speech production, two were left-handed;
however, they  both showed  left-hemisphere preference
in the dichotic listening  task. That the same subjects
showed consistently larger CNVs in the left hemisphere
in both the WSP  and WBP conditions  suggests  that
these asymmetries are not related to speech production
perse.   Rather, the results  may  be peculiar to any
paradigm employing a word as  SI. The data presented
by   Weinberg et  al. (this  volume) support  this sug-
gestion.

   Although the relationship between right-handedness
and left hemisphere lateralization  for speech  is geh-
erally   accepted,  the  nature of the  relationship for
non-right-handers  and  persons  with  mixed-handedness
is unclear (Newcombe and  Ratcliff 1973,  Branch et al.
                                              277
      Table 2. Grand Mean CNV Amplitudes
        Obtained Across Subjects and Days for
          Each Electrode in Each Condition
Electrode!
IF3-A2

IF4-A1

IF3-L

IF4-L

T1-A2

T2-At

CZ-A1

CZ-A2

CZ-L

Condition
Std
•4.62
±3.338
-4.86
±3.40
•4.02
±2.96
-2.98
±2.74
•1.46
±2.31
-1.54
±2.84
-17.45
i 7.69
-16.78
± 7.74
-16.72
± 4.41
WBP
-1.23
±4.23
-0.01
±3.27
-238
±3.57
-0.62
±1.72
3.47
±3.07
2.49
±4.25
-9.45
±4.09
•8.78
±4,49
-9.67
±3.07
WSP
-1.87
±2.98
-0.38
±2.89
-2.16
±3.30
-0.07
±4.22
0.44
±2.28
0.60
±2.70
-10.70
± 4.77
-10.18
± 4.58
-11.11
± 4.31
"Bottom line of each row is the standard deviation for each mean.
1964). Similarly, Briggs and Nebes (1976) have shown
a high  incidence  of  right-ear superiority for dichotic
listening in a population of left-handers. As a result, a
priori  assignment  of subjects  into groups  or  data
analysis performed on the  basis of these variables may
result  in more  artifact than  fact. We propose  that
subsequent  research   would  be  more  profitable  if
analyses were  done first on a single-subject, single-trial
basis. For  each  subject the direction, magnitude, and
variability   of  any   asymmetries  would  then   be
quantified. As a  second  step,  single-trial  data of  all
subjects could be subjected to a classification  analysis
to form groups based on clusters of dependent variables.
Electrocortical variables relating to cerebral dominance
for  speech production would  then be  determined
empirically rather  than being forced into the mold of
pre-existent  assumptions.   As  was  recently  noted,
perhaps we should be reasoning  from  the  waveform
outward rather than from the event inward.

   This  study definitely points to the necessity  for
further  detailed investigation  of  the  methodological
considerations raised.

-------
                           STD
IF4-L
IF3-A2
IF4-A1
Cz-L
Cz-AI
 Cz-A2
 T1-A2
 T2-A1
oe*^

                                                                           WBP
                                                                                        S2
                                                                                                                            WSP
                                                                                                                                                                  00
                                                                                                                    SI
           Fig.  I. Overlay of averaged waveforms obtained on successive runs for subject LAT5. Negativity at the active electrode is seen as an upward

           deflection. See text for location of reference electrodes.
                                                                                                                                                     s
                                                                                                                                                     e.

-------
CNV, Electrode Site, and Speech                                                                   279

Summary                                             asymmetries  were found when speech and  nonspeech
                                                       paradigms were  compared regardless of active electrode
   CNV  amplitudes were  evaluated at inferior  frontal    site  or reference  montage. Significantly smaller  CNV
and  inferior temporal  recording sites employing  both    amplitudes were seen at all electrodes  when a  word
contralatcral and  linked  mastoid  references in  three    was  used as  SI. The need  for further methodological
different   conditions.   No   significant  hemispheric    studies is emphasized.

-------
 LANGUAGE  REFERENCES
 Albert,  M.L. Auditory  sequencing  and  left cerebral
    dominance  for  language.   Neuropsychol.,  1972,
    10:245-248.

 Anderson, S.W. and Jaffe,  J. Eye movement bias and
    ear preference  as indices of speech lateralization in
    brain.  Scientific report #15,  Department of Com-
    munication  Sciences,  New  York  State Psychiatric
    Institute,  1973.

 Annett,  M. The binomial distribution of right, mixed
    and left handedness. Quart.  J. Exp. Psychol., 1967,
    19:327-333.

 Barlow,  H.B. Single  units  and  sensation:  A neuron
    doctrine   for  perceptual  psychology?  Perception,
    1972, 1:371-394.

 Barnet, A.B. and Lodge, A. Click evoked responses in
   normal   and  develop men tally  retarded  infants.
    Nature, 1967, 214:252-255.

 Barnet, A.B., De Sotillo, M.V. and Campos, M.  BEG
   sensory evoked  potentials  in  early  infancy  mal-
   nutrition.  Paper presented  at Society  for Neuro-
   science, St. Louis, 1974.

 Bartlett,  F.C. Thinking. Basic Books, New York,  1958.

 Beagley,  H.A.  and Knight, JJ.  Changes in auditory
   evoked response with intensity. J.  Lar. Otol.,  1967,
   81:861-873.

 Begleiter,  H. and Platz,  A.  Cortical evoked potentials
   to  semantic    stimuli.   Psychophysiol.,   1969,
   6:91-100.

 Bever, T.J. and Chiarello, R.J.  Cerebral dominance in
   musicians   and   nonmusicians.    Science,  1974,
   185:537-539.

 Blank, M. Mastering the intangible through language. In
   D. Aaronson and R.W. Rieber (Eds.), Developmental
   psycholinguistics  and communication disorders. New
   York Acad. Sci.,  1975, 263:44-58.

Bogen, J. Wernicke's  area-where is it?  Presented  at
   New York Academy of Sciences Conference on  the
   Origins and Evolution of Speech and Language, New
   York, Sep  1975.

Bogen, J.E. and Bogen, G.M. Wernicke's area-where is
   it? In  S.  Harnad,  H.D.  Steklis  and  J.  Lancaster
   (Eds.),  Origins   and  evolution of language   and
    speech.  Ann. New  York  Acad. Sci.,  New York
    1976, 280:834-843.

 Branch, C., Milner, B.,  and Rasmussen, T.  Intracarotid
    sodium-amytal  for  the lateralization  of  cerebral
    speech dominance. J. Neurosurg., 1964, 21:399-405.

 Briggs, G.G. and Nebes, R.D. The effects  of handed-
    ness, family history  and sex on the performance of
    a  dichotic listening  task. Neuropsycholotia  1976
    14:129-133.

 Brown, W.S., Marsh,  J.T.  and Smith,  J.C.  Contextual
    meaning  effects on speech-evoked potentials Behav
    Biol., 1973,9:755-761.

 Brown,  W.S.,  Marsh,  J.T. and  Smith, J.C.  Evoked
    potential  waveform  differences produced  by  the
    perception of different  meanings of an ambiguous
    phrase. Electroenceph.  clin.  Neurophysiol   1976
    41:113-123.

 Buchsbaum,  M.  and Fedio, P. Visual information  and
    evoked responses  from   the  left and right  hemi-
    spheres.  Electroenceph.  clin.  Neurophysiol.  1969
    26:266-272.

 Buchsbaum,  M.  and Fedio, P. Hemispheric  differences
    in evoked potentials  to  verbal and nonverbal stimuli
    in the  left and right  visual fields. Physiol. & Behav
    1970, 5:207-210.

 Butler, R.A., Spring, M.   and Keidel, W.D.  Stimulus
   repetition rate factors which influence the auditory
   evoked  potential  in  man.  Psychophysiol.   1969
    5:665-672.
Butler,  S.R.  and Glass, A.  Asymmetries in the CNV
   over  left and right hemispheres  while subjects await
   numeric information. Biol. Psycho., 1974, 2:1-16.

Butler,  S.R.  and  Norrsell,  U.  Vocalization possibly
   initiated by  the minor  hemisphere.  Nature,  1968
   220:793-794.


Calearo, C. and  AntoneJli, A,R. "Cortical"  hearing tests
   and   cerebral   dominance.  Acta  Otolar    lOfii
   56:18-26.

Callaway,  E. and  Buchsbaum, M. Effects  of cardiac
   and  respiratory  cycles  on  averaged  visual  evoked
   responses.  Electroenceph. clin.  Neurophysiol
   19:476-480.                               "

-------
Language References
                                              281
Callaway,  E. and  Halliday,  R.A.  Evoked  potential
   variability: Effects of age, amplitude, and methods
   of measurement. Electroenceph. clin. Neurophysiol.,
   1973,34:129.

Cannon, A. and Nachshon, I. Effect of unilateral brain
   damage on  perception  of temporal order. Cortex,
   1971,7:410-418.

Chapman,  R.M.  ERPs and  Language. Transcript  of
   panel at  Fourth International Congress on  Event
   Related Slow Potentials  of  the  Brain  (EPIC IV),
   D.A. Otto, Program Chairman, Hendersonville, N.C.,
   April 1976.

Chapman,  R.M. Evoked potentials of the  brain related
   to  thinking. In F.J. McGuigan  & R.  Schoonover
   (Eds.),  The psychophysiology of thinking:  Studies
   of  covert  processes.  Academic   Press,  New York,
   1973:69-108.

 Chapman, R.M. Latent components of average evoked
   brain responses  functionally  related to information
   processing. In International symposium on cerebral
   evoked potentials in man (precirculated abstracts).
   Presses    Universitaires    de   Bruxelles,   Brussels,
    1974a:38-42.

 Chapman,  R.M.  Semantic  meaning  of  words  and
   average  evoked  potentials.  In   International  sym-
   posium on  cerebral evoked potentials in man (pre-
   circulated   abstracts).  Presses   Universitaires  de
   Bruxelles, Brussels,  1974b:43-45.

 Chapman, R.M., Bragdon, H.R., Chapman, J.A. and
   McCrary,  J.W.  Semantic  meaning of words  and
   average  evoked  potentials.  In  J.  Desmedt (Ed.),
   Prog.  clin.  Neurophysiol., Vol.  3: Language  and
   hemispheric specialization in man: Cerebral event-
   related potentials. Karger, Basel, 1977:3647.

 Chapman, R.M.,  McCrary,  J.W., Bragdon,  H.R. and
   Chapman, J.A. Latent components of  event-related
   potentials  functionally  related to  information  pro-
   cessing. In  I.E.  Desmedt (Ed.),  Progr.  clin. Neuro-
   physiol.,  Vol. 6: Cognitive components in  cerebral
   event-related  potentials  and  selective  attention.
   Karger, Basel, (in press).

 Chapman, R.M., McCrary, J.W.,  Chapman,  J.A.  and
    Bragdon,  H.R. Brain responses related  to  semantic
   meaning.  Brain and Lang., 1978,  5:195-205.

 Chflders, D.G., Ross, J.J., Perry, N.W. and Nevis, A.H.
    Dyslexia  and cortical  correlates. VIII  Iscerg Sym-
    posium, Pisa  (Italy), September 7-12,  1970:35-51.

 Chomsky, N.  Aspects of the theory of syntax. MIT
    Press, Cambridge, Mass., 1965.

 Cohen, B.D., Noblin,  C.D., Silverman, A.J. and Penick,
    S.B.  Functional asymmetry  of  the human brain.
    Science, 1968, 162:475-477.

 Cohen,  G.  Hemispheric  differences  in  serial versus
   parallel   processing.   J.  exp.   Psychol.,   1973,
   97:349-356.

Cohen,  J. Very slow brain potentials relating to ex-
   pectancy:  the  CNV.  In  E.  Donchin  and  D.B.
   Lindsley   (Eds.).   Average   evoked   potentials:
   Methods,  results  and evaluations. NASA  SP-191,
   Washington, D.C., 1969:143-198.

Cohen,  J.  and  Walter, W.G.  The interaction of re-
   sponses in the  brain  to semantic stimuli. Psycho-
   physiol., 196, 2:187-196.

Cohn,  R. Differential cerebral processing of noise and
   verbal stimuli. Science, 1971, 172:599-601.

Connors, C.K. Cortical visual evoked response in chil-
   dren with learning disorders. Psychophysiol., 1971,
   7:418-428.

Conrad,  R. Speech  and reading. In J.F. Kavanagh and
   I.G.  Mattingly  (Eds.),  Language by ear  and eye.
   MIT Press, Cambridge, Mass., 1972:205-240.

Critchley,  M.  Developmental  dyslexia.   Heinemann
   Medical, London, 1964:40-51.

Crowell,  D.E.,  Jones,  R.H.,  Kapuniai,   L.E.  and
   Nakagawa, J.K. Unilateral cortical activity in  new-
   bom  humans:   An   early   index   of   cerebral
   dominance. Science, 1973, 180:205-208.

Darwin, C.J. Ear differences and hemispheric specializa-
   tion. In F.O. Schmitt  and F.G. Worden (Eds.), The
   Neurosciences:  Third  study program.   MIT  Press,
   Cambridge, Mass., 1974:57-63.

Davis,  A.E.  and  Wada, J.A. Hemispheric  asymmetry:
   Frequency analysis  of visual and auditory  evoked
   responses to non-verbal  stimuli. Electroenceph. clin.
   Neurophysiol., 1974, 37:1-9.

Davis,  P.A.  Effects of acoustic stimuli on the waking
   human brain. J. Neurophysiol., 1939, 2:494-499.

DeVilliers, J. Quantitative  aspects of a grammatism in
   aphasia. Cortex,  1974, 10:36-54.

Dirks,  D. Perception of  dichotic  and monaural verbal
   material  and cerebral dominance  for speech. Acta
   Otolar., 1964, 58:73-80.

Dixon,  W.J.  (Ed.)  BMDP  Biomedical Computer Pro-
   grams. Developed at  Health  Sciences Computing
   Facility, UCLA; sponsored by  NIH Special Research
   Resources  Grant  RR-3.  Univ.  of   Cal.  Press,
   Berkeley,  1975.

Donchin, E.  A multivariate approach to the analysis of
   averaged   evoked potentials.  IEEE trans, biomed.
   eng., 1966, BME-13, 131-139.

Donchin, E.  Data analysis techniques in avenge evoked
   potential research. In E. Donchin and D.B. Lindsley
   (Eds.), Average evoked  potentials: Methods, results,

-------
  282

    and  evaluations.  NASA  SP-191,  Washington,  D.C.,
    1969:199-236.

  Donchin, E., Kutas,  M. and  McCarthy,  G.  Electro-
    cortical  indices of hemispheric  utilization.  In S.
    Hamad, R.W. Doty, L. Goldstein, J. Jaynes and G.
    Krauthamer (Eds.),  Lateralization  in  the  nervous
    system. Academic Press,  New York, 1977:339-384.

  Dorman, M.F. Auditory  evoked  potential  correlates of
    speech  sound  discrimination.  Perc.  Psychophys.,
    1974, 15:215-220.

  Doyle,  J.C.,  Ornstein,  R.  and  Galin,   D.  Lateral
    specialization of cognitive mode: II. BEG frequency
    analysis. Psychophysiol., 1974, 11:567-578.

  Eason, R.G., Groves,  P.  and Boneli, L. Differences in
    occipital evoked potentials recorded simultaneously
    from both cerebral hemispheres in man. Proc.  75th
    Ann. APA Conv., 1967:95-96.

 Ellis,  L.  Slow EEC  changes  preceding speech  pro-
    duction. Dissert. Abst. Inter.,  1972,  32 (8-B):4920.

 Ertl,  J. and Schafer,  E.W.P. Cortical  activity preceding
    speech. Life ScL, 1967, 6:473-479.

 Fedio,  P. and Buchsbaum, M.  Unilateral temporal
   lobectomy  and  changes in evoked responses during
   recognition  of verbal and  nonverbal material in the
   left and right  visual  fields.  Neuropsychol.,  1971,
   9:261-271.

 Feldman, R.M. and  Goldstein,  R. Averaged  evoked
   responses to synthetic  syntax  sentences. J.  Speech
   Hear.  Res., 1967,10:689-696.

 Freeman, W.J. Waves,  pulses  and  the  theory of neural
   masses. In  R. Rosen and  F.M. Snell (Eds.), Progress
   in  theoretical biology.  Academic Press,  New York,
   1972:88-166.

 Freeman,  W.J. Mass  action  in  the  nervous system.
   Academic Press, New York, 1975.

Friedman,  D.,  Simson, R.,  Ritter, W.  and  Rapin,  I.
   Cortical evoked  potentials elicited by real  speech
   words  and  human  sounds.   Electroenceph.  clin.
   Neurophysiol., 1975a, 38:13-19.

Friedman, D., Simson, R., Ritter,  W. and Rapin, I. The
   late positive component  (P300)  and  information
   processing in sentences. Electroenceph. clin.  Neuro-
   physiol., 1975b,  38:255-262.

Friedman, D.,  Simson,  R.,  Ritter, W. and Rapin, I. CNV
   and P300  experimental paradigms  for the study of
   language.   In   J.   Desmedt   (Ed.),  Progr.   clin.
   Neurophysiol.,  Vol 3: Language  and hemispheric
   specialization  in   man:   Cerebral   event-related
   potentials. Karger, Basel, 1977:205-211.

 Furth, H.G.   On  the nature of  language  from  the
   perspective of research with profoundly  deaf chil-
                             Language References

   dren. In  D.  Aaronson  and  R. W.  Rieber (Eds.),
   Developmental psycholinguistics and communication
   disorders.   New  York   Acad.   Sci,   Vol.   263,
   1975:70-75.

 Galambos,  R.,  Benson,  P.,  Smith, T.S.,  Shulman-
   Galambos, C. and  Osier, H. On hemispheric differ-
   ences in  evoked   potentials  to  speech  stimuli.
   Electroenceph.    clin.    Neurophysiol      1975.
   39:279-283.                          "         '

 Galin,  D.  and  Ellis,  R.R.  Asymmetry  in evoked
   potentials  as  in index of lateralized cognitive pro-
   cesses: Relation  to  EEC alpha asymmetry. Neuro-
   psychologia, 1975,  13:45-50.

 Galin, D. and  Ornstein, R. Lateral specialization  of
   cognitive  mode:  An  EEC study. Psychophyriology,
Gardiner, M.F., Schulman, C. and Walter, D.O. Facults-
   tive  EEC asymmetries  in  infants  and adults.  In
   Cerebral  dominance,  BIS Conference Report  1973
   34:37-40.

Gardiner,  M.F. and Walter, D.O.  Evidence of  hemi-
   spheric  specialization  from  infant  EEC.  In  S.
   Hamad,  R.W. Doty, L. Goldstein, J. Jaynes and G.
   Krauthamer  (Eds.), Lateralization  in the  nervous
   system. Academic Press, New York, 1977:  481-500.

Gazzaniga, M.S. The bisected brain. Appleton-Century-
   Crofts, New York,  1970.

Gazzaniga,   M.S.  and Hillyard, S.A.  Language  and
   speech  capacity  of the right hemisphere. Neuro-
   psychol.,  1971,9:273-280.

Geschwind,  N.  The organization of language and the
   brain. Science, 1970, 170:940-944.

Geschwind,  N.  and Levitsky,  W. Human brain: Left-
   right  asymmetries  in  temporal   speech  region
   Science,  1968,  161:186-187.                     *

Gibson, J.J. Principles of perceptual  learning and de-
   velopment. Appleton, New York, 1969.

Goff, W.R.,  Matsumiya, Y., Allison, T. and Goff, G.D
   Cross-modality   comparisons  of   average  evoked
   potentials. In E.  Donchin and D.B.  Lindsley (Eds.)
   Average   evoked  potentials: Methods, results  and
   evaluations.  NASA   SP-191,   Washington   DC
   1969:95-141.

Goldberg,  H.K.  and  Arnott,  W.  Ocular motility m
   learning   disabilities.  J.  Learn.   Disabil     l«n/\
   3:4042.                                      '°»

Goldstein, R. and Rodman, L.B. Early components of
   averaged  evoked  responses  to rapidly  repeated
   auditory  stimuli.  J.  Speech  Hear.  Res.,   IQ<.?
   10:697-705.                                   °7»

Goldstein,  R.,  Rodman, L.B.  and  Karlovich,,  R«

-------
   Language References

   Effects of stimulus rate  and  number  on the early
   components   of  the  averaged  BEG  response.  J.
   Speech Hear.  Res.,  1972,  15:559-566.

Grabow, J.D. and Eliot,  F.W.  The electrophysiological
   assessment  of  hemispheric  asymmetries  during
   speech. J. Speech Hear. Res., 1974, 17:64-72.

Greenberg, HJ.   and  Graham, J.T.  Electroencephalo-
   graphic changes  during learning of speech  and non-
   speech stimuli.  J.  Verbal Learn.  Verbal Behav.,
   1970, 9:274-281.

Greenfield, P.M. The grammar of action in  cognitive
   development.   In D.O. Walter  et al. (Eds.),  Con-
   ference on human brain function.  BRI Publ., UCLA,
   Los Angeles,  1976:67-73.

Gross, M.M.,  Begleiter,  H., Tobin, M. and Kissin,  B.
   Auditory  evoked   responses   comparison  during
   counting clicks  and  reading. Electroenceph. clin.
   Neurophysiol., 1965, 18:451-454.

Grozinger,  B.,   Kornhuber,  H.H.  and   Kriebel,  J.
   Methodological  problems  in  the investigation  of
   cerebral  potentials  preceding  speech:  Determining
   the onset and suppressing the artifacts caused  by
   speech. Neuropsychologia, 1975, 13:263-270.

Haaland,  K.Y. The effect  of dichotic, monaural, and
   diotic  verbal  stimuli on auditory  evoked potentials.
   Neuropsychologia,  1974,  12:339-345.

Haggard,  M.P.  Encoding  and  the  REA  for speech
   signals. Quart. J. Exp. Psychol., 1971, 23:34-45.

Hartley,  L.R.  The  effect of stimulus relevance on the
   cortical evoked  potentials.  Quart. J. Exp. Psychol.,
   1970, 22:531-546.
Hubel,  D.H.,  and Wiesel,   T.N.   Receptive  fields,
   binocular interaction  and functional architecture in
   the cat's visual cortex. J. Physiol. (London),  1962,
   160:106-154.

Jasper, H.H.  The ten twenty  electrode system of the
   international  federation. Electroenceph.  clin. Neuro-
   physiol., 1958, 10:371-375.

Jewett, D.L., Romano, M.N. and Williston,  J.S. Human
   auditory  evoked  potentials:  Possible  brain  stem
   components  detected on the scalp. Science,  1970,
   167:1517-1518.

John, E.R., Herrington,  R.N.  and Sutton, S.  Effects of
   visual form on the evoked response. Science,  1967,
   155:1439-1442.

John, E.R.,  Ruchkin,  D.S.  and  Villegas, J. Experi-
   mental  background:  Signal analysis and behavioral
   correlates of  evoked potential configuration in cats.
   Ann. N.Y. Acad. Sci., 1964,112:362420.

John, E.R.  and  Thatcher,  R.W.   Functional  neuro-
   science,  Vol. I. Lawrence Erlbaum Assoc., New
   Jersey, 1976.
                                               283

Johnston, V.S. and Chesney, G.L.  Electrophysiological
   correlates of meaning. Science,  1974,  186:944-946.

Kelly,  D.L., Goldring, S. and  O'Leary,  J.L.  Average
   evoked  somatosensory   responses  from   exposed
   cortex of man. Archs.  Neurol., 1965, 13:1-9.

Kimura,  D. Functional  asymmetry  of  the  brain  in
   dichotic listening. Cortex, 1967, 3:163-178.

Knott,  J.R.  Some effects  of  "mental  set"  on  the
   electrophysiological processes  of the human cerebral
   cortex. J. exp. Psychol.,  1939, 24:384-404.

Konorski, J. Integrative  activity  of the brain.  Univ. of
   Chicago Press, Chicago, 1967.

Kostandov,  E.A.   and   Bailing,  E.G.  (Hemispheric
   lateralization  of the slow negative potential during
   verbal  activity.]  Zh.  vyssh.  nerv.   Deyat.  I.  P.
   Pavlova, 1973, 23:938-946.

Kuhl, P.K. and  Miller, J.D. Speech perception by the
   chinchilla. Science, 1975, 190:69-72.

Ladefoged,  P.  Preliminaries to  linguistic  phonetics.
   Univ. of Chicago Press, Chicago, 1971.

Lamb,  L.E.  and  Graham,  J.T.  Influence of signal
   variables  on evoked  response to sound. J. Speech
   Hear. Res.,  1967, 10:257-267.

Langer, J. The  ontogeny of cognitive function  func-
   tions. In E.H. Lenneberg (Ed.),  Language and brain:
   Developmental  aspects.  Neurosci.  Res.  Prog. Bull.,
   1974, 12:511-656.

Lenneberg,  E.H.  Biological  foundations of language.
   Wiley, New York, 1967.

Lenneberg, E.H. The neurology  of language. Daedalus
   1973,  102:115-134.

Lenneberg,  E.H.  (Ed.)  Language  and brain:  Develop-
   mental  aspects. Neurosci. res. prog. Bull.,  1974>
   12:511-656.

Levy,  J., Nebes,  R. and Sperry,  R.W.  Expressive lan-
   guage in the  surgically separated minor hemisphere.
   Cortex, 1971, 7:49-58.

Liberman, A.M. The grammars of  speech and language.
   Coi/iitive Psychol., 1970, 1:301-323.

Liberman, A.M.  The specialization  of  the  language
   hemisphere.  In P.O.  Schmitt  and  F.G.  Worden
   (Eds.),  The  Neurosciences:  Third study  program.
   MIT Press, Cambridge, Mass., 1974:43-56.

Liberman, A.M., Cooper, F.S.,  Shankweiler, D.P. and
   Studdert-Kennedy,  M.   Perception of the  speech
   code. Psychol.  Rev.,  1967, 74:431-461.

 Liberman, AM.,  Delattre, P.C.  and Cooper, F5. The
   role of selected stimulus variables in the perception of

-------
  284

    the unvoiced stop consonants. Am. J. Psychol., 1952,
    65:497-516.

  Ufshitz, K. The averaged  evoked cortical response to
    complex  visual   stimuli.   Psychophysiol.,  1966,
    3:55-68.

  Lindsey, J.W. The auditory evoked potential in man: A
    review. T.I.T. J. Life Sci., 1971, 1:91-110.

  Lindsley,  D.B.  Bilateral differences in brain potentials
    from  two   cerebral  hemispheres  in  relation  to
    laterally and stuttering.  J.  exp. Psychol.,  1940,
    26:211-225.

 Lisker, L. and  Abramson,  A.S. Across-language  study
    of voicing in  initial stops: Acoustical measurements.
    Word, 1964a, 20:384-422.

 Lisker,  L.  and  Abramson, A.S.  Stop categories and
    voice onset time. Proc. 5th Int. Cong. Phonetic Sci.,
    Munster, 1964b.

 Low, M.O., Wada, J.A. and Fox, M.  Electroencephalo-
    graphic  localization  of  certain  aspects  of sound
    localization in the human brain.  Electroenceph. clin.
    Neurophysiol., 1974, 36:563-564.

 Low, M.D., Wada, J.A. and Fox, M.  Electroencephalo-
    graphic localization of conative  aspects of language
    production in  the human brain. In W.C.  McCallum
    and J.R. Knott (Eds.), The responsive brain. Wright,
    Bristol, England, 1976:165-168.

 Luria,  A.R. Higher cortical functions in  man.  Basic
    Books,  New   York.  Consultants   Bureau,  1966.
    (Original Russian text,  1962, Moscow Univ. Press).

 Luria,  A.R. The working brain.  Penguin,  Middlesex,
    England, 1973.

 Luria,  A.R. Language and brain. Brain &  Lang., 1974,
    1:1-14.

 MacKay, D.M.  Information, mechanism and meaning.
   MIT Press, London, 1970.

 MacNeilage, P.  Motor  control  of  serial  ordering  of
   speech. Psychol., Rev., 1970, 77:182-196.

Madell, J.R. and  Goldstein,  R. Relation between loud-
   ness and  the amplitude of the early components of
   the  averaged  EEC response.  J.  Speech  Hear. Res.,
   1972,  15:134-141.

Marsh,  G.R. and Thompson, L.W. Hemispheric asym-
   metry  in slow  cortical potentials as a function  of
   verbal or non-verbal set. Soc.  Neurosci. Prog. Abst.,
   1972:170.

Marsh,  J.T. and  Brown, W.S.  Evoked  potential  cor-
   relates of meaning  in the perception of language. In
   J. Desmedt (Ed.), Progr. clin. Neurophysiol., Vol. 3:
   Language  and  hemispheric  specialization  in man:
                             Language References

    Cerebral  event-related   potentials.  Karger,  Basel,
    1977:60-72.

 Matoumiya,  Y.,  Tagliasco, V., Lombrosco, C.T. and
    Goodglass,  H.  Auditory evoked response: Meaning-
    fulness of stimuli and interhemlspheric asymmetry.
    Science, 1972,  175:790-792.

 McAdam, D.W.  and  Whitaker, H.A.  Language  pro-
    duction: Electroencephalographic localization in the
    normal human brain. Science, 197la,  172:499-502.

 McAdam, D.W.  and  Whitaker, H.A.  Electrocorttcal
    localization of language  function - reply to Morrell
    and Huntington. Science, 1971b, 174:1360-1361,

 McKee,  G., Humphrey,  B.  and McAdam, D W. Scaled
    lateralization of alpha activity during linguistic and
    musical tasks. Psychophysiol., 1973, 10:441-442.

 Megela,  A.L.,  Teyler,  T.J.  and Hesse, G.W. ERP re-
    sponse  decrement and  recovery as  a function of
    stimulus type and  scalp location. Physiol.  Behtv
    1977,  19:15-22.                                *'

 Menyuk,  P. The language impaired  child:  Linguistic or
    cognitive  impairment.  In  D. Aaronson  and  R W
    Richer (Eds.), Developmental psycholinguistlcs and
    communication  disorders.  Ann.  N.Y. Acad  Sci
    1975, 263:59-69.                            **•

 Michalewski, H.J. Lateralized cerebral processing and
    the   development   of   hemispheric   slow   scalp
   potentials-the CNV. Unpublished  doctoral dissert*.
   tion, Simon Fraser Univ., 1975.

 Molfese,  D.L. Cerebral  asymmetry  in infants, children
   and adults:  Auditory  evoked potential  responses to
   speech and nonspeech stimuli. Unpublished doctoral
   dissertation, Pennsylvania State  Univ., Univ  P»i*
   Pa., 1972.                                    "*•

 Molfese,  D.L. Neuroelectrical  correlates of categorical
   speech  perception in adults. Brain and  Lang.. IQ/?*
   5:25-35.                                 ••i*'»,

 Molfese,  D.L. The ontogeny of cerebral asymmetry in
   man:  Auditory evoked  potentials  to linguistic and
   non-linguistic stimuli in infants  and children.  Inj
   Desmedt (Ed.), Progr. clin. Neurophysiol., Vol  3*
   Language  and hemispheric specialization in  man'
   Cerebral  event-related  potentials.  Karger   Baui'
   1977:188-204.                           '   ™e|«

 Molfese,  D.L.,  Freeman, R.B.  and  Palermo,  D.S. Thft
   ontogeny  of brain   lateralization  for  speech  and
   nonspeech stimuli. Brain  &  Lang., 1975, 2:356-350

 Molfese,  D.L. and Hess,  T.M. Hemisphere specialization
   for  VOT perception  in  preschool children. J  »v-
   Child Psychol., 1978, 26:71-84.              '  Xp*

Molfese,  D.L.  and  Molfese, V.J.  Infant  speech De
   ception: Learned or  innate. In  H.A. Whitaker and

-------
Language References
                                               285
   H. Whitaker (Eds.). Perspectives in neurolinguistics
   and  psycholinguistics,  Vol.  4,  Academic  Press, in
   press.

Molfese,   D.L.,   Nunez,   V.,  Seibert,   S.M.    and
   Ramanaiah,  N.V. Cerebral  asymmetry:  Changes in
   factors  affecting its  development.  In S.R. Hamad,
   H.D.  Steklis  and J.  Lancaster  (Eds.),  Origins  and
   evolution of language and speech.  Ann. N.Y.  Acad.
   Sci., 1976, 280:821-833.

Monett, L.K.  and  Huntingdon, D.A.  Electrocortical
   localization of  language  production.  Science,  1971,
   174:1359-1360.

Morrell, L.K. and  Huntington,  D.A. Cortical  potentials
   time-locked  to  speech  production:  Evidence for
   probable   cerebral   origin.   Life    Sci.,   1972,
   11:921-929.

Morrell, L.K. and  Salamy, J.G.  Hemispheric asymmetry
   of  electrocortical  responses   to  speech stimuli.
   Science, 1971,  174:164-166.

Naatanen,  R.   Evoked   potential,   EEC,  and  slow
   potential  correlates   of  selective  attention.  Acta
   Psychol., 1970,33:178-192.

Nebes,   R.D.  Hemispheric  specialization  in  commis-
   surotomized man. Psychol. Bull., 1974, 81:1-14.

Neisser, U. Cognitive  psychology. Prentice-Hall, New
   Jersey,  1967.

Nelson,  D.A.  and Lassman, P.M.  Combined  effects of
   recovery period  and stimulus intensity on the
   human  auditory evoked  vertex response.  J. Speech
   Hear. Res., 1973, 16:297-308.

Neville,  H. Electroencephalographic correlates of lateral
   asymmetry  in  the processing  of  verbal  and non-
   verbal  auditory  stimuli.  J.  Psycholinguist.  Res.,
   1974, 3:151-163.

Newcombe, F.  and  Ratcliff,  G.  Handedness,  speech
   lateralization  and  ability.  Neuropsychol.,   1973,
   11:399-407.

Ohlrich, E.S.  and Barnet,  A.B. Auditory evoked re-
   sponses during  the first year of life. Electroenceph.
   clin. Neurophysiol., 1972, 32:161-169.

Osgood, C.E.  Semantic differential  technique  in the
   comparative study of cultures. Am. Anthrop., 1964,
   66 (3:2): 171-200.

Penfield, W. The excitable cortex in  man.  Liverpool
   Univ. Press, Liverpool, England, 1958.

Perkel, D.H. and  Bullock,  T.H. Neural coding.  Neuro-
   gci. Res. Prog. Bull.,  1968,  6:221-348.
Piaget, J. Biology and knowledge.  Univ.  of  Chicago
   Press,  Chicago, 1971b.

Picton, T.W., Goodman, W.S. and Bryce, D.P. Ampli-
   tude of evoked responses to tones of high intensity.
   Acta Otolar.,  1970,  70:77-82.

Picton, T.W., Hillyard, S.A., Krausz, H.I.  and Galam-
   bos, R. Human auditory evoked potentials:  I. Evalu-
   ation  of. components.  Electroenceph.  clin. Neuro-
   physiol., 1974,36:179-190.

Posner,   M.I.   Cognition:   An   introduction. Scott,
   Foresman, Glenview, 111.,  1973.

Pribram,  K.H. Languages of the  brain. Prentice-Hall,
   New Jersey,  1971.

Ratliff,  S.S.  and  Greenberg,  H.J. The averaged  EEC
   response  to  linguistic and nonlinguistic stimuli.  J.
   auditory Res., 1972, 12:14-25.

Reed,  S.K.  Psychological processes  in pattern recog-
   nition. Academic Press, New York, 1973.

Regan,  D. Evoked potentials in psychology,  sensory
   physiology and clinical medicine.  Wiley, New York,
   1972.

Rieber,  R.W.  The role of language  and thought  in
   developmental  psycholinguistics-An  historical  re-
   view.  In  Developmental  psycholinguistics and com-
   munication disorders. Ann. New York  Acad.  Sci.,
   1975,263:7-11.

Ritter, W., Simson, R. and Vaughan, H.G., Jr. Associa-
   tion cortex potentials and  reaction  time in auditory
   discrimination. Electroenceph. clin. Neurophysiol.,
   1972, 33:547-555.

Robbins, K.I.   and  McAdam, D.W.  Interhemispheric
   alpha asymmetry  and imagery mode. Brain & Lang.,
   1974, 1:189-193.

Roemer, R.A.   and  Teyler,  T.J.  Auditory  evoked
   potential asymmetries  related to word meaning.  In
   J. Desmedt (Ed.), Progr.  clin. Neurophysiol., Vol. 3:
   Language and hemispheric specialization  in  man:
   Cerebral   event-related   potentials.  Karger, Basel,
   1977:48-59.

 Roth, W.T., Kopell, B.S. and Bertozzi, P.E. The effect
   of  attention  on  the averaged evoked response  to
   speech sounds. Electroenceph. clin.  Neurophysiol.,
   1970, 29:38-46.

 Roth, W.T.,  Kopell,   B.S., Tinklenberg, J.R., Hunts-
    berger, G.E.  and  Kraemer, H.C. Reliability of  the
    contingent   negative variation   and  the   auditory
    evoked  potential.  Electroenceph.   clin.  Neuro-
    physiol., 1975, 38:45-50.
 piaget, I. The construction of reality in the  child.
    Ballantine Books, New York, 197la.
 Schafer, E.W.P.  Cortical  activity  preceding  speech-
    semantic specificity. Nature, 1967, 216:1338-1339.

-------
 286

 Schwartz, E.L. Spatial mapping in the primate sensory
    projection: Analytic structure and relevance to per-
    ception. Bid. Cyber.,  1977, 25:181-194.

 Schwartz, G.E., Davidson,  R.J., Maer, F. and  Brom-
    field,  E.  Patterns  of hemispheric  dominance in
    musical, emotional, verbal, and spatial tasks. Paper
    presented  at the  Society for  Psychophysiol.  Res.
    Meetings, Galveston, October, 1973.

 Seamon,  J.G.  and Gazzaniga,  M.S.  Coding  strategies
    and cerebral laterality effects. Cognitive  Psychol.,
    1973, 5:249-256.

 Seitz, M.R.  The effects  of  response requirements and
    linguistic context  on  averaged  BEG responses to
    clicks. Brain & Lang., in press.

 Seitz,  M.R.  and  Mononen,  L.J.  Averaged electroen-
    cephalic response  (AER)  differences  between  ipsi-
    lateral  and  contralateral  hemispheres  in a  dichotic
    listening paradigm.  Paper presented at the Annual
    meeting of  the Amer. Speech  Hear. Assoc.,  Las
    Vegas, 1974.

 Shagass, C.  Evoked  brain  potentials in psychiatry.
    Plenum, New York, 1972.

 Shankweiler,  D. and Studdert-Kennedy,  M. Identifica-
    tion of consonants and vowels presented to left and
    right ears.  Quart. J. Exp. Psychol., 1967, 19:59-63.

 Shamrd,  G.A.W.   Evoked  responses  -  Introductory
    studies  in  their relationship  to the  perception of
    language. J. Lar. Otol.,  1969, 83:947-958.

Sharrard,  G.A.W.  Further conclusions regarding  the
   influence of  word meaning on the cortical  averaged
   evoked  response in  audiology. Audiology,  1973,
   12:103-115.

Shelbume, S.A., Jr. Visual evoked responses  to  word
   and nonsense syllable  stimuli. Electroenceph. elm.
   Neurophysiol., 1972, 32:17-25.

Shelbume,  S.A., Jr. Visual evoked  responses  to lan-
   guage  stimuli in normal  children. Electroenceph.
   clin. Neurophysiol.,  1973,34:135-143.

Shepherd, D.C.  Individual specificity  of the late com-
   ponents of averaged EEG  responses (AERs) evoked
   by  repeated  clicks.  J. Speech Hear.  Res.,  1969,
   12:762-777,

Skinner,  P.H.  and  Antinoro, F.  The  effects of  signal
   rise time and duration  of the early components of
   the auditory  evoked cortical  response.  J.  Speech
   Hear. Res.,  1971, 14:552-558.

Smith, L.L. and Goldstein, R. Influence of background
   noise on the  early components of the averaged EEG
   response  to  clicks.  J.  Speech  Hear.  Res.,   1973,
   16:488-497.
                              Language References

 Spoor, A., Timmer, F. and Odenthal, D.W. The evoked
    auditory  response (ear) to intensity modulated and
    frequency modulated  tones and tone bursts. Inter-
    national Audiology, 1969, 8:410-415.


 Stevens, K.N.  and House,  A.S. Speech  perception. In
    J.V. Tobias (Ed.), Foundations  of modern auditory
    theory,   Vol.   2.   Academic   Press,   New   York,
    1972:1-62.


 Stevens, K.  and  Klatt, D. Role of formant transitions
    in  the   voiced-voiceless distinction  for stops.  J.
    acoust.  Soc. Am., 1974,  55:653-659.


 Straumanis,  J.J.,  Jr.,  Shagass, C.  and Overton,  D.A.
    Problems  associated  with  application of  the con-
    tingent  negative variation to psychiatric  research. J.
    nerv. ment.  Dis., 1969, 148:170-179.


 Studdert-Kennedy, M.  and  Shankweiler,  D.   Hemi-
    spheric  specialization  for  speech  perception.  J.
    acoust.  soc.  Am., 1970, 48:579-594.


 Sutton, S. The specification of pscyhological variables
    in an averaged  evoked  potential experiment. In E.
    Donchin and D.B. Lindsley (Eds.), Averaged  evoked
    potentials: Methods, results,  and evaluation  NASA
    SP-191, Washington, D.C., 1969:237-298.


 Sutton,  S.,  Brazen, M.,  Zubin, J. and John,  E.R.
    Evoked-potential correlates of stimulus uncertainty
   Science, 1965,  150:1187-1188.


 Sutton, S.,  Tueting, P., Zubin,  J.  and  John,  E.R.
   Information  delivery and the sensory evoked  poten-
   tial. Science, 1967, 155:1436-1439.


 Szentagothai, J.,  and Arbib, M.A.  Conceptual models
   of neural  organization.   Neurosci.  Res. Prog  Bull
   1974, 12:307-510.                                '


 Szirtes, J.  and Vaughan, H.G., Jr. Characteristics  of
   cranial and  facial potentials associated with  speech
   production.  In  J. Desmedt (Ed.), Progr. clin.  Neuro-
   physiol., Vol. 3: Language and hemispheric speciali-
   zation  in man:  Cerebral event-related  potential.
   Karger,  Basel, 1977:112-126.                    W'


 Szirtes, J. and Vaughan, H.G., Jr., Topographic analysis
   of speech-related cerebral potentials. Electroencenh
   clin. Neurophysiol.,  1973, 34:754.

Taurozzi, N. L'influenza  della durata e  del ritrno  di
   stimolazione sulla risposta ud  itiva  evocata (ERAl
   Nuovo Arch. Laringol., 1973, 1:102-106.

Teyler, T.J.,  Roemer, R.A.,  Harrison, T.F. and Thomo.
   son,  R.F. Human  scalp-recorded  evoked-potentiai
   correlates  of linguistic  stimuli. Bull.  PsychonQJr
   Soc., 1973, 1:333-334.                         m*

-------
Language References
                                               287
Thatcher,  R.W.   Electrophysiological  correlates  of
   animal and  human memory.  In  R, Terry  and S.
   Gershon (Eds.), The neurobiology  of aging,  Vol. 3.
   Raven, New York, 1976:43-102.

Thatcher, R.W. Evoked potential correlates of  delayed
   letter matching. Behav. Biol.,  1977a, 19:1-23.

Thatcher, R.W. Evoked potential correlates of hemi-
   spheric  lateralization  during  semantic information
   processing. In S.  Harnad, R.W. Doty, L. Goldstein,
   J. Jaynes  and G.  Krauthamer (Eds.).  Lateralization
   of the nervous Astern.  Academic  Priss, New York,
   1977b:429448.

Thatcher, R.W. and  April, R.S.  Evoked potential cor-
   relates of  semantic  information processing in nor-
   mals  and  two  aphasic patients.  In  R.W.  Rieber
   (Ed.), Neuropsychology  of  language:  Essays  in
   honor  of  Eric  Lenneberg.  Plenum, New York,
    1976:95-124.

 Thatcher,  R.W.  and John,  E.R.  Functional neuro-
   science,  Vol. 1,  Lawrence  Erlbaum  Assoc.,  New
   Jersey,  1976.

 Vaughan, H.G., Jr.  and Hull, R.C. Functional relation
   between  stimulus intensity  and   photically  evoked
   cerebral    responses    in   man.   Nature,   1965,
    206:720-722.

 Vaughan,  H.G., Jr.  and  Ritter, W. The  sources  of
    auditory evoked  responses recorded from the scalp.
    Electroenceph.    clin.    Neurophysiol.,     1970,
    28:360-367.

 Vella,  E., Butler,  S.R. and  Glass,  A.  Electrical cor-
    relates  of right hemisphere function. Nature, 1972,
    236:125-126.

 Wada,  J.A., Clark,  R. and  Hamm,  A. Cerebral hemi-
    spheric asymmetry in humans. Archs. Neurol., 1975,
    32:239-246.

 Wanner, E., Teyler, T.J.  and  Thompson,  R.F.  The
    psychobiology of speech and language-an overview.
   In J. Desmedt (Ed.), Progr. clin. Neurophysiol., Vol.
   3: Language and hemispheric specialization in man:
   Cerebral  event-related  potentials.  Karger,  Basel,
   1977:1-27.

Weinberg, H. and  Cole,  R.E. Averaged evoked poten-
   tials  to classes  of visual stimuli. Psychonom. Sci.,
   1968, 11:71-72.

White,   M.J.  Laterality   differences  in  perception.  A
   review. Psychol. Bull., 1969, 72:387-405.

Whitaker, H.A. and Seines, O.A. Anatomic variations in
   the  cortex:  Individual differences and the problem
   of  the localization   of  language function.  In  S.
   Harnad, H.D. Steklis and J. Lancaster (Eds.), Origins
   and  evolution  of  language  and speech. Ann. N.Y.
   Acad. Sci.,  New York, 1976, 280:844-854.

Witelson, S.F.  and Pallie, W. Left hemisphere speciali-
   zation for language in the newborn:Neuroanatomical
   evidence   of   asymmetry.   Brain   Res.,   1973,
   96:641-646.

Wood, C.C. Auditory and  phonetic levels of processing
   in speech perception: Neurophysiological and infor-
   mation-processing  analyses. J. Exp. Psychol., Human
   Perc. and Perf., 1975, 104:3-20.

Wood,  C.C. Average  evoked  potentials  and  phonetic
   processing  in   speech perception.  In  J.  Desmedt
   (Ed.),  Prog. clin. Neurophysiol,,  Vol.  3: Language
   and  hemispheric  specialization  in  man:  Cerebral
   event-related potentials.  Karger, Basel,  1977:73-86.

Wood,  C.C.,  Goff,  W.R.   and Day,  R.S.  Auditory
   evoked potentials during speech perception. Science,
   1971, 173:1248-1251.

Zimmerman, G.N. and Knott, J.R.  Slow potentials of
   the  brain  related  to speech processing  in  normal
   speakers and stutterers.  Electroenceph. clin.  Neuro-
   physiol., 1974, 37:599-607.

 Zurif,   E.B.  Auditory  lateralization:  Prosodic  and
   syntactic factors. Brain and Lang., 1974, 1:391-404.

-------
V. DEVELOPMENT AND AGING
    Section Editor:

    Rathe Karrer
    Illinois Institute for Developmental Disabilities
    Chicago, Illinois, U.S.A.

-------
DEVELOPMENT AND  DEVELOPMENTAL DISORDERS1
 R. KARRER

 Illinois Institute for Developmental Disabilities and Department of Psychology, University of
 Illinois, Chicago, IL, U.S.A.
   This section  is a summary  of postconference cor-
respondence  concerning  development   as  an  inde-
pendent variable in ERF  research. An exchange  of
correspondence  on development  was solicited from a
number of investigators taking  part in the conference,
as well as from some  not participating. Those contri-
buting to this correspondence are  listed below:

   1. J.  Cohen,  Northwestern   University   Medical
      School, Chicago, Illinois.

   2. R. Dustman, Veteran's Administration  Hospital,
      Salt Lake City, Utah.

   3. B. Fenelon,  University of Newcastle, New South
      Wales,  Australia.

   4. R. Halliday, Langley Porter Neuropsychiatric  In-
      stitute, San Francisco, California.

   5. H.  Kohn, New  Jersey Rutgers Medical School,
      Piscataway, New Jersey.

   6.  G. Lelord,  Regional  Hospital Center  of  Tours,
      Tours, France.

   7. G.  Marsh,  Duke University,  Durham,  North
      Carolina.
 'Editor's note: This section, unlike others in the volume, did
  not  emerge  by  design from pieconfeienoe correspondence
  among selected panelists. The interest of ERP investigators in
  problems of development  and  its  disorders, however,  was
  clearly apparent from the spontaneous submission of the ten
  data  papers  assembled here. These papers were originally
  included in the  psychopathology plenary session chaired by
  J.R. Knott and J. Tecce at EPIC IV. Dr. Kami accepted the
  responsibility ex post facto of  organizing  and editing these
  papers.  In  order  to  provide some continuity with other
  sections, Dr. Kaner also  undertook  a postconference  cor-
  respondence  to define the  critical problems of development,
  aging, and  related disorders that  might  be  amenable to
  resolution by ERP methods.
   8. C. Warren,  Illinois Institute for  Developmental
     Disabilities, Chicago, Illinois.

   Discussion in the correspondence was  confined to
major  problems that arise whenever the  investigator
injects  change  over  age  or  maturation into an ERP
study. Age takes on significance whenever processes are
changing rapidly (as  at the extremes of the  life span),
whenever  age is a variable, or whenever rate, or lack,
of development is  of concern. Some investigators con-
sider development  a  tool for  the study of process and
view it as  lacking  its  own substantive information;
from this  viewpoint,  the papers included in this section
may seem more appropriate  to  other  sections of this
volume. Other  investigators,  in contrast, consider de-
velopment per se  as  a substantive area of investigation
in its own right,  one that raises unique questions and
provides useful information.  However  the investigator
views this question,  there are common problems that
need to be considered. The following issues were  posed
as the  framework for critical  evaluation  of develop-
ment and  ERP.
 Issues
 The paucity of developmental information

   The main body  of information  concerns sensory-
 evoked potentials to relatively simple stimuli (Dustman
 et  al.  1976,  EUingson  1968).  Furthermore,  most
 evoked  potential data  relate  to infancy  with little
 effort to collect data between age 2 and adolescence.
 Studies spanning middle and old age are also needed to
 complete  the longitudinal picture.  We  need further
 longitudinal and cross-sectional studies  of contingent
 negative variation (CNV), Beiettschaftspotential (BP) or
 readiness  potential,  and  P300  that encompass  the
 entire life span. The studies at this conference reflect a
 broadening  of  thinking and  efforts. Such efforts are
 intimately related to the following issue.

-------
292

Conceptual basis of developmental studies

   Most developmental  studies have been descriptive.
While this approach Is necessary and important, investi-
gators  need also to conduct studies that utilize  pre-
diction* from known events In behavioral and neuro-
logical development.  Such  work could  advance our
understanding of the relationship between event-related
potentials (ERPs) and behavior. For example, It would
be  of interest to  relate the development  of P300 to
changes in  cognitive  growth. If current  formulations
about  the significance of P300 for cognitive processing
of  stimulus  events  are correct, changes  in  size and
topography  should occur with the infant's  cognitive
development. Since P300 reflects subjective probabili-
ties of events, it should reflect the Infant's and child's
development of hypotheses  concerning   event  prob-
abilities and expectancies (cf.  Bower 1974,  Stevenson
1972). Such an index would add  specificity to our
present generalized markers  (e.g., eye movement, heart
rate)  of model building,  habitation  processes, and
orienting of infant and young child.

   As another example, consideration could be given to
the question of whether the  development  of motor
differentiation and skill lead to change*  In the wave-
form  and  topography  of the  potentials  that precede
movement. These data have  consequences  for the inter-
pretation  of motor-related  potentials  and  the  Inter-
action of  their components.  The  P300  and motor-
related potentials  may  prove clinically uwful for the
evaluation  of  cognitive  and  motor  development  In
children  or adults that challenge  our present nemo-
psychological tests.

   Since  the  study of ERPs is a major  source of
Information  concerning human  brain  function,  It is
necessary  to establish their relationship to meaningful
segments  of behavior  and to their neurological under-
pinnings.  Questions concerning  neurological develop-
ment  could  be  generated comparable to  those posed
for behavioral  development.  It  would be  of interest,
for example, to know whether CNV is present in some
form prior to or in the absence  of the normal matura-
tion of dendritic processes, e.g., in infants, mentally
retarded,   or  sensory-deprived  animals.  Correlative
studies of slow  ERPs to known developmental data in
humans may be  fruitful.

Methodological  issues

   Useful developmental Information is dependent on
good methodology. In  work with  an infant or young
child,  the time  spent in putting on electrodes becomes
extremely  critical. Since the attention span  and "co-
operation span" of the infant  or  child  is  short, the
amount of data  that can be  collected in a single session
is  severely  limited.  Repeated  recording  sessions  to
collect sufficient trials for  reliable  waveforms are not
                                            Karrer

always possible, particularly if one is dealing with an
outpatient  population.  There  has been some experi-
mentation  with  electrode harnesses  or  helmets  for
quick and  easy  application, but the  results  have met
with little  success. Development of  an effective and
durable rapid-mount electrode harness would facilitate
developmental ERP research.

   A closely  related problem  concerns the limitations
of restricting the subject by direct wiring of electrode
leads to the  amplification system. The innovative  use
of commercially available telemetry  systems offers  a
possible solution to this problem.

   More imaginative paradigms and stimulus events that
can hold the Infant's or child's attention during  the
trial  are  also needed. Such improved techniques would
assist in  controlling changes in state during the session
and  minimize  movement artifact.  Data  recovery is
usually about 50% or less with children and infants.
How does this affect the interpretation of the data?

Variability

   The amplitude of the average response is  dependent
upon both the  amplitude and latency variability (jitter)
of single-trial responses.  That is, averaged  responses are
larger  when  variability  is reduced.  This factor is  a
major  one  in the large-amplitude evoked responses of
Down's  Syndrome individuals.  The  determination  of
variability  requires trial-by-trial analysis.  Since  ampli-
tude and variability  are  functions of underlying neural
systems, some  additional understanding of the neural
systems is gained by considering both measures.

Sensory system function changes with  age

   At  a given  light intensity, the  amount of light
reaching the  retina  of a 70-year-old is about one-
quarter  of that  reaching the retina  of a 20-year-old.
Some  consideration  should  be given  to determining
sensory  thresholds of subjects of different  ages  and
adjusting stimulus intensities so  that  the effective
intensity used  across ages is reasonably constant. This
factor  may also be of importance  when studying ERPs
in certain  patient or  handicapped  groups  in whom
generalized  or  specific  sensory  loss  is  suspected  or
known.

Discussion

   The response, concepts, and recommendations that
emerged in  the postconference correspondence  were
focused  on  these problems  and  future  directions  of
developmental  ERP work rather than the  critical evalu-
ation of present or past work. This emphasis is in part
due  to  the  paucity  of developmental ERP research.
The  next conference should foster a critical discussion
of findings.

-------
Development and Developmental Disorders
                                               293
Conceptual issues

   ERP investigators  are  fond of touting their  tech-
niques as  a  means of investigating  the intact  brain.
Thus far,  the  validity  of this assumption  is more a
matter of  faith than  empirical evidence. The value of
the ERP in developmental research may lie in the fact
that  different  components appear  to  bit  related to
different features  of cognitive processing  and motor
performance. ERPs should be  considered an adjunct to
anatomical and biochemical studies for obtaining infor-
mation on the  integrity of physiological and behavioral
systems.  Further  ERP  studies in animal models and
clinical   populations   with  known   lesions,   central
nervous  system   dysfunction, or  genetic  anomalies
should be  undertaken to expand  our knowledge of the
brain mechanisms that underlie specific ERP compon-
ents in humans.

   Most   developmental  studies  have been  cross-
sectional (cf. Cohen 1973, Dustman et al.  1976, Elling-
son  1968). Evidence  of the long-term  stability of the
ERP measures is lacking. That  is,  are the measures
made at a single  point during development predictive
of measures taken later in the life span?  Longitudinal
studies  that  relate ERPs,  particularly  slow  potentials,
to anatomical,  physiological, and  psychological changes
during development are needed.  There is an increased
interest in the relationship between  ERP development
and the plethora of information on cognitive (cf. Flavell
 1977) and perceptual development.  No reliable  body
of data relates ERPs  to the development of functional
abilities (e.g., reading,  motor  skill),  much less to cog-
nitive, perceptual, motor,  or psychosexual development.

   Parametric studies  during development are essential.
For  example,   variations  in   stimulus intensity can
accentuate  differences between   normal  and patho-
logical development, differences which  in turn  may be
related  to the  modulation  of  environmental  input
(Buchsbaum  1976, Galbraith et al. 1976, Karrer 1976).
Stimulus  intensity effects within or across modalities
are dependent upon sensory threshold and the subjects'
subjective  magnitude  function. Yet  there has been no
attempt to study this problem developmentally.

   The need for parametric studies is demonstrated by
Courchesne's finding (1977)  of  no  N1-P2  differences
between children (aged  5-8)  and adults in a visual-
attending  paradigm.  Since N1-P2 differences between
attend  and  nonattend channels depend  on a  rapid
presentation  of stimuli (Schwent and  Hillyard  1975),
the  ISI used by Courchesne (1250 msec) may not have
been rapid enough to elicit developmental differences.
In  children,  however,  Courchesne   observed  large
negative waves (30 /LtV) with latencies of 400 to 500
msec and  large positive waves that peaked around 970
msec. These potentials, absent in adults, occurred only
to novel stimuli and  were frontally distributed. What
the  potentials  signify in  terms  of development is
unclear;  parametric studies of  late components  and
slow ERPs during development  are needed to resolve
this question.

   It  is an  oversight  that major changes in anatomy
and metabolism during the early years have never been
tied  to ERP development. So far, the alpha rhythm
and other bands have been the main variables studied.
The  only important  findings  to  emerge  from  such
studies are the breakpoint at about 3 months (when
alpha starts) and the slow speeding of the rhythm. The
full  realization  of these  efforts  will require better
information on correlative anatomy. Such  data would
help, for example,  to elucidate  the mechanism under-
lying  disparate   latencies  of  ERPs  recorded   over
homotopic sites in  the  two hemispheres. In addition,
sensory deprivation and  other experimental manipula-
tions such as hypothyroidism (Eayrs 1971) in newborn
animals  result in  well-known  anatomical  (reduced
dendritic density) and biochemical changes. There have
been  no studies   of concomitant effects  on  slow
potentials that should theoretically be related to such
changes, e.g., the  presumed  dendritic basis  of  CNV
(McSherry   1973).  ERP  data  from  multiple  sites,
gathered in  different  laboratories, may provide norms
relevant  to the  sparse information on  dendritic de-
velopment  and myelination  that  could be related to
retarded development.

   Research  on the  normal  development  of  ERPs in
children should be  directed to factors that have proved
to be significant in the early development  of the EEC
itself.  Factors of  ERP  topography,  symmetry, and
complexity  are more germane in development than in
the  adult.  The literature suggests that the  function of
CNS structure alters during early development and that
the processes of myelinization and synaptic elaboration
are  major  aspects of  neural development.  Rate of
conduction  and complexity of  synaptic  potentials are
clearly correlated to, if not causal for, the ERP.

   The  development  of paradigms  that reflect normal
behavioral  growth  and  permit concomitant  ERP re-
cording  is  necessary to understand  developmental
disorders. ERPs would, it is hoped, shed light on the
mechanisms  underlying the disorder in  question. For
instance,  myelinization   deficiencies  may  result in
slowed  neural   conduction   creating  a  nonoptimal
temporal confluence  of  CNS  activity  and,  conse-
quently,  long   ERP  latencies.  Presently,  there  is
insufficient  information  for  other than  the crudest
interpretation of possible deviant responses in patients.
It  is  imperative  to  determine means  and  standard
deviations of amplitude, waveforms, and latencies over
the  life  span in  order to detect differences associated

-------
 294
                                             Karrer
 with learning  disability,  brain  trauma, infectious dis-
 orders, toxins affecting the brain, and demyelinating
 disease.

   While the  papers in this section clearly  show that
 various  ERP  measures in  developmental^  disabled
 children are  different from those found  in  normal
 children, there is no evidence concerning the reliability
 or reproducibility of these differences. Studies  of inter-
 hemispheric  differences  related to skill  deficiencies,
 handedness,  and cognitive development are  also neces-
 sary  (see  Fenelon,  as well  as John and Prichep, this
 volume). Evoked  response  patterns of individuals may
 be idiographic to the extent that different conditioning
 histories,   variations   in   perceptual   and   cognitive
 strategies,  and motivation affect them. Group data may
 yield only general nomothetic features of ERPs; never-
 theless, normative  studies  provide the  empirical basis
 for  defining  differences   in  the  ERP  patterns  of
 aberrant populations.

   Behavioral techniques presently can "discriminate"
 developmental syndromes better  than  ERP measures.
 While there  is disagreement on  the diagnostic role of
 ERPS, it seems that these highly technical, complicated
 and  unvalidated procedures are mainly appropriate for
 sensory problems (e.g., color vision, acuity in certain
 subject populations,  brain stem  EP for the  auditory
 pathway),  problems not easily assessed by other means.
 The   relationship of ERPs,  psychometric, and neuro-
 psychologic measures should be examined.

   We  tend  to  treat  experimental  manipulations as
 though they were synonymous  with the labels used to
 characterize  them.  A  current favorite  is  "attention."
 Three recent studies  of auditory  EPs  in hyperactive
 children (Halliday et al. 1976, Prichep et al.  1976, Hall
 et al. 1976) have manipulated  attention, although the
 experimental  procedures  used  were  different  in each
 case. It may well be that all these studies tap the same
 process, but that remains to be seen.  One way around
 this   equivalence  problem  is  to  record  ERPs while
 varying task difficulty. Such an approach has  a better
 likelihood   of  unraveling  the  relationship  between
 cognitive abilities and  changes in brain states.


Methodological issues

   Electrode application  fatigues  some  children  and
 uses  up precious experimental time. Helmets and other
 headgear seem to increase rather  than decrease experi-
 mental time. Three methods of electrode  application
 were  recommended:   bentonite   and  cotton  balls,
 collodion,  and electrode  collars  with surgical tape.
 (Data on   the  comparative  effectiveness  of  these
 methods is not available.)  Anxiety and boredom may
 be reduced by free  play between recording  conditions
 or  the use  of  cartoons   and TV  programs.  These
 strategies may serve as reward at appropriate times or
 be an integral part of the experimental paradigm.

    Selection of  subjects  on  the  basis  of skill  and
 behavioral  maturation  criteria may  provide a  greater
 payoff than selection  based on  chronological age, IQ,
 or clinical  category. Most studies  concerned with de-
 velopmental  abnormalities   focus  on   comparisons
 between pathological and age-matched normals. While
 these  comparisons are often  informative, it may  be
 difficult  or  impossible  to  predict  a priori   which
 variables  within  the  normative  population are most
 critical to  "control."  For  instance, it  may   not  be
 sufficient, depending on  the question asked, to match
 for  age  alone  when  the experimental population in-
 cludes mental retardates. The  experimenter may need
 to control  for mental, as well as  chronological, age-
 which means that two  separate control groups may be
 required.  Similar  problems exist with the behavioral
 and physiological  parameters that  may be of interest.
 Pathological groups differ on many variables, and it is
 difficult to  specify what the observed ERP differences
 represent.  One  recommended  procedure  is, first,  to
 define general  groups  (e.g.,  retarded of age x and IQ
 y),  even  though  these groups  may be  too   hetero-
 geneous, and then to sort on the basis of  performance
 in  the  ERP  task (e.g.,  fast  RT or  slow RT, low
 threshold or high threshold). This  further sorting pro-
 vides a  more  homogeneous and  functional grouping
 that  allows better analysis  of  the  ERP-behavior re-
 lations  and of the  aberrant  vs.  normal  development
 data. One can determine  if a normal subject with slow
 RT is similar to a retardate with fast or slow RT. The
 measures  obtained (e.g.,  topography, amplitude, and
 latency) may  allow one to infer similar,  or different,
 underlying processes that can  then be further  tested.
 Any  study  that attempts  to predict aberrant develop-
 ment from  EEC, evoked  potential,  or slow potential
 data  must  attend precisely to  the  criteria measures
 used  to determine the aberrant individual or  group.
 The criterion  measures are the weakest link in these
 studies. Relying on specific  performance  in a  specific
 and meaningful task is  presently  the best tactic  (unless
 one has  data on the presence  of localized  brain
lesions).

   The  Musso and Harter study  (this section),  which
compared normal  children with two groups of learning
 disabled  children  (objectively  defined), demonstrates
 that more specific information can be obtained when
 more than one experimental  group is included in  the
 design.  Their  hypothesis that  children  with   visual
 reading disabilities allocate too much attention to  the
 task deserves investigation.

   Developmental  data exacerbate the analysis prob-
lems of ERP peak identification. Latency, polarity, and
topography   are   features   used  to   discriminate
homologous waveform  components, yet each of these

-------
Development and Developmental Disorders

parameters is subject to change with development. This
problem  demands  close  attention   to   behavioral
measures of the process being manipulated.  Variability
is also an issue here. Too often, we tend to forget the
variability and multiplicity of waveforms described in
the  literature and  speak of  the  CNV, BP, or P300
waveform. It seems reasonable to expect greater wave-
form variation during development,  especially during
aberrant  development.  Analysis  of  variability could
facilitate  our  understanding  of  different  waveform
components.

   It is equally important to examine the interrelation-
ships of changes in the development of sensory-evoked
responses, CNV, readiness potential, and other aspects
of ERPs in  different modalities. Developmental studies
can  be  considered another strategy to gain information
on  the  independence (or  interdependence) of these
measures. With studies  utilizing  paradigms based in
developmental theory, the promise of ERP  research in
relation to developmental neurobehavior theory can be
realized.

   There  was agreement  that sorting on the basis of
pretrial  EEC  characteristics   should  be   attempted,
although  this procedure  may be  complicated  and
severely reduce  the number of trials. Which recording
site  does one use  for  pretrial  sorting when one is
interested in symmetry  or topography  issues?  EEC
amplitude  and  frequency  predominance  vary  with
recording site and  task. Should  one  parcel out  the
underlying  EEC of each site? HaUiday et al. (1976)
sorted on the basis of positive- or negative-going base-
line and  found no difference  in evoked potentials. A
recent  study (Trimble  and  Potts 1975) found  that
alpha affected early (< 120 msec) components of the
visual ERP  while Tanquay and  Ornitz (1972) found
little correlation between EP  amplitude and EEC in-
tensity in sleeping children.

   These  interrelationships, as well  as many of  the
issues discussed above, require analysis epochs appro-
priate to the task. The developmental studies in  this
volume  make it  clear that  expansion of sampling
intervals  is  necessary if the details of  children's ERPs
are  to be unraveled. The utility  of  a  500-msec post-
stimulus  interval  is  severely  limited  (see  Callaway's
comments in the Psychopathology section, this volume,
 for  counterpoint).

   Developmental ERP investigators use, of necessity, a
 multivariate approach since  they must contend  with
 the  three dimensions  of stimulus, behavior, and ERP.
 One must  maintain  rigid stimulus (task) control  and
 analyze the  covariation between behavior and  ERP
 dimensions. This  triangular paradigm (Sutton  1969)  is
 best analyzed by  consideration  of the variability ex-
 tremes  along  each dimension. To  tease  apart ERP-
 behavior   relationships   under   specific   situational
                                               295

demands, one  can sort ERP variables (e.g., CNV ampli-
tude) along behavioral dimensions (e.g., as a function
of slow vs. fast RT). The reverse procedure of sorting
behavior (RT) along ERP  dimensions (e.g., as a func-
tion of high vs. low CNV  amplitude) provides further
information on the ERP-behavior linkage. Development
introduces   a   fourth  dimension  to  this  triangular
paradigm  (Karrer 1976).  Behavior,  physiology,  and
stimulus condition may  be  sorted  along the  time
dimension  of  development. A first approximation for
dealing with these complex relations is to determine
the neurobehavioral  relation at various levels  of de-
velopment.

Synopsis  of  Papers

   Papers  in this section (and  other sections) provide
an  indication  of the  potential  use  of neuroelectric
measures in the study  of normal and aberrant develop-
ment over the life span. At one end of the life  span,
Harter  et  al.  show that visual ERP components may
indicate the operation and development of two visual
systems in the  infant. A  PI30-160  component may
reflect subcortical visual processes prior to 2 months of '
age, while  a  P320-400 component  may  reflect the
development  of  cortical  visual processes  during the
second month of life.

   There   are   four  papers that  report ERP  charac-
teristics of reading-disabled or normal  children  during
cognitive  processing. Shelburne (Language section, this
volume)  examines ERPs  to the  third  character  of
consonant-vowel-consonant  words   and   nonsense
syllables.  He reports that reading-disabled children not
only  exhibit  poorer  discrimination  of  these  two
stimulus categories than  do normal  readers, but also
that  they fail to elicit enhanced ERPs to the third
uncertainty-reducing  character.  Fenelon  adds to  the
picture by showing  an  intriguing  difference in  the
pattern of ERP activity between good and poor readers
that seems intuitively  related to language processing.
Problem  readers appear to generate  the CNV weakly
 over left  parietal regions. Musso and Harter report that
 reading-disabled  children  with a visual (in contrast to
 auditory)  perceptual  problem have greater differences
 in P300  amplitude to relevant compared  to irrelevant
 stimuli.  Reading-disabled   children  also  had  longer
 latencies  and, perhaps, slower information processing.
 Friedman  et al. find that children's cognitive processing
 affects the amplitude of components in the P300 range
 in  a  fashion  similar  to  adults,  i.e.,  increased P300
 amplitude as  a function of increased cognitive activity.

    Two papers deal with  multidimensional and factor
 analytic methodologies for determining neurobehavioral
 characteristics of specific  clinical populations. John et
 al.  (this  volume)  report their  development   of  an
 ambitious  neurometric  battery  of  EEC  and  ERP
 measures  that promises  new  power in discriminating

-------
296
                                            Karrer
differences  in  brain functioning  underlying impaired
(or enhanced) behavior.  Laffont  et al. also report a
novel  methodology  for  determining the  association
between  clinical  characteristics  and   neuroelectric
events.

   Three papers report characteristics of CNV and BP
associated with motor  performance and its presumed
or explicit  impairment. Papini  et  al. found that the
CNV  of  hemiplegjc  children  is  reduced  over  the
impaired hemisphere. Karrer et al. report that the BP
waveform may change with development and take on a
different waveform in aberrant populations, such as the
mentally  retarded. Deecke  et al.  show that  the  BP
decreases  in  amplitude in  old  age  and  may even
become  a  positive-going  waveform.  The  final paper
(Marsh) reviews the sparse work on ERPs during aging
and  suggests  ERP  research  could be  the interface
between  psychological  and  physiological research  on
aging.

-------
MATURATION OF PATTERN EVOKED POTENTIALS AND
VISUAL PREFERENCE IN  6-  TO 45-DAY-OLD INFANTS:
EFFECTS OF CHECK SIZE


M.R. BARTER, F.K. DEATON, AND J.V. ODOM

University of North Carolina at Greensboro, NC, U.S.A.
   The investigation of pattern vision in normal human
infants has been restricted due to the limited ability of
infants to make behavioral responses to visual stimuli.
Harter and White (1970) and others have shown that
visual evoked potentials (VEPs) are sensitive to the
sharpness and spatial frequency of patterns  and that
VEPs may be  used  to  estimate refractive error and
visual acuity in adults.  Harter  and Sultt (1970) and
Harter et al.  (1977b) have presented  data indicating
that VEPs to patterned  light  flashes may  be used to
investigate the development of pattern vision in infants
during the first 6 months of life. In the present study,
VEPs and the percentage time fixated (PTF) were used
to  estimate  spatial  frequency  sensitivity  and  visual
acuity in 6- to 45-day-old infants.

Method

   Ten  infants  (seven male,  three female)  from an
adoption agency participated in  this study. Half of the
infants were  6 to 26 days when tested and half were
between  27  and 45  days.  The experiment  was con-
ducted  in the  nursery  with  ambient lumination of
about  4 mL. Infants were held by an experimenter
while visual  stimuli were presented 63 cm  (25 in.)
from the infants' eyes and subtended  22° x  22°. (See
Harter et al.  1977a  for a complete  report  of these
data.)

   Flashed checkerboard transparencies were used to
elicit VEPs  (recorded  monopolarly  from an  active
electrode placed 1 cm above the inion on  the midline
referenced to  the  right earlobe). PTF  served  as  a
behavioral  measure   of visual  discriminability and
preference. Checkerboard (individual checks subtending
 11,  22,  45,  90,  and  180 min of arc)  and diffuse
transparencies, all with equal luminance transmittance,
were back illuminated  with a  10-jusec flash (2.5 log
units in intensity above  adult threshold) every 1.1 sec.
Each checkerboard was presented for 64 consecutive
trials, with 4  to 6 replications of each trial series being
necessary.  Infants  were  tested every  other day until
sufficient data  were gathered.  Recordings were  made
only when the infant was sitting quietly  and fixating.

   Data  were quantified by measuring VEP amplitude
relative to baseline (mean amplitude during the 60- to
80-msec  postflash  epoch)  at   the  following   four
latencies  after  the flash: the most  positive portion
between  130  and  160 msec  (P2), the most negative
portion between 190 and 210 msec (N2) and between
240 and 300 msec (N3), and the  most positive portion
between 320 and 400 msec (P4).

Results

   Samples of  VEPs from  all infants to the diffuse
and  22-min check flashes are shown  in  Fig. 1. Each
tracing  reflects  an average  of the  four replications
obtained nearest to each infant's 25th  day of life.

   The relationship between VEP amplitude and check
size  depended  on  the VEP component measured (Fig.
2, p < .03). This  relationship tended to be bimodal
for P2 amplitude. The 11-  and 22-min checks elicited
responses greater in amplitude than diffuse light or the
45-min  checks (p<.05). The  fact that P2  amplitude
to the 11-min check and  diffuse light differed indicates
a visual acuity of 20/220 or better for these infants. A
second  mode  was suggested by the increase  in P2
amplitude (p < .06) as check size was  further increased
from 45 to 180 min.

   In contrast, the effects of check size  on P4 ampli-
tude  and  PTF varied with age (p <.03). The 6- to
26-day-old  infants  indicated  little  change  in  P4
amplitude or PTF in response to the different check
sizes. In contrast,  the 27- to 45-day-olds showed  a
linear increase in P4 amplitude and PTF  as check size
was  varied from diffuse light to the largest check (180
min). The correlation between P4 amplitude and PTF
was  highly significant (r = 0.92, p < .01). Regression

-------
 298
                                      Barter et al.
lines  were  best-fit  to  these  measures  and  indicate
extrapolated visual acuities (arrows in Fig. 2) of 6 and
10 min, respectively.  These  values  are  reasonably
similar  and  correspond  to  acuities  of 20/120  and
20/200, respectively.


Discussion

   What  optical or  neural factors  might underlie the
shape  of the function  between VEP  amplitude  and
check size and the changes in this  function during the
first 2 months  of life? A  review  of optical changes
during the first 2 months (Maurer 1975) indicated  that
such changes may  not account  for  the  present results.
Maurer's review and others, in conjunction with  the
finding  that  age selectively influenced the later VEP
components,  suggest the effects of age may be attri-
buted to neural changes.

   The two types of functions between VEP amplitude
and  check size, the first being reflected by P2 ampli-
tude in  response to small check sizes and the  second
by  P4  amplitude  in  response to  larger  check sizes,
suggest   one  interpretation.  Inverted  U-shaped func-
tions, associated with  changes  in  spatial frequency,
appear to reflect the receptive field and spatial tuning
characteristics of the neurons activated. This interpreta-
tion has been applied to psychophysical data (Blakemore
and  Campbell  1969),  human EP  data  (Harter  and
     BR
     HA
 3   TR
 3
     PE
     Ml
                                                            AB
                                                            JO
                                                            Jl
                                                            DA
                                                           MA
                                             LATENCY, sac x 0.1

Fig. 1' Visual evoked potentials to the 22-min check size and diffuse flash (D)from each infant that participated in
the experiment. Each tracing i* an average of 128 potentials obtained nearest to each infant's 25th day of life.  The
extra dashed VEP for JO was in response to  the 180-min checkerboard. The vertical dashed lines indicate latencies
vfhere P2, N2, N3, and P4 amplitude measures were made.

-------
Maturation of Pattern EPs
                                              299
                     N2 N3
                               AGE (DAYS):
                                 iT
                                 8-26  »	..
                                 2745 •	.
          D  11  22  45 90 180    D  11 22 45  90 180
                     CHECK SIZE, min of arc

 Fig. 2.  Effects of the size of checks in checkerboard
 flashes and of diffuse flashes (D) on the amplitude of the
 four VEP components (first four rows) and the percent-
 age time fixated (bottom row).  Solid tines indicate the
 means based on all subjects (left column), dashed lines
 indicate the means  of data collected from 6 to  26 and
 27 to 45 days after birth (right column).  Regression
 lines  were best-fit to P4 amplitude and the percentage
 time fixated in the 27- to 45-day-old group; the X-axis
 intercepts (vertical arrows) begin at  10 and  6  min.
 respectively.

 White  1970,  Harter 1970,  Armington et  al.  1971,
 Karmel et al.  1974, Karmel and Maisel 1975),  and
 animal single-unit data (Campbell et al. 1969). If this
 interpretation is applied to the present data, it suggests
 that at least two aggregates of neurons contributed to
 the  effects   of check size  on  VEP  amplitude,  one
primarily reflected by the early VEP components and
tuned  to  relatively high  spatial  frequencies and  the
other  by  the  late VEP  components  and  tuned  to
relatively low spatial frequencies.
   Reviews of behavioral, physiological, and anatomical
data related  to infant development (Bronson  1974,
Cohen  and  Salapatek  197S)  also  suggest that two
neural   systems  differentially  contribute  to  visual
capacities manifested in the first 2 months of life, one
subcortical   and   the   other   cortical.   Subcortical
mechanisms may account  for  visual capacities mani-
fested  in the first month, whereas cortical mechanisms
may account for  the  abilities that become manifest
starting in the second month.
   A  number  of  factors suggest  that  the  bimodal
 nature  of  the present data may be interpreted within
 the  framework of subcortical and cortical mechanisms.
 The  first  mode  was primarily  due to changes in P2
 amplitude,  a  relatively early component of the VEP
 and   more  coincident  in  time  with  the  arrival of
 subcortical activity at the cortex. The early component
 was  sensitive  to  small  check sizes (higher spatial fre-
 quencies) as are the subcortical  (geniculate) neurons of
 animals (Campbell  et  al.  1969) and were poorly cor-
 related with the  behavioral PTF measure. In contrast,
 the  second mode was  primarily due to changes in P4
 amplitude,  a  later component of  the VEP and more
 likely to  reflect  the  processing of information within
 the  cortex. The late  component was most sensitive to
 larger check  sizes (lower spatial  frequencies) as  are
 cortical, particularly complex, neurons (above studies).
 P4   amplitudes  were   highly   correlated  with  the
 behavioral PTF mea«ure, and were not influenced by
 check size  until  the  second month of age. The above
 indirect line  of evidence  suggests that the spatial
 frequency sensitivity  of the  early VEP  components
 may reflect the contribution of subcortical mechanisms
 and the late VEP components may reflect the contribu-
 tion of cortical mechanisms. If this interpretation is cor-
 rect,  the  change  in  responsivity  of  the late  VEP
 components and the  transition from passive  to more
 active and discriminating visual preference indicate an
 acceleration of cortical function between 28 and 45 days
 after birth.

-------
 CONTINGENT  NEGATIVE  VARIATION, EVOKED
 POTENTIAL,  AND PSYCHOPHYSICAL MEASURES  OF
 SELECTIVE  ATTENTION IN  CHILDREN WITH LEARNING
 DISABILITIES1
 M.F. MUSSO AND M.R. HARTER

 Department of Psychology, University of North Carolina at Greensboro, Greensboro, NC, U.S.A.
   This investigation was conducted to specify some of
 the  electrocortical correlates  of learning disabilities in
 children. A reading  disability may be defined as "the
 failure to learn to read with normal proficiency despite
 conventional  instruction, a culturally adequate home,
 proper  motivation, intact senses, normal intelligence,
 and  freedom from gross neurological effects" (Eisen-
 berg 1964, p. 65). A problem in assessing the etiology
 behind  learning disabilities  is that the contribution of
 specific  component  skills  that  characterize  the  dis-
 ability  has  not  been  identified.  Learning disabilities
 have  been  attributed  to  a  number of interrelated
 processes-e.g.,  perceptual  and  psycho-motor  (Wold
 1969),  neurological  (Ingram  1971), and  attentional
 (Anderson et al.  1973). Attentional effects on electro-
 cortical activity  and performance in reading disabled
 and normal children  were compared in this study with
 the aid of the visual evoked  cortical potential (VEP),
 contingent negative variation (CNV), and reaction time
 (RT).

 Method

   Twenty-seven  children,  divided into three equal
 groups matched for age (7 to 12), IQ (above 90), and
 sex,  participated. One  group  consisted  of  children
 classified as normal (N) in reading ability and the other
 two  groups  consisted of children classified as having
 reading  disabilities  attributed respectively to  visual
 (VRD)  and  auditory  (ARD)  perceptual  problems.
 Classifications were made on  the basis of the Slinger-
 land  Screening  Test  for Identifying Children  with
'These data  are  portions  of M.  F. MUMO'S dissertation,
 "Piychophysical Performance, Contingent Negative Variations,
 Visually  Evoked Cortical Potentials, and Selective Attention:
 a  Behavioral and Neurophytiological Assessment of Learning
 Disabilities in  Children," conducted in the  Department  of
 Psychology,  University  of  North  Carolina  at Greensboro,
 1975.
 Specific Language Disabilities. All  children were free
 from gross  neurological  damage  and physical dis-
 abilities, according to medical records, and were taking
 no  form of medication. The children were  solicited
 through the Guilford County School  System. Visual
 acuity, vertical and lateral phorias, and color blindness
 were  evaluated  by subjective reports  in  conjunction
 with  an  Ortho-Rator   and  Ishihara  color  plates.
 Spherical refractive error was assessed by both VEPs
 (Barter  and White 1968) and subjective  report. The
 frequency of detected visual  anomalies did not differ
 significantly between the three groups.

   The  subjects'  abilities were  assessed  in  a  visual
 discrimination   attentional  task  that  consisted  of
 flashing  a  warning  stimulus, SI  (a  clown's  face),
 followed after  1100  msec  by one of two randomly
 presented flashes, a  relevant  (S2rel)  or  an  irrelevant
 (S2irrel)  stimulus.  The  entire  S1-S2 sequence was
 repeated once every 5 sec. The subject was required  to
 lift  his finger off a microswitch  key  within a critical
 period of time. If  the subject  responded  to S2rel
 within this period (Hit)  or  withheld a response  to
 S2irrel (Correct  Rejection), he was given a token; if  he
 responded to S2rel  late (Miss)  or to  S2irrel (False
 Alarm), he lost a token and  was given immediate
 negative feedback in  the  form  of a  loud click. The
 click was given  1000 msec after S2 in the  case of a
 Miss and at the  time of response (mean latency of 395
 msec  after  S2)  in the  case of a False  Alarm.  Any
 auditory  evoked potentials to these clicks were suf-
 ficiently  after  S2  to ensure  they  would not  con-
 taminate  the  VEP measures employed. Tokens  were
 exchanged for rewards after each experimental session.
 The  False  Alarm  and  Hit  rates were  used  as  a
behavioral measure of attention or discriminability (d')
 according to Signal Detection Theory (Swetts  1964). It
may be  noted that the critical period of time for the
purpose  of feedback  was the  median  RT  for the
problem  and  subject (as determined  from  previous

-------
CNV, EP and Attention Measurement
                                                                       301
 data) whereas for the purpose of determining d', it was
 1000 msec. The response accuracy  measure,  therefore,
 was independent of response speed.

    Four pairs of S2s were presented, each representing
 a different level of complexity in terms  of discrimina-
 bility:  red  and green diffuse light; vertical  and hori-
 zontal  lines; the letters  b and d; and the words was
 and saw. Each stimulus of the  pair was both relevant
 and irrelevant for  one condition, giving a total of eight
 different problems,  two for each pair. Each  condition
 consisted of at least  32  presentations of each stimulus
 of a pair in random order.  The eight problems were
 presented to each subject twice, each pair being pre-
 sented  four times  (replications). Stimuli were presented
 on a LVE Model 1346 Multiple Stimulus Projector and
 were 40 msec in duration and 2.5  degrees in subtense.

    Cortical  potentials were  recorded, amplified, and
 averaged starting at  the  onset of SI. Active  electrodes
 were  placed approximately  over  Oz  and Cz  and
 referenced to the  right earlobe. The one-half amplitude
 low and high frequency filters of  the amplifiers were
 set at 0.15 and 35.00 Hz. A signal averaging computer
 was programmed  to average activity during  the entire
 SI  to  S2 interval and to sort the  32 presentations  of
 S2 relevant and irrelevant. Therefore, CNVs prior to S2
 and VEPs  following S2 were recorded  and averaged.
 CNVs weie quantified by measuring the  area under the
 CNV  deflection in  terms of a voltage-time  dimension
 (Tecce  1972).   VEP   latency  was   quantified  by
 measuring the time  between onset  of S2 and the peak
 of the  prominent  positive  component  occurring
                         between 270 and 435 msec (P300) after onset of S2.
                         The  effect of selective  attention  to  relevant and ir-
                         relevant stimuli on  VEP amplitude was quantified by
                         measuring the peak-to-trough amplitude  of the surface
                         negative and positive components at about  200 and
                         300  msec, respectively,  after S2  and finding  the dif-
                         ference in this  measure  when a given S2 was  relevant
                         or irrelevant.

                         Results and discussion

                            The  problem  type  differentially  influenced  d1,
                         CNVs, and VEPs. Color and line orientation discrimina-
                         tions  yielded higher d'  scores (p t .01) than letter or
                         word  discriminations (Fig.  1).  The  largest  Oz  VEP
                         difference between relevant and irrelevant stimuli (Fig.
                         1) was also observed for colors (p < .05). VEP and d'
                         results indicate that colors were  discriminated more
                         readily than words.  Smaller CNVs (Oz and Cz)  were
                         obtained to color than word stimuli (p  < .05)  (Fig. 2),
                         which indicated a  greater  state  of  preparedness  or
                         arousal  for  the  word  problems.   Together,  these
                         measures  show  that  the  word  problems  were more
                         difficult and arousing to the children.

                            The relevancy of stimuli had  a pronounced effect
                         on VEP  amplitude:  all  children gave  larger VEPs to a
                         stimulus  when  it was relevant as compared  to irrele-
                         vant.   The  VRD  group  showed  greater   relevant-
                         irrelevant  differentiation in their  Oz VEPs  than the N
                         group (p ( .05) (Fig. 3> This greater  differentiation in
                         the VRD  group suggests  that  they were  selectively
                         attending  more than the other groups. As the children
                 1.4
             J
             u
                 0.6
an
                        C  Li  L» W
                                    ju
1.4

1.2

1.0
nHll
                                               C  Li   Li  W
                                                PROBLEMS
   Fig. I, Effect oj'problem type on behavioral response accuracy (
-------
 302
540
f 520
* 500
| 480
| 460
440,
•
• Oz
•
•
.
[n
••







IM



^^





C Li Le W
430
410
390
370
350
Cz
•
»
•
P
-


nn
-



                                                 PROBLEMS
                                                                            C  Li  Le W
                  Fig. 2.  Effect vj problem type on CNV amplitude. Same abbreviations as Fig. 1.
>
}
UJ
1
UJ


2.0
1.5
1.0

0.5
0.0
•
•

•n
•Ml








N ARD VRC
GROUPS
Fig.  3. Difference  in  VEP amplitude due  to  selective
attention  in the normal (N), auditory (ARD). and visual
reading disability (VRD) groups.
in the three groups were all trained to the same level
of performance on the  visual discrimination tasks, it
appears that the  VRD  group compensated for their
deficiency  by greater selective attention in  the visual
discrimination task.

   The  latency of the  P300  component  of VEPs
recorded from  both the vertex and occiput indicated
that  VRD  children had longer latencies than ARD
children, who, in  turn, had  longer latencies  than N
children  (p < .05) (Fig. 4). These  latency  differences
were interpreted as suggesting that the reading disabled
child  processes sensory  information  at  a slower  rate
than the normal  child, which may be indicative of a
neural deficiency.

   The  VEP  measures  of  component  latency  and
selective  attention  to  relevant and  irrelevant  stimuli,
which yielded significant group differences, support the
notion  of  a  sensory-specific  deficit  in the  reading
disability syndrome. The separation of reading disabled
children  according  to  modality-specific   perceptual
capabilities is amenable to the present  study's findings.
8
g
>*
u
Z
Ul
3
i

310 r

305

300
295
Oz


•
r[]
^•M



M




                          N   ARD  VRD
                                                   GROUPS
395
390

385
380
375
4
•

Cz
-
-
•
n
^^H







M—»






                                                                             N   ARD  VRD
 Fig. 4. Mean latencies of the P300 component in the normal (N), auditory (ARD), and visual reading disability (VRD)
 groups.

-------
HEMISPHERIC EFFECTS OF STIMULUS  SEQUENCE AND
SIDE OF STIMULATION ON SLOW POTENTIALS  IN
CHILDREN  WITH READING PROBLEMS1


B. FENELON
The University of Newcastle, New South Wales, Australia
   It has generally been accepted that CNV amplitude
is  independent of the  modality of stimuli employed.
The  assertion was originally made by Walter (1967),
but  formal  experimental  tests  were  not carried  out
until  recently, presumably because experimenters, in
manipulating  stimulus  mode, obtained  generally con-
firmative results.  Supportive experimental evidence was
put forward by Blowers et al. (1976) who, using nine
combinations of auditory, tactile, and visual stimuli
and measuring vertex CNV, found no modality-related
differences.

   There have been opposing  findings. Gaillard  and
Naatanen (1976) reported modality-related differences
in CNV measured from three scalp sites (vertex, mid-
left temporal, right occipital). They drew attention to
the differences in effects at different sites, in different
parts  of  the  waveform,  and  for the warning  and
imperative stimuli.

   The  question  of CNV asymmetry is also of great
interest  in  relation  to  clinical   problems,  such as
localising the motor speech area (Low et al.  1976) or
throwing light on  the brain  responses of stutterers
(Zimmerman  and  Knott  1974). It  is  of  special
importance in assessing cerebral dominance and hemi-
spheric lateralization of problem readers.

   Earlier studies of CNV features of dyslexic children
suggested  that the  nature of  S2 has a determining
effect on electrical  activity that  develops before and
after  S2 (Cohen  et al. 1965, Fenelon  1968). Reading,
particularly  in interaction with a tutor, involves the
rapid sequencing of visual  and auditory identifications
of words,  or elements  of words.  Alternatively,  a
particular  sound  constellation   is  required  to  be
matched  by  a  visual  identification.  In  physical
elements, the child is  experiencing  a  sequence of
 1 Computer programs used in this study were provided by the
  Burden Neurological  Institute, Bristol, England. The author
  wishes to thank Mis. S. Byron for typing the manuscript
stimulation that is visual-auditory (V1-A2) or auditory-
visual (A1-V2). These  sequences are fundamental  to
certain techniques of teaching. In most instances, the
visual performance involves eye fixations followed  by
saccadic movements to fixate a visual feature appearing
in the right hemifield.

  The aim  of the present experiment was to examine
the   effects   of  lateralised  stimulation  on  certain
measures of  the  event-related slow  potential  (SP),
comparing  reading  problem  children  with  normal
readers  in unimodal and bimodal stimulus sequences.
On  the basis of previous neuropsychological studies  (to
be reported),  it was  hypothesised (1) that the problem
readers  would develop CNV comparatively weakly in
the left parietal region and  (2) that right hemisphere
responses would  be more  highly correlated in  the
problem group than in the normal group.

Method

Subjects

   Seven problem  readers,  mean  age  114  (±  11.6)
months, mean IQ 113.4 (±  12.8), were referred by a
special class supervisor as being retarded 12 months or
more in reading achievement  on  at least two out of
three  group   reading  tests  (Australian Council  for
Educational Research, Melbourne) and on the Schonell
Rl word recognition test. The  mean reading age of the
problem readers was 92 (± 7.2) months compared with
 118 (± 5.4) months in the normal readers  group. The
seven normal readers were of comparable chronological
age (114 ± 5.4 months) and intelligence (mean IQ  117
± 13.4).

   The  handedness of the subjects, assessed by several
different  methods,   was  uniformly  righthanded.
Although hand of response  has  been shown to  affect
SP   symmetry  (Otto  and  Leifer  1973,  Kutas  and
Donchin 1974), response laterality was  not used as an

-------
 304
                                          Fenelon
 independent variable because of time limitations in the
 full experimental sequence.

 Apparatus

   A Devices M-19 unit was interfaced with a remotely
 located PDP-12 computer. Low band-pass filters were
 set  at  15  Hz, and time  constant at 3  sec.  Ag/AgCl
 electrodes  were applied to the scalp at sites iF4,  iF3,
 iP4, and iP3 and were referred to common  mastoids.
 The  bipolar vertical  EOG was also recorded.  Audio
 signals  were 700-Hz or 1000-Hz pure tones, duration
 100  msec, delivered  through  earphones at 70  dB.
 Visual   signals  were   low-intensity  flash   stimuli
 subtending a retinal angle of 1° and displaced approxi-
 mately  9  from  a central unilluminated  fixation  spot
 that was also employed for eye fixation when auditory
 stimuli  were presented. Visual trials  were  conducted in
 semidarkness in a sound-attenuated room.

 Procedure

   The  paradigm  employed  was  the simple  S1-S2
 motor  response  sequence (ISI 1400 msec  and  ITI
 greater  than 10 sec). The  subject responded to S2 with
 a  button held in the  right hand. The parent and an
 assistant remained in the  room to keep up motivation
 and alertness and to provide company for  the child.

   Following training  trials, the experiment  was con-
 ducted  in  two stages:  (1) On-line  averaging of eight
 eye-movement-free  responses  in each of the  following
 four  conditions  (where  Al  = 700-Hz  tone;  A2 =
 1000-Hz tone;  V = visual stimulus; L = left ear or left
 visual field; R = right side):

   LV1 - LV2,  RV1 - RV2, LAI - LA2, RA1 - RA2

 Sequencing  of conditions was  varied from subject to
 subject  to  avoid systematic  arousal  effects.  During
 averaging runs,  the subject was advised of his reaction
 time  following each trial and encouraged to improve
his performance.  (2)  Each subject  received  16 trials
each  for  Al  • RV2  and RV1 - A2 (A = binaural
 1000-Hz  stimulation)   in   predetermined   random
sequences.  Acceptable  single   trials   were stored  for
off-line   processing.  At  this  stage,  the  instructions
informed  the subject  that the SI modality predicted
the other-modality stimulus at  S2.

   Various  SP  measures were taken using off-line
cursor and integration  programs. The analyses reported
below were conducted on maximum  peak-to-peak CNV
amplitudes  measured  during  the interval from  400
 msec post-Si  to the onset of S2.  Maximum positive
 peak was  determined  for the unimodal  sequences in
 the  interval 400  to 480 msec post-Si. In the bimodal
 tequences, post-Si positivity  was prolonged and the
 corresponding  point was taken in the 650- to 730-msec
 interval.  Maximum  negative  peak in both  sequences
 was measured in the 80-msec interval prior to S2. The
 stimulation   sequence  and  measurement  interval
 (unimodal condition) are illustrated in Fig. 1.

 Results

   Sample CNV  waveforms  (auditory unimodal con-
 dition) are illustrated in Fig. 1.  Means  and standard
 deviations for the unimodal conditions are represented
 graphically  in  Fig.  2.  Mean  maximum  CNV  and
 standard  deviation in the  bimodal  stimulation con-
 ditions are shown in Table  1.

   Data  from unimodal  and bimodal  conditions were
 submitted to 2 (groups)  x  2 (stimulus conditions) x 2
 (hemispheres)  x  2  (anterior-posterior  locations)
 analyses  of variance  with repeated measures on factors
 2 to 4. Factors with  p < .10 are  grouped in Table 2.

   To  assess  intra- and interhemispheric  symmetry of
 waveforms, correlation  coefficients  were  computed
 from  group-average data for  each  waveform pair. Cor-
 relations   were computed for each waveform pair by
 shifting in one-point  increments ±64 points from the
 origin. The maximum correlation observed for each
 pair is shown in Table 3  for each unimodal stimulation
 condition and group. (This analysis was not undertaken
 for bimodal stimulation data.)  Maxima occurred at or
 near zero  displacement.  Correlations coefficients r >
 0.26,  t(62) = 1.98 are significantly different from zero.
 Correlation coefficients were  Z-transformed  and  differ-
 ences  between pairs  of coefficients were  tested for
 significance in relation to the standard error (SEZ)  =
 0.18.

   Significant differences between  pairs of coefficients
 in a row  or in a  column  are  indicated in Table 3.
 Between-group differences for right hemisphere associa-
 tions  support the hypothesis that  right  hemisphere
 responses  would  be  more  highly correlated in  the
 problem  group than  in the normal group.  Some cor-
 relations  support  differential  hemispheric  organization
 of responses in problem and normal readers.

 Discussion

   The data  for  unimodal  visual  stimulation and for
bimodal  stimulation   do  not  support the hypothesis
 that CNV develops  comparatively weakly in the left
 parietal  region  of problem   readers.  Some support,
 however, is derived from the auditory stimulation data
 if one  is  willing  to accept trends (p < .10) obtained on
 a  stringent analysis  applied  to  the data from  small
 groups of problem readers who were not extreme cases
 of disability. The  first-order  interaction for groups x
 location  reflects  the  strong  development  of parietal
 CNV  in   the  normal group.  The  higher  order  inter-
 actions (p  <  .10)  result  from complex  differences

-------
Hemispheric SP in Dyslexic Children

                IF3S1TE
                                              305
          IF4 SITE
                                                                                                10MVJ.+
      h'ig. I. Group average CNV. Auditory SI-S2 sequence id fight ear.  Problem readers, heavy trace. Negative up.
                Table 1. Group Mean (M) CNV Maximum Amplitude and Standard Deviation (SD),
                                     in Microvolts, for Bimodal Condition
Stimulation
condition*
A, - RVF,
RVF.-A,
Experimental
group
Problem
Normal
Problem
Normal
Recording tites
1F3
M
11.7
22.6
17.9
19.9
SO
5.37
6.19
6.31
4.62
IF4
M
16.1
19.9
14.6
21.6
SO
6.26
4.93
3.93
6.22
IP3
M
20.3
26.6
21.1
32.2
SD
7.76
12.27
10.74
11.03
IP4
M
19.7
26.9
22.9
343
SD
6.93
8.36
4.96
8.39
                   •RVF - right visual field; A - both ear*.
 between groups in parietal CNV amplitudes when the
 right ear receives the signal. In absolute amplitude, the
 left  parietal  CNV  of  problem  readers  is  only SS
 percent of that for the normal group in both auditory
 stimulation conditions. Right  parietal CNV of problem
 readers actually exceeds  that  of the normal group
 when the  left  ear  receives  the  signal.  Over  all six
 stimulation conditions, the problem group averages 63
percent of normal group CNV amplitude at the  left
parietal site, while at the right parietal site the figure is
81 percent. This indicates a relative left-sided weakness,
in accordance with the first hypothesis.

   Prom  Fig.  2,  a  difference in response  between
groups is  observable for left- and right-sided  auditory
stimulation.  At the left parietal site, CNV amplitude in

-------
306
                                                                                          Fenelon
         LEFT VISUAL FIELD STIMULUS
                                                             RIGHT VISUAL FIELD STIMULUS
30
21
26
24
22
20
IB
16
14
12
10-
 I

 41
                                                        30
                                                        28
                                                        ?6
                                                        24
                                                        22.
                                                        20-
                                                         II
                                                         16
                                                         14-
                                                         I?
                                                         10-
                                                         |.
                                                         A
                                                         4.
                                                         V-
                                                                              "
                                                                                  I     I
          LF
                          LP
                                  RP
                                                               LF
                                                                       RF
                                                                               LP
                                                                                        RP
            LEFT EAR STIMULUS
                                                                   RIGHT EAR STIMULUS
>
5.
   )6
   (4
  J2
  30
  28
  26
  24
  {2
  20
   8
  16
  10
   4
   2-
         LF
           1
                '
                            i     ;
                            i     i
                            i     i
                 RF

               L-LEFT
               R-RIGHT
                          LP
/•'if;. 2.  Mean maximum p.
UH)n>l4lM>niH'<-i>i>M-Sl
                               RP

                               F-INFERIOR FRONTAL SITE
                               P-INFERIOR PARIETAL SITE

                          i<> nt-Kuiive ami'liiiitii' (mnrwolix) in C
                                                                                       RP
                                                                          PROBLEM READERS
                                                                          NORMAL READERS
                                                                                  o 490 msec ami

-------
Hemispheric SP in  Dyslexic Children
                                               307
     Table 2. AIMOVA Summary of CNV Measures
                (df = 1/2 All Factors)



Groups
Hemispheres
Location
G x L
GxSX L
G x H x L
Stimulus conditions
Unimodal
Visual
F

5.38
8.96



P

.05
.05



Auditory
F


20.22
4.24
4,57
3.40
P


.001
.10
.10
.10

Bimodal
F
10.82

29.48



P
.01

.001



both  groups  is  more  than  30  percent greater  for
right-sided stimulation. At  the  right parietal site,  the
problem group  loses 45  percent  amplitude  compared
with  an  increase of  more  than  80  percent  in  the
normal  group.

   Other  features of  interest  are: (1)  The problem
group CNV amplitude is generally less than that of the
normal  group, the difference  being significant (p < .01)
under bimodal stimulation (Table 1). (2) Frontal CNV
amplitudes in the problem group exceed those of the
normal group for unimodal visual stimulation (Fig. 2).
The  differential  tends  to  disappear  when  auditory
stimulation is used (Fig.  2) and is reversed  under the
more  complex bimodal  stimulation conditions (Table
1). (3) Parietal CNV amplitude greatly exceeds that for
frontal CNV in both  groups, as shown by the signifi-
cant locations main   effect across  all  stimulus  con-
ditions (Table 2).

   The second hypothesis  receives partial support  from
the correlations.  Right hemisphere responses are  more
highly  correlated  in  the   problem group  than  in the
normal  group, but only in the visual  stimulation con-
ditions.

   In  CNV studies, the association between physical
stimuli confers symbolic meaning on the simple stimu-
lation  sequence.  Even in  the absence  of instructions
that  give   specific  meaning   to   SI  and  S2, the
conditional linkage between the  two  stimuli evokes
some  level of language mediation  of the contingency
involved,  leading  to  the  initiation  of  the  required
motor  response.  The clectrophysiology  of cerebral
response to this simple configuration may  provide clues
to the topography and dynamics of language processing
and  production in the brain of problem learners and
retarded persons (Karrer  and  Ivins 1976c)  as  well  as
normals. The present  data raise a number of questions
concerning  the  organization of  electrical activity  to
                    Table 3. Intra and Interhemispheric Correlation Maxima of Waveform Pairs
                                      in Unimodal Stimulation Conditions3

                                                            Sites intercorrelated
               Modality     Field    Group
                                                Interhemispheric
                   Intrahemispheric
                                           iF3-iF4
     iP3-iP4    IF3-SP3
                                                                                      iF4-iP4
Problem





Left
Normal

Problem
Right
Normal
Problem


Auditory


Left
Normal

Problem
Right
Normal
.20

.73

.50

.74
.71

.76

.63
*
.78
.87

.83

.87

.86
.64

.68

.47
*
.89
.09

.08

.20

.36
.55
*
.25

.63

.49
47
*
.20

.57
b
.29
.61

.67

.29

.38
                a|f r > 0.26, p < .05. The difference between adjacent pairs of coefficients in rows and columns
                 was also evaluated (see text) and significant differences are indicated by aster isk(s).
                ^Trend: p <.10.
                *diff: p < .05.
               •*diff: p <.01.

-------
 308
                                           Fenelon
laterallzed  auditory  stimulation. Intrahemispheric cor-
relations show  an  interesting difference  in pattern
between groups.  In  the right hemisphere during right*
sided stimulation, frontal-parietal correlations are lower
in problem readers  than in  normals, Conversely, in the
left  hemisphere during left-sided stimulation, frontal-
parietal correlations  are lower in normals than problem
readers.   Also   the   cross-hemisphere   correlations
(especially  parietal) are strikingly different between the
groups when stimulation is right-sided.

   In view of the small sample  and limited  range  of
reading disabilities in  the present study, it is prudent
to withhold  attempts at  explanation until more data
are available.
Summary
   SPs were recorded from inferior frontal and parietal
sites over  both hemisphere!  of problem and normal
readers of equal age, schooling, and intelligence. Stimu-
lation  was  given  in  visual,  auditory,  and  mixed
modality sequences. Correlations  between  SPs in the
right  hemisphere were  relatively high in the problem
group,  but  differences  between   groups   were  not
observed in  auditory  conditions.  The results support
the hypothesis  that  problem readers generate  CNV
relatively weakly in the left parietal region.

-------
TASK-RELATED CORTICAL POTENTIALS IN CHILDREN
IN TWO  KINDS OF VIGILANCE  TASKS


D. FRIEDMAN, H. VAUGHAN, JR., AND L. ERLENMEYER-KIMLING

New York State Psychiatric Institute, New York, and Department of Neurosciences,
Albert Einstein College of Medicine, Bronx, NY, U.S.A.
   The existence of long-latency positive waves record-
able  from the scalp of adult subjects, generally referred
to as Association Cortex Potentials (Ritter et  al. 1972)
or P300 (Sutton  1969),  has  been  known for over a
decade. These  late  positive  waves have  been  little
studied in  children.  With renewed  interest in  the
neurophysiologjc bases of childhood learning  disorders
and  in neurophysiologic development in general, data
have been published recently on "task-related" cortical
potentials in 6- to 8-year-old children (Kurtzberg et al.,
in press); "event-related"  slow potentials  in aphasic,
dyslexic and normal children (Otto et al. 1976); visual
event-related   potential   correlates   of  informational
linguistic content in normal children (Shelburne 1973);
and  visual event-related potential correlates of meaning-
ful stimuli  in  normal children (Symmes and  Eisengart
1971).

   Kurtzberg et al. (in press)  presented data showing
the  existence  of late  positive  component morphology
similar to that  of the adult by the age of 8. Shelbume
(1973) replicated  in  children his  earlier  study  with
adults (Shelburne  1972), showing  an  enhanced  late
positivity to a third consonant,  which  delivered task-
relevant  information. Symmes and  Eisengart (1971)
reported a  late vertex negative wave, peaking at about
500 msec, to meaningful visual stimuli. Other investiga-
tors have also  reported the occurrence  of late negative
deflections to complex  visual  patterns   in  children
(Kurtzberg et  al., in  press; Neville 1975), as  well as in
adults (Cohen and  Walter  1966,  Courchesne  et al.
 1975, Lifschitz 1966). No general  hypothesis  as to
what this  negative wave  represents  has  been  formu-
lated.

   The present study  is concerned with initial analyses
of late positive waves recorded in response to numeric
stimuli while   children were involved  in  information
processing  in two  different vigilance tasks. In one task,
the  subject had to respond to the same target through-
out a block of trials, while  in the other, the subject
had to respond to the  repetition of any immediately
preceding  stimulus. The  processing demands  of  the
latter task were greater, since the subject had to hold
the stimulus in memory and wait to  see if it would
repeat.

   These children  formed  part  of a  normal  control
group  used  to study  a  large  number  of  biological,
sociological, and  psychological  variables in  a longi-
tudinal  project concerned with children at high risk for
schizophrenia (Erlenmeyer-Kimling 1971, 1975). The
data from  the  high risk sample  and  from additional
control  children  will  be  discussed in  a subsequent
report.  Data from an earlier version of one of the tasks
have  been  reported  elsewhere  (Rutschmann  et  al.
1977).

Method

   Six  children, aged  11-14, served  as  subjects. This
project  was their  second  round  of testing, occurring
approximately 2  years after they were  initially seen.
Each child received a psychophysiologic battery, whicn
took approximately 2.5 hours  and included  Tasks A
and B. Task A  was  a vigilance task  (Orzack  and
Kornetsky  1966 standard task) in which the signal was
the  number 08 (15 per block  of 60  stimuli)  and the
nonsignals  were  15  of  numbers from 02-19, each
occurring 3 times during a block. Thus, the  signal to
nonsignal ratio was 1:4. Task B (similar to  the Orzack
and Kornetsky 1966 A-X task), which always followed
Task A, was a vigilance  task with different demands.
The  signal  in Task  B  was the  repetition  of  any
immediately  preceding number  (which  occurred  16
times per block),  and the  nonsignals  were 12 of the
numbers from  02-19, occurring  4 times each within a
block  (48  per  block). This procedure yielded a signal
to nonsignal ratio of 1:4. Any bias that may have been
introduced by the fact that Task B  always  followed
Task A is opposite to the hypothesis,  since the  pre-
diction was  that  late positive  component  amplitude
would  be larger to nonsignals in Task B (i.e., opposite
to habituation effects).  In  both tasks the  required

-------
310
                                    Friedman et al.
 response  was  a  brisk  extension  of  the right wrist.
 Reaction  time  was measured at the onset of the EMG
 burst.

   For  each task, a practice  block of trials was given
 before each set of seven to eight blocks to ensure that
 the  subject  was  responding  correctly  and  was pro-
 ducing  usable  EMG pulses (due  to  equipment  mal-
 function,  two  subjects  received only 7 blocks of each
 task). A short  rest between blocks and a  5  to  10 min
 rest  between tasks were given.  Stimuli (50-msec dura-
 tion) were presented on a  slave  scope, transmitted to a
 video monitor in  the subject's room by a video camera,
 and  were  flashed  at moderate  intensity at a visual angle
 of  2 deg 20  min.  The  ISI was  1.5 sec.  EEC  was
 recorded  from  Fz, Cz, Pz,  and Oz, referred to the right
 ear;  and vertical  EOG was recorded from  an electrode
 located above  the right eye. Physiologic  signals  were
 recorded  by means of Ag/AgCl electrodes and  were
 amplified  with a time-constant  of 1 sec and a  gain of
 approximately  20,000.  Data  acquisition  and stimulus
 presentation were  under  control  of a  PDF  11/10
 computer. Data were digitized at 4-msec intervals for a
 total of 1100 msec  (100 pre- and  1000 post-stimulus)
 and  were  stored  on 9-track  digital tape for  off-line
 analyses.  Ideally,  there  were   120 signals  and 360
 nonsignals for  Task A  and 128 signals and 384 non-
 signals for Task B. Because of blinking and  eye move-
 ment artifact, however, roughly  one-quarter to one-half
 the  trials  for a given subject had  to be eliminated to
 produce an artifact-free average. Great care  was taken
 to eliminate these artifacts due  to  their known effect
 on   event-related  potentials,  especially   in children
(Eisengart and Symmes  1971,  Shelburne 1973).
   4. P319-352 of greatest  amplitude in the parietal
region, but  also present in  the  central and occipital
areas,  and larger to signal than to  nonsignal stimuli.

   5. P462484  largest  at   the   parietal  placement,
identifiable   in  all  loci, and larger for signal  than
nonsignal stimuli. This  late positive  activity returned
slowly to baseline in the parietal and central areas.

   A late negative  wave, peaking at about 600 msec,
was  seen only in  response  to signals  and was  more
negative in Task A than in Task B.
        NONtlOMALS
                                       SIONALt
                     	TAIKA


                     ....TAfK •
                                                                            5>iV
Results

   Fig. 1 presents the grand mean waveforms for the
six subjects for each type of stimulus  from each task.
The occipital  visual event-related potential consisted of
three  initial  peaks, P143,  N187, and  P231,  which
remained essentially constant in amplitude  and peak
latency  in all  conditions.   Grand  mean  waveforms
revealed  several  additional  peaks,  each  of  which
possessed a characteristic  timing and topography and
varied to differing degrees across conditions. In  order
of increasing  latency, the principal positive peaks can
be identified as follows:

   1. P176 identified solely  in the parietal placement
in all conditions.

   2. P220 most clearly defined in frontal and central
placements of nonsignal averages.

   3. P264 present  in the signal averages of both tasks
in the frontal  and central recordings.
 Fig. 1. Grand mean event-related potentials across six
 subjects. Stimulus  onset  at 100  msec. Time lines
 every 110 msec.
   Using the grand mean waveforms as a template, two
of the investigators (DF and HV) independently deter-
mined the presence or absence  of these  late positive
components  in  the  data  of  individual  subjects.  A
component  was  considered  present  only  if it  had a
clearly defined peak at a  given electrode  and  if its
topography  and  latency  were  similar to  that of the
grand  mean.  Amplitude  of  a given component was
scored by  fust determining  the electrode at which it
was most clearly  defined and then measuring amplitude
at that latency  at each  of the other electrode  sites.
Thus, although P200-264 was present as a well-defined
peak  rnainly at the frontal  and central electrodes, an
amplitude gradient across electrodes was  obtained by
means of this method.

-------
Cortical Potentials in Children
                                                311
   Table 1 presents (1) mean latencies, as measured in
the data  of the individual  subjects,  for each  of the
peaks (labeled for their latencies, taken from the grand
mean waveforms) and  (2) the number of  subjects in
whom the investigators scored a  given peak as present.
T-tests  for correlated  means  revealed no significant
latency differences between tasks for any of  the com-
ponents.

   Fig.  2  presents  the mean baseline-to-peak  topo-
graphical  distribution   for these  four positive  peaks.
Unconnected  points  represent  P143 of  the  visual
evoked potential. T-tests for correlated means showed
no significant amplitude differences in P143 between
tasks for either signals or nonsignals.

   Analyses  of  variance  for repeated measures (task
and electrode location) were  used to assess significance.
With  the  exception  of PI 76 to signals, electrode loca-
tion  was  a significant source  of  variation  for  all
components to both signals and nonsignals (p  < .05 for
P176  to  nonsignals; p <.001  for all  others). P176
amplitude was larger to Task B nonsignals than to Task
A nonsignals [F (1,5) = 9.8, p < .05]. For P220-264,
no significant  task effects  were observed for  either
signals  or  nonsignals.  For  P319-352,  and  only for
nonsignals, Task  B produced larger  amplitudes than
Task A [F (1,5) = 4.8, p < .07]. For P462-484, and
again  only  for  nonsignals,  Task B  produced  larger
amplitudes than  Task  A [F  (1,5) = 4.8, p <  .07]. The
only  significant task effect for signal responses was for
P462484,  where  the  task x electrode location  inter-
action was significant  [F (3,15) = 5.4, p < .01]. Tests
for  simple  effects  (Winer  1962)  showed that this
interaction was due  to greater amplitudes at Fz and  Pz
during Task  A than Task B  (p < .01  at both), while
there was no difference  at Cz or  Oz. This finding is
possibly due  to longer  and more variable reaction times
in Task B than in Task  A. This factor would lead  to
increased  late-component jitter in Task B relative  to
Task A and, thus, a decrease in  amplitude. It is also
possible that decreased temporal overlap  of these late
components in Task B, relative to Task A, contributed
to this effect.

   The  results  demonstrate  an  effect of  processing
complexity   on  late  positive component  amplitude
elicited  by  nonsignals  of  Task   B. This  electro-
physiologic effect was paralleled by behavioral changes
in that mean reaction  time was longer in Task B (544
msec) than in Task A (466 msec).

Discussion

   These  data  show an effect of cognitive processing
demands  on late positive component amplitude and
replicate the findings  for adults (Friedman et ah, this
volume). As in our interpretation of the adult data,  the
most parsimonious explanation of the children's data is
that late component amplitude to nonsignals in Task B
is increased, relative to  those in Task A, because of
more complex  processing demands  for the nonsignal
                           Table 1. Mean Peak Latencies, Standard Deviations 
-------
312



   IS-

   M-

    S'
  NONSIGNALS
              P176
SIGNALS
       • TASK A
       > TASKS
   15-

   10-
   ,..
              P220-264
                       P319-3S2
   •3J

   15-

   10
    0
   •3
              P462-484
Fz  Cz  Pz  Oz
                                    Fz  Cz  Pz  Oz
  Fig,  2.  Mean baseline-to-peak signal and nonsignal
  amplitudes for the four positive peaks from the two
  tasks.
stimuli. In Task B, the subject must store  each non-
signal  and wait to see  if it recurs, while in  Task A, a
simple  target-present-or-absent  decision  is  made  on
each  trial. These  data  support  the  findings of others
who  have  shown  increased  P300 amplitude  as  a
function of increased cognitive activity (Donchin et al.
1973,  Friedman et al. 1975, Poon et al. 1976), as well
as those  who  have shown  the  presence of late com-
ponents within the  P300 range in tasks involving the
storage of information  for  task-relevant signal stimuli
(Chapman 1973,  Posner et al. 1973).  The main  dif-
ference between this study and these others is that a
comparison was made between physically identical but
task-differentiated nonsignals.

   There are  few data on late component morphology
and  topography  in  children. The morphology of the
responses  obtained in this study appear very similar to
those of the adult, marked by, in  the  case  of signals,
large-amplitude  late positive components  and, in  the
case  of nonsignals,  smaller, but reliable late  positive'
                                    Friedman et al.

components  in  the latency  range reported for P300
(e.g., Ritter et al. 1972, Squires et al. 1975).

   There  appeared to be  three  distinct  late  positive
components, similar in latency and scalp distribution
to the P3a, P3b and SW responses seen by Squires et
al. (1975). P220-264, of similar latency  to P3a, did not
differ  in  amplitude  between  tasks, although  it  was
larger  to  the signal (infrequent stimulus)  than to the
nonsignal  (frequent stimulus) in both tasks. Squires et
al.  (1975)  found  P3a larger  to infrequent  than to
frequent  stimuli, whether or  not these  stimuli  were
attended  or  ignored. P319-352 (latency range  of P3b)
and  P462484 (latency range  of SW) did differ between
tasks,  but consistently only  for  nonsignals. In  unpub-
lished  data,  Petrasek and Vaughan have  also  seen a
relatively  constant latency late  positive wave (latency of
300  msec)  and a  second positive component, which
varied with reation time.

   In  addition,  a late negative wave, maximal at the
occiput, was observed, but  only  in averages of signal
stimuli. Late negativity has been  reported by Symmes
and  Eisengart (1971) to be maximal  at  vertex  (500
msec to peak) in response to a cartoon stimuli in 5-11
year old  children  and by Neville (1975) at left and
right  temporal   electrodes   in response  to  pictorial
representations of common objects in 9-13 year olds; it
can  also  be  seen  in  the  data  of Kurtzberg et al. (in
press) to  signal  stimuli in a  visual discrimination  task,
maximal between temporal and occipital electrodes. A
late  negative wave was also seen in  two adult  subjects
of Courchesne et al.  (1975)  when the  subjects viewed
quasi-random color patterns  and easily  recognizable
black  and white patterns. The functional significance
of  this late negative wave  is not  known, nor  is it
known whether these reported negative waves  share a
common generator.

   In  conclusion, our data  are  consistent with  con-
clusions drawn from the studies of adult subjects, both
in general morphology  and  in the cognitive correlates
of   the  late  positive  components.  These  types  of
vigilance tasks are appropriate for the study of electro-
physiologic correlates of cognitive activity in children
and  are  probably  suitable  for  the investigation  of
perceptual disorders in school-age children.

Summary

   Visual  ERPs  were recorded from six  early  adoles-
cents in response to numeric stimuli during two kinds
of vigilance tasks. In Task A, subjects responded to the
same signal throughout a block of trials. In Task B, the
signal was the repetition of any immediately preceding
number.   The processing demands  of Task B  were
greater  than  those  of Task  A,  since in Task B the
subject had to  retain a short-term memory of  each
nonsignal  in  order  to determine  if it recurred  on the

-------
Cortical Potentials in Children
                                               313
next  trial. Three late positive components,  P220-264
with  a  centro-parletal focus, P319-352 and P462484,
both  with parietal  maxima, were observed  to signals
and nonsignals of both tasks. Nonsignals elicited larger
amplitude late positive components In Task  B than in
Task  A. P462484  amplitude was larger for signals in
Task  A than Task B, a difference that may  be due to
longer and more variable reaction times in Task B.

   These  results  indicate an  effect  of cognitive  pro-
cessing complexity  on late positive component ampli-
tude   elicited  by   physically  identical,  but  task-
differentiate   nonsignals,  and  demonstrate  that the
morphology and  cognitive correlates  of the responses
obtained in  children of this age appear very similar to
those of the adult.

Acknowledgments

   We are grateful to Dr. Jacques Rutschmann for his
participation and help in the design  of  the tasks. We
also  wish  to express  our appreciation  to  Mr. Jim
Hollenberg for computer programs used in this study
and  to  Ms.  Rebecca Jacobsen  for her aid in running
subjects and data analysis. This  study was supported
by  Giant MH-19560 from the National Institute  of
Mental  Health to L. Erlenmeyer-Kimling  and by the
Department of Mental Hygiene of New York State.

-------
 EVOKED  AND  SLOW POTENTIALS  DURING SENSORY
 CONDITIONING  IN  AUTISTIC,  MENTALLY  RETARDED,
 AND  NORMAL CHILDREN1
 F.  LAFFONT, N. BRUNEAU, Ph. JUSSEAUME, AND G. LELORD
Groupe Cerveau et Circulation, Tours, France
   Contingent  negative  variation  (CNV)  has  been
studied  in  retarded  children  and  children  with be-
havioral disorders. Walter (1969)  reported the absence
of CNV in  autistic children. Small et al. (1971), using
photographs of subjects and their parents, found CNV
in autistic as well as control children.  Lelord  et al.
(1973),  coupling sensory stimuli,  observed large CNVs
in autistic  children.  Laffont (1973),  coupling  sound
and  ankle  reflex,  reported similar results. Low and
Stoilen  (1973) showed  that  CNV did not  reflect
minimal brain  damage, school failure, or specific dis-
orders. Cohen  (1973) reported  CNV was present to  a
greater degree in control  children  than  in children with
learning disabilities. Karrer and Ivins (1976b) found  a
small  frontal  CNV  and lack  of  topographical dif-
ferentiation in retardates. Otto et al. (1976) demon-
strated that  aphasic children showed greater positivity
than control children during warning and encodement
intervals. Andreasen  et  al. (1976) reported lack of
CNV  in hyperactive  children.  The purpose  of this
study was to  compare  the characteristics of  evoked
potentials (EPs) and  slow potentials (SPs) in autistic,
mentally retarded, and normal children.

Apparatus and  procedure

  Material  and  methods were similar  to  those em-
ployed by Lelord et al. (1976). The nurse who usually
took care of the child was present and sometimes was
obliged  to  hold  his  hands. Ag/AgCl  electrodes were
placed at the vertex  and over the right  occipital area.

  Two  sessions  composed of  10 series of 20  trials
each  were   used  for every  subject.  The auditory
stimulus, a  brief sound (S) of 4  msec,  1 kHz, 25 dB
'This study was supported by 1NSERM (C.L. 1975, 1, 205-6)
 Fondation pour la Recherche Medicale  (S.O. 5.75), Caisse
 Regionale de Securite Sociale du Centre.  We wish to thank J.
 Goldstein for his assistance in the development of this text
 above adult threshold was presented alone for the first
 2 series (habituation: SHI, SH2). The visual stimulus, a
 1200-lux flash of light (L) from a lamp 40 cm in front
 of the subject, was presented 800  msec  after S for 8
 series per session (coupled SL). S was again presented
 alone during the final  2 series of  the second session
 (extinction: SE1, SE2).

   Stimuli were given only when the BEG was free of
 artifact  from head movements or muscle contractions.
 Many traces had to be  discarded because of artifacts.
 Vertical and horizontal averaged EOGs were recorded,
 and  series in which ocular potentials were larger than
 vertex potentials were discarded. It was not possible to
 record  EOG in more than half the patients who would
 not  tolerate electrodes  near the eyes. Three or four
 sessions were  required  in a number of autistic and
 severely retarded children.

 Subjects

   One-hundred-twenty-five children (60 girls, 65  boys,
 average  age = 8) were  examined.  Seventy  had been
 hospitalized in a psychiatric ward for  problems of
 varying intensity. Thirty-five had severe disorders: total
 or nearly  total  absence of speech, serious emotional
 problems,  and IQ <40. Thirty-five were moderately
 retarded (IQ >40) with more or less severe behavioral
 problems.  Twelve  children  were pupils  in  a special
 center  for  mental retardates  (IQ  between 40 and 60)
 with few emotional  problems. Forty-three normal chil-
 dren were brought to the laboratory by their parents.
 Most were students  in  public  schools; the youngest
 were still at home with their parents.

 Clinical characteristics

  The   clinical  characteristics  of  each  child  were
studied by  a  medical and sociologic team using ques-
tions that included several response alternatives  (e.g.,
"present,"   "absent,"    "intense").   Each response

-------
EP, SP during Sensory Conditioning
                                               315
alternative constituted a characteristic. Sixty-five clinical
characteristics were delineated for each child including:
lack of meaningful interpersonal relationships; need for
sameness; withdrawal   from  contact  (when intense,
resembled catatonia); psychomotor symptoms; speech
difficulties; anxiety (with agitation); perturbed activity
(when incoherent); total lack of initiative; paradoxical
agility;  sensitivity to  noise;  mood difficulties (with
excitement,  with depression, with  atonia); eating prob-
lems;  problems  with  bladder and  bowel  control;
perturbed sleep; problems  in sexual behavior;  aggres-
siveness  absent  (directed toward others,  toward self);
schooling possible (impossible);  absence of psycho-
logical or psychomotor  anomalies;  difficulties beginning
after birth   (at  birth);  neurological  signs; mental
deficiency  (complicated, isolated); adaptation  to  the
examining situation; IQ ( > 80, 60 to 80, 40 to  60, 20
to  40, <20,  impossible  to examine, homogeneous,
nonhomogeneous, intermediate); and age (0 to 2 years,
2 to 4 years, 4 to 6 years, 7 to  13 years).

Electrophysiological criteria

   Event-related  potentials  (ERPs) were  averaged for
each set  of 20  trials.  ERPs  in the habituation (SH),
coupled SL, and extinction (SE) conditions exhibited
different  waveforms,  amplitudes,  and scalp localiza-
tions.

   For each series,  ERPs were assessed  at  the vertex
and occipital area in three time windows:  (1)  from  0
to  100 msec after stimulation (early phenomenon), (2)
from  100  to  400  msec  after  stimulation (evoked
potential-EP),   (3)  from  400  to  800 msec  after
stimulation   (slow  potential-SP). The   amplitude of
each  potential  was   measured  peak   to  peak  for
responses  found in the  first  and  second windows.
Amplitude relative to a horizontal baseline that passed
equidistant  to the extreme  peaks recorded  during the
first 100 msec  was  the measure  of response  for the
third window. In the  first two  windows,  the potentials
were generally polyphasic. In certain subjects, however,
these  potentials were replaced  in the second  window
by a single  early wave whose amplitude  was measured
from the baseline previously defined.

    Amplitude values for each child were compared to
 mean values (m) for normal children  calculated during
 an earlier study (Lelord et al. 1976). The amplitude of
 each response was taken as equal to the mean ampli-
 tude if it was within ± 2 a of m; larger if it fell above
 or smaller if it fell below m by  more than 2 a.

    For each child, amplitudes of  responses to sound in
 coupled SL and extinction  series were  compared to
 those in the  second  habituation series (SH2).  Con-
 ditioning  was   considered  to  have  occurred  if SL
 coupling augmented  the amplitude  of responses to
 sound  as compared to SH2 responses. The amplitude
of responses to sound and light stimuli in SL series was
compared.  The positive or negative polarity of SPs in
the 400- to 800-msec window was also scored.

   EP  and  SP  responses at  the  vertex  or in the
occipital region were scored as "present" if the ampli-
tude measured more than 4 jiV. Responses to sound
alone were considered present  if they  appeared in
either SHI  or SH2. Conditioned EPs and SPs were
defined  as occurring where the  response to click was
larger in coupled SL trials than in SH2. Frequency of
conditioned response was scored  as "absent," "few" (n
< 6 out of 16 tracings) or "many" (n > 6). Evoked SP
of the  same  polarity at the vertex  and occiput were
considered  "generalized."   EP  responses  to  light
(window 2) were "regular" (n > 12) or "irregular" (n
< 12);  SP  responses  to light were "absent," "few" (n
<6), or "many"  (n >6).  Some  subjects evidenced
rhythmic phenomena (from 7  to 13  c/sec) in response
to stimulation. In  certain subjects  a "conditioning to
time" defined as a  response in the absence of light was
observed during extinction. In this manner, a total of
125 electrophy&iological measures were scored for each
child.

Analysis of  relationships

    A factorial  analysis  of  correspondence (Benzecri
 1976; also cf. Bruneau et al., this volume) of electro-
physiological  and clinical  data was  done  for  the 125
children with a  UNIVAC 1108  computer. Data were
extracted from a frequency  table with values between
0 and  1. The analysis then considered two multidimen-
sional  spaces. Each electrophysiological measure was
situated in relation  to  m clinical axes;  each clinical
 trait was situated  in relation  to n electrophysiological
 axes. Then a  search of the best space directions (inertia
 axis) was performed. The representations of clinical
 traits  and electrophysiological data  were  made sym-
 metrical to allow the passage from one representation
 to another. The two symmetrical representations were
 then superimposed on  a two-dimensional space. The
 relationships  between   electrophysiological data  and
 clinical traits were expressed  by  the  proximity  of
 numbers and letters, which represented  the  two data
 domains, respectively.

 Resulvs

    Fig. 1  illustrates  the two-dimensional  space  of
 factorial Axes 1  and 2, which represented 53.5% of the
 information.   A  progression  of  electrophysiological
 factors and  clinical traits  could  be  seen  along the
 second axis.  This  progression, depicted  for  Area I  in
 Fig. 2, demonstrated four approximately  equal bands
 distributed from top to bottom.

    Area I was  characterized  by: (1) many generalized
 conditioned SPs to  S  of coupled SL, (2) conditioned

-------
  316

  SPs larger  than  unconditioned SPs evoked by  L,  (3)
  SPs and EPi conditioned to time during extinction, (4)
  many vertex visual SPi, and (5) visual EP small at  the
  vertex and irregular on the occipital region. Associated
  clinical  trait*  were withdrawal from social  contact,
  need  for  tamenesi,  catatonla,  complicated  mental
  deficiency, and IQ between 20 and 40.

    Area II  was characterized  by:  (1) the  absence of
 conditioned  EP  or  SP  response to  S  during  SL
 coupling;  (2)  no rhythmical potential evoked by S of
 coupled  SL; (3) no  conditioned  auditory  EP during
 extinction; (4) during extinction,  SPs and  rhythmical
 potentials conditioned to  time; and (5) small visual EPs
 and many visual  SPs. Clinical traits included: schooling
 impossible,  lack  of meaningful interpersonal  relation*
 ships,  agitation,  speech  difficulties,  excitation,  non-
 homogeneous IQ, and neurological signs.

    Area III could be divided in two subareas:  superior
 (a) and inferior  (b). In  subarea (a), there were (1)
 many conditioned SPs localized either at the vertex or
 the occiput, (2) few  conditioned to time after  SL.
 Clinical  traits were  IQ between 40 and 60,  depression,
 atonia, perturbed activity,  and aggressiveness  directed
 toward  others. In  subarea  (b), there  were  (1) few
 conditioned EPs  and few  conditioned SPs, (2) early
 phenomena, (3) regular visual EPs  with average ampli-
 tude. Clinical traits included isolated mental  deficiency,
 absence  of need for  sameness,  absence  of aggres-
 siveness, and no perturbed sleep.

    Area  IV was characterized  by:   (1) many  con-
 ditioned rhythmic potentials and EPs at the vertex and
                            AREA!

                  Miny pmrolizMl condition** SPi.
                   Whhdrnwl from wcW eonttct.

                           AREA II

               Abttnct of condition** mponn (EP or IP).
                Schooling Imponlblo; omotloml trouble*.
 AX 181
  I.OS
                AREA HI

(i) Miny conditioned loctUitd SP; ftw condition** EP.
10 botwoOT 40 ind BO; portHrtid octlvltv; inrmlvomM.

     (b) Ftw condition** rtipoimi (EP « SP).
    Mintil dcflelmcy without imolitod troublw.

                AREA IV

  Miny condition** tvoktd md rhythmic potintlih.
  Abnnet of mtntal difletanoy; obttnot of troublo.
                           AXIS 2
                           44.SS

Fig.  I. Analysis of relationships. Two-dimensional space
depicting relations between elect rophysiological and clinical
measures clustering as four bands or areas.  EP: evoked
potentials; SP: stow potentials.
                                                                                     Laffont et al.

                                               In the occipital region, (2) generalized  conditioned  SP
                                               absent and conditioned SPs localized at the vertex, (3)
                                               conditioned  rhythmical  potentials  present  on  the
                                               occiput,  (4)  no conditioning to  time, (S) visual EP
                                               amplitude large both  at vertex and in occipital region
                                               Clinical traits included: IQ > 80, presence of meaning
                                               ful interpersonal relationships, absence of paychomotor
                                               symptoms, absence of speech difficulties, and absence
                                               of anxiety,

                                                 The results are  summarized in Fig. 3 by representa-
                                               tive profiles  from three 8-year-old  children. The left
                                               column  reflects Area I electrophysiological character-
                                               istics associated with clinical  traits of an autistic child.
                                               The  middle   column   represents   Area  IV  electro-'
                                               physiological  characteristics  of a  normally adjusted
                                               child.  The  right  column corresponds to  Ana  III
                                               electrophysiological  characteristics  associated   with
                                               clinical traits  of a  moderately mentally  retarded child.
 Discussion

   If one considers the frequency of occurrence of EPs
 as  a  reverse indicator  of  variability,  then  Area III
 retarded  subjects were  more  variable than  the other
 groups of children studied.  This  result may  support
 Callaway's finding (1973) in dull subjects. The absence
 of EP conditioning also confirms the observations  of
 Shipley (1970) who recorded a facilitation in normal
 but not in retarded subjects when sound and light were
 paired.

   The enhancement of SPs in autistic children was not
 observed in the CNV experiments of Walter (1969) and
 Small  et al.  (1971). This difference may be due to the
 absence of  motor  response  contingencies  and the
 motivational variables related to  motor involvement.
 For instance, enhanced SPs have also been observed in
 experiments  using  ankle  reflex as an  unconditioned
 stimulus (Laffont 1973). Enhancement of SPs does not
 seem to be related to extracerebral artifact because the
 EEC during  the trials was without movement or EMG
 artifact, and ocular movements were  minimal.  Eye
 movements were less frequent after sound (seen in 15%
 of the  EEC  tracings) than after light (seen in 80% of
 the  tracings).  Moreover,  averaged  SPs after  coupled
 sound  were  larger  than averaged SPs  after light.  It
 should  also  be  noted that  ocular potentials are gen-
 erally  small  or absent  in the occipital region. Gen-
 eralized conditioned SPs, however, were recorded both
 at the vertex and occipital region.

   A comparison  between Areas I  and  H  shows that
 the clinical  traits  of autism  are related to  generalized
 conditioned  SPs. However, the clinical traits  of severe
 mental  retardation,  impossibility of schooling, lack of
meaningful interpersonal relationships, and excitation
 corresponded to  the  absence  of conditioned SPs.

-------
  », SP during Sensory Conditioning
                                                                                             317
         14
       11
              17
             II
             8
                                 [LECTKOPHYIIOLOaiCAI.
                                   CMARACTtmiTICI
                 AX It!
                   18
                   HI
                   Al
                   42
                   Ml
                   Wl
                                   tf»lillll«|«tliMiruL
                                   mnywitMriiwIV
                                    CLINICAL T HAITI
                 41
                   It
                   14 :  MMltMlyMMlH
                   01 I  Mri (ormMM
                   17
                   11 :  tawtunmiMMty
                   13 :  lMkof»k!MUf Mitral
                   II
                   M
                               a
                               10
                                   IfMlMktltoMMln
Fig.  2. Analysis of relationships (Area I of Fig. I). The
relationships  were given hy the proximities  of letters and
numbers,  which respectively  represented electrophysio-
logical  characteristics  and  clinical  traits.  EP: evoked
potentials;  SP: slow potentials:  S: sound:  L:  light; SL:
coupled sound and light.
                      AUTISTIC
Similar ocular  artifacts  are  observed  in both  cases,
while conditioned  SPs, large  in Area I, are missing in
Area II.
   Polarity  of  SPs tends  to  be negative in autistic
children  and  positive in mentally deficient (Lelord et
al.  1976). Such a positivity, observed by Karrer and
Ivins (1976b) in mental retardates and by Otto et al.
(1976)   in   aphasic   children,   suggests   differing
mechanisms related to maturation or motor overflow
(Cohen 1973).

   The marked presence of EPs conditioned to time is
striking  in Areas  I and II. This particular  form of
acquisition,   mentioned  by   Popov  (1948),  Rusinov
(1959),  and  Lelord et al. (1967) in  mental retardates,
has  been studied more recently under  the  guise of
"emitted potentials" (Sutton  and Paul  1973,  Picton
and   Hillyard   1974)  or    "congruent  potentials"
(Buchsbaum  et  al.  1974a)  in  normal subjects.  It
appears   as   an enhancement  of some capacity to
reproduce temporal sequences in young patients.

Summary

    A  factorial  analysis of clinical data  and  EEC data
 obtained during conditioning with sound and light was
                                            NORMAL
                                                                                        RETARDED
     SH
                                                                              oc Wk
     SL
OC
                                             OC
      8E
            OC
                                      I

                                   V


                                  oc  <•»-•*
                                                      vv
                  I    CLICK
                           I   FLASH
           FLASH ABSENT        2Q MV
L
 Fig. 3. Individual tracings. The three tracings summarize the principal clinical traits and rlfctrophysiological characteristics
 exhibited by the analysis oj relationships. To the left; an autistic child (Area I. Fig. I); in the middle: a normal child (Area IV,
 Fig. 1); to the right: a mentally retarded child (Area lll(b). Fig. I). Top traces: habituaiion (SH); middle: conditioning (SL);
 bottom: extinction (SE).  V: vertex; Oc: occipital.

-------
318                                                                                          Laffontetal.

described in  125 children (43 normals,  82 psychiatric    localized  conditioned  SPs.  Clinically,  three  major
patients).  From  an  electrophysiological  perspective,    groups  appeared:  autistic,  mentally   deficient,  and
three  major groups could be defined: (1)  generalized    normal  children. The  relationship  of electrophysio-
conditioned SP without conditioned EP, (2) few con-    logical  and clinical   profiles  derived   from  factorial
ditioned  EPs or SPs, (3) conditioned EPs  and vertex    analysis was discussed.

-------
CNV  AND  EEC  PATTERNS IN CHILDREN WITH
CEREBRAL PALSY  AND  KNOWN  BRAIN LESIONS
M. PAPINI, R. ZAPPOLI, A. PASQUINELLI, M.G. MARTINETTI, AND S. GUERRI

Nervous and Mental Disorders Clinic, University of Florence, Florence, Italy
   Weinberg  and  Papakostopoulos (1975)  have  em-
phasized the need to study ERF patterns in relation to
the utilization of signal information for motor control,
a subject to  which an  entire section is devoted  in this
volume. Patients with  asymmetric brain  damage, such
as  hemiparesis  or hemiplegia  which impairs  motor
skills, constitute a unique clinical population for such
studies. Goto et al. (1973) found that  slow negative
potentials  recorded from the Rolandic area of hemi-
plegjc  adults were greater preceding voluntary move-
ment of the impaired limb than the  healthy limb.
These authors, however, did not report  the extent of
brain lesions, the degree of motor impairment, or the
symmetry of potentials on the two hemispheres.

   CNV patterns in children with focal  paralysis have
not been previously reported. Since  the frequency of
EEG abnormalities in children with cerebral palsy is
high (e.g., Winfield et  al.  1955), the present study was
undertaken  to   examine  the  relationship  of EEG
patterns  and the amplitude, shape and symmetry of
CNV in children with  this disorder. Detailed considera-
tion of the nature and extent of lesions which may
affect the symmetry of CNVs (McCallum 1972) will be
given.


Methods

    Six children  (7 to 12 years old; four males and two
females), four of whom were affected by hemiparesis,
one by slight upper  monoparesis, and  one by slight
upper monoparesis and triplegia, were examined. In all
cases   X-ray    contrast   investigation-pneumoen-
cephalography,   positive   contrast   ventriculography,
angiography, and computerized axial tomography-was
performed  to  determine  the  extent  and  cause  of
cerebral lesions. Clinical assessments of each child are
summarized in Table 1.

    A foreperiod reaction time  paradigm was used: SI
(click)-1.5  sec ISI-S2 (moderate intensity repetitive
tones)—response (terminating S2 as quickly as possible
by  a button press). Thirty-five trials were performed
using the more  skillful hand and 35 using die paretic
hand.  During the session,  children  gazed  at  a  black
fixation point on a mirror. A Grass 10-channel ac-dc
polygraph having a 6-sec time constant was  used to
amplify EEG data. Ag/AgCl  electrodes were placed
over C3, C4, and Cz and referred to linked mastoids.
Electrodes  above and below  the  eye were  used to
monitor vertical  EOG. In some cases, EEG recordings
were obtained from C3  and C4 referred to ipsilateral
mastoid and Cz  referred to linked ears. Finger photo-
plethysmograms,  electromyograms  from  the  muscles
and limbs involved  in  synkinetic movements, and
reaction times (RT) were also recorded. Polygraph data
were averaged on-line with a signal analyzer and stored
on  analog tape  for off-line averaging of selected trials
(without eye movement artefacts). The amplitude of
each CNV (8 to 16 trials)  was measured from baseline
to  points at 50,  100, and 150  msec preceding  the
onset  of  S2.  Baselines  were  defined as  the average
amplitude  of EEG activity 800  msec preceding  SI.
Experimental procedures have been detailed elsewhere
(Zappoli et al. 1973).


Results and discussion

   In spite of restlessness and motor impairment, CNVs
were obtained from all children, and more than 35% of
the trials, without movement or other artefacts, could
be  averaged.  In  some  cases,  the  response  proved
difficult owing  to the hemiparesis, and was  only
performed with motor involvement of the whole limb.
S6  was able to respond  only with the paretic hand as
the  other   was  plegic.  S3  presented  pathologic
synkinetic movements of the  hemiparetic limbs when
responding with the good hand.

   Vertex CNV voltage ranged from 8.3 to 40 juV. The
group  mean amplitude of artefact-free vertex CNV was
about  24 MV during unimpaired limb performance and
26.5 juV during impaired hand performance.  This dif-
ference  resulted  from two  subjects who showed  a

-------
320
                                      Papini et al.
                                   Table 1. Clinical Description of Subjects
8
1
2
3
4
6
6
Sex
M
M
F
F
M
M
Age
9
10
9
7
12
12
Motor impairment
R hemiparetii
R hemiparesis
L hemiparesit
R hemiparesis
R slight upper
monopa resit
Triplegia;
L slight upper
monoparetit
Epllepiy
R focal
R focal
R focal
	
R focal
L focal
EEC abnormal itiai
R normal;
L focal paroxysm.
activity
R normal;
L focal paroxysm.
activity
R flat;
L focal paroxysm.
activity
Almost continuous
slow wave pattern
Diffuse paroxysm.
activity
R focal paroxysm.
activity;
Lflat
Diagnoiis
Hemispheric
atrophy
Hemispheric
atrophy
Hemispheric
atrophy, especially R
Tumor of the
brain stem
Slight 1. Cerebrovasc.
disorder
Hydrocephalus;
malfunction of the
spinal cord
marked difference in skill between the two hands (slow
RT  with  the  paretic and  fast RT with  the  healthy
hand). These subjects showed no synkinetic movement
when  using the unimpaired hand.  No  differences  in
vertex CNV or in RT were apparent in subjects who
presented synkinetic movements.

   Analysis of amplitude asymmetries between C4 and
C3 indicated that the less impaired hemisphere showed
higher mean voltage  CNV (x = 29.8 vs. 18.2 juV) when
the  less  impaired  limb  responded (e.g., Fig.  1). The
only  subject who  presented no asymmetry in C3 and
C4 CNV  amplitudes during less impaired hand per-
formance  was a child (SS) with slight  monoparesis and
diffuse paroxysmal EEC abnormalities  (who  success-
fully  recovered). During trials in which the impaired
hand  was  used   (five  cases-S6  excluded  due   to
paralysis), mean  CNV voltage  was much lower in the
impaired (14 jiV)  than  less impaired  hemisphere (31
jt\0 in every subject.

   Goto  et  al.  (1973) have  suggested that  voltage
enhancement occurs  in  the less impaired  hemisphere
due  to compensation. Alternatively,  the  asymmetry
may  be  due  to  the  brain  lesion   and  consequent
dysfunction  in the  electrogenesis of slow potentials (cf.
McCallum 1972).

   CNV morphology  often showed a marked difference
between  C3  and  C4.  The  shape of the  CNV in a
hemisphere  was  often constant and  independent of
amplitude.  Three  CNV  patterns  were  apparent: (1)
normal CNV (Fig.  2A); (2) a small, smooth CNV with
a alow ascending limb, often preceded  and  followed by
smooth poststimulus positivities (Fig. 2B); and (3) a
large,  irregular saw-toothed and  sharp pattern  with
deep and rapid high voltage (up to 40 ^V) poststimulus
positivities (Fig. 2C). Differences were more evident in
raw recordings than averaged CNVs.

   The first  pattern was recorded  over  apparently
healthy hemispheres. The second  CNV pattern was
recorded  mainly  from  hemispheres with neuroradio-
logical signs of marked cerebral atrophy and flat EEC
activity, or signs  of a subcortical tumor and continuous
slow wave EEC patterns. The  third  type was recorded
from  hemispheres with marked EEC paroxysmal dis-
charge and, at times, slight atrophy of cerebrovascular
origin. Thus,  asymmetries  in CNV  shape  seem to be
related to the type of pathology and spontaneous EEC
activity of the two hemispheres.

   In  normal adults, symmetric CNVs are usually pro-
duced  by  the two  hemispheres  regardless of which
hand  performs (Weinberg and  Papakostopoulos 1975).
In children  with  asymmetric  brain  lesions  and motor
impairment,  a  marked   asymmetry   in  shape  and
amplitude was recorded. Asymmetry in shape seems to
be related  to  the spontaneous bioelectrlcal activity of
the hemispheres, while  asymmetry of amplitude seems
to be  related to the impaired hand performing.

   Central areas  seem to be Involved in the  efficient
utilization  of information  for skillful  performance
(Papakostopoulos, this volume). The  above data suggest
that processes of less impaired areas  are involved in the
functional balancing of more impaired areas in order to
enhance motor control when the more Impaired limb is
required to perform.

-------
CNV, EEC in Children with Cerebral Palsy

                81      S2
                                                                                                321
                                                                                          [EOjuV
                                      I2BMV
 C3-2M
Cz-2M
                                                C3-2M
Fig. 1. Amplitude asymmetries^ in subject 4, right hemtparests. A: Averaged responses to eight artifact-free trials using
left (unimpaired)  hand.
                              followed by  postimperative negative variation  (PINV) in all  derivations,  CNV
                        CNV  oowed  y  p
 amplitudes: C4 - 22.2 nV, C3 " 19,4 uV, Cz • 25.2 uV, Mean RT - 325 msec. B: Avenge of eight right (impaired)
 hand trials, CNV amplitudes: C4 - 24.8 uV, C3 - 18,8 uV, Cz - 34, 3 uV. Mean RT' 480 msec. Less evident PINV.
 VEOG » vertical electrooculogram  with superimposed SI and S2; 5-sec epoch, C4, C3, Cz referred to linked mastoids
 (2M). Baseline * 800 msec before SI; Negative up.
            SI
    V-EOG
               sec
                               I50]UV V-EOQ
    C4'2M
    Cz-2M
   A
 Fig, 2. Shape differences. A: Subject 5. Marked difference in CNV shape between the two hemispheres. In C4, note
 slightly sharper CNV of the healthier hemisphere and in C3, slightly smooth CNV of the more impaired hemisphere.
 The main difference seems to be in the abundance of superimposed rhythms. B: Subject 4. Smooth CNV with a slow
 ascending  limb  preceded and  followed  by  evident  smooth  poststimulus positivities,  C.  Subject  3. Irregular,
 saw-toothed, and sharp CNV with deep and rapid high voltage post-Si positivities, (Same abbreviations as Fig. 1),

-------
 SLOW POTENTIALS OF THE  BRAIN PRECEDING CUED
 AND NONCUED  MOVEMENT:  EFFECTS OF
 DEVELOPMENT  AND MENTAL RETARDATION
 R. KARRER, C. WARREN, AND R. RUTH

 Illinois Institute for Developmental  Disabilities, Chicago, IL, U.S.A.
   There  has  been  little  study  of slow  potential
(SP)-behavior relationships during normal development
or in  the mentally retarded (MR). Such information
could  provide  insight concerning cognitive and motor
development, as well as further specify the linkage
between scalp  recorded potentials and behavior. Karrer
and  Ivins (1976a,  1976c) demonstrated an inverse
relationship of post-warning signal positivity (PWSP) at
Cz and Fz to age and to the manipulation of reaction
time (RT). Preadolescent children showed larger PWSP
than adolescents, and  warning signals that resulted in
faster  RTs  were  associated with  larger PWSP.  We
speculated that the larger PWSP in young children may
be  a  reflection  of  inhibitory  processes  utilized  to
organize a motor act and facilitate a subsequent motor
response   set.   Papakostopoulos  and  Crow (1976)
hypothesized  that  a component after  the  response
signal,  the late positivity  (P300), may also be related
to inhibitory motor processes.

  Karrer  and  Ivins (1976b, 1976c) also  found  that
adolescent MRs  had more PWSP  than  age-matched
normals, but less than preadolescent children. Follow-
ing  the above interpretation,  the MR may need  to
mobilize inhibitory processes to  a greater extent than
normal age mates. On the other hand, no differences in
CNV amplitude were  found  between  these groups in a
RT  or perceptual  task. The MR, however, did not
always  reflect  the  same  relations of SP  activity  to
performance   as   did  age-matched  normals.   Most
notably, the timing and topography of SP activity were
different from  that of normals,  suggesting a different
organization   of  cortical   events   associated   with
behavior.

  Otto et al. (1976) reported SP data in normal and
learning disabled (LD) children (aphasics and dyslexics)
performing pictorial  and letter-matching  tasks.  The
results  appear  consistent  in some respects  with the
findings cited  above.  Following  a  warning  interval,
information  was  presented  at S2 and  S3,  to  which
 subjects  made  same/different  matching  responses.
 During  both the warning and encodement intervals,
 positivity at Cz  was found to be  inversely related to
 age in normal groups. Moreover,  LD children showed
 slower RT and  greater  positivity  at Cz  than normal
 children  during  warning,  encodement, and response
 intervals. Positivity  became greater  prior  to S3 (the
 point at which  letter matching and decision was re-
 quired)  and became even larger prior to the response,
 by which time matching and decision may be assumed
 to have  been completed. The slower  RT and increased
 errors in the LD group  indicated that the matching
 decision was more difficult (i.e., more uncertain), and
 took longer than for normals. Probably because of task
 requirements, the locus of SP differences  was late  in
 the epochs, instead  of early  as found by Karrer and
 Ivins (1976c).

   The present paper is an interim report of SP activity
 preceding  simple movement  during development  in
 normal and retarded children. A simple RT task with
 well-known   developmental  differences   was  used
 (normal  children and MR  adolescents are slower than
 normal preadolescents or adolescents). In addition, the
 traditional motor readiness potential, or Bereitschafts-
 potential (BP),  task  of  a simple  noncued voluntary
 movement was employed. If the young child's prepara-
 tion  to  respond  requires greater  inhibitory effort  to
 control irrelevant motor behavior, then there should be
 positive components  within the BP not apparent in the
 motorically  developed adolescent.  Since  the child  is
 slower than the  adult or  MR in  RT, reflecting less
 organized   motor   processes,   promotion  negative
 potentials were expected to begin later in the interval
 preceding movement  (Karrer and Ivins 1976c).

   Analogous  differences in positivity were predicted
 for the MR. Further, the MR should exhibit a different
organization of neural events reflected by differences
in topography  and   timing  of  potentials  preceding
movement  when compared to age-matched normals.

-------
SP Preceding Movement

 Finally, it was predicted that SP  differences between
 the  MR and  normal subjects would be  larger  when
 performing a cued response as opposed  to a voluntary
 noncued  response.  This expectation  was based  upon
 previous  work  (Berkson   1960,  Clausen 1966),  im-
 plicating response initiation as an important factor in
 slow reaction time of the mental retardate.

 Method

 Subjects

    Data were  gathered from 12 (8  in cued  task)
 children (X age = 7.4 yr), 11 preadolescents (X age =
 12.6),  13 adolescents  (X  age = 18.0),  and 11  MR
 adolescents (X age =  17.7). All were male. Mean  IQs of
 the  group were  105, 110, 118,  and 65, respectively.
 MRs had no diagnosed neurologic impairment (except
 one  Down's  Syndrome) and  were students at special
 education facilities.  All but one preadolescent and one
 adolescent were right handed.

 Procedure

    Activity preceding movement was  recorded during a
 nonwamed  simple  reaction  time (RT) task  and a
 noncued  voluntary  button  press task. In  both  con-
 ditions, subjects responded (R)  by pressing a button
 with his dominant thumb.  In the RT  task, the response
 cue  was a flash from a Grass Photostimulator (intensity
 =  8).  The flash was reflected off a  38-in.  diameter
 opaque Plexiglass  hemisphere.  The  average  intertrial
 interval was approximately 14 sec, and varied roughly
 ±  4 sec.  Subjects were instructed to  respond as fast as
 possible  to  the flash. They  sat  2 ft  in front  of the
 hemisphere   and  were  requested   to  refrain  from
 blinking  or making other eye or body movements. For
 the  noncued response condition, subjects were asked to
 make  a series of responses separated by  about  2 to 5
 sec. They were required to  fixate on a point  at the
 center of the hemisphere  while at least 50 trials were
 collected for each task in one or two sessions.

 Recording

    Beckman  nonpolarizing electrodes  were placed at
  Oz,  Cz, C3, C4,  and   Fz referred  to  linked-ear
  electrodes on the inner earlobes for  dc recording with
  a Grass (Model 7) Polygraph and Ampex (Model 1300)
  FM recorder. Electrodes above  and below the left eye
  were separately referred to linked ears for two mono-
  polar  recordings. Simultaneous deflections  of opposite
  polarity  on these two channels indicated an eye  blink
  or  movement. EMG activity was measured  from the
  responding  and contralateral  thumb and summed  by
  means of a Grass integrating ac amplifier.

  Data reduction and analysis

     The data were digitized, edited,  and  averaged on a
  PDF  11/10  computer at  4 msec/point,  with  trials
                                               323

contaminated by excessive drift, unstable baselines, or
sudden  dc  changes  excluded. Excluded  trials  were
tallied   separately  if  an  eye  movement  or  blink
occurred:  (1) during the epoch prior to or  200 msec
after R; and (2) from 200 msec after R to the end of
the epoch (600 msec after R, cued; 320 msec after R,
noncued). Trial  salvage  rate  for  cued and noncued
response was 42% and  21% for children, 49% and 36%
for preadolescents,  72% and 63% for normal adoles-
cents, and 70% and 44% for MRs, respectively.

   Averages  were  generally  time-locked  to EMG  onset
in the  responding thumb. The absence of thumb EMG
in five MRs necessitated locking to the response, and
for all groups there  were occasional trials with severe
EMG attentuation. These trials were assigned the mean
EMG onset time of the remainder of trials. Comparison
of   EMG  and  response-locked  averages  indicated
essentially  identical  waveforms. Averages  time-locked
to stimuli were also computed for the cued response
condition for comparison of stimulus-locked and  EMG-
locked activity.

   In the  cued  response  task, the mean voltage of the
200-msec  interval  before the  stimulus was  used  as
baseline. Positive and  negative voltage  (V) deviations
from baseline between stimulus  onset  and 500  msec
after the response for each trial were calculated by the
formula, SV/4N, where N  is the number of points
summed,  which gave the mean fzV/msec. In the non-
cued task, the juV/msec was  computed  from 600 msec
to EMG for each subject; the voltage was relative to a
baseline  calculated  from 800  to 600  msec prior  to
EMG.
 Results

 Noncued response

   Inspection of waveforms (Fig. 1) of the young child
 and the retardate showed considerable deviation from
 the  usual BP of the adult. There was also considerable
 variation  within groups. The  normal adolescent's pre-
 dominant waveform was similar to an  adult BP; a slow
 negative shift commenced about 500-700 msec before
 EMG onset and reached a steeper peak about 150 msec
 after the button press. The amplitudes were smaller, or
 positive,  in  Fz  and Oz. One-third of  the  adolescents
 did not show a discernible response over central areas.

   The waveform  of young children was more complex
 than the adolescent  waveform with at least  two to
 three  prominent features, depending on the recording
 site,  prior  to  the  response.  There   was an initial
 positivity (Oz,  Cz, Fz) starting 600-650 msec prior to
 EMG, followed  by a prominent negative shift (3-5 ;iV),
 and peaking 150-350 msec prior to  EMG. A  second
 prominent (Cz  and C4) positive shift (5-6 nV)  peaked

-------
 324
                                                Karrer et al.
                 AGE:
10-13
          18-18
           EMG
                          21
  28
38
26
F/£. ./.  Average activity for one subject in each group, showing representative waveforms and distributions for the
noncued task. Note lack of associated activity in the eye below lead (Eb). Positive is up.  Gain for eye above lead (Ea
and Eb) Is '4 EEC, N * number of trials in average. NML m normal.
near the onset of  EMG and  was followed by a slow
return to baseline. Some averages  appeared to reflect
low frequency activity, predominant at age 6-8 (Lairy
1975), and,  therefore,  probably  were  not time-locked
to  the response. Only a small proportion exhibited  the
adult waveform (Fig. 2).

   Preadolescents  exhibited  the  greatest  waveform
variation:  27%  exhibited  the   complex   waveform
characteristic  of  young  children; 35%  exhibited a
unique monophasic positive wave, peaking at or before
EMG onset; and  18% showed normal adult waveforms.
This  mixture reflects  the  transitional nature of  the
group  In motor development and  emergence of asso-
ciated  EEC patterns.

   The  predominant  MR waveform was  a long slow
positivity (4-10 ptV), beginning about 800 msec  prior
to  EMG  and returning  slowly  to  base  level  after
response. Sometimes, a slight negativity occurred mid-
way in the positivity,  resembling  the waveform of a
young child. As with adolescents, one-third of the MRs
failed to show a discernible response over central areas.

   Fig. 2 shows idealized representations of observed
waveforms, the  proportion  of subjects showing each,
and  a  representative  average of one subject from the
modal  group characterized by  the waveform. Waveform
classifications  were derived from visual  inspection of
Cz  recordings  of all subjects by two judges. Chi square
         analyses of  the proportion  of the four  waveforms
         (ignoring   the  rhythmic   activity  and  no-response
         subjects) were performed between groups.  Adolescents
         and MR were significantly different (x2 » 10.5, df * 3,
         p < .01) as were the three normal groups (x2 = 33.7,
         df-6,p<.001).

            Statistical  evaluation  (Mann-Whitney  U-Test) was
         performed  on the /uV/msec measure of those MRs and
         normal adolescents showing the predominant MR and
         adolescent  waveform (excluding  all other waveform
         types). Fig.  3 depicts  these  relations. These selected
         groups  differed  significantly,  indicating  that  the
         retarded had greater positivity in all leads (U-3, 0, 8,
         2,  5,  p < .05,  respectively, for Oz, Cz, C3, C4, and
         Fz).

            As  a measure of irrelevant  motor behavior, the ratio
         of the number of trials discarded as a result of  eye
         movement  divided by the number of trials completely
         free of eye-movement  contamination  was  determined
         for  each eye-movement category and individual. This
         ratio of preresponse eye  movement to  otherwise good
         trials  for children, preadolescents, and retardates was
         greater than for adolescents  (U-17, 26, 34,  p < .05,
         respectively).  Children,  preadolescents  and retardates,
         however,  did not  differ on  this ratio. EMG of  the
         nonresponding thumb did not provide a useful measure
         of Irrelevant  motor behavior because  of the low in-
         cidence of measurable activity.

-------
SP Preceding Movement
                                                       325
           JVW     •
                                                       ,16-18X
                              AGE:  6-8     10-13   NmL      Mr

                                     16      18       54        0
                                     54       27        8
                                     0       36        Q
 0       0
                                     0       18       31       36
                                     0

                                  N.12
11
 8

13
55

11
1.25 «c
 /y#. 2. Idealized Cz waveforms (lefi)< percentage of subjects showing waveform in each group (middle), and a representative
 subject's average (right) for the noncued button press, N - number oj subjects per group. Positive, up.
   Correlations of  preresponse  and postresponse  eye
movement ratios to juV/msec, prior  to the  response,
were  calculated for each  recording  site  within  each
group.  None  were  significant  for postresponse  eye
movement.  All  significant correlations were for the
preresponse eye-movement exclusion  category.  One
correlation was  found  to  be significant in the 6- to
8-year-old  group (r *  0.60, p  < .05 for C4). There
were two significant correlations in the MR (r  « 0.64,
•0.65, p < .05, for Oz, Fz) while the correlation for C4
approached significance but in the opposite direction (r
• -0.54). There were no significant correlations  for the
preadolescents or adolescents. These relations indicated
that,  for children, the greater  the ratio  of rejected
trials  to  retained  trials, the larger the positive 0ess
negativity) at  C4 on retained trials. In the MR, larger
ratios  were  associated  with  larger  positivity  (less
negativity) at  Oz,  but less positivity (more negativity)
atFz.
          Cued response

            An  ANOVA  of  RT  between  groups revealed a
          significant difference  (F 3/44-5.27, p < .003). RT was
          slower  in MR (X-417 msec^and children $-451
          msec)  than  in adolescents  (X=275  msec). Children
          (N-8)  were  also  slower  than preadolescents (X»322
          msec).

            Examples  of waveforms in each group are shown in
          Fig, 4. The  averages of activity time-locked to  EMC
          onset show different  amplitudes and waveforms across
          ages and  MR.  There  is little evidence  of the VEP in
          these   averages,  indicating  that  the  variability  of
          response is sufficient  to cancel the  VEP.  Children and
          preadolescents   exhibited   a   definite   slow-rising
          positivity, peaking at or after EMG  onset, although RT
          ranged  from  250  to 750  msec.  This positivity was
          often  preceded by a negative-going component  (not

-------
326
      •H).86T-
                  .• •.
                                    A
       0.00
      -0.66-L

                                              MR
                                             16 - 18
                                C3   Cz
Fig. 3.  Mean nv/msec SP value for those subjects exhibiting
modal waveforms for normal and retarded adolescents
across recording sites for noncued task. Pre-R refers to
measurement epoch.

seen in Oz) that was larger in children than the other
groups. This  negative-positive waveform is somewhat
similar to that  found in the  children's noncued wave-
form.  In general,  the  MR had waveforms similar to
normal  adolescents, who  had  only small positives
associated with EMG onset.

   A  repeated-measures ANOVA (Groups and Leads)
of the juV/msec measure indicated that  a leads effect
                                                       and  groups x  leads  interaction were significant  (F
                                                       4/152=26.4; F  12/152=3.8, p < .001, respectively. The
                                                       predominant  polarity  exhibited  by  the  child  was
                                                       significantly more positive than  that of the adolescent
                                                       or MR at Oz and Cz (U=6, 5, Oz; 3, 6, Cz).

                                                         Significant   correlations   between   SP  and   RT
                                                       measures were found in adolescents at C4 (r=0.70), in
                                                       preadolescents at Fz (1=0.66), and in MR at Oz and Fz
                                                       (r=0.86, -0.85, respectively). There  were no significant
                                                       correlations in  children, but  all were the  same sign as
                                                       in the MR and approached significance in Oz and  Fz.
                                                       Further, correlations of RT  across tasks  to  SP in  the
                                                       noncued  task  indicated no significant correlations in
                                                       the  normal groups.  For the retarded, however, activity
                                                       in Cz, C4 and  Fz was related to RT (r=0.65,  0.61,
                                                       0.65, p <  .05, respectively).

                                                         Comparisons of the ratio of rejected eye-movement
                                                       trials to good trials indicated that  children,  preadoles-
                                                       cents, and MR  had greater ratios than adolescents, but
                                                       only preadolescents were significantly greater (U=20, p
                                                       < .02). Groups  were again compared  on correlation of
                                                       these ratios of rejected  eye-movement  trials (as an
                                                       index of irrelevant  motor activity) to SP/zV/msec. As
                                                       in the noncued task,  there  was, again,  a significant
                                                       correlation in children for C4 (r=0.62). There were no
                                                       other significant correlations.

                                                         In contrast to the  average time-locked to EMG, the
                                                       average time-locked to the stimulus showed large VEP
           AGE:  6 • 8
                                          10 • 13
          16- 18
NmL   '        ^   MR
     EMG
Fig. 4. A verage activity of one subject in each group showing representative pairs of waveforms from each recording site for
the cited task. Upper waveform of each pair is stimulus time-locked; lower waveform of pair is EMC time-locked. Positive, up.
            Upper waveform ofeachpa
Vertical lines indicate stimulus and EMG events.

-------
SP Preceding Movement

components  (Fig.  4).  This  finding  indicates  that
locking to  the  EMG (or response) maximizes com-
ponents  associated  with  movement and  minimizes
stimulus events. There was a lateral difference in the
size  of the  positivity  found  by response-locked vs.
stimulus-locked  averages  that  changed with  age and
MR. Table 1 shows the difference between  C4 and C3
in maximum  peak negativity to peak positivity asso-
ciated with  EMG. Children had maximum positivity at
C3 when activity was EMG-locked  but  at C4 when
stimulus-locked.  Preadolescents,  MR,   and   normal
adolescents  had  greater  positivity   at C4  than  C3,
regardless of time-locking.

       Table 1. Difference in Peak Negativity to
                   Peak Positivity

                    (C4 - C3 in /uV)

STM-I
Rip-L
Normal children Retardate*
6-8
3.0
-1.4
10-13
1.9
2.2
16-18
3.8
1.7
16-18
1.4
1.6
Discussion

   The  main features  of adolescent movement-related
potentials  were  consistent  with  adult  waveforms
(Gilden  et  al.  1966;  Deecke  et al.  1969,  1973;
Vaughan  et  al.  1968).  There  was  a  clear,  slowly
increasing negativity (Ml), beginning about  800 msec
prior to EMG.  Peak negativity  (N2) usually occurred
well after EMG onset.  No obvious p re motion positivity
(PMP or PI) was apparent, although there were small
positive  components  (noise?)  in some  averages.  Al-
though  there was a postresponse  positive-going com-
ponent  (P2)  that terminated the waveform around the
baseline, it did not consistently overshoot baseline. It
is probable that the lack of a well-defined SP in a third
of  the  adolescents was  a function of the  reduced
signal-to-noise ratio obtained with the reduced number
of trials employed. There is considerable variability in
waveform within, as well as between, age groups.

   The  literature  indicates  that  there  is  considerable
variability in the presence of PMP,  timing of N2, and
amplitude of P2.  P2 varies in size  and  slope and may
be dependent upon the nature of the response (Deecke
et al.   1969,  Fig. 6)  and  the  time constant of  the
recording  system. Variability  seems  to  have  been
ignored  in  favor of  "the" BP waveform (compare
waveforms and timing  of components in  Deecke et al.
1969, Fig. 2, 5; Gerbrandt et al. 1973,  Fig. 4; Haze-
mann  et al.,  this volume,  Fig.  1;  McCaUum, this
volume, Fig. 2; Papakostopoulos, this volume, Fig. 1;
Vaughan et al.  1968, Fig. 2, 3).
                                               327

   The  presence of  positivity in children's waveforms,
the  similarity  of  waveforms in both  tasks,  and  the
existence   of  significant  positivity-irrelevant  motor
activity relationships for both tasks confirms our initial
hypothesis (Karrer and  Ivins 1976a, 1976c).  Irrelevant
motor activity in  the child reflects  the lack of motor
differentiation.  Positivity  seems  to accompany  suc-
cessful inhibition of such activity. The  more the child
exhibits eye  movement when instructed not  to do so,
the greater inhibition (positivity) he must develop to
control eye movement to produce a  "good trial." More
inhibition  of  contralateral  than  ipsilateral-irrelevant
movements  may be required  since correlations were
significant  only for  C4. Right-hand response  required
inhibition of response  on the left  side.  Contralateral
inhibition of "motor overflow" develops slowly and is
often  undeveloped  in  adolescents  for various  hand
movements  (Abercrombie et al.  1964, Cohen et  al.
1967, Stern et al. 1976).

   Some portion of  SP positivity in  children may serve
the same function as the PMP described by Deecke et
al.  (1973). A requirement of greater  activity in  the
parietal and  other association areas for initiation  of a
coordinated button  press could account for the larger
positive potentials observed in children  than in  ado-
lescents. This waveform may have phylogenetic, as well
as  ontogenetic, significance. Donchin  et al. (1971)
observed in monkeys a transcortical SP complex, pre-
ceding a noncued lever  press (cf.  their  Fig. 10), which
bears a striking resemblance to the modal waveform
observed in children. The second positive component
observed by  Donchin  et al. was more prominent in
central  than  in frontal  areas. In children, this com-
ponent  was more noticeable at  Cz and  C4 than at Fz
and  was  more pronounced than  that  observed  by
Donchin et al.

   The  predominantly   positive  waveform   of   the
retarded in the noncued task is even  more divergent
than  that  of the  child  from the normal  adult wave-
form.  Lelord et al. (1976) and  Laffont  et  al. (this
section)  also  report  positivity  in  young,   severely
retarded children  during sensory conditioning. These
prominent  positive waveforms,  therefore,  may reflect
immaturity  or  impairment  of sensorimotor  develop-
ment,  although it  is  not  clear what sensory con-
ditioning tasks have  in  common with self-paced motor
tasks.  Deecke  et  al. (this  section) report  a  similar
positive-going BP  in elderly subjects.  These  authors
suggest  that  this waveform may be  due to the pickup
at the linked-ear reference of activity arising from basal
temporal  cortex.  We have  observed that, in normal
children, the  ear is  essentially quiet when referred to
nose-tip or midchest. Therefore, it is unlikely  that the
positivity in  the child is due to an active ear reference.
Although we have not directly assessed this possibility
in retardates, the  monopolar infraorbital lead may be
considered  a  control for  ear  activity.  This lead  was

-------
 328
                                        Karrer et al.
 essentially  flat, indicating an indifferent reference. It is
 obvious that inhibitory processes express themselves via
 different waveforms and  topography as a function of
 age, retardation, and performance.

    An intriguing possibility is that positivity in the MR
 may  reflect subtle abnormalities  in  pyramidal motor
 neurons.  For  instance,  differentiation  into mature
 dendritic spines is arrested in  the retarded  individual
 (Huttenlocher  1974;  Purpura  1974,  1975).  Primitive
 spines are  retained, and there  is a severe reduction in
 the total  number  of spines. The retarded child  may
 also  have   a  significantly  greater  number  of total
 cortical  synapses (Cragg 1975). These  facts suggest a
 different, less efficient  synaptic geometry in retarded
 children. Since negative SP activity  is considered  to
 reflect,  in  part,  activity  in  apical  dendritic  fields
 (McSherry  1973), the  atypical positive BP of the MR
 may  reflect the altered  spatial  distribution of synapses
 in central  motor areas.  If  these  speculations  are
 correct,  the BP could yield information about cortical
 synaptogenesis  in  the   retarded  and  the   normally
 developing  child.

    The  data confirm  our predictions  of a  different
 neurobehavioral organization in MR; i.e., topography
 and  SP  relations  to  behavior were  different. MRs
 apparently  do not generate contralateral inhibition:  the
 correlation  between eye movement and  SPs was not
 significant  and was negative   in  C4.  In  contrast  to
 age-matched normals, MRs generated more positivity in
 posterior regions and less  positivity in anterior regions
 when inhibiting eye movement. This  more diffuse and
 topographically different pattern of SP activity, accom-
 panying  inhibition  of  irrelevant movement  was asso-
 ciated with poor performance. Compared to normals,
 MRs  exhibited higher correlations of RT to posterior
 and anterior SP activity in the cued task; greater Oz
 positivity and  greater  Fz negativity accompanied slow
 RT.  Poor  performance  in the  cued  task  was  also
 related  to a different diffuse pattern of  positivity at
 central, right,  and anterior areas in the noncued task;
 dower subjects  had greater positivity. The relations of
 SPs to  behavior are obviously  different from normal
 groups (although  the trends in  children were  often in
 the same direction).

   The  presence  in the  cued  task  of  positives in
 conjunction  with   response  onset  may  be similar  to
 transcortical positivities  found by  Donchin et  al.
 (1971) to  accompany  cued lever  press.  The central*
 post central   area   was   also  found   to  elicit  larger
positivities  than the frontal area.  The different lateral
 topography  of  positivity with  age and event-locking
 may reflect  the role of the nondominant hemisphere in
sensory  spatial appraisal of the flash  (larger positivity
in C4 for  stimulus-locked  averages)  and  the role  of
 motor components in EMG-locked  averages (positive in
C3 for children but more negative in C3 for  normal
 adolescents).  This lateral relationship was  essentially
 the  same  for preadolescent normals  and  adolescent
 MR. These  data  again indicate the value  of comparing
 forward and  backward averages (Karrer  and  Ivins
 1976c).

   It was assumed that the retardate would be more
 similar  to normal in the noncued task, where there was
 no requirement to initiate a movement quickly. In the
 cued task, by contrast, a response had to be quickly
 initiated upon demand,  and differences were predicted
 between SP activity of normal and MR adolescents.
 Counter to these expectations, the MR differed most
 from the normal adolescent in the noncued task, but
 were similar in the cued task.

   Contrary  to  previous  findings  (Karrer  and  Ivins
 1976c), the amplitude  of positivity correlated directly
 with RT in  the cued  task. This difference  may  be
 explained if one  assumes that  the process of actively
 inhibiting  irrelevant  movement  is  incompatible  with
 rapid   response.   A  preparatory   interval,  in which
 subjects  could   inhibit irrelevant  movement  after
 receiving  the  warning  signal,  was employed  in  the
 earlier study.  In  the present nonwarned task, both the
 inhibition of  irrelevant movement and the preparation
 for response had to  occur  after the  imperative signal.
 Hence,  the  greater the  effort needed to  inhibit irrele-
 vant movement (larger positivies), the slower  the RT.

   Our  results and those of Otto et al. (1976) indicate
 that differences in level of  functioning affect SP wave-
 form and topography. The  range  and type of idiosyn-
 cratic or idiopathic variables that influence SP activity,
 however,  is   poorly   understood.  The   moderately
 retarded, who  exhibit  great   variability  in behavior
 (Baumeister 1968), must utilize a variety of combina-
 tions of underlying  processes, which  are reflected,
 presumably,  in  great  variation in  the  polarity  and
 topography of slow potentials from task to task.

   The  large positive shifts  observed in young children,
 elderly  subjects,  retardates, and  other  primates  may
 reflect  a common underlying inhibitory phenomenon.
 In   primates,  children, and  MR,  the   inhibitory
 phenomenon could be associated with the phylogenesis
 and  ontogenesis of motor control mechanisms.  In older
 adults   and  the  mentally   retarded,  the  inhibitory
 phenomenon  could be  associated with  declining  or
 deficient motor  control. The  relationship  of positive
 SPs  and inhibitory processes is further  discussed  by
 Marczynski  in   the  Electrogenesis  section  and   by
 Papakostopoulos  in the  Motor  Control section of this
volume. Further  research is needed  to  elaborate  and
validate these  hypotheses.

 Acknowledgments

   We  wish to express  our appreciation  to Anthony
 Morrelli and  Dr.  Robert Baldauf of the Oak Park,

-------
SP Preceding Movement                                                                             329

Illinois,  schools,  Ronald  Sterrett  of  Samuel  Kirk    Armstrong-Scharf  for  assistance  in  data  collection,
Center, Palatine, Illinois, and Donald Zingrabe of West    organization,  and analysis. This work was supported by
Leyden  Township  schools,  all  of  whom  facilitated    the Illinois Department of Mental  Health and Develop-
access  to children. We are also grateful to Joan Gillet,    mental Disabilities and NICHD  Grant No. HD 08265
Dick  Mills,  Jim  Griffith, John  Helton, and  Cynthia    to the senior  author.

-------
AGE-DEPENDENCE OF THE BEREITSCHAFTS-
POTENTIAL1


L. DEECKE, H.-G. ENGLITZ, AND G. SCHMITT

Department of Neurology, Ulm University, Ulm, West Germany
   The slow negative shift of the cortical dc potential
that occurs in the  foreperiod of voluntary, i.e., self-
paced, movement (Bereitschaftspotential, BP, or readi-
ness potential, Kornhuber and Deecke  1964, 1965) has
been  studied  mainly  in young adult subjects. CNV
research  has  been conducted  on an  older age group
(Loveless  and Sanford  1974), but no such investiga-
tions  have been reported concerning the BP  develop-
ment  in  these groups.  The present  study was con-
ducted  in  order  to obtain a normal  control group,
similar in age  distribution to a group  of patients with
Parkinson's disease in  which the  BP was nearly absent
(Deecke et al.  1977).

Methods

   Subjects were neurological clinic patients with non-
cerebral  diseases, e.g.  slipped  discs.  All  were right-
handed. Five  age  groups  were selected: I derived from
34  previous   experiments  (17-29 years);  II  (30-39
years); III (40-49  years);  IV (50-59  years); and V
(60-69 years). Groups  II-V each consisted of six sub-
jects  (usually  three females  and three males), three
performing right-sided  and  three left-sided voluntary
index finger flections by pulling the trigger of a pistol.
Movement  onset  was  defined as the  very  first EMG
activity  in the agonist muscle (M. flexor digitorum
communis,  pars  indicis).   EMG, recorded  through
bipolar surface electrodes, was rectified and averaged.
The index  finger  was rapidly moved, held in position,
and returned  only  after analysis time. Four  to five
hundred finger movements at irregular intervals of 4-5
sec  were  averaged per  experiment. Subjects  fixed their
gaze on  a given  spot  during  trials and  avoided eye
blinking.  In addition, oculomotor  and other  artifacts
were  eliminated  by a  preaverage editing  procedure.
Beckman Ag/AgC I electrodes were affixed to locations
C'3, C'4  (defined in Deecke et  al. 1969), Cz, P3, P4,
and Pz (10/20 system) and  referred to linked earlobes.
MonopoJar recordings   and  two bipolar  derivations
'Supported by the Deutsche Forschungsgemelnschaft, SFB 70.
(C'3-C'4 and C'3-Cz) were stored on tape and reverse-
averaged off-line. BP amplitude  was calculated  as the
arithmetic mean of the six monopolar leads (five leads
in group I because there was no vertex recording). Two
points of measurement were selected: BPo = amplitude
(with respect  to  a  pre-potential base line) at the first
EMG activity, and BP150 = amplitude  150 msec prior
to first EMG  activity.  Statistical analysis included  an
analysis of variance  and correlation tests.

Results

   BP amplitude  was  found to  be  relatively constant
until  the end of the  fourth decade (39 years) after
which the  amplitude gradually declines.  In Fig.   1,
typical  examples are  given of  the  cerebral potentials
preceding voluntary finger movement.  In  a group  II
subject  (left), aged 31  years, all three potentials are
clearly discernible (right-sided movement):  (1) the BP
in all monopolar recordings,  with a maximum  at the
vertex; (2) the promotion positivity (PMP) in  all  mono-
polar recordings  with a  maximum  midparietally; and
(3) the  motor potential  (MP) in the bipolar recordings
C'3  versus  C'4  and C'3 versus Cz. In the C'3-C 4
derivation, a slight contralateral preponderance starts
about 300  msec  prior to the onset of EMG activity
(0).  In  a group  III subject  (right), the BP recorded
under  similar  conditions  was  smaller.  Its  amplitude
further  decreased  with  older age  groups. Above 60
years  positive  BPs occurred occasionally in all mono-
polar  recordings (Fig. 2). In  younger adults, positive
BPs are  found only in frontal leads.

   In  Fig.  3,  the  mean  BP150 (hatched)  and BPo
(white)  amplitudes  of the  different  age  groups are
shown. A gradual decline of the two BP amplitudes  is
seen.  The analysis of variance of BP150 for  groups  II
to V was marginally  significant (F=2.9, p=.06). The
analysis of variance of BPo was significant  (F  = 5.1,
p < .01). The slightly larger BPo amplitude in group II,
as compared to  I,  can be explained by the missing
vertex recording in group I.

-------
 Age Dependence of BP
                                                                                                      331
                                                                               43 yr       N-ABO
                                                               c,
                                    0
                                                                             -1
                                        sec
Fig.  1.  Typical  examples of movement-related potentials at different ages. Cortical activity preceding right-sided
voluntary finger movement in  a 31-year-old subject (left) and in a 43-year-old subject (right). The vertical line
represents EMG onset.  In the left column, three potentials are clearly discernible: (I) the slow negativity oftheBP,
starting 1 sec before EMG onset and maximum at the vertex; (2) premotion positivity, starting about 80 msec prior to
EMG onset, maximum midparietally; and (3) the motor potential as additional negativity over the motor cortex about
60 msec prior to EMG onset, typically seen in the bipolar recordings C3 vs. C'4 and C'3 vs. Cz.  In the 43-year-old
subject at right, the BP is markedly smaller. N, number of trials.
       significant   negative   correlation  was   found    was not significant (coefficient of determination r2
   A   significant   negative  correlation   was
between BP amplitude (averaged from all six mono-
polar  leads  in each subject) and  age. The correlation
coefficient  (r) for BP1SO  was 0.58  (2p  < .01);  the
correlation coefficient for  BPo was 0.66 (2p < .001).2
Correlations, like  analysis  of variance, were computed
only on the data  from groups  II to V ranging from 30
to 69 years  of age.

   Fig.  4  shows  the regression lines  at the  different
electrode  locations, as  indicated for  BPo  (solid lines)
on the  left, and for BP150 (dotted lines) on the right.
All correlations were significant  on   the basis of the
two-tailed   hypothesis,  except   for  BP150   at  P3
matching only the one-tailed test.

   Unlike  the  BP, the  regression  line  for   the  MP
(estimated as the  difference BPo • BP1SO  in a bipolar
recording, C'3 versus C'4, cf. Deecke et al.  1969, 1976)
 'Two-tailed test
0.066). This finding indicated that the motor potential,
which  reflects the  activity in the  contralateral motor
cortex immediately prior to the  onset of EMG activity,
is relatively constant with age.

Discussion

   A  gradual decline  with  age  of  the  average  BP
amplitude  after  the  fourth decade  of life  is  docu-
mented here. The reduction  of  BP as a result of aging
is probably a true diminution of this cortical potential.
It  seems unlikely that the reduction is solely caused by
an  increase in skull  thickness  or in the  amount of
cerebral  spinal  fluid  between the cortical surface  and
the skull. Such factors would affect other potentials as
well,  and  the  amplitude  of the  alpha rhythm,  for
instance, is  not markedly reduced with  age. Also, the
fact that the motor potential  was  found not  to be
significantly affected  by aging (Fig. 4) is not consistent
with  the assumption of  a  general  reduction  of all
cortical  potentials with  age.  Late  components  of

-------
 332
                                                        Deecke et al.
                         65 yr
n-426
   C4-C3
   C4-C,


   P4


   '3
  LEFT INDEX
  FINGER
                                                 lOjuV
                     -0.5
    +0.5  sec
 Fig. 2. Positive BP in a 65-year-old subject. C'4, C'3, Cz,
 P4, P3,  and  Pz  electrodes  were  referred to  linked
 earlobes.  The subject performed 425 left-sided finger
 movements, but positivity occurred also with right-sided
 movements.
CJ
LU

5
4
2
1


I
BP1
50

B
f 1
Po


—






1(1
17-29 30-39 4049 50-59
1 II III IV


r-|2SE
60-69 yr
V
 Fig. 3. Mean BP amplitudes in the different age groups.
 Blank columns: BP amplitude at movement onset (first
 EMG  activity); hatched columns: BP amplitude  150
 msec prior to EMG  onset. Group I (17-29 years) con-
 sisted of 34 previous  experiments (therefore  smaller
 standard errors, SE,  than in groups II to V). Groups II
 to V consisted of six  subjects each.
average evoked  responses  and  the  CNV have  been
reported by some  authors  to be reduced with  age,
while others found little or no reduction in amplitude,
if a certain  level of motivation was maintained (Marsh
1975,  LUders 1970,  Loveless and Sanford 1974). The
factors responsible  for the age-dependent BP reduction
are   not yet  known,  but  among  others  a  general
decrease in motivation  as  well  as the rarefaction  of
cortical  elements  have to  be  considered.  In senile
dementia, the  late  components of  the  visual evoked
response are even increased (Visser  et al. 1976).

   The  observation  of occasional  positive BPs in the
older age  group (60-69 years) is  difficult to explain.  In
younger adults, positive  BPs  only  occur  over the
frontal  region referred to either linked ears or  mastoids
(Deecke  et al.  1969).  In  patients  with  bilateral
Parkinsonism, an all-positive BP is often found (Deecke
et al. 1977). The present results show that it can also
occur  in healthy elderly subjects, although the signifi-
cance  of  this finding is not clear. Karrer et  al. (this
section and  personal communication) reported positive
BPs in mentally retarded children. Since mastoid or ear
references have  been used  in BP studies of the elderly
and the retarded, observed positivity could result either
from absolute   positivity  over  the  convexity of the
skull,  or  from  greater  negativity  at  the  reference
electrodes,  which  reflect  activity  from  temporobasal
cortical structures.
                                                                10
                  Fig.  4.  Regression lines at the different recording loca-
                  tions.   Solid regression  lines for BPo  with electrode
                  locations  indicated on the left; dotted  regression lines
                  for BP150  with  electrode  position  indicated  on  the
                  right. For simplicity, subscript 3 stands for contralateral
                  electrode  positions, subscript 4 for ipsilateral locations
                  for left-sided movements as well as for right-sided move-
                  ments.  Double regression lines for the motor potential
                  (MP), measured as the difference BPo - BP150 in a bi-
                  polar  recording  contralateral  versus   ipsilateral pre-
                  centralfcf. Deecke et al. 1969,1976).

-------
AGING EFFECTS  ON THE  HUMAN  EVOKED POTENTIAL1
G.R. MARSH

Center tor the Study of Aging and Human  Development and Department of Psychiatry, Duke
University  Medical Center,  Durham, NC, U.S.A.
   In  the quest for understanding  how  aging affects
brain  mechanisms, the  averaged evoked potential (EP)
has been an effective tool. The reverse may also prove
to be  the  case—that  aging  effects  may  prove  to be
useful  tools in  understanding mechanisms  underlying
the generation of EPs. The following review  is intended
to show how  EPs have  been used  in  studying the
effects of aging and to note how changes seen in EPs
may  be related to  changes in physiology and anatomy.

Passive stimulation

   Contrasting  EPs  from  young, mature  subjects with
EPs  from  older  subjects  lias shown  the early  com-
ponents (20-80 msec latency) to be larger in the  older
group.  Shagass  and  Schwartz   (1965a)   measured
somatosensory  (S)  EPs  from  electrodes  placed  on  a
parasagittal  line 7  cm left  of the midline with the
"active"  electrode  2 cm posterior to a line from vertex
to the auditory meatus and  a. "reference" electrode 6
cm anterior. Peaks  falling at about 20-,  40-, 50- and
80-msec  latencies were smallest in amplitude  in a 20-
to 39-year-old  age group and  were increasingly larger in
two groups, 40-59  and 60-80 years of age. The latency
of the peaks  at 20 and  40  msec  was  found  to be
longer in the older  groups. One set of positive-negative
peaks  at about 30 msec  was  omitted  from analysis
since  they  were not found  in  the oldest two groups.
Peaks falling at longer latencies were not explored.

   Lu'ders (1970) recorded SEPs from  the same sites as
Shagass  and  Schwartz (1965a), but  against an  ipsi-
lateral  ear  reference.  He  measured SEPs in four dif-
ferent  age  groups  (X  =  23.6, 36.4,  51.6, 62.0) and
found  somewhat  similar  changes  in  amplitude, but
opposite  changes in latency.  A peak falling at about 25
msec increased in amplitude, and  a  succeeding peak at
about  30 msec decreased in latency across  age groups
(45-70).  Peaks  falling at  about 50, 70, and 110 msec
'This work was supported in part by training grant AG 00017
 and  research grant AG 00364  from the National Institute of
 Aging.
increased  in  both  amplitude  and  latency  with age
(again, past 45 years). Later peaks falling at approxi-
mately 250 and 400  msec decreased  in amplitude as
early  as   30  years  of  age and  remained  relatively
unchanged with  age  thereafter. These later peaks also
showed no changes in latency across age groups.

   Schenkenberg (1970) was unable to replicate either
the increased  amplitude  or longer  latency  of early
components when  recording from C3 and C4 to linked
earlobes.  Instead, a marked decrease in amplitude was
found for some late  components (100-200 msec) after
the teenage years and a  rather  remarkable stability was
found for the early components. The latencies to peaks
were found to be increased for some  of the early and
all of the late  components. Since many  of the early
SEP  components  are only  seen  contralateral to the
stimulated limb, the  above measures pertain only to
contralateral  measures.  However,  those  components
that  could be measured on the ipsilateral hemisphere
showed similar changes.

   Shagass (1968,  1972) has also presented data on EP
"recovery."  In  this  paradigm,  an  initial stimulus  is
given, and then a second identical stimulus is presented
after  some short,  fixed delay.  The EP to this  double
stimulation is obtained, and then usually the  EPs to
single stimuli  are  subtracted so mat a "residual" EP
elicited by the  delayed stimulus is obtained. Such EPs
show greater  amplitude  with shorter delays  for older
subjects  at peaks  falling at 20, 30, 40 and  80 msec.
The  latency  of these peaks is delayed, however, and
does not recover at as short an interstimulus interval in
the older subject as  it  does in the young. This delay
was  found at peaks with  the  following latencies: 32,
35, 40, 45, 50, 80, and  110 msec.

   There   has  been  only  one  investigation   of  aging
effects on the auditory  (A) EP. Schenkenberg (1970),
recording  at  C3  and C4, found  no changes  in the
components  of the AEP  from  the  late teenage years
through old age.

-------
  334

    More studies have been carried out on the visual (V)
 EP. One early study (Shagass et al.  1965) reported no
 age differences during the first 200 msec from an Oz
 •Cz  derivation.  The age  range, however,  was  from
 19-45.  Studies extending  further  into  old  age with
 careful  attention  to  sex  differences have  shown a
 number  of age-related differences.  Straumanis et al.
 (1969)  found  that a group of subjects with an average
 age of about 70 had greater amplitude components at
 30-,  40-, and 90-msec latencies and were longer in
 latency  on the first  six of eight components measured
 when compared  to a  younger  group  (1945  years,
 average  24  years).  These  EPs  were  obtained  from a
 midline  site approximately at Oz and referenced to a
 midline  electrode  12 cm anterior.

    Kooi and Bagchi (1964) recorded EPs from midline
 occipital, parietal, and central locations against  a linked
 ear reference and  found similar  data.  The age range of
 their population  (N  =  130) extended from 28  to  72
 years, with only six  females in  the group. They found
 a peak-to-peak measurement of components in  the 80-
 to 90-msec  latency  range  to correlate  at a low (r =
 0.21) but  significant  level (p<.05) with age. The
 latency  of a wave occurring with a latency of about 70
 msec was also positively correlated with age at a  low
 (r  = 0.26) but significant level (p < .05). They could,
 however, show no relationship of a later  wave (latency
 about 140 msec)  to age.  They  also reported no  re-
 lationship between  pupil   size  and  EP  amplitude  or
 latency.
   Dustman and  Beck (1966,  1969) using subjects over
a  wide age  range, reported similarly that  early com-
ponents (before  100  msec) of the VEP recorded from
01  and 02 (to linked earlobes) increased in amplitude
from age  20 to 30 years.  Peaks occurring  later, espe-
cially after 200 msec, decreased markedly in amplitude
after the teenage  years. The same pattern was reported
by Schenkenberg (1970), who found a]] peaks increased
in latency starting in  middle age. Some acceleration of
the amplitude  loss was seen in old  age for the  latest
peaks, about 300 msec.  Similar trends were observed
for frontal (F3 and F4) and central (C3 and C4) record-
ing sites, with one exception.   Central sites showed an
increase in  amplitude of  a component  occurring  at
about 100 msec.
   Schenkenberg   also   used   a  cumulative  voltage
measure  to  analyze  the amount  of deviation  from
baseline  over  set  intervals  (e.g.,  0-100,   101-200,
202-300, and 302-600 msec).  These values showed the
same trend for all modalities and all recording sites-an
increase  in  voltage  deviation  in  the post-300-msec
range,  starting  at  about  age 40. Some increases were
noted  in earlier  components, usually in the primary
region  for that  modality and in the  frontal region.
                                              Marsh

    A notable  exception was the VEP recorded from 01
  and 02 where cumulative voltage decreased  sharply
  after age  40. In  this case,  voltage  diminished  over
  primary visual  areas, but increased over the central
  region  in  older  subjects.  The   cumulative  voltage
  measure in the post-primary  time period may, there-
  fore, reflect  alpha-like  after-discharge  in  the visual
  modality.

    Changes in cumulative voltage,  in general,  paralleled
 changes in EP components occurring within the same
 time  intervals. In  the late  (post-300-msec)  interval,
 there often were  no  specific EP  components  against
 which  to  compare  this activity. The extent  to which
 differences between  cumulative voltage and  peak-to-
 peak measures are indicative of brain aging remains to
 be determined.
    Buchsbaum  et al.  (1974b)  also reported a decrease
 in VEP for peaks falling between 100 and 200 msec in
 male  subjects. These  VEPs were   recorded  from  the
 vertex rather than occipital area. Notably, they found
 that older subjects had considerably greater stability in
 VEP amplitude as  the stimulus varied over four  dif-
 ferent   intensities.   Conversely,   these   investigators
 observed less stability  over replications in the auditory
 EP of subjects over 40-especially in the longer  latency
 peaks (beyond 256 msec).

    A light  source with intensity modulated by a con-
 tinuous  sine wave function can be used to elicit a  sine
 wave-like  EP.   Age-related  changes  in  the  phase lag
 between  stimulus  and  EP  peaks  have  been studied
 (Perry  and Childers  1969).  In  the stimulus frequency
 range of 12-30 Hz,  the average subject under 45 years
 of age demonstrates a  latency  of 77 msec, while older
 subjects have an average latency of 102 msec.

    Another variable  that needs  to  be considered is the
 subject's sex. Most reports of aging effects on EPs have
 dealt  almost  exclusively with  male subjects. Shagass
 and  Schwartz  (1965  a,b)  and Schenkenberg (1970),
 however, have  shown  that  older  women have higher
 amplitudes  and shorter latencies than men in most EP
 components for  all modalities. These effects could
 result from hormonal differences,  although this  ex-
 planation seems less likely as females exhibit the same
 or even  larger  differences after menopause. Buchsbaum
 et  al. (1974b) have  reported higher amplitudes  and
 shorter  latencies  in  visual  EPs   of children  before
 puberty,  although  Schenkenberg  (1970)  found  the
 reverse:  larger  VEPs  in males  than  females  before
 adolescence. Evidence  for any  mechanism other than
 CNS  responsivity to gonadal hormones is lacking at the
 present time.

Cognitive influences: CNV and the late
components of the EP

   Despite  the   diminishing amplitude with  increased
age reported  for  late  EP  components under passive

-------
Age Effects on EP
                                               335
stimulation conditions,  the same loss of amplitude is
not as marked when the subject must deal cognitively
with stimuli. Marsh and Thompson (1972) found that
late  components of auditory  EP of young and  old
groups  did not  differ  when  subjects  were judging
stimuli  for slight  differences in pitch.  Schenkenberg
(1970), however,  has   reported that AEPs were  less
sensitive than VEPs to age differences.

   Marsh  (1975) has  demonstrated VEP  differences
between young and old. The task  required searching a
mental  list to ascertain if a low-intensity  illuminated
digit, presented at the fixation point, matched  any of
the mentally stored digits.  The average amplitude of
the late positive component of the older group  was
smaller  than  that of the  young group  at Cz, Pz,  and
C6. No difference was  found at Fz or  C5. No  differ-
ence  in latency for this late  component  was  found
between groups.

   A  study of  CNV amplitude in  a  task in which the
pitch of tone pips was  being judged showed no  differ-
ence  between  the old and the  young (Marsh  and
Thompson 1973). Loveless  and Sanford (1974), in a
study employing long  intervals  (up  to  15  sec),  found
the elderly lower in CNV amplitude when performing
in the  usual  fixed  foreperiod  RT tasks.  The older
group, however, did not  differ  from the young when
irregular foreperiods were used. The authors noted  that
the waveform appeared to have two components, an
early  negative "orientation" wave  at about 1  sec and a
later negative CNV response shortly before  the impera-
tive stimulus. The young group, in  particular, showed a
sudden  increase in  CNV just  before  the  imperative
stimulus. Such activity might have  resulted from  motor
readiness potential activity. Since  reaction times were
longer in  the  older  group, especially  at the  longer
intervals, readiness potentials would not have been as
likely to  appear in  the  old group  during the  "CNV
interval."  Deecke  et al.  (this  section), however, re-
ported decreasing Bereitschaftspotential  with increasing
age past 40 and almost no potentials in subjects over
60 years,  findings which could also  account  for lower
amplitudes at long foreperiods as observed by Loveless
and  Sanford  (1974).  Moreover,  the  latter authors
reported that,  about 1  sec  after the  warning  signal,
"orientation" potentials appeared which were larger in
the younger group. Further  attention should be given
to this  phenomenon since other work  has often con-
fused this  earlier waveform with the final level  of the
CNV.

   Thompson and Nowlin (1973) found  age differences
in RT, but only marginal differences  in CNV amplitude
between an old and young  group.  The young group
demonstrated  increased  CNV  amplitude and  slowed
heart rate  on fast RT  trials. The  older group did not
demonstrate  this concordance  of response.  Froehling
(1974)  was able to show a correlation among CNV
amplitude, heart  rate, and  reaction time in an older
group of males. Her subjects, however, were especially
chosen  to be in  excellent health and were  also well
practiced  on the task (for several days).  By far the
strongest relationship was between CNV amplitude and
RT,  with  heart rate and heart rate x CNV amplitude
being weak predictors.

Anatomical and physiological correlates

   That the body undergoes change with  increasing age
has never been  disputed. The difficulty has  been  to
ascertain  those   changes  resulting  from  disease  or
accident not directly  applicable to the study of aging.
The  loss of neurons in certain parts of the neocortex
with  age  was shown many years ago (Brody 1955).
This  has  been more  finely  detailed in recent years
(Brody  1973) to  show the greatest loss in layers 2 and
4. Since  these layers have  neurons with  inhibitory
effects  on underlying neurons,  this finding  could  be
interpreted  to support the observed increase  in  early
EP components.  Furthermore, Scheibel  et  al. (1975)
have demonstrated a marked decrease in the number of
dendrites  and dendritic spines with age, evidence that
may be related to  the  reduction of amplitude in late
EP components.

   CNS reactivity is also controlled to some extent  by
hormonal  influences. Changes  with  age in  how the
body handles major  hormones  are now  being un-
covered. Such alterations could produce  positive  feed-
back, triggering cascading effects  throughout  the body
(Finch 1973). Ten  years ago theories of aging changes
in the CNS emphasized  ischemic  or anoxic effects of
reduced blood flow. Such effects,  however, can  no
longer be  seen as the sole mechanism behind changes
in the  aging brain. The  interface  between  neuro-
chemistry  and psychophysiology in the gerontology of
the CNS could be fruitful over the next decade.

Conclusion

   The  evidence  reviewed here suggests  that the CNS
continues  to change throughout the life cycle. Changes
in various EP parameters have been observed across the
life span,  although  the  mechanisms  underlying these
changes are  not understood. Animal models for study-
ing  the  neurophysiology  of  aging  are problematic.
Since primates are  long-lived, old animals are difficult
to obtain. On the other hand, short-lived species, such
as the  rat,  are too remote from man  on the phylo-
genetic scale to provide an effective model for human
aging processes. Therefore, homo  sapiens appears to be
the best "model"  for such studies!

   Experimental   psychologists have provided a  num-
ber  of  profitable  routes of attack on  the  perceptual
and  cognitive changes  with age  (Birren  and Schaie
1976).   This  body   of  knowledge  may  enable

-------
 336
                                              Marsh
psychophysiologists to act as synthesizers between psy-
chology and  physiology.  The use of pharmacologic  in-
tervention  may  allow  the testing of proposed physio-
logic  mechanisms in man.   Even without  the  use  of
drugs, if some cognitive  processes are slowed with age,
then parameters of the EP associated with such processes
should demonstrate such slowing. It may be possible to
tease  apart EP  components that overlap in younger
subjects but  are separated in  latency  with  increasing
age.
   One  of the difficulties in research on  aging popula-
tion is to separate  the effects of disease, which appear
more frequently  in older age groups,  from the effects
of aging per se. The matter of healthiness is always one
of  degree,  but  can  be  crucial when  subjects  from
old-age  groups are used, as noted in  the Froehling
(1974) study. A balanced emphasis on  health, cognitive
and  perceptual factors,  and physiological mechanisms
should lead to a useful role  for psychophysiology in
the study of aging processes.

-------
DEVELOPMENT AND  DEVELOPMENTAL  DISORDERS:
DISCUSSION  SUMMARY*
   Harter  questioned how one can meaningfully inter-
pret ERP  changes during a task unless performance in
the task is  directly assessed. The question is  further
complicated if a clinical  population  is used, subjects
who may  not cooperate or follow instructions,  as well
as those who  are normal. The question becomes: What
happens to  ERP  differences  between  normals  and
abnormals  when performance  on a  task is directly
measured  or  matched? Knott argued for standards
against which  to assess brain activity.

   Harter  felt that the presence  or  absence of com-
ponents P2 and P4 might  have  considerable  clinical
importance, A  lack of change  in P4 to check  size
might indicate that the cortex of the  infant older than
3  months was not developing normally. Knott asked if
Harter's data  were  due to maturation changes  in the
optic  tract,  the  retinal system,  or the geniculate. It is
important to  try to get an idea of  just  where these
changes occur. Knott also encouraged the analysis of
the motor potential in the young child. Many tech-
niques now  used in infant research, such as an operant
conditioning  paradigm,  could be  fruitfully applied in
ERP research.

   Marsh stressed that  aging provides  a natural  experi-
ment  that begins at birth and ends at  death. Viewed in
this  light,  the  life span  constitutes  a continuum in
which   the  brain  is changing  in ways that  we  are
beginning to understand.

   Knott asked  if  experimental manipulations  of the
normal  subject that mimic the abnormal could provide
clues  for  changing  the abnormal  to  normal. Fenelon
answered  that it was  uncertain  how far  that model
could go but that it was necessary to pursue it.

   Fenelon stated that  his  studies with  dyslexics by
means  of spectral analysis confirmed work done else-
where, i.e., greater activity in theta and beta bands and
less in the  alpha range during rest. Overall, the energy
output  for dyslexic children is considerably lower.  The
evoked  responses   are lower  in  amplitude  than in
'Editor's  note: This  summary was prepared by Dr. Katrer
 from the transcript of EPIC IV discussions related to develop-
 mental issues and the  foregoing papers. The discussion and
 papers were originally  part of the Psychopathology plenary
 session.
normal subjects  and have less lateral asymmetry.  It is
uncertain  if  these  data reflect lesions, problems  in
interhemispheric   transfer,  or  developmental  lags  in
critical  areas  of the  brain  such  as  the  secondary
posterior  association  area  in   the  left  hemisphere.
Fenelon further  stated  that  the clinical objective  of
ERP  research must be  prescriptive  treatment rather
than  diagnosis or prognosis alone. We already know
who  the  problem  children  are.  CNS  measures  or
evoked response  measures should provide clues on  how
to tailor procedures for individuals,  perhaps feedback
procedures or some special sensory input coupled  with
lateralization.  Fenelon  felt  that  John's  procedures
would go far  in this direction.

   Stamm, noting that  Arthur  Rubenstein still played
the piano well at 87, asked Deecke  about the signifi-
cance of the  decline in  the Bereitschaftspotential  (BP)
in the elderly. Stamm wanted to know whether there
is  any  relationship  to motivational  factors  or  to
response  characteristics.  Deecke replied that it is  very
difficult  and  dangerous to  say  that  motivation  is
affected by  age,  that older subjects  have  less motiva-
tion.  The diminution  is  a matter  of differences  in
performance. BP amplitude  was larger preceding  slow
movement than  rapid  movement  and,  therefore, was
related  to response  characteristics.  Knott  felt  that
motivation can obviously be  affected by age and  that
one needs to  establish some standard  for motivation.

   Dongier  reported  studies  on  BPs in  depressed
patients in his  laboratory  that  showed no  amplitude
effects.   In   subcortical recordings  in  Parkinsonian
patients, there was  a positivity preceding the onset  of
EMG in  an  area rostral to the thumbs in a voluntary
motor task.  He  wondered  if  the Bristol  group  had
observed this phenomenon also. Dongier also pointed
out that BP  will be larger if there  is  some outcome.
Simple finger flexions give  relatively  small BPs. It  may
be  that as one  gets older,  one becomes  more bored
with just flexing  one's finger 400 times.

   Deecke   thought  that   Dangler's   recording  of
negativity  on the cortical surface (or scalp), positivity
in  the  white matter,  and  negativity  again  in the
thalamus  could  be understood in  terms  of volume
conduction, i.e.,  a sink  in the white  matter. McCallum
thought that one would  see a possible positive effect in
the caudate nucleus.

-------
  338
                           Development Summary
    Rosen  wondered  if structural changes in the skull,
 which  occur with  age,  such as calcification of bony
 structures, or thickening of membranes,  affect BP via
 increased   resistance  or  passive   effects.  Deecke
 responded that, if this were so, not only  BP but all
 other potentials likewise should be  diminished.  This
 was  not  the case;  i.e.,  the motor  potential was less
 affected  than the  BP  and  there was no  reduction in
 alpha  activity.  Therefore, the  selectiveness  of the
 reduction  cannot be explained in that way.

    Otto pointed out that Grant (this volume) presented
 a  schematic  diagram  of "functional brain  capacity"
 that  corresponded  well  to the changes  observed by
 Karrer in children and by Deecke in the elderly.

    Otto recalled that Loveless used  a very long fore-
 period   and  found  an  interesting  phenomenon. In
 younger adults,  there was no sign  of negativity until
 very  close to the point where the onset of movement
 should  be, where  the  BP occurs  associated with the
 movement. But  in  older adults, there  was not much
 negativity,  unless there was perhaps a  long sustained
 (baseline)  negativity. The  old  adults appeared to be
 having trouble gauging  the time  period so that the late
 negativity  seen  in younger adults was not seen. There
 was a great difference in the ability  to  prepare. Knott
 felt that  this  issue  should be  further  explored with
 appropriate experimental designs. Loveless (in  a com-
 ment submitted after the  conference) stated that, while
 he  was convinced that the E wave  of  his data was a
 BP, he  was  not certain  that Deecke's  data  would
 explain   the elderly's poor  RT  with long predictable
 foreperiods. The obvious  possibility is that the  elderly
 show a  BP at short  foreperiods,  but  not at long ones.

   Teece said that the fact there is a dissociation in the
 aged on  hard/easy items on the Wechsler memory scale
 and other kinds of  tests,  particularly  tests involving
 short-term  memory and distraction,  suggests that there
 is  more  to the  poor performance problem  than just
 motivation. With praxolone, a vasodilator,  it is possible
 to increase the  capacity  of the  aged to do better on
some  of these  memory items. It would  be of interest
to see what sort  of effects could be obtained with the
 BP. To Tecce it is clearly not just motivation because
 the aged can  do as  well  as  the young on the task.
 Deecke injected the comment that  there  has  been a
 paucity of  pharmacological studies of BP.  One should
 first  establish  the  response  of the  normal and the
 younger subjects and then go  on to older populations.


   Deecke commented that, in Karrer's data, positivity
 was early and that  he (Deecke*) has  also  found such
 curves in  the evoked potential  but did not schematical-
 ly  indicate  it.  Deecke did not know what it was—
 perhaps a different  distribution  of negativity. Perhaps
 the negativity first arrives at the convexity and then is
 more  involved  with basal structures. He  also  com-
 mented  on  Karrer's recording  of   EMC  from  the
 opposite,   nonresponding  thumb.   Certain  cerebral
 palsied patients  are unable to make a unilateral  move-
 ment; unilateral  efforts  in these patients  are  always
 accompanied with similar  movements on  the  other
 side.   Shibasaki   and   Kato   (1975)  studied   this
 phenomenon in normals  by comparing the BP during
 unilateral  and bilateral movements. They proposed that
 any unilateral movement is actually  a unilaterally in-
 hibited bilateral  movement.
   Karrer responded  that  motor  overflow  and  the
development of inhibition  of overflow in children  is a
clinical sign  commonly used by  pediatric neurologists.
Even on simple movements, some overflow is present.
The more complex the motor path, the more overflow
is present, a disparity that is inversely related to  age.
Individuals  differ  in  the  ability  to  inhibit  extraneous
movement, especially across  the midline, even in  adult-
hood.  The complexity of  the positive-negative-positive
BP observed by Karrer in  children  might be related to
extraneous movements occurring simultaneously. Motor
response  patterns  and the  inhibition of  extraneous
movements  are poorly differentiated in children. To
obtain data relevant to that problem, he recorded the
thumb contralaterally  and also  saved those  trials that
had  a  coincident eyeblink  within a  defined   time
window  at  thumb   press  to  see  if  simultaneous
extraneous motor  responses make a difference in  wave-
form or amplitude.

-------
 DEVELOPMENT REFERENCES
Abercrombie,  M.L.J., Lindon,  R.L.  and Tyson, M.C.
   Associated  movements  in  normal  and  physically
   handicapped children.  Develop. Med.  Child Neurol.,
   1964, 6:573-580.

Anderson,  R.P., Halcomb,  C.G. and  Doyle, R.B. The
   measurement of  attentional  deficits.  Exceptional
   Children, 1973, 39:434-439.

Andreasen, N.J.C., Peters,  J.F.  and Knott, J.R.  CNVs
   in hyperactive children: Effects of  chemotherapy. In
   W.C. McCallum and J.R. Knott (Eds.), The respon-
   sive brain. Wright, Bristol, 1976: 178-182.

Armington, J.C., Corwin, T.R. and Marsetta, R. Simul-
   taneously recorded retinal and  cortical  responses to
   pattern   stimuli.   J.   op.    Soc.   Am.,    1971,
   61:1514-1521.

Baumeister, A.A. Behavioral inadequacy and variability
   of  performance.   Am.   J.  ment.   Defic.,   1968,
   73:477-483.

Benzecri,   J.P. I.  Taxinomie,  II.  Analyse des cor-
   respondances. Dunod, Paris, 1976.

Berkson, G. An analysis of reaction time in normal and
   mentally deficient young men.  III.  Variation  of
   stimulus and response complexity. J. ment.  Defic.
   Res., I960, 4:69-77.

Birren, J. and  Schaie, K.W. (Eds.), The handbook  of
   the  psychology of aging. Van Nostrand Reinhold,
   New York,  1976.

Blakemore, C. and Campbell, F.W. On the existence of
   neurones in the  human visual system  selectively
   sensitive  to the  orientation  and size  of  retinal
   images. J. Physiol., 1969, 203:237-260.

Blowers,  G.H., Ongley, G.C. and  Shaw,  J.C. Implica-
   tions  of cross-modality  stimulus  permutations  for
   the CNV. In W.C. McCallum and J.R. Knott (Eds.),
   The responsive  brain.  Wright, Bristol, 1976:  20-25.

Bower,  T.G.  Development  in infancy.  Freeman, San
   Francisco, 1974.

Brody, H.  Organization  of the  cerebral cortex.  III. A
   study  of  aging in the  human cerebral  cortex.  J.
   comp. Neurol., 1955, 102:511-556.

Brody, H.  Aging of  the vertebrate brain. In M.  Rock-
   stein (Ed.),  Development and  aging in the  nervous
   system.  Academic Press, New York,  1973: 121-133.
Bronson, G. The postnatal growth  of  visual capacity.
   Child Dev. 1974,45:873-890.

Buchsbaum, M. Self-regulation of stimulus intensity. In
   E.G.  Schwartz and D. Shapiro (Eds.), Consciousness
   and  self regulation.  Vol.  1. Plenum,  New York,
   1976.

Buchsbaum, M.,  Coppola, R.  and  Bittker,  T.E.  Dif-
   ferential effect of congruence, stimulus meaning and
   information  on early and late components  of  the
   average  evoked response. Neuropsychologia,  1974a,
   12:533-545.

Buchsbaum, M.S., Henkin, R.I. and Christiansen, R.L.
   Age  and   sex  differences  in  averaged  evoked
   responses in  a normal population, with observations
   on patients with gonadal dysgenesis. Electroenceph.
   din. Neurophysiol., 1974b, 37:137-144.

Callaway,   E.  Des  correlations  entre  des  potentiels
   evoques  moyens et la mesure de I'intelligence. In A.
   Fessard  and   G.  Lelord  (Eds.), Average  evoked
   responses and  their conditioning  in normal subjects
   and  psychiatric  patients.  INSERM  (Paris),  1973:
   213-222.

Campbell,  F.W.,  Cooper,  G.F. and Enroth-Cugell, C.
   The spatial selectivity of the visual cells  of the cat.
   J. Physiol., 1969, 203:223-235.

Chapman, R.M. Evoked  potentials of the brain related
   to thinking.  In F.J. McGuigan and R.A. Schoonover
   (Eds.), The psychophysiology of thinking. Academic
   Press, New York, 1973: 69-108.

Clausen, J. Ability  structure  and subgroups  in mental
   retardation. Spartan, Washington,  D.C., 1966.

Cohen,  H.J.,  Taft, L.T., Mahadeviah,  M.S.  and Birch,
   H.G.  Developmental  changes in overflow in normal
   and  aberrantly  functioning  children.  J.  Pediat,
   1967, 71:39-47.

Cohen,  J. The  CNV in  children with special reference
   to learning disabilities. Electroenceph. clin. Neuro-
   physiol., Suppl. 33. 1973:  150-154.

Cohen,  J.  Learning  disabilities and conditional brain
   activity.  In  R. Karrer (Ed.), Developmental  psycho-
   physiology of mental retardation. Thomas, Spring-
   field, 1976.

Cohen,  L.B.  and Salapatek,  P. (Eds.),  Infant  per-
   ception: from sensation to cognition. Vol. 1. Basic
   visual processes.  Academic  Press, New York, 1975.

-------
 340
                          Development References
 Cohen,  J.  and  Walter,  W.G.  The  interaction  of
    responses in the brain to semantic stimuli. Psycho-
    physiol., 1966,2:187-196.

 Cohen,  J.,  Offner,  F.  and  Blatt, S.  Psychological
    factors in the production and  distribution of the
    contingent  negative  variation (CNV).  Proc. Sixth
    International   Congress   of   Electroenceph.  clin.
    Neurophysiol. Vienna, 1965: 251-254.

 Courchesne, E. Event-related brain potentials: compari-
    son between children and  adults.  Science, 1977,
    197:589-592.

 Courchesne,  E.,  Hillyard,  S.A.  and  Galambos,  R.
    Stimulus  novelty,  task  relevance and  the visual
    evoked  potential  in  man.   Electroenceph.  clin.
    Neurophysiol., 1975,  39:131-143.

 Cragg,  B.C. The  density of synapses and neurons in
    normal,  mentally  defective, and aging brains. Brain
    Res., 1975, 98:81-90.

 Deecke, L., Scheid,  O.  and Komhuber,  H.H. Distri-
    bution of readiness potential, promotion positivity,
    and  motor  potential  of the human cerebral cortex
    preceding voluntary finger movements.  Exp. Brain
    Res., 1969, 7:158-168.

 Deecke, L., Becker, W., Grozinger,  B., Scheid,  P., and
    Kornhuber, H.H. Human brain potentials preceding
    voluntary  limb  movements.   Electroenceph.  clin.
    Neurophysiol.,  1973, Suppl. 33:87-94.

 Deecke,   L.,   Grozinger,  B.  and  Kornhuber, H.H.
    Voluntary   finger  movement  in man:   cerebral
    potentials   and   theory.  Biol.  Cyber.,   1976,
    23:99-119.

 Deecke,   L.,   Englitz,  H.G.,  Kornhuber,  H.H.  and
    Schmitt, G. Cerebral  potentials preceding voluntary
    movement  in patients with  bilateral  or unilateral
    parkinsonakinesia.  In J.E. Desmedt (Ed.) Progr. clin.
    Neurophysiol.  Vol.  1,  Attention, voluntary  con-
    traction  and  event-related  cerebral   potentials.
    Karger, Basil, 1977: 151-163.

Donchin,  E., Otto, D.,  Gerbrandt,  L.K.  and  Pribram,
   K.H. While  a monkey waits:  Electrocortical events
   recorded  during the  foreperiod  of a  reaction time
   study.  Electroenceph.   clin.  Neurophysiol.,  1971,
   31:115-127.

Donchin, E., Kubovy, M., Kutas, M., Johnson, R. and
   Hern ing,  R.I. Graded  changes in evoked  response
   (P300) amplitude as a function of cognitive activity.
   Perc. Psychophys.,  1973, 14:319-324.

Dustman,   R.E.  and  Beck,  B.C.  Visually  evoked
   potentials:  Amplitude  changes  with  age.  Science,
   1966,  151:1013-1015.

Dustman,  R.E.  and Beck, B.C. The  effects of matura-
   tion and  aging on the  waveform of visually evoked
   potentials. Electroenceph. clin. Neurophysiol., 1969,
   26:2-11.
 Dustman, R.E., Schenkenberg, T. and Beck, E.G.  The
    development of the evoked response as a diagnostic
    and evaluative procedure. In R.  Karrer  (Ed,), De-
    velopmental psychophysiology  of mental  retarda-
    tion. Thomas, Springfield, 1976: 247-310.

 Eayrs,  J.T. Thyroid and developing brain: Anatomical
    and behavioral effects. In M.  Hamburgh and E.J.W.
    Barrington   (Eds.),   Hormones   in   development.
    Appleton-Century-Crofts,   New    York,    1971*
    345-355.

 Eisenberg,  L.  Behavioral  manifestations of  cerebral
    damage  in  childhood. In  H.G. Birch (Ed.),  Brain
    damage in children. Williams and Wilkins,  Baltimore.
    1964: 61-73.

 Eisengart, M.A. and Symmes, D. Effect of eye blink on
    the  visual  evoked  response  in children.  Electro-
    enceph. clin. Neurophysiol., 1971,  31:71-75.

 Ellingson, R.J. The  study of brain electrical  activity in
    infants.  In  L.P.  Lipsitt and  H.W.  Reese  (Eds.),
    Advances in child  development and  behavior. Vol.
    3. Academic Press,  New York, 1968.

 Erlenmeyer-Kimling, L. A prospective study of children
    of schizophrenic  parents. USPHS Grant MH-19560
    1971.

 Erlenmeyer-Kimling, L. A prospective study of children
    at risk for schizophrenia: Methodological  considera-
    tions and some preliminary findings. In R.D. Wirt,
    G,  Winokur and  M.  Ruff (Eds.),  Life history  re-
    search in psychopathology. Vol. 4, Univ.  of Minne-
    sota Press,  Minneapolis, 1975:  23-46.


 Fenelon,  B.  Expectancy waves  and  other complex
    cerebral  events in dyslexic  and  normal  subjects
    Psychon. Sci., 1968, 13:253-254.

 Finch,  C.E. Monoamine metabolism in the aging male
    mouse. In  M.  Rockstein  (Ed.), Development  and
    aging in  the nervous system.  Academic Press, New
    York,  1973:  199-218.

 Flavel,   J.H.   Cognitive  development.   Prentice-Hall
    Englewood Cliffs, N.J., 1977.

 Friedman, D., Simson,  R., Ritter, W.  and  Rapin, I.
   Cortical  evoked potentials elicited by real speech
   words  and  human  sounds.  Electroenceph.  clin
   Neurophysiol., 1975, 38:13-19.

 Froehling, S.D. Assessment of behavioral and  physio-
   logical functioning in  elderly  males. Unpublished
   doctoral dissertation, Univ. of Miami,  1974.
Gaillard, A.W.  and Naatanen, R. Modality  effects on
   the   contingent  negative  variation  in  a  simple
   reaction-time task.  In  W.C.  McCallum and  J.R.
   Knott (Eds.), The responsive brain. Wright,  Bristol,
   1976: 4045.

-------
Development References
                                               341
Galbraith,  G.C., Gliddon,  J.B. and Busk, J. Electro-
   physiological  studies of  mental retardation.  In R.
   Karrer (Ed.),  Developmental  psychophysiology of
   mental  retardation.   Thomas,   Springfield,  1976:
   311-333.

Gerbrandt, L., Goff, W.R. and Smith,  D.B. Distribution
   of the human average movement potential. Electro-
   enceph. din. Neurophysiol., 1973, 34:461-474.

Gilden, L., Vaughan,  H.G. and Costa, L.D. Summated
   human  EEC  potentials with voluntary movement.
   Electroenceph.     clin.     Neurophysiol.,     1966,
   20:433-438.

Goto,  H.,  Adachi,  I.  and Katayama,  S.  Negative slow
   potentials  preceding voluntary  movement of healthy
   and paralyzed persons.  Electroenceph. clin. Neuro-
   physiol., 1973, 34:712.

Hall,  R.A., Griffin, R.B., Mayer,  D.L., Hopkins, K.H.
   and   Rappaport,  M.   Evoked  potential,   stimulus
   intensity,   and   drug  treatment  in  hyperkinesis.
   Psychophysiol.,  1976, 13:405418.

Halliday,  R.,  Rosenthal, J.H., Naylor,  H.  and  Callaway,
   E. Average evoked potential  predictors of clinical
   improvement  in hyperactive  children treated with
   methylphenidate: an  initial study  and  replication.
   Psychophysiol. 1976, 13:429-440.

Harter,  M.R. Evoked cortical responses to  checker-
   board patterns:  Effects of check size as a function
   of   retinal   eccentricity.   Vision   Res.,   1970,
   10:1365-1376.

Harter,  M.R.  and  Suitt,  C.D. Visually evoked cortical
   responses  and pattern vision in the infant: A longi-
   tudinal study. Psychon.Sci., 1970,  18:235-237.

Harter, M.R.  and White, C.T. Evoked  cortical  responses
   to checkerboard patterns: Effect of check size  as a
   function   of  visual   acuity.   Electroenceph.  clin.
   Neurophysiol., 1970, 28:48-54.

Harter, M.R.  and White,  C.T. Effects  of contour sharp-
   ness  and  check size  on visually  evoked cortical
   potentials. Vision Res., 1968, 8:701-711.

Harter, M.R., Deaton, F.K. and Odom, J.V. Maturation
   of evoked potentials and visual preference in 6-45
   day old infants:  effects of check size, visual acuity,
   and  refractive  error.   Electroenceph. clin.  Neuro-
   physiol., 1977a, 42:595-607.

Harter, M.R., Deaton, F.K. and Odom, J.V. Maturation
   of pattern visual evoked potentials in infants from
   six  days  to  six months: Visual  acuity,  refractive
   error  and  visual attention. In  I.E. Desmedt  (Ed.),
   New developments  in visual evoked potentials  of the
   human brain. Oxford  Univ. Press, London, 1977b.

Huttenlocher, P.R. Dendritic development in neocortex
   of children with mental defect  and infantile spasms.
   Neurology, 1974, 24:203-210.
Ingram, T.T.S. Specific  learning difficulties  in child-
   hood:  A  medical  point  of view.  Br.  J. educ.
   Psychol., 1971,41:6-13.

Karmel,  B.Z.  and  Maisel,  E.B.  A  neuronal  activity
   model for infant visual attention. In L.B. Cohen and
   P.  Salapatek (Eds.), Infant perception: From sensa-
   tion  to  cognition.  Vol.  1. Basic  visual processes.
   Academic Press, New York, 1975: 78-131.

Karmel, B.Z.,  Hoffman, R.F. and Fegy, M.J. Processing
   of  contour  information by human infants evidenced
   by pattern-dependent evoked potentials. Child Dev.
   1974,45:39-48.

Karrer,  R. Developmental psychophysiology of mental
   retardation:  Some  problems and trends.  In  R.
   Karrer  (Ed.),  Developmental  psychophysiology of
   mental  retardation.   Thomas,  Springfield,   1976:
   454473.

Karrer,  R. and Ivins, J. Post-warning signal positivity in
   relation  to development,  RT  performance,  and
   warning signal  compounding. In W.C. McCallum and
   J.R.  Knott  (Eds.), The  responsive brain.  Wright,
   Bristol, 1976a: 132-135.

Karrer,  R. and Ivins, J. Event related slow potentials in
   mental retardates. In W.C. McCallum and J.R. Knott
   (Eds.),  The responsive brain. Wright, Bristol,  1976b:
   154-157.

Karrer,  R.  and  Ivins,   J.  Steady  potentials  accom-
   panying  perception   and   response  in  mentally
   retarded  and normal  children. In  R.  Karrer (Ed.),
   Developmental psychophysiology of mental  retarda-
   tion. Thomas, Springfield, 1976c: 361-417.

Kopi, K.A.  and Bagchi,  B.K. Visual evoked responses
   in  man:  Normative  data. Ann.  N.Y.  Acad. Sci.,
   1964, 112:254-269.

Kornhuber,  H.H.  and  Deecke,  L.  Hirnpotentialand-
   erungen beim  Menschen vor und  nach Wilkurbew-
   egungen, dargestellt mit Magnetbandspeicherung und
   Ruckwartsanalyse. Pflugers Arch., 1964, 281:52.

Kornhuber,  H.H.  and  Deecke,  L.  Himpotentialand-
   erungen  bei   WUlkurbewegungen  und   passiven
   Bewegungen des Menschen,  Bereitschaftspotential
   und  reafferente  Potentiale. Pflugers  Arch., 1965,
   284:1-17.

Kurtzberg, D., Vaughan, Jr.,  H.G.  and Kreuzer, J.
   Task-related cortical  potentials  in  children, in J.
   Desmedt (Ed.), Progr. clin. Neurophysiol. (in press).

Kutas,  M.  and  Donchin,  E. Studies of  squeezing:
   Handedness, responding  hand, response  force, and
   asymmetry  of  readiness  potential. Science, 1974,
   186:545-548.

Laffont,  R.  Role  du  movement  dans  revocation
   d'activites lentes  corticales. In A.  Fessard  and G.
   Lelord (Eds.),  Average evoked responses  and their

-------
   342
                          Development References
    conditioning  in  normal  subjects  and  psychiatric
    patients. INSERM (Paris), 1973: 405-420.

  Lairy, G.C.  (Ed.).  The evolution of the  EEC  from
    birth to adulthood. In A. Remond (Ed.), Handbook
    of  electroencephalography  and   clinical  neuro-
    physiology.  Vol. 6B.  Elsevier,  Amsterdam,  1975:
    73.

  Lelord, G., Gousset,  A.  and Henin-Ribeyrolles, D.
    Interet  des  methodes de conditonnement electro-
    physiologique dans 1'etude de la debilite mentale. In
    B.W. Richard  (Ed.),  First  Congress of Association
    Internationale    Etude   Scientifique   Arrieration
    Mentale.  Michael Jackson Reigate,  1967: 368-376.

  Lelord, G.,  Laffont, F.,  Jusseaume, Ph. and Stephant,
    J.L. Comparative  study of conditioning of AER by
    coupling  sound  and  light  in  normal  and autistic
    children. Psychophysiol,  1973, 10:415-425.

 Lelord, G.,  Laffont,  F. and Jusseaume,  Ph.   Con-
    ditioning of evoked  potentials  in  children of dif-
    fering intelligence. Psychophysiol.,  1976, 13:81-85.

 Lifshitz, K.  The averaged evoked  cortical response to
    complex  visual   stimuli.  Psychophysiol..  1966,
    3:55-68.

 Loveless, N.E.  and Sanford,  A.J. Effects of age on the
    contingent negative variation and preparatory set in
    a reaction time task. J. Geront.,  1974, 29:52-63.

 Low, M. and Stoilen, L. CNV  and EEC in children:
    Maturational characteristics and findings in the MCD
   syndrome. Electroenceph. clin. Neurophysiol., 1973,
   Suppl. 33:139-143.

 Low, M.D., Wada, J.A. and Fox, M. Electroencephalo-
   graphic  localization of conative  aspects of language
   production in  the human brain.  In W.C.  McCallum
   and J.R. Knott (Eds.), The responsive brain. Wright,
   Bristol,  1976: 165-168.

 Luders, H. The effects of aging on the wave form of
   the   somato-sensory   cortical   evoked   potential.
   Electroenceph.     clin.     Neurophysiol.,    1970,
   29:450-460.

Marsh, G.R.  Age differences in evoked potential  cor-
   relates of a  memory scanning process.  Exp. Aging
   Res., 1975, 1:3-16.

Marsh,  G.R.  and Thompson, L.W.  Age differences in
   evoked potentials during  an auditory  discrimination
   task. Gerontologist, 1972, 12:44.

Marsh, G.R. and Thompson, L.W. Effects of age on the
   contingent  negative variation in  a pitch  discrimina-
   tion  task. J. Geront., 1973, 28:56-62.

 Maurer, D. Infant visual perception: Methods of study.
   In  L.B. Cohen  and  P.  Salapatek  (Eds.), Infant
   perception:  From sensation  to  cognition. Vol. 1.
   Basic visual  processes.  Academic Press, New York,
   1975: 1-76.
 McCallum,  W.C. Some  psychological, psychiatric and
    neurological aspects  of the CNV. In A. Fessard and
    G. Lelord (Eds.), Activites  evoquees  et  leur  con-
    ditionnement chez 1'homme normal et en pathologic
    mentale. Colloque INSERM,  Tours, 1973:  295-326.

 McSherry, J.W. Physiological origins: A review. In W.C.
    McCallum and J.R. Knott (Eds.), Event-related slow
    potentials of the brain. Electroenceph.  clin. Neuro-
    physiol., 1973, Suppl. 33: 53-62.

 Neville, H.  The development  of cerebral specialization
    in  normal  and  congenitally  deaf children:   An
    evoked  potential and  behavioral study. Paper  pre-
    sented    to   the  International   Neuropsychology
    Society, Tampa,  Florida, February  1975.

 Orzack, M.H. and Kornetsky,  C. Attention  dysfunction
   in  chronic  schizophrenia.  Archs.  gen.  Psychiat.,
    1966, 14:323.

 Otto,  D.A.  and Leifer,  L.J. The  effect  of modifying
   response  and performance  feedback parameters on
   the CNV in humans. Electroenceph.  clin. Neuro-
   physiol.,  1973, Suppl. 33:  29-37.

 Otto,  D.A., Houck, K.,  Finger,  H., and   Hart, S.
   Event-related  slow  potentials in aphasic,  dyslexic
   and  normal  children  during pictorial  and letter
   matching. In W.C. McCallum  and J.R. Knott (Eds.),
   The  responsive   brain.   Wright,   Bristol,   1976:
   172-177.

Papakostopoulos,  D.  and Crow,  H.J. Electrocortico-
   graphic studies of the  contingent negative variation
   and  'P300'  in  man.  In  W.C. McCallum  and J.R.
   Knott (Eds.), The responsive  brain. Wright, Bristol
   1976: 201-204.

Perry, N.W., Jr. and Childers, D.G. The human visual
   evoked  response:  Method  and  theory.   Thomas,
   Springfield, 1969.

Picton,  T.W.  and   Hillyard,  S.A.  Human  auditory
   evoked potentials. II.  Effect  of attention.  Electro-
   enceph. clin. Neurophysiol., 1974, 36:191-199.

Poon, L.W., Thompson,  L.W. and Marsh, G.R.  Average
   evoked potential  changes as a  function of processing
   complexity. Psychophysiol.  1976, 13:43-49.

Popov,  N.A.  Etudes  de  psychophysiologie,  I.  La
   cyclochronie.  II.  La  dominance. Vol.  1.  Ed  du
   Cedre, Paris,  1948.

Posner, M.I., Klein,  R.,  Summers, J. and Buggie, S. On
   the selection of signals. Mem. Cogn., 1973,  1:2-12.

Prichep,  L.S., Sutton,  S. and  Hakerem,  G.  Evoked
   potentials in hyperkinetic  and  normal   children
   under certainty  and  uncertainty:  A  placebo  and
   methylphenidate   study.   Psychophysiol.    1976
   13:419-428.

Purpura, D.P. Dendritic  spine  "dysgenesis"  and mental
   retardation. Science, 1974, 186:1126-1128.

-------
Development References
                                               343
Purpura,  D.P.  Dendritic   differentiation  in  human
   cerebral cortex:  Normal and aberrant developmental
   patterns. In G.W. Kreutzberg (Ed.), Physiology and
   pathology  of  dendrites.  Raven, New  York, 1975:
   91-116.

Ritter, W., Simson, R. and Vaughan, Jr., H.G. Associa-
   tion  cortex potentials and reaction time in auditory
   discrimination.  Electroenceph.  clin.  Neurophysiol.,
   1972,33:547-555.

Rusinov, V.S. Electroencephalographic studies in  con-
   ditional reflex formation in  man. In M.A.B. Brazier
   (Ed.),  The  central nervous  system and behavior.
   Josiah   Macy   Foundation,   New  York,  1959:
   249-256.

Rutschmann, J., Cornblatt, B. and Erlenmeyer-Kimling,
   L. Sustained  attention  in   children  at risk  for
   schizophrenia.   Archs.    gen.   Psychiat.,   1977,
   34:571-575.

Scheibel,  M.E.,  Lindsay,   R.D.,  Tomiyasu,  (J.  and
   Scheibel, A.B. Progressive dendritic changes in aging
   human cortex. Exp. Neurol., 1975, 47:392-403;

Schenkenberg, T.  Visual, auditory, and somatosensory
   evoked responses of normal subjects from childhood
   to senescence.  Unpublished doctoral dissertation,
   Univ. of Utah, 1970.

Schwent,  V.L.  and Hillyard,  S.A.  Evoked potential
   correlates  of selective attention with multi-channel
   auditory  inputs. Electroenceph. clin. Neurophysiol.,
   1975, 38:131-138.

Schwent, V.L., Hillyard, S.A. and Galambos, R.  Selec-
   tive attention and the auditory vertex potential.  I.
   Effects of  stimulus delivery  rate.  Electroenceph.
   clin. Neurophysiol., 1976,40:604-614.

Shagass, C.  Averaged  somatosensory evoked potentials
   in  various psychiatric disorders. In J. Wortis (Ed.),
   Recent advances in  biological  psychiatry. Vol. X.
   Plenum, New York, 1968: 205-219.

Shagass,  C.  Evoked  brain potentials  in psychiatry.
   Plenum, New York, 1972.

Shagass, C.  and  Schwartz, M. Age, personality  and
   somatosensory  evoked  responses.  Science,  1965a,
   148:1359-1361.

Shagass, C.  and Schwartz,  M.  Visual cerebral evoked
   response  characteristics  in a psychiatric population.
   Am. J.  Psychiat., 1965b, 121:979-987.

Shagass,  C.,   Schwartz,  M. and Krishnamoorti,  S.R.
   Some psychological correlates of cerebral responses
   evoked by light flash.  J. psychosom. Res., 1965,
   9:223-231.

Shelburne, Jr., S.A. Visual evoked responses to word
   and nonsense  syllable  stimuli.  Electroenceph.  clin.
   Neurophysiol., 1972, 32:17-25.
Shelburne,  Jr., S.A. Visual evoked  responses  to  lan-
   guage stimuli  in normal  children.  Electroenceph.
   clin.  Neurophysiol.,  1973, 34:135-143.


Shibasaki,  H.  and  Kato,  M.  Movement-associated
   cortical  potentials  with  unilateral  and  bilateral
   simultaneous  hand  movement.  J.  Neurol.,  1975,
   208:191-199.

Shipley,  T. Evoked brain  potentials and sensory inter-
   action in the retarded  child. Am. J. ment. Defic.,
   1970, 74:517-523.

Small,   J.G.,   DeMyer,  M.K.  and  MUstein, V. CNV
   responses of autistic and normal children. J. Autism
   Child. Schizophrenia, 1971, 1:215-231.

Squires,  N.K., Squires, K.C.  and Hillyard, S.A. Two
   varieties of long-latency positive waves  evoked by
   unpredictable  auditory stimuli in  man.  Electro-
   enceph. clin. Neurophysiol., 1975, 38:387-397.

Stern,  J.A.,  Gold,  S., Hoine, H. and  Barocas,  V.S.
   Toward  a more refined analysis of  the "overflow"
   or "associated movement"  phenomen. In D.V.  Siva
   Sankar (Ed.), Mental health in  children. Vol. II.  PJD
   Publ., Westbury, New York, 1976.

Stevenson,  H.W. Children's learning.  Chapt. 14, Prob-
   ability  learning.   Appleton-Century-Crofts,   New
   York, 1972.

Straumanis, Jr., J.J.,  Shagass, C. and Overton,  D.A.
   Problems  associated with  application of the  con-
   tingent negative variation to psychiatric research. J.
   nerv. ment. Dis., 1969, 148:170-179.

Sutton, S.  The specification  of psychological  variables
   in an average  evoked  potential experiment. In  E.
   Donchin and D.B.  Lindsley (Eds.), Average evoked
   potentials.   NASA,   Washington,   D.C.,   1969:
   237-297.

Sutton, S.  and Paul, D.D. Evoked potential correlates
   of the detectability of low intensity signals. In  A.
   Fessard   and  G.  Lelord   (Eds.),   Average  evoked
   responses  and their conditioning in normal  subjects
   and  psychiatric  patients.  INSERM  (Paris), 1973:
   79-92.

Swetts,  J.A.  Signal  detection  and   recognition  by
   human observers. Wiley, New York, 1964.

Symmes,  D.  and  Eisengart,  M.A.  Evoked  response
   correlates  of meaningful visual stimuli in  children.
   Psychophysiol., 1971, 8:769-778.

Tanquay,  P.E.  and  Ornitz,  E.M. Two measures  of
   auditory evoked  response  amplitude  and their  rela-
   tionship to background EEC.  Psychophysiol., 1972,
   9:477-483.

Tecce,  J.J.  Contingent negative variation and  psycho-
   logical  processes  in  man. Psychol.  Bull., 1972,
   77:73-108.

-------
 344
                         Development References
 Thompson,  L.W. and  Nowlin,  J.B.  Relation  of in-
   creased attention to central and autonomic nervous
   system  states. In L.F. Jarvik, C. Eisdorfer and I.E.
   Blum  (Eds.),  Intellectual  functioning  in  adults:
   Psychological  and  biological influences.  Springer,
   New York, 1973: 107-124.

 Trimble,  J.L.   and  Potts,  A.M.   Ongoing  occipital
   rhythms and the VER. I. Stimulation  at  peaks  of
   the   alpha    rhythm.  Invest.   Ophthal.,   1975,
   14:537-546.

 Vaughan, H.G.,  Costa, L.D. and Ritter, W. Topography
   of the human motor potential.  Electroenceph. clin.
   Neurophysiol., 1968, 25:1-10.

 Visser, S.L., Stain, F.C., van Tilburg, W., op den Velde,
   W., Blom, J.L. and  de  R«ke,  W. Visual evoked
   response in senile  and presenile dementia. Electro-
   enceph. clin. Neurophysiol.,  1976, 40:385-392.

 Walter,  W.G. Slow  potential  changes in the human
   brain  associated with expectancy, decision and in-
   tention.  Electroenceph. clin.  Neurophysiol.,  1967,
   Suppl. 26: 123-130.

Walter,  W.G. Can "attention"  be  defined in physio-
   logical terms?  In C.R. Evans and T.B. Mulholland
   (Eds.), Attention in neurophysiology. Butterworths,
   London, 1969:  27-39.

Weinberg,  H. and Papakostopoulos,  D. Pattern dis-
   crimination in  the  spatial  distribution of  the  con-
   tingent   negative  variation.   Electroenceph.   clin.
   Neurophysiol,, 1975, 39:21-28.

Winer, BJ. Statistical principles in experimental design.
   McGraw Hill, New York, 1962.

Winfield, D.L., Hughes, J.G. and Sayle, W.E. Electro-
   encephalography-sleep  findings in cerebral  palsy.
   Pediatrics, 1955, 16:88-92.

Wold,  R.M.  Visual and perceptual aspects  for the
   achieving  and underachieving  child.  Special Child
   Publications, Seattle, 1969.

Zappoli,  R.,  Papini, M.,  Briani,  S.,  Benvenuti, P. and
   PasquineUi, A.L.  L'"onda  d'aspettative" (CNV)  in
   lesioni  accertate  dei lobi  frontali. Riv.  Pat  nerv.
   ment., 1973, 44:341-357.

Zimmerman,  G.  and Knott, J.R.  Slow potentials of the
   brain  related  to  speech  processing  in  normal
   speakers and  stutterers. Electroenceph. clin. Neuro-
   physiol., 1974, 37:599-607.

-------
VI.  PSYCHOPATHOLOGY
     Section Editors:

     John R. Knott and Joseph J. Tecce
     Tufts University School of Medicine
     Boston, Massachusetts, U.S.A.

-------
 EVENT-RELATED POTENTIALS  AND
 PSYCHOPATHOLOGY12
 J.R. KNOTT3
 Tufts University School of Medicine and New England Medical Center Hospital, Boston, MA,
 U.S.A.
 J.J. TECCE4

 Tufts University School of Medicine and Boston State Hospital, Boston, MA, U.S.A.
    In 1966 Grey Walter commented that studies of
 intrinsic electrical activity  of the human brain have
 contributed little to an understanding of either psy-
 chopathologjc processes or clinical diagnosis (Walter
 1966). In his opinion, event-related brain potentials
 including  the contingent negative variation (CNV),
 held more promise.

    In pre-Congress (EPIC W) correspondence among
 members of this panel, Knott raised  the question:
 "Why is slow potential research in psychiatry unpro-
 ductive?"  His question evoked a flurry of defensive
 responses, but it became obvious that investigators of
 event-related potentials and behavior (normal or path-
 ologic) have concentrated, often exclusively, on one
 of four types of ERPs (Fig. 1). Some study  early or
 late potentials; others study surface negative  or posi-
 tive activity. In effect, many ERP investigators develop
 a  self-limiting type of visual field defect. One group
 has a right homonymous hemianopsia and sees only
 early processes. Another has a left-field defect  and sees
 only late processes. Some have further combined de-
 fects: upper field, who perhaps see only late positivity,
or lower field defect, who see only late negativity.
Development of  such types of blindness impedes
achif vement of a fuller understanding of event-related
                   NEGATIVE
              EARLY
              N1.N2
  FAST
           III
               P1.P2
   LATE
   CNV
             SLOW
IV
   P300
                    POSITIVE

Fig. 1. Four classes of event-related brain potentials
based on polarity at recording site and early versus late
occurrence.
 Panel participants: John R. Knott and Joseph J. Tecce, Co-chairmen; Peter Abraham; Enoch Callaway; Maurice
Dongier; Bernardo Dubrovsky; Gilbert Lelord; Martine Timsit-Berthier; Roberto Zappoli.
2
 The chairmen note that another review of CNV, other late slow potentials, and psychopathology, which has
borrowed quite heavUy from this panel's EPIC IV precircuiited material,  pre-Congress correspondence and
Congress presentations, has been presented elsewhere (Roth 1977).

 Dr. Knott is Professor-Emeritus, Departments of Neurology and Psychiatry, University of Iowa. At the time of
this Congress, he was Professor of Neurology, Boston University School of Medicine, and presently he is Clinical
Professor of Neurology, Tufts University School of Medicine.
4
 Dr.  Tecce is Director of the Laboratory of Neuropsychology, Boston State  Hospital, and Associate Professor,
Department of Psychiatry, Tufts University School of Medicine. Preparation of this chapter was in part supported
by USPHS Research Grant AG-OOS1S from the National Institute on Aging and by Research Scientist Develop-
ment Award 1-K02-00016 from the National Institute of Mental Health (Dr. Tecce).

-------
348
                                                                                     Knott and Tecce
potentials and their relation to behavior, and limits
productivity in the area of psychopathology.

    Pre-Congress correspondence also indicated the
need  to  discuss five  particular topics if the central
question were eventually to be resolved. These included:

     1. Subject populations  and relations between
        ERPs in normal and pathologic groups. Path-
        ology itself is in need of rigorous operational
        definition for proper classification of subject
        populations.
     2. Procedures.   Basic methodology  must be
        established for eliciting and scoring  ERPs
        and manipulating  them by pharmacologic
        and other parameters.
     3. Response measures.  Such critical terms as
        CNV, PINV and P300 must be defined opera-
        tionally and logical methods determined for
        interrelating  these and early potentials (N2,
        P2, etc.). Response measures such as breath-
        ing  and EKG should  be correlated  with
        ERPs.
     4. Usefulness of explanatory hypotheses. In par-
        ticular, the  ceiling hypothesis (Knott and
        Irwin 1967, 1973), distraction-arousal hypo-
        thesis (Tecce et al. 1976), the concept  of an
        inverse  relationship between stress and CNV
        magnitude (Knott and Irwin 1973), and pos-
        sibly shape (Tecce 1972,  van Veen  et al.
         1973) should be evaluated.
     5. Electrogenesis. Especially as deduced  from
        topographic distribution of ERPs in relation
        to pathology and to task set, a more rigorous
        investigation of electrogenesis should be pro-
        fitable.
    Each of these areas should be attacked through
studies of normative as well as psychopathologic mater-
ial. Indeed, we were of the opinion that experimental
analogs of psychopathology  could be established in
presumably normal subjects. Such analogs could In-
clude the use of psychotropic drugs (how else can the
effect of drugs upon psychiatric patients  be under-
stood?) as well as manipulation  of tasks or external
stimulating conditions. Delaunoy and his colleagues
(this section), in fact, describe a simple method to in-
duce PlNVs in normal adults. Attempts to create such
models may help us to  understand  the mechanisms
that  produce  abnormal ERPs  in psychopathologic
states and to discover methods for re versing both the
effect and the cause.

    Returning  to Fig. 1, and the classification of ERPs,
 the newly  developing area of "far-field potentials"
(Jewett et al. 1970) will require further incorporation
Into the full matrix  of CNV data processing.  These
events, which occur within 10 msec of stimulus onset,
 appear to yield clinically  useful information  about
 brain stem integrity.  Although  they may not relate as
 dependent variables to attentional or behavioral varia-
bles, they may serve as indices of Independent varia-
bles subserving such processes. Since far-field poten-
tials indicate the intactness of sensory pathways pre-
ceding higher processing, the applied investigator may
be able to utilize brainstem EPs in assessing the environ-
mental insults that induce neuropathy. (See Seppalain*
en, this volume.)

    One may be critical of the fractionation of efforts
by ERP  investigators with  upper-  and lower-field
defects who study either CNV or P300 and ignore the
opportunity  to experimentally manipulate (possibly
simultaneously)  both the slow negativity (CNV) and
the late positivity (P300). A combined approach would
help us understand what each may signify in terms of
psychophysiological correlates. Upon this groundwork
were  laid the  discussions that took place during EPIC
    The section  of "Psychopathology" was opened
by Tecce, who summarized experimental and clinical
studies being conducted with his colleagues at Boston
State Hospital. The experimental work involved the
alteration of CNV development by two environmental
challenges, distraction and drugs. The clinical research
included the change in CNV development in schizo-
phrenics and former heroin addicts by drug therapy,
and alterations of CNV development in the aged. He
described disruption in CNV development in the aged
as  reduced amplitude associated  with lengthened
(slower) reaction time to S2 during short-term memory
tasks  and termed  this a "CNV distraction effect"
(Tecce et al.,  this  volume). Associated increases in
heart and eyeblink rates were considered evidence of
elevated levels  of arousal. Four measures showed reli-
able changes'. CNV (decrease), reaction time (increase),
heart rate (Increase), and eyeblink rate (increase) dur-
ing sustained  distraction (continuous mental  arith-
metic) and during both visual and auditory phasic dis-
traction. These four measures were viewed as reflecting
a  distraction-arousal  coupling  and as indexing  an
experimental  analog of psychopathology.  He then
presented the  distraction-arousal hypothesis (Fig. 2)
to explain  psychopathology,  particularly in schizo-
phrenics (Tecce and Cole 1976). He also showed that
CNV  prolongation of the  type ascribed to schizo-
phrenics (see later discussions by Dubroviky, Dongjer
and Timsit-Berthler) can occur in a normal individual
in  the state produced by distraction-arousal (Fig. 3)
and inferred  that  post-imperative negative  variation
(PINV)  may  be a reflection of distraction-arousal
couplingin schizophrenics (Tecce and Hamilton 1973).

     Tecce  pointed out that, although eyeblinks are
 usually regarded as an undesirable artifact, they may
 well be a sensitive and accurate indicator of disturbed
 psychological functioning. (See Tecce et al. 1978).

    He also reported  a CNV  rebound phenomenon!
that was unexpectedly found in a study of short-term

-------
ERPs and Psychopathology
                                                                  349
                                                                  FUNCTION ELEVATING
                                                                  FUNCTION LOWERING
             ASSOCIATION FUNCTIONS

                            I	
                                  ENERGY FUNCTIONS
   DISTRACTION
 FOCUSSEO
ATTENTION
 INHIBITORY
DE-ENERGIZING
  FUNCTIONS
EXCESSIVE
AROUSAL
Fig. 2.  Schematic diagram of a model showing a distraction-arousal association and interrelationships among
attention, arousal, and distraction processes. Distraction normally elevates arousal levels. Focussed and narrowed
attention, by reducing distraction, can attenuate excessive arousal. Excessive arousal levels heighten distractibility
and, therefore, increase the likelihood of distraction. Inhibitory functions, such as those produced by neuroteptic
drugs, reduce distraction by decreasing excessive arousal levels. Reprinted from Tecce and Cole (1976) by cour-
tesy of author and Prentice-Hall
memory effects on CNVs of 36 normal volunteers. In
this experiment, the control condition consisted of
light,  tone, key-press trials, and the distraction task
consisted  of two  types  of randomly  mixed trials,
(1) those identical with the control condition (no-letter
trials) and (2) those with  auditory letters presented
within the light  tone (S1-S2) interval for short-term
memory (letter trials). The reliable and expected CNV
decrease was found for letter trials, but an unexpect-
ed CNV increase above control CNVs was found for no-
letter  trials.  The enhancement in CNV amplitude in
no-letter  trials above values found in control trials is
approximately equal in magnitude to  the  reduction
in CNV amplitude in letter trials compared to control
trials. This rebound effect is absent in elderly individu-
als and patients  subjected to prefrontal leucotonmy
and, therefore, may be an indicator of plasticity of
brain  functioning.

    Tecce also reviewed data on the influence of psy-
chotropic drugs on CNV development in normal and
clinical populations. CNV amplitude was an accurate
indicator  of early  paradoxical drowsiness and later
alertness and excitation produced by dextroamphet-
amine in normal volunteers. Individuals characterized
                      by basal CNV shapes with a slow rise time (Type B)
                      typically showed the paradoxical drowsiness (and, In
                      some cases, increases in dysphoric mood), but individ-
                      uals having basal  CNV shapes with a fast rise time
                      (Type A) did not (Tecce and Cole  1974). Phenobar-
                      bital tended to  produce greater dysphoria and CNV
                      disruption in Type B than in Type A subjects (Tecce
                      et al. 1977), and former heroin addicts having Type B
                      basal CNVs tended to become less agitated when given
                      methadone and showed increased  CNV amplitudes
                      compared to  Type A patients (Tecce et al.,in press).
                      Evaluation of drug effects on ERPs and clinical state
                      in chronic schizophrenics indicated  increases in CNV
                      amplitude and clinical improvement  following chronic
                      administration  of mesoridazine  and  thioridazine
                      (Tecce et al.  1978)  and decreases in CNV amplitude
                      and clinical deterioration in patients receiving fluphen-
                      azine enanthate and decanoate (Tecce and Cole 1976).
                          Finally, in a study of the effects of psychosurgery
                      carried out for the National Commission for the Pro-
                      tection of Human Subjects of Biomedical and Beha-
                      vioral Research, it was found that amplitude of CNV
                      recorded at Cz and Pz tended to be higher in patients

-------
  350
                                                                     Knott and Tecce
                             SUBJECT 1
                                                    SUBJECT 2
            CNV

CONTROLj   EOG

            RT (msec))
 ADDING
            CNV

CONTROL2   EOG
RT (msec) )
                                 288

                                   I
                             SI     S2   MR
                                                     **w™»v^
                                                                           20/uvL
                                                                                 0.5 MC
                                                                           20jUVl_
                                                                                 0.6 sec
                                         ^^^^y^^vvv
                                         vvw^-v^^
                                                         202
                                                     SI     S2   MR
 Fig. 3. Example of CNV reduction during sustained distraction produced by sustained cognitive activity (adding
 sevens) in two young normal subjects. For subject 2, distraction produced CNV prolongation accompanied by
 lengthened reaction time to S2 and elevated levels of heart rate andeyeblink rate (distraction-arousal coupling).
 The occurrence of similar CNV prolongations in schizophrenics  may also reflect distraction-arousal processes.
 CNV (six trials per average) was recorded from vertex (Cz) and linked mastoid processes. EOG was recorded from
 above and below the right eye. Relative negativity at Cz and the supra-orbital site is upward. Reprinted from
 Tecce and Hamilton (1973) by  courtesy of author and Elsevier Publishing Company.
 having a more favorable outcome than for those having
 a less favorable outcome  (Tecce  1977). (Eyeblink
 frequency was the most sensitive measure differentiat-
 ing the two groups: the less favorable outcome group
 blinked 35 times per minute compared to 25 times
 per minute for the more favorable outcome group.)
 The one CNV  deficit in the more favorable outcome
 group was a reduction in frontal CNV, presumably an
 indication of the cognitive cost these patients paid for
 the amelioration of their emotional disturbance.

    On the basis of these experiments, Tecce  con-
 cluded that (1) distraction is a reliable source of CNV
 disruption and impairment in reaction time perfor-
 mance, (2) distraction-arousal coupling is  useful in
 studying ERPs and psychopathology, (3) eyeblink
 frequency is a useful adjunct indicator of disturbed
 psychological functioning, and (4) Type A and B CNV
 shapes may be useful  predictors of the therapeutic
 effectiveness of psychotropic drugs.

    While these experiments demonstrate that models
 of "abnormality" can be created by environmental
 manipulation of normal subjects, Knott believed that
 distraction is a type of stress  induced by the need to
 perform simultaneously conflicting tasks. Because the
 motivational system that normally enables the subject
                                       to carry out the instructed task is being interfered
                                       with, more energy in the form of higher motivation
                                       level is required in the distraction than in the control
                                       condition to equal the same level of performance.


                                           Dubrovsky  delineated  three  methodological
                                       problems pertinent to ERP studies of psychopathol-
                                       ogy. First, EOG "cancellation" techniques (previously
                                       advocated by McCallum and Walter 1968 - vide infra),
                                       used to remove ocular contamination of the abnormal
                                       prolongation beyond CNV and known as post-impera-
                                       tive negative variation (PINV), could result in subtrac-
                                       ting out the actual potential coming from the brain as
                                       well as that coming from the eye. A second problem
                                       concerned  the making of electroclinical  correlations
                                       based  on single-session recordings. He suggested that
                                       several  recording  sessions were  needed  to establish
                                       adequate rapport  between subject and experimenter
                                       to approach optimal  performance. Third, the  type.
                                       of instruction given to the individual, whether patient
                                       or normal volunteer, has received insufficient attention
                                       in the study of CNV types.

                                           Dubrovsky also  reported  substantive data on
                                       PINV and phobias to specific animals or  objects (see
                                       Barbas and Dubrovsky, this volume). These findings show
                                       how instructional  sets can be used as an independent

-------
ERPs and Psychopathology
                                             351
variable in the manipulation of the amplitude and
temporal  character of CNV in psychopathological
research. (Ed: Another relevant experiment reported
by Me Adam et al. in 1969 demonstrated that shape is
affected in various ways by both external and internal
instructional variables.) After patients had been treat-
ed by desensitization to previously phobogenic stimu-
li, the CNV takes' on a "normal" shape.

    Dubrovsky thought that too much concern was
being given  to relationships between early and late
potentials and acted that the study of different types
of brain potentials implicitly assumes that the nervous
system works in an integrated, serial manner. Poten-
tials measured in the frontal cortex reflect information
that has already been processed in primary receiving
areas. He suggested that  the nervous system be con-
sidered as parallel, interacting subsystems that process
information.  For instance, visual input is processed
initially in visual  cortex  and later  in frontal cortex,
but there is  a modification of the  information by
other parallel subsystems as well.  In studying CNV
and psychopathology, Dubrovsky recommended  the
measurement and classification of amplitude and shape
of slow ERPs in different clinical populations.

    Knott commented that studies of late (or even
later) brain potentials cannot ignore earlier potentials
(including the far-field  potentials)   because early
potentials carry information that can affect later acti-
vity. Conversely, the tonic level of the cortex created
by processes  related to slow late potentials associated
with one stimulus  can, in turn, affect early potentials
evoked by the next stimulus. This has been shown by
McAdam  (1968,  1969).  Tonic levels created  by or
associated with late potentials may be what "instruc-
tional set" is all about.

    Knott also commented on the surprising lack of
research on the association of P300 to psychopathol-
ogy. If this potential can  be modified experimentally,
then  whatever psychopathological  states  that have
been mimicked in normals should make it useful in
clinical populations both  as a dependent and indepen-
dent  variable. Despite   the substantial number of
studies on P300 and psychological processes, very little
has been done to establish the possible usefulness of
P300 as an indicator of psychopathology.

    Callaway  questioned whether it would be simpler
to assess schizophrenics by their verbalizations rather
than distration-arousal measures. He focused on short-
latency (less  than  200-msec) ERPs  and psychopathol-
ogy (left  half of Fig. 1)  and pointed out two advan-
tages to the use of these potentials. One  was their
apparent value in developing empirical diagnostic pro-
cedures, as exemplified by the work of E. Roy John
(this volume). He also noted that earlier components
may yield other methods of developing and testing
theories about psychopathology and other aspects of
individual differences. For example, schizophrenia is
not a single disease entity. Some schizophrenics may
have structural defects in short-term memory (involv-
ing information which  can decay with a half-life of
2SO to 200 msec and which does not enter conscious-
ness). This type of memory is analogous to a buffered
system used to construct filters. Callaway cited early
studies (Callaway 1975)  of segmental sets in schizo-
phrenics that indicated variability in evoked responses
within 90 to 200 msec. These results contrast sharply
with work by Shagass  (1976), which showed hyper-
stable  evoked responses within the first 90 msec.
Shagass has suggested that schizophrenics do not pre-
process data properly so that when the stimulus comes
in, other cerebral processes operate  and no compensa-
tion occurs. This could account for the initial stable
response  and for the later variability. Callaway suggest-
ed that the later responses are being "gummed up" by
other things happening in the brain  in addition to
processing of  the  stimuli. He saw the early evoked
responses as not having been "tweaked up," and thus
leading to highly variable late responses. Normal sub-
jects, on  the other hand, make immediate adjustments
to  incoming stimuli to compensate for whatever the
brain may be  doing at that moment and, therefore,
compared to schizophrenics  show more variability in
earlier responses and more regularized or "tweaked"
later responses.

    Callaway noted reports that if a subject pressed a
switch to perceive a stimulus, the evoked response to
the stimulus (click) was  depressed and  the  recovery
cycle for full amplitude was about a second. This lag
time  suggests  neural inhibition  associated,  perhaps,
with  ultra short-term (iconic) memory functions. He
wondered whether schizophrenics would show persev-
eration of this type of neural inhibition. "Sick schizo-
phrenics" (off medication) showed decreased ampli-
tude  of  evoked responses in  the  200- to 500-msec
post-stimulus  period. On the  other  hand, patients
who were not as acutely ill did not show this decrease
in amplitude (Braff et al. 1977). Although these tech-
niques are not useful  for diagnosing schizophrenia,
they  provide a  good example of  how early evoked
potentials  can  contribute  to our understanding of
CNS functioning, particularly short-term memory.

    Timsit-Berthier  represented  the  right  side of
Fig.  1 in  discussing CNV, its prolongation (PINV),
and P300. (For fuller  discussion see Timsit-Berthier
et al., this section.) Briefly, four types of CNV shape,
based on duration and resolution (return to baseline),
were  described: (1) rise of CNV between SI and S2,
resolving with slight positivity  after S2; (2) rise of
CNV between SI and S2 and declining to baseline
before S2; (3) rise of CNV between SI  and S2,  con-
tinuing without diminution  for seconds after S2; and
(4) rise of CNV between SI and S2, declining more

-------
  352
                                Knott and Tecce
 slowly than Type II, but more rapidly than Type III.
 Also described was a "non-CNV," termed "flat." She
 suggested that CNV prolongation might represent an
 inversion of P300, and perhaps an extension out to
 "P600."

   Positivity and negativity were expressed as a ratio,
 and those  smaller than minus 0.1 with the inversion
 of P300 tended to be characteristic of psychotic indi-
 viduals. Timsit-Berthier also asserted that these CNV
 types  permitted discrimination of psychotic subjects
 from a control group. This classification permitted
 identification of 93 percent of neurotics (as opposed
 to psychotics).

   Timsit-Berthier viewed the clinical application of
 ERPs as involving the differentiation, by cumulative
 curve  measurements  including shape,  of  psychotic
 and neurotic populations rather than the differentia-
 tion of psychotic and normal individuals. (No evidence
 of discrimination  of neurotic from control subjects
 was apparent, which makes an either-or discrimination
 of "psychotic" from  "neurotic"  rather problemat-
 ical!)

    Abraham advocated the use of the "Present State
 Examination" (Wing et al. 1967), a method of psychi-
 atric assessment reputed to have cross-cultural valid-
 ity. This twofold assessment provides both  a qualita-
 tive  description and a quantitative measure of the
 condition. It permits bypassing diagnostic labels and
 provides assessment of different areas of total person-
 ality disturbance, thus yielding  a measurement of
 functional impairment and psychopathology, indepen-
 dent  of psychiatric diagnosis. The  Present State
 Examination  was reported  to have  inter-observer
 reliability and to be quantitatively related to physio-
logical  variables. Indeed, Abraham  wondered why
Timsit-Berthier stopped her measurements at physiol-
ogy and did not also measure psychiatric  state. He
 thought it more important to study types of psycho-
pathological disturbances that can be measured consist-
ently  in different laboratories rather than nebulous
diagnostic categories. In  his own work, he reported
that  the joint use of the Spiral After Effect (SAE)
test and the Present State Examination has improved
discrimination among diagnostic categories of psychi-
 atric patients.

    In  a post-Congress memorandum  of consensus
 prepared after discussions with Dongier and Timsit-
 Berthier, Abraham stated: "CNVs can and do assist in
 psychiatric assessment, and an understanding of the
 processes underlying  psychiatric disorders, provided
 that there  is  standardization  of  procedures (within
 and between laboratories), adequate quantification of
 data, and (election of suitable differential diagnoses."
Abraham claimed,"... when  the CNV values  are
 extremely abnormal, the probability that the subject
 is suffering a psychotic opposed to a neurotic illness
 may be as high as 20:1." He further stated "... given
 a substantial sample of data from  the individual sub-
 jects: (1) the pre-imperative amplitude of the CNV,
 (2) the post-imperative negativity divided by  that
 amplitude value, (and) (3) the duration of the Spiral
 After Effect... objectively, consistently, and with
 substantial confidence when taken together, discrim-
 inate between severely disturbed and healthy individ-
 uals."

     Abraham further pointed out that Small and Small
 (1971),  Abraham et al. (1976), and Timsit-Berthier,
 Delaunoy et al. (this section) agree that  severely ill
 psychotic patients have (l)lower amplitude pre-imper-
 ative negativity  and (2) higher amplitude post-imper-
 ative negativity  than do "healthy subjects"; and  that
 (3)  these findings are  persistent  over  an extended
 period of observation. Thus, in his  opinion, there was
 greater agreement on the use of CNV in psychiatry
 than generally supposed.  He also pointed out  that
 electrophysiological data   from  the realm  of slow
 potentials yield  objective information  not dependent
 upon  subjective observations based upon patients'
 verbalizations or psychiatrists' qualitative opinions.

     Lelord  described a sensory  (sound-light) condi-
 tioning paradigm that  requires  no overt responses
 (detailed elsewhere in this section). Subject groups
 include  schizophrenic  and  neurotic  adults,  schiz-
 phrenic  children,  and normal  controls.  Auditory
 potentials (to SI) were less obvious in schizophrenics
 than in neurotics and normals. Visual evoked poten-
 tials (to S2 at occiput) were also  less obvious in schiz-
 ophrenics than  in  normals; however, there  were no
 group differences  in interstimulus slow potentials
 (CNVs).

     Autistic children showed CNVs and conditioned
evoked potentials that were smaller  than those in
normal children. Mentally retarded children had fewer
evoked potentials at the occiput and more generalized
positive slow potentials.

     Thus, the late potentials in this paradigm do not
discriminate psychotics, but they  may  discriminate
autistic  from normal   children. Earlier  potentials
appeared to be discriminative. The similarity of neurot-
ics and normals  in  the Timsit-Berthier paradigm for
CNVs thus holds in this experiment. Different effects
in children and adults may require further experiment-
al consideration.

     There is a paucity of information in man on the
electrogenesis of ERPs, possibly because of difficulty
in obtaining suitable clinical material. ZappoU cited
data from six clinical cases involving Freeman-Watts
frontal plane sections for  lobotomy operations (see
Zappoli et al., this section) as being incompatible with

-------
ERPs and Psychopathology
                                             353
the findings of Skinner (this volume) that medio-thal-
amic fronto-cortical systems were  important for the
genesis and regulation of anterior frontal negative slow
potentials.   The CNV development of lobotomized
patients indicated that prefrontal brain regions, when
isolated from normal connecting pathways with thalam-
ic nuclei, continue to receive and process information
of relevant stimuli, possibly under the control of cort-
ical and nonspecific subcortical pathways outside the
thalamic projection system. (Ed: Despite the problems
of phylogenetic differences, some  progress has been
made in the study of ERP electrogenesis and neuro-
chemistry in subhuman species).

     Marczynski  (this volume) has reviewed a wide
range of electrophysiological, pharmacological, behav-
ioral, and psychopathological data and has proposed a
model that may be useful to clinical investigators relat-
ing drugs and changes in slow potential patterns.
                    *     *    *
    The Symposium was thrown open to general dis-
cussion.

     Weinberg inquired why measurements of reaction
time were necessary to make inferences about proces-
sing irrelevant stimuli in Tecce's distraction paradigm.
Tecce replied that without performance decrement in
a central task, the efficient processing information in
two tasks is a matter  of divided attention, whereas
distraction implies some  type of performance decre-
ment in a central task. He then offered a formal defi-
nition of distraction as  a hypothetical organismic
process that directs attention toward irrelevant stimuli
in the environment (internal or external) and  inter-
feres with the selection of relevant  stimuli resulting in
response decrement to  the  relevant stimuli. Thus,
Tecce emphasized the importance of impaired perfor-
mance in a central task  as a critical attribute of the
distraction process and as one of two criteria suggested
earlier for  a "CNV distraction effect," the  other re-
quirement being evidence that the stimuli extraneous
to the central task have been  processed (Tecce  and
Hamilton 1973). He suggested that  Knott would prob-
ably regard "distraction" as causing "stress" but that
the concept of "distraction-arousal"  might  be more
experimentally testable and that Callaway's notions
of  "gummed  up" or "tweaked up" had too  much
surplus meaning to be useful as theoretical constructs.

     Knott responded that stress was a more inclusive
term and that the environmental  manipulation that
causes  "distraction" introduces  competing stimuli
and causes  stress as an organismic response. He  felt
that Tecce's own data suggest the validity of this argu-
ment, since  the  stress occurring  in consequence of
distracting environmental  stimuli  is  associated with
elevated heart rate and increased eyblink rate, which
Tecce calls  "arousal" (Tecce 1971). "Stress" may be
Induced by a variety of internal and external manipu-
lations; "distraction" is one of these.

    The  issue  whether distraction and  arousal are
subconstructs of stress is similar to a previous contro-
versy whether attention and arousal are subconstructs
of motivation (Tecce  1972). Since distraction con-
notes a disruption in the steering functions of atten-
tion (stimulus selection) and since arousal is devoid of
such associative properties, stress, in encompassing
both distraction and arousal, would appear to include
both steering and energizing properties, and conse-
quently,  have broader explanatory value. Distraction
and  arousal, on the  other hand, being narrower in
scope appear to have less imprecise meaning and lend
themselves more readily to experimental testing. The
question of whether stress connotes only impairment
in performance, as postulated for distraction, or also
connotes beneficial effects on behavior, requires clari-
fication both conceptually and experimentally. Tecce
viewed concomitant elevation in heart rate  and eye-
blink rate as reflecting a negative (stressful?) hedonic
state.

     Abraham's strategy of direct observation  of pa-
tients  rather than reliance upon their verbalizations
was  regarded by Tecce as important. He pointed out,
however, that electrophysiological measures are gen-
erally  more reliable than psychiatric observations. In
response to Callaway's suggestion that simple verbal-
izations of patients might be as  useful as electrophysi-
ological  measures, Tecce  reiterated the  marked dis-
agreement among  psychiatrists  in Interpreting verbal-
izations  of patients and pointed out that Callaway's
proposal could not  be  followed with mute patients.
Included under the mantle of "electrophysiological
measures" would be the topographic distributions uf
CNVs  in normal  states  and their disruption under
conditions of altered environmental manipulation or
in association with psychopathological states.

     Knott raised the point that the shape and topog-
raphy  of the CNV process required further definition,
experimentation, and discussion. Tecce has  proposed
two "types" of CNV based upon morphology of the
ascending negativity (Tecce 1971) and added a third
type with a triphasic time course (Tecce  and Cole
1976). Timsit-Berthier (this section)has proposed four
types of CNV. Earlier, van Veen et al. (1973) showed
that there were three morphological types based upon
the  development  of pre-S2 negativity. These were
related  to  personality  factors  defined  by Witkin's
"perceptual mode" of field dependency (Witkin et al.
 1962).

     The  suggestions introduced  by  Timsit-Berthier
(this section) regarding the  inversion of P300 as  a
causal factor in prolonging the CNV (the PINV) clear-
ly  need  experimenting testing.   This  provides an

-------
 354
                                Knott and Tecc^
 opportunity  for  investigators (whose field  defects
 leave only the right lower quadrant intact) to discover
 whether such inversion is possible in normal subjects and
 to define the instructional/environmental conditions
 eliciting the proposed effect (cf. Delaunoy et al,, this
 section). Conceivably this could  serve as a  "model"
 for a psychopathologic process.

      Dongier discussed the problem of defining patient
 groups  by challenging  Abraham's position  on the
 "Present State Examination." He felt that other meth-
 ods might prove more profitable in the exploration of
 diagnostic and descriptive correlates of electrophysio-
 logical processes. Dongier also questioned the change
 of CNV in schizophrenics subsequent to administra-
 tion of drugs. Tecce  responded that abnormally low-
 amplitude CNVs are increased by mesoridazine and
 thioridazine in  chronic schizophrenics and  that psy-
 chopathology,  particularly  hallucinations,  shows  a
 concomitant  decrease.  These pharmacologically pro-
 duced  alterations in  CNV were different for frontal,
 central, and parietal areas. He called this phenomenon
 "topographical discordance."

     Knott added that variability in the topography of
 slow potentials  has been well  documented  with evi-
 dence that indicates  "the CNV" represents a multi-
 plicity  of brain  potentials having different spatial and
 temporal characteristics.  A  complete  account  of
 "processing" in  CNS awaits analysis of these variables
 and correlation with well-defined operational variables
 ("normal" and  psychopathologic) on the behavioral
 side.

     Returning to problems  of methodology, Rosen
 asked to what extent eye movement potentials contri-
 buted to what is regarded as cortically generated CNV.
 Tecce replied that the combined use of visual fixation,
 careful  off-line exclusion of trials with clearly present
 eye  movements, and  evaluation of averaged vertical
 EOGs for averaged CNV trials provided collective pro-
 tection  against ocular contamination of CNV measure-
 ment. Peters et al. (1970), for instance, have described
 very  precise measures for the identification of eye
 movements. This method takes account of the fact
 that potentials of cerebral origin spread into deriva-
tions below  the eye  and  may be recordable from
points  quite  remote  from  scalp  derivations.  The
  method lor "cancellation" of eye movements mention-
  ed by Dubrovsky (vide supra) may also "cancel" the
  CNV. Whether one uses manual or programmed com-
  puter methods for acceptance or rejection of individ-
  ual trials, it is indeed necessary to keep the baby clean
  as well as to guard against throwing out the baby with
  the bath  water, as has been reported for the potenti-
  ometer method (Waszak and Obrist 1969).

     In summary, the  panel  agreed  that psychiatric
 subject populations required more rigorous definition,
 although the precise means of such determinations
 could not be established. There appeared to be con-
 siderable innovation in the manipulation of CNV as a
 dependent  variable with experimental analogs of
 psychopathology. Evaluation  of experimental  and
 clinical data will  continue  to require close  attention
 to the exact experimental methodology employed.
 Indeed, basic conclusions regarding process may be
 entirely dependent upon these independent  variables.
 Divergence of methods in measurement of slow poten-
 tial data persists, no unanimity of choice has emerged,
 and it is an open question whether early or late ERPs
 carry  the  most relevant information with respect to
 normality versus pathology.  Although  explanatory
 hypotheses are  alive  and  competing,  some future
 rapproachment  seems  probable,  as  evidenced  by
 comparisons  of the  "stress-ceiling"  hypothesis  and
 the "distraction-arousal" hypothesis. Such hypotheses
 make  it possible to construct testable models of elec-
 trocortical pathology. Finally,  the  usefulness  of
 temporo-spatial descriptors of  the  total process of
 event-related potentials (note the plural) may make it
 possible to obtain a more precise electrical brain picture
 (the literal translation of  "electroencephalograph")
 in relation to behavior.

    This section was  initiated  with a  provocative
question: Why is slow potential research in psychiatry
unproductive? The contents of  this  section  indicate
that the research has been more productive than initial-
ly believed, particularly in terms of possible hypo-
thetical constructs. (There were  only eight papers
directly addressed to psychopathology presented  at
the previous two Congresses, and only six at this one!)
The utility of slow potential techniques in individual
psychiatric  assessment, however,  requires   further
validation and awaits further definitive investigations.

-------
EXPERIMENTAL PRODUCTION OF POST-
IMPERATIVE NEGATIVE VARIATION
IN NORMAL  SUBJECTS  *
J. DELAUNOY, A. GERONO, and J. C. ROSSEAU

Department of Medical Psychology and Psychosomatic Medicine, University of Liege,
Liege, Belgium
    Several years ago, the Liege laboratory (Timsit-
Berthier et al. 1972, 1973) reported that contingent
negative variation (CNV) is abnormally prolonged for
several seconds in most psychotics and some neurotics.
This phenomenon is called  postimperative negative
variation (PINV). Transitory PINVs have occasionally
been observed in normal subjects under stressful con-
ditions. Gauthier and Gottesman (1976), for instance,
produced PINVs by superimposing labyrinthine stim-
ulation and a mental reckoning task on a classical
CNV paradigm.

    What psychophysiological parameters control the
appearance of PINVs in normal subjects? The present
study was designed to elicit PINVs experimentally in
normal subjects by subtle modification of a classical
CNV paradigm.

Methods

    The subjects were 15 women from 25 to 55 years
old (mean age: 31.5 ± 6.2) who were applicants for a
marriage counselor position. The experiment was con-
ducted  during job  interviews,  which ensured high
motiviation of the subjects and, presumably, CNVs of
high amplitude. To eliminate neurotic or psychotic
subjects, clinical interviews were conducted and pro
jective tests were administered.

    Vertex EEC referred to the left mastoid was recor-
ded using Ag/AgClelectrodes and an  11-sec  time con-
stant. The electrooculogram (BOG) was recorded from
right infra- and supra-orbital electrodes and the galvan-
ic skin response (GSR) from electrodes applied to the
thenar eminence (active) and dorsal surface (inactive)
of the left hand. Atropine (0.25 mg for two electrodes)
was injected  subcutaneously at vertex and mastoid
electrode sites in three subjects to control for GSR
artifact. Electrophysiologjcal data were averaged in
8-sec epochs at 128 Hz with an Enhancetron Comput-
'This work was supported in part by Grant 20397-
FRSM.
er.
    Two paradigms were used to elicit the CNV and
PINV:
    1.   Situation A. In a standard CNV paradigm, a
        click (SI) was followed after 1.5 sec by a
        train of light flashes (S2) (0.28 joule, 18 Hz)
        terminated by pressing a button.
    2.   Situation B.  Stimuli were presented as  in
        Situation A except that the flashes continued
        for 1 sec,  regardless of response, in one-third
        of the trials selected at random.

    A modified ABA sequence was used as follows:
A (36 trials), B (24 trials), and A (12 trials). Six CNVs
were thereby obtained for each subject by averaging
successive blocks of 12  trials. Trials contaminated by
GSR or EOG artifact were rejected from averages.

    PINV duration was measured from S2 onset to the
point at which the signal returned to baseline. It should
be noted that averages obtained during conditions A
and B were not strictly comparable since the duration
of flashes was longer in the four B trials not terminated
by button pressing.

Results

    Distinct prolongation of the CNV beyond S2 was
observed in ten subjects. In these subjects, PINV dura-
tion increased from less than 2 sec (situation A) to
4.5 to  6 sec (situation B) and then decreased again
when situation A was reinstated. Fig. 1 illustrates the
experimental production of a transitory and reversible
PINV in one subject.

    Situation B did not produce PINVs in the remain-
ing five subjects. Four subjects in the group, however,

-------
  356
                                                                            Delaunoy et al.
                                         SITUATION
            J_
   CLICK    SLI
    _L
 i????
   CLICK    SLI
                            SITUATION
                                 B
                                         ITUATION
    _L
_L
   CLICK    SLI
  _
 3v( _
    1

 Fig. 1. Representative waveforms observed in a normal
 tubjtct  during a classical CNV paradigm (situation A)
 and a modified paradigm in which the key press did
 not terminate the imperative stimulus train on one-
 third of the trials (situation B). Postimperative nega-
 tive variation appeared during situation B and then
 disappeared when the normal stimulus-response con-
 tingencieswere reinstated (situation A, bottom trace).

 exhibited high  amplitude  postoperative  positive
 wavei in situation A, a feature that was not present in
 the PINV group.  Fig. 2 illustrates this distinctive fea-
 ture. In these subjects, there was a reduction of post-
imperative positivity in situation B.

    Upon  completion  of  the  experiment, subjects
 were  asked what they thought had happened during
 situation B. Subjects in whom PINV was observed did
 not understand what had happened or could not pro-
 vide an explanation. Non-PINV subjects rationalized
 that the experimenter had interfered with the apparat-
 us or that an equipment malfunction had occurred.

 Discussion

    Two-third* of the subjects in this  experiment
showed  substantial prolongation of PINV following a
imall modification of the experimental protocol. It is
 possible that the prolonged negative wave as an electro-
 dermal  artifact elicited by  the change in stimulus-
                                                                                                ITUATION
                                                                                                   A
                                                                                              .SITUATION
                                                                                              >    B
                                                                                              SITUATION
                                           Fig. 2. Waveforms observed in one of four subjects
                                           who did not exhibit PINV during situation B. Note
                                           the large positive slow potential shift following S2
                                           (SLI) in this subject during situation A. This positive
                                           shift  was considerably reduced in the four subjects
                                           during situation B.
                                          response contingencies of situation B. Raskin et al.
                                          (1969) and others have shown that the GSR is an im-
                                          portant component of the orienting reaction to novel-
                                          ty. Averaged palmar skin responses, however, showed
                                          no resemblance to waveforms recorded simultaneous-
                                          ly at the vertex. As a further precaution, scalp record-
                                          ing sites were atropinized  in  three subjects. PINVs
                                          were observed in all of these subjects. It is unlikely,
                                          therefore, that the experimentally produced PINVs
                                          were an electrodermal artifact.


                                              Two subgroups of subjects were distinguished on
                                          the basis of the appearance or absence of PINV in sit-
                                          uation B and by the presence of absence of postimper-
                                          atlve  positivity in situation A, The latter positivity
                                          was not present in any PINV subjects, although the
                                          significance of this observation is not presently under-
                                          stood. Results of the a posteriori interrogation of sub-
                                          jects provide  a possible clue. Prolongation of the PINV
                                          was observed in subjects who were unable to resolve

-------
PINVs in Normal Subjects
                                             357
the ambiguity of the experimental intervention, where-
as, PINVs did not appear in subjects who were able to
resolve the uncertainty of the situation, regardless of
the correctness of the explanation.


    Donchin (1968) proposed  that a large  positive
wave, peaking about 250  msec  after stimulus onset,
reflects the resolution of uncertainty. Avast literature
(summarized by Tueting,  this volume) associates this
positive component (P300) with  a variety of cognitive
events and decision-making processes. The large post-
imperative  positive wave  observed in non-PINV sub-
jects  is probably  related to the  P300  component.
Absence  of this waveform in subjects who failed to
resolve the uncertainty of the experimental interven-
tion is consonant with Donchin's (1968) proposal.

    PINVs have been observed  in at least four situa-
tions: (1) in normal children less than 10 years old
(Timsit-Berthier and Hausman 1972), Low and Stoilen
1973); (2) in normal  adults under stress (Gauthier
and Gottesman 1976); (3) in psychotic patients with
poor  prognoses (Timsit-Berthier et al.  1972, 1973);
and (4) in patients with organic brain damage (Paty et
al., in press). PINVs tend to be transitory and reversi-
ble in normal adults under stress, but irreversible in
psychotics.

    Results of the present study confirm the findings
of Gauthier and Gottesman (1976) that PINV can be
experimentally produced in normal adults. The diver-
sity of conditions and populations in which the PINV
has been observed, however, suggest that it  probably
is not a unitary phenomenon. Further study is needed
to establish the functional significance of component
processes contributing to postimperative negative vari-
ation.

Summary

    Fifteen normal subjects participated in two exper-
imental paradigms: (1) a classical CNV paradigm and
(2) a modified CNV paradigm in which motor response
did not terminate S2 in one-third of the trials. In the
second  situation,  a  prolongation of postimperative
negativity was observed in two-thirds of the subjects.
A posteriori questioning suggested that subjects show-
ing prolonged PINV were unable to resolve the ambigu-
ity of the experimental  intervention, while subjects
showing no prolongation were able to rationalize what
had happened.

-------
 EFFECTS  OF VISUAL DISTRACTION ON
 CONTINGENT NEGATIVE  VARIATION
 AND TYPE A AND  B  CNV SHAPES*

 J.J. TECCE,2 J. SAVIONANO-BOWMAN, AND J.B. KAHLE
 Laboratory of Neuropsychology, Boston State Hospital, Boston, MA, U.S.A., and
 Department of Psychiatry, Tufts University School of Medicine, Boston, MA, U.S. A.
    The  auditory presentation of letters within the
interstimulus  interval of  a  classic CNV  paradigm
(Walter et al.  1964) produces a reduction in CNV
amplitude and lengthened  reaction time (RT) to S2
(Tecce et al.  1976). This  CNV distraction effect is
accompanied by elevated levels of psychophysiological
arousal (increased heart rate and eyeblink rate). The
purpose of the present study was to test the generality
of this distraction-arousal association in a short-term
memory task  for visual letters and to assess the CNV
distraction effect in two groups of individuals charac-
terized by basal CNVs with fast (Type A) and  slow
(Type  B) rise times. The A-B typology has been shown
to be related  to individual  differences in response to
amphetamine  (Tecce and Cole 1974) and phenobarbi-
tal (Tecce et al. 1977).

Method

Subjects


    Twenty-four male college undergraduates served
as paid volunteers. Age range was 18 to 25 (X=20.3).
Twenty individuals were right-handed and four were
left-handed. All participants were screened for possible
medical and psychiatric problems.

Experimental procedures

    A simple reaction time paradigm with a constant
foreperiod constituted the basic  experimental  pro-
cedure. A preparatory signal (SI) consisted of a brief
(0.15-sec) flash of a black "X" (2 cm in height) appear-
ing on  a circular patch of dim light (2.5 cm in dia-
meter), which was projected onto a translucent plastic
panel 5 cm in height and 4 cm in width. The projector
was located 1 m from the subject's eyes at an approxi-
mate angle of 25° from the horizontal and stood 44 cm
1 Supported by U.S. Public Health Service Research
 Grants MH-19211 and AG-00515.
2Research  Scientist Development  Awardee  1-K02-
 00016 from the National Institute of Mental Health.
from  the  floor. The second stimulus  (S2) was  a
1000-Hz tone of approximately 70 dB (SPL) presented
through earphones 1.5 sec after the "X" and terminat-
ed by a telegraph key press. All but one subject pressed
the key with the right hand. A control run (C) consist-
ed of 16 trials of Sl-S2-motor response and lasted
approximately 4 min. Inteitrial intervals varied ran-
domly from 8 to 14 sec (X=l 1) within a rectangular
distribution of values 1  sec apart. In addition to the
control run, there were two letter conditions - visual
letters-recall (R) and visual letters-no recall (NR). The
letters-no recall task was similar to the  control con-
dition except that four letters were presented through
the visual  projector within  the S1-S2 interval. The
letters A, C, E, H, K, L, N, P, S, and U could appear
with equal probability. Letter exposure  time was
0.2 sec and interletter interval was 0.1 sec. The time
interval between the fourth letter and tone (S2) onset
was 0.15 sec.  Subjects were instructed to ignore the
letters in this letters-no recall condition and to concen-
trate only on terminating the tone. The letters-recall
task was similar to the no-recall task except that upon
hearing "OK" spoken by  the experimenter, subjects
repeated the four letters in the same order as presented.
The "OK" signal for information feedback occurred
irregularly from 1 to 4 sec after the key press.


    Subjects were randomly assigned to two groups.
The  first group (n=13) received the test  sequence:
C1-C2-C3-R-C4-NR-C5;  the second group (n=ll) re-
ceived  C1-C2-C3-NR-C4-R-C5.  (Two subjects  were
dropped from the second group  due  to technical
problems.) The first two control conditions were not
considered in  data analysis.  Prior to the sequence of
test runs, subjects received a 15-min rest period. A test
run was given every 10 min (4-min test and 6-min rest).
Session duration was approximately 1H hours. The
testing room (3.8 by 4.5  m) was dimly lighted by a
floor lamp with a 60-W incandescent bulb. Subjects
were seated in a semlreclined position and were encour-
aged to keep their eyes on the stimulus panel and to
avoid blinking during presentation of stimuli. Behavior
was  monitored  on  closed-circuit television. Verbal
communication took place through a two-way inter-
com system.

-------
Visual Distraction and CNV Shape

Recording procedures

    EEC was recorded from Fz, Cz, and Pz with link-
ed earlobes  as a reference. EOG  was recorded from
3.0 cm above and 2.0 cm below the right eye as mea-
sured from center  to center  of pupil and electrode.
Electrodes were Beckman  Ag/AgCl  standard (EEC)
and  miniature (reference and EOG) types.  A Grass
Model 7B polygraph was  used to record EEC and
EOG. The time constant for EEC  and EOG was 8 sec.
High-frequency cutoff  was 75 Hz  (50% amplitude
reduction) with  12 dB  per octave roll-off. CNV and
EOG data were recorded on magnetic tape. Trials with
eye movements (16% of total trials), premature key
presses within the S1-S2 interval (<1% of total trials),
and extraneous baseline shifts, such as those produced
by body movements (27% of total trials), were omitted
in off-line averaging with a CAT 1000. Average CNVs
were based on 6 to 12 trials per run, the number being
constant for a given  individual.  CNV amplitude was
measured as the difference in average voltage (sampled
every 16 msec) between the 256-msec epoch pre-S2
and  the 512-msec  epoch pre-Sl; this difference was
referenced to an on-line 25-fiV calibration pulse. The
EKG for cardiotachometric analysis was recorded from
sternum  to lower left chest. Overall heart rate (beats
per  minute) for each condition  was determined  by
obtaining the mean of a random sample of 20% of in-
dividual momentary  heart rate in a 4-min run (20%
and  100% samples yield  comparable results). Eye-
blinks were  defined as an  EOG excursion of at least
50 fjV and  of less than 1000-msec duration  (usual
duration: 150 to 300 msec) as measured at the EOG
baseline. Eyeblink  rate  (blinks per minute) was based
on  the number of blinks occurring during an entire
4-min run. In the letters-recall condition, accuracy of
recall  (percent  correct trials)  was determined  by
dividing  the number of trials in  which recall  of  the
four letters  was correct  by 15 (total number of trials
scored). A trial was correct if letters were repeated in
the same sequence given.

Type A and B CNV shapes

     Subjects were classified  into two groups  on  the
basis of CNV shapes. Fig.  1 shows two types of CNV
determined  by a fast rise  time (Type A) and  a slow
rise  time (Type B) (Tecce 1972). The slope  of  the
negative-going (ascending) limb  of CNV was deter-
mined by comparing a 112-msec EEC epoch occurring
760 to 872 msec  after SI (point L in  Fig. 1 is  the
midpoint of this segment) with  an epoch occurring
256 msec before S2 (point H in Fig. 1 is the midpoint
of this segment). The value of L represents the mean
of eight voltages sampled 16 msec apart. A small volt-
age difference (less than 6 uV) between L and H (HD
on left side of Fig.  1) indicates a quick rise time and  a
Type A shape. A  large voltage  difference (6  nV or
more) between L and H (HD on  right side of Fig. 1)
                                            359

indicates a gradual rise time and a Type B shape. As
in previous work (Tecce and Cole 1974, Tecce et al.
1977), this quantitative  determination of the  A-B
typology is based on Cz recordings made during the
control conditions (C3 run in this study).

Statistical analyses

    For comparisons between letter and no-letter con-
ditions, mean differences were evaluated by correlated
Mests  with 23 degrees of freedom. For comparisons
between Type A and Type B groups, mean differences
were evaluated by  independent  Mests with 22 df.
Unless otherwise  specified, reported differences are
significant at p < .05.

Results

    Preliminary  evaluation of CNV amplitude, RT,
heart rate, and eyeblink frequency  for control  runs
occurring before and  after  letter conditions  showed
that the three controls were not different from one
another on each response measure. Consequently, data
for these three control conditions  were  pooled for
statistical evaluation. Tables 1 and 2 contain summaries
of means, standard deviations, and significant compar-
isons among experimental conditions for the response
measures. For simplicity of presentation, CNV negativ-
ity appears as algebraically positive in the tables.

Measurements

    Table  1 contains mean values of CNV amplitude
for pooled controls and for letter conditions for Fz,
Cz,  and Pz locations. For Fz  and Cz, mean CNV
amplitude  was lower for the  letters-recall than the
letters-no recall condition (p < .05). For each record-
ing site,  mean amplitude of CNV was lower in  both
the letters-recall and letters-no recall condition com-
pared to the pooled controls (p < .001). Fig. 2 shows
representative CNV data for one subject.


    An evaluation  of topographical differences  for
pooled controls indicated that  CNV amplitude was
larger at Cz than at Pz  (p < .001) and larger at Pz than
at Fz (p_< .001).  The mean difference between Cz
and Fz (Xrjjff = 8.35  /iV; 80^= 6.18) was signifi-
cantly  (p < .001) larger than  the   mean difference
between Cz and Pz (Xrjiff = 4.33 juV; SDrjtff = 4.37).
    As shown in Table 1, mean RT was slower for the
letters-recall condition than for either the letters-no
recall condition  or pooled controls (p < .001). Mean
heart rate and eyeblink rate were higher in the letters-
recall  condition than in either the letters-no recall
condition or pooled controls (p < .001). For the lett-
ers-recall  condition,  mean  percent  of  trials with
correct recall of letters was 75.58 (SD = 24.01).

-------
  360
     Si
                                                                                         Tecce et al
                 TYPE A
                                                               TYPEB
                                                    Si
 I   k
S2  MR
                                                                                       10/itV
0.5 MC
  Pig.  1'.  Two types of CNV shapes based on fast (TypeA) and slow {Type B) rise times.  Type A has a difference
  of less than 6 n V between land H (shown as HD).  The L-H difference is 6 pV or greater for Type B. The record-
  ing site is Cz with linked earlobes as reference. Relative negativity at Cz is upward.
               Table 1. Electrophysio logical Measures and Reaction Times - Means
                                       (Standard Deviations)
Condition
Letters-
recall
Letters-
no recall
Pooled
controls
CNV amplitude, jiV
Fz
-1.12a'b
(6.43)
2.30a- b
(4.98)
4.82°
(3.20)
Cz
5.55a-b
(6.30)
8.69a' b
(7.33)
13.17°
(7.12)
Pz
2.16b
(5.24)
4.29b
(4.74)
8.84C
(5.63)
Reaction
time,
msec
392a- b
(135)
239a
(47)
234
(42)
Heart
rate
76.20a- b
(10.26)
71.618
(9.61)
71.07
(9.93)
Eyeblink
rate
20.15a-b
(11.33)
12.408
(8.78)
12.57
(8.32)
'Letten-recall and lettere-no recall «lgniflcantly different from each other.

 Significantly different from pooled contrail.

 Significant difference*: Cz> Pz > Fz.
      Table  2.  Electrophysiological Measures and Reaction Times for Type A and B Groups
                 during Pooled Control Conditions-Means (Standard Deviations)
Response groups
Group
TypeA
TVP«B
n
12
12
CNV amplitude, jiV
Fz
4.37
(3.29)
5.27
(3.17)
•p<4» bp<.oe.
Cz
10.85
(7.60)
15.49
(6.04)
••••••••i^M
Pz
6.26"
(5.36)
11.438
(4.80)
™^"»i^^™.
Reaction
time,
msec
250.33 b
(52.64)
217.92b
(20.65)

Heart
rate
74.02
(9.35)
68.12
(9.99)

Eyeblink
rate
11.66
(9.11)
13.48
(7.74)


-------
 Visual Distraction and CNV Shape
                                                                             361
 Cz
                                          1$ec
 Pz
        »     t
       SI    S2
       CONTROL
SI    S2
LETTERS
RECALL
  SI    S2
LETTERS NO
  RECALL
Fig. 2.  Examples of CNV traces of one Individual for
control, letters-recall, and letters-no recall conditions
for frontal (Fz}, central (Ci), and parietal (Pz) record-
Ing sites.  The SI -S2 (flash-tone) interval was 1. 5 sec,
The on-line calibration signal at the left of each trace
ii 25 pV .   Relative negativity at Fz,  Cz, and  Pz
(referred to linked earlobes) is upward.

Type A and B CNV shapes

    The criteria described above for classification of
CNVs by difference in CNV slope resulted in  12 Type
A and 12 Type B individuals. For the C3 run, which
was the basis for determining these types, three A's
and two  B's had  absolute Cz  CNV amplitudes leu
than 6 fiV while nine A's and ten B's had amplitudes
greater than 6 MV. Table 2 indicates that the Type B
group shows a pattern of higher CNV amplitudes (for
pooled  control conditions) than does  the  Type  A
group (p < .03  for the Pz recording site). These points
are discussed further below.

    Fig. 3 presents  the  differences in  CNV ampli-
tude for control minus  letters-recall. The larger this
difference, the relatively  smaller is the value  of CNV
amplitude  in the letters-recall condition.  For  all
three recording sites, Fig.  3  indicates that there
is  a relatively greater decrease in CNV amplitude for
the Type B  group in the  letters-recall condition than
for the  Type A group.  For Fz, this difference ap-
proaches statistical significance (p < .07).


     Table 2 indicates that Type B subjects tended to
have a faster mean RT (218 msec) than did the Type
A group (250 msec) (p  < .08). Means of heart rate
and eyeblink rate for the two groups also appear in
Table 2. Mean percent of trials with correct recall of
letters was  76.26  (SD - 24.02) for the B group and
74.89 (SD • 23.99) for the A group.
Discussion

     There were two main.findings in this study. First,
CNV amplitudes were reduced  during  a short-term
memory task as compared to a simple RT paradigm.
The criteria for demonstration of a distraction effect
appear to have been fulfilled: (1) stimuli in the irrele-
vant (distraction) task were  processed (75%  recall of
letters was  achieved) and (2) a performance  decre-
ment in the central (nondistraction) task resulted (sig-
nificant slowing of RT when  recall was required). This
finding replicates the distraction effect on CNV observ-
ed with a similar short-term memory task for  auditory
letters (Tecce  et al. 1976). Other studies have shown
CNV reduction in a visual short-term memory match-
ing-to-sample task (Roth et al. 1975) and in  a task
involving extraneous  auditory  (music) stimulation
(Miller etal. 1973).

     The second main finding was that visual stimuli
produced a distraction-arousal association that repli-
cates previous results with  auditory  distraction:  re-
duced CNV, increased RT,  elevated heart rate,  and
increased eyeblink  frequency.  Consequently, the«e
data extend the generality of the  distraction-arousal
hypothesis  and suggest  that this hypothesis pro-
vides a useful model to account for disruption in
CNV development.

    In addition, the four measures (CNV, RT, heart
rate, and eyeblink rate) are  considered to covary re-
liably for both auditory and visual distraction, com-
prising a reliable pattern of disturbed neuropsycho-
logical functioning that might be of value in the as-
sessment of abnormal neurological and psychiatric
conditions. In a recent study of psychosurgery, pa-
tients having  a more favorable  postoperative out-
come showed  a  pattern of higher CNV amplitude
(at  Cz and Pz), faster RT, and lower  heart and eye-
blink rates  compared to  patients  having  a less fa-
vorable  outcome (Tecce 1977). Eyeblink rate differ-
entiated the two patient groups most clearly.

     Although  eyeblinks are  a serious source of arti-
fact in the recording and measurement of CNV, they
appear to have value, per se, as a dependent variable
and indicator of psychological disturbance. The joint
occurrence of elevations in eyeblink rate and heart rate
in  the letters-recall condition was accompanied  by
subjective evidence (verbal reports) that the short-term
memory task was a demanding and unpleasant experi-
ence. We infer, therefore, that elevations in heart rate
and eyeblink rate, considered jointly,  are potentially
useful indicators of negative hedonic experience. Pre-
vious work has shown that a dissociation of these two
measures (increase in heart  rate  and decrease in eye-
blink rate) following dextroamphetamine administra-
tion is related to euphoria and feelings of well-being-
i.e., a positive hedonic experience (Tecce  and Cole

-------
 362
                                      Tecce et al.
           •  TYPE A GROUP
           '///,  TYPE B GROUP
> 12
a.
2 10
 a   6
 UJ
    2
 Ffc.  J.   CNV amplitude differences (control minus
 letters-recall) for subjects having basal CNVs with a
 fast rise time (Type A group-n=12) and a slow rise
 time (Type B group-n=12). Larger difference scores
 indicate greater CNV reduction in the letters-recall
 task  compared to pooled control values. For the let-
 ters-recall condition, means (and standard deviations)
 of CNV amplitude  values  were; Fz, 1.03 (5.54); Cz,
 5.00 (3.85); and Pz, 1.47 (2.75) for  the  Type A
 groups and Fz, -2.44 (6.19); Cz, 6.10 (8.21); and Pz,
 2.85 (6.99) for  the Type  B group. For simplicity,
 negative CNV values appear as positive.
 1976). These findings suggest that elevated heart rate
accompanied by decreased  eyeblink frequency may
reflect an  emotional state  associated with  positive
hedonic tone (as in an amphetamine experience), but
that increased  heart rate accompanied by increased
eye blink frequency probably indicates an emotional
state associated with negative hedonic tone (as during
distraction).  Further  discussion  of   the  eyeblink-
hedonic  hypothesis appears  elsewhere  (Tecce  et
al. 1978).

    The anterior-posterior distribution of amplitude
of CNV recorded  in control (no-letters) condition
showed a  steeper fronto-central than parieto-central
gradient of CNV development. This ordering of CNV
amplitude values (Cz > Pz > Fz) and the greater steep-
ness in the fronto-central gradient of  CNV develop-
ment in comparison with the parieto-central gradient
are in  agreement with  previous work (Tecce et al.
1976)  and suggest two  indices of normal CNV top-
ography of potential value in the assessment of brain
dysfunction in individuals  having neurological and
psychiatric  problems.

    The present results on the CNV distraction effect,
distraction-arousal coupling, and topographical CNV
distribution agree  with previous work, while other
aspects of the present findings do not. First, in the
earlier study (Tecce et al. 1976), when auditory letters
 were presented in the S1-S2 interval with instructions
 that no recall would be required and  that they could
 be ignored (letters-no  recall), amplitude of CNV was
 not changed. In the present study, on  the other hand,
 visual letters  significantly  reduced CNV amplitude
 even when  no recall was required. This unexpected
 decrease in CNV  amplitude appears to be related  to
 the use of the four successive visual letters as a chain
 of anticipatory  stimuli, the fourth  letter appearing
 consistently 150 msec before S2 (tone) and, therefore,
 serving as a cue  for tone onset. In  postexperiment
 reports, subjects indicated attending to the extraneous
 letters (despite instructions to ignore them) in order
 to facilitate preparation to respond to S2. Such a use
 of the extraneous letters was inadvertently facilitated
 by asking the subjects to fixate on the projection screen
 in order  to avoid unnecessary eye movements. The
 auditory-letters task did not lend itself as readily  to
 such usage of extra-task cues, partly because the spok-
 en letters wer6 characterized by asynchronous proper-
 ties that are intrinsic to human speech. In the present
 task, the automated presentation of  the extraneous
 visual letters produced a predictability that was lacking
 for the auditory letters.

     The foregoing analysis suggests that in the visual
 letters-no recall task self-instructed sets to process let-
 ters occurring within  the  S1-S2 interval may have
 made a distraction task out of what was intended  to
 be a nondistracting one. If so, the question arises  as
 to why the use  of letters as facilitating cues did not
 result in faster reaction times in the letters-no recall
 condition (239 msec) compared  to the control condi-
 tion (234 msec). One possibility is that whatever facil-
 itation in reaction time resulted  from the usefulness
 of extraneous letters as cues was offset by the disrup-
 tion of a unified  attention  set to S2. In addition, it
 may not be possible to improve upon reaction time in
 the control condition, where an individual may already
 be at an irreducible physiological limit. In any case,
 the fact that a decrease in CNV amplitude occurred in
 the letters-no recall condition without changes in heart
 rate and  eyeblink  rate suggests that part of the reduc-
 tion in CNV magnitude occurring in both visual and
 auditory letters-recall tasks may be reflecting interfer-
 ence with a cognitive process independent of changes
 in psychophysiological arousal level. Furthermore,  to
 the extent that  reaction  time reflects  a process  of
 motor readiness, the reduction of CNV amplitude  in
 the  letters-no recall  condition   without associated
changes in reaction time suggests that part of the
complex called "the CNV" is independent of motor
 preparation (Tecce 1972).

    Despite the unexpected reduction of CNV ampli-
tude in the letters-no recall condition  (where distrac-
tion should  have been minimal  or nonexistent since
letters could be ignored), there was significantly great-
er CNV reduction  during the letters-recall task (where

-------
Visual Distraction and CNV Shape
                                             363
recall of letters was intended to heighten distraction).
We interpret the greater disruption of CNV develop-
ment in the short-term memory task (letters-recall) as
due  to greater  distraction produced by the complex
cognitive activity required in processing the substan-
tive  properties (meaning) of the extraneous letters,
i.e.,  remembering the letters themselves, in contrast
to the simple cognitive functioning involved in proces-
sing the  formal properties of the letters (as cues or
events to facilitate response to S2) as in the letters-no
recall task. By similar logic, since psychophysiological
arousal levels were selectively increased in the letters-
recall task  but  not in the  letters-no recall task, it ap-
pears that  a necessary condition for the joint occur-
rence of distraction and arousal changes (distraction-
arousal coupling) is the kind of higher level brain func-
tioning involved in processing information  of a sub-
stantive nature, such  as the lexical  content of letters
in the present study.
     The pattern of lowered CNV amplitudes  and
slower RT shown by Type A compared to Type B sub-
jects in control conditions is interpreted as reflecting
a chronic disruption in attention functions for these
individuals, possibly  from internal distractions (self-
instructed  strategies to facilitate response speed, co-
vert  verbali/.ations about  performance success,  and
the  like). The possibility  that the determination of
Type A shape (less than 6/jV for  d  in  the relative
difference  in  voltage values between points H and L
in Fig.  1) was associated with CNV amplitude values
of less than 6fiV appears unlikely in view of the com-
parable proportion of Type A and Type B individuals
having  Cz  amplitude values above and below 6  |iV.
The  pattern  of  relatively greater reduction in CNV
amplitude  by Type B subjects in the letters-recall  task
suggests that they may be more reactive to external dis-
traction (letters) than Type A subjects. If these inter-
pretations have any merit, it is  possible that Type A
subjects are more chronically distracted by internal
sources and Type B subjects by external factors. The
possibility  that chronicity-reactivity differences  in
distraction characterize individuals classified by CNV
shapes having fast and slow rise time deserves further
study.

Summary

    Twenty-four normal volunteers  were tested  in
three conditions: (1) a constant-foreperiod simple re-
action time task consisting of a  flash-tone-key press
sequence (control), (2) the same task with the addition
of a short-term memory  task consisting of four visual
letters presented within  the flash-tone interval with
the requirement  that they be repeated after the key
press to tone (letters-recall), and  (3) the presentation
of letters without the requirement of recall (letters-no
recall). CNV amplitude was reduced in both the letter
recall  and no-recall  conditions compared to control.
Reaction  time increased  only  in the recall condition.
Decreased CNV amplitude  associated with lengthened
RT was interpreted  as a CNV distraction effect. Indi-
viduals with a fast CNV rise time (Type A) showed a
pattern of lowered  CNV amplitudes and lengthened
RTs and appear to be chronically distracted. Subjects
with a slow CNV rise time (Type B) showed a trend
toward reduced CNV amplitudes  during the letters-re-
call task and appear to be  reactive to external distrac-
tion. The pattern of CNV  amplitude, heart rate, eye-
blink, and RT seems to provide a useful index of psy-
chological disturbance.

 Acknowledgments

     The assistance of Debbie Meinbresse, Jean Nigro,
 Gene Y. Chen, Mary Beth  Boehner, Linda Fuss, Baiba
 Liepins, and Andrea Schoening is gratefully acknow-
 ledged.

-------
 PERSONALITY TRAITS AND ELECTRO-
 PHYSIOLOGICAL  FACTORS  DURING SENSORY
 CONDITIONING  IN NORMAL AND PSYCHIATRIC
 POPULATIONS


 N. BRUNEAU, P. DUBOST, P. JUSSEAUME, F. LAFFONT, AND G. LELORD
 School of Medicine, Tours, France
    Sensory evoked (EP)and slow potentials (SP)may
 vary with individual personality traits (e.g., Callaway
 1975), Shagass (1972) reported that somatosensory
 EP amplitude is  inversely related to extraversion
 scores on the Maudsley Personality Inventory. Knott
 and Irwin (1968) found that stress and constitution-
 al anxiety in combination depress CNV amplitude.
 Buchsbaum  and  Pfefferbaum  (1971) distinguished
 two types of subjects  on the  basis of response to
 photic stimulation: visual EPs  either increased (aug-
 menter) or decreased (reducer) proportional to stimu-
 lus Intensity. Tecce and Cole (1974) also observed di-
 vergent electrophysiological patterns associated with
 behavioral alertness or  drowsiness  in two groups of
 normal adults following amphetamine administration.

    Dongier (1973a)  suggested that  contradictory
 report* of the relationship of aberrant  CNV patterns
 and clinical diagnosis may be a consequence of Impre-
cise diagnostic categories. In order to avoid this noso-
logical pitfall, the present itudy was designed to assess
 the relationship of personality traits and electrophysi-
ologlcal patterns irrespective of psychiatric diagnosis.
A sensory conditioning paradigm (Lelord etal. 1958,
Beglelter and Pletz 1969, Laffontetal. 1972) was used
to minimize intersubject variability in normal  and
psychiatric populations.

Method

 Subjects

   Forty-one normal  (females-28, average «ge-23
 years) and 93 psychiatric patients(neurotic-55, piy-
 chopsthic-19, pjychotic-13, unolmifted-8, females
 -90, and average age-30 years) were examined, Psy-
 chiatric subjects were  outpitlants or confined briefly
 in a piyohUtrlo ward. All subjects were intelligent and
 cooperative enough to complete a personality question*
 naire.
Questionnaire

    Fifteen questions were selected from a 52-item
personality inventory previously administered to 144
subjects (including 86 psychiatric patients).  These
questions represented 15 personality  traits derived
from a principal component factor analysis.  Three
choices were possible for each question: 0=no re-
sponse; l»false;2=true.

Neurobehavioral procedures

    Subjects  were isolated  in a dark, soundproof
room and sat in a comfortable armchair. Electrodes
were placed at the vertex and occipital region referred
to the left earlobe.  EEC was recorded with ac am-
plifiers  set at  a bandpass down 3 dB between 0.1 and
100 Hz (cf. Lelord et al. 1976). Vertical and hori-
zontal EOGs were also recorded.

    Auditory and visual stimuli described by Laffont
et al, (this volume) were used for sensory condition-
ing. Subjects completed two 1-hour sessions consisting
of 10 series of 20 trials presented at random intervals
varying  from 4 to 30 sec. The auditory stimulus was
presented alone during the initial two aeries of the first
session to establish unpaired  control values, A visual
stimulus was presented 700 msec after the auditory
itlmului during conditioning (eight series per session).
The auditory stimulus was again presented alone dur-
ing the  final two series of the second session to study
extinction of the conditioned response.

Data analysis

    Signal averages were constructed for each series
of 20 trials, Records were discarded If BOG deflec-
tion exceeded 8 \N, Eleetrophysiologlcal and clinical
dsts were  then factor  analyzed  as  described by
Uffont et al, (this volume),

-------
ERPs and Sensory Conditioning
                                                                                  365
                                                      SOUND ALONE
                                             VERTEX
                                                       OCCIPITAL REGION
                                                    CONDITIONED SOUND
      OCCIPITAL REGION
       |FLASH

      J CLICK
10/uVl
VERTEX
                    100 msec
                                                                       OCCIPITAL REGION
Fig. 1. Rhythmical potential conditioning. Each tracing represents average of responses to 20 stimuli. Left: visual
evoked potential. Right: auditory evoked potential before conditioning (upper traces) and  during conditioning
flower traces).
Results

Electrophysiological data

    Fig. 1 to 3  illustrate three basic electrophysiolog-
ical patterns observed during sound-light (SL) condi-
tioning relative to sound alone (S) control waveforms.
These patterns may be characterized as:
   1.  Rhythmic  potentials associated with increased
      amplitude  of the auditory evoke'd potentials
      (AEP) in the occipital region (Fig. 1).
   2.  Increased AEP amplitude at the vertex (Fig. 2).
   3.  Slow potentials (SP) at the vertex and/or occipi-
      tal region (Fig. 3).


     Evoked or rhythmical potentials within 400 msec
 of stimulus onset were counted if peak-to-peak ampli-
 tide  was > 4 u V. Slow potentials, measured  relative
 to a  100-msec poststimulus baseline, were counted if
 the  amplitude was  > 8 uV. The frequency (n) of
 occurrence of each pattern within the 16 conditioning
 series was classified as: absent (n < 2); few (2 < n < 6);
 many (7 < n <  11); or great  many (n >  12). AEP
 amplitude was  characterized as large if  it exceeded
 control amplitude during either conditioning  session,
 or small if it equaled control amplitude.

      Visual evoked  potentials (VEP) were classified
 separately as moderately regular (6 < n < 12) or very
                                      regular (n > 12) with matt (< 19 uV) or large O 19
                                      MV) amplitude. VEPs were  sometimes followed by
                                      rhythmic aftereffects. "Conditioning to time," defined
                                      as a response to "the absence of light," was sometimes
                                      observed during extinction series.

                                      Factor  analysis

                                          Fig. 4  illustrates the  results  of factor analysis.
                                      Four factors emerged, represented as horizontal bands
                                      with  characteristic  electrophysiological  (left)  and
                                      psychological (right) features. Factors I and IV clearly
                                      reflected opposite extremes in terms of both electro-
                                      physiological measures and  personality traits. These
                                      factors may be  summarized as follows:
                                         I.  Trust and tolerance associated with rhythmic
                                            conditioning  predominant occipitally without
                                            SPs.
                                         II  Reflection  and stability associated with large
                                            conditioned EPs.
                                        III.  Restlessness  associated  with few conditoned
                                            EPs.
                                        IV.  Distrust and intolerance associated with local-
                                            ized  or generalized SPs without rhythmic  or
                                            conditioned EPs.
                                       Discussion

                                           Factor I was characterized by conditioned rhy-
                                       thmic and evoked potentials in the occipital region.

-------
  366
                                                                        Bruneau et al.
           OCCIPITAL REGION

          (FLASH
         {CLICK
                 10/uV
                     100msec
                                               SOUND ALONE
                                      VERTEX               OCCIPITAL REGION


                                              CONDITIONED SOUND
                                      VERTEX
OCCIPITAL REGION
                    Fig. 2. Evoked potential conditioning. (See legend in Fig. 1.)

        LIGHT                     i            SOUND ALONE
                                  IL
                                      VERTEX                OCCIPITAL REGION

                                             CONDITIONED SOUND
VERTEX
                 200msec

                    Fig. 3. Slow potential conditioning. (See legend In Fig. 1.)
                                                             OCCIPITAL REGION

-------
ERPs and Sensory Conditioning
                                            367
           «2
           x
                        ELECTROPHYSIOLOGICAL FACTORS

                                                           AREA!

                     Many conditioned evoked responses to S of coupled SL,

                                  > of rhythmical conditioned potentials,
                     Predominance )
                     Absence of conditioned slow potentials,
                     Presence of auditory evoked potentials only at
                     the vertex before SL,
                     Presence of conditioned evoked potentials on one
                     tricing efter SL,
                   -  Moderately regular)
                                       visual evoked potentials,
                                                           AREA II

                   - Large conditioned evoked potentials at the vertex
                      end/or on the occiput to S of coupled SL;
                      greater in the first or equal in both sessions.
                   - Conditioned evoked potential greater than visual
                      evoked potentiil.
                   - Presence of evoked potential conditioned to time on
                      one or both tracings after SL.
                   - Visuel evoked potentials greater during the second
                      session; followed by rhythmic after effects.
                   - Greeter during the second session.
                                                           AREA III

                       Few conditioned responses to S of coupled SL.
                       Presence of conditioned slow potentials.
                       Predominance of conditioned evoked potentials.
                       Small conditioned evoked potentials.
                       Conditioned evoked potential smaller than visual
                       evoked potential.
                       Pretence of auditory evoked potential on both the
                       vortex and the occiput before SL.
                       Presence of conditioned evoked potentials on both
                       tracings after SL.
                       Presence of slow potentiils conditioned to time
                       after SL.
                       Absence of evoked potential conditioned to time
                       after SL.
                       Very regular visual evoked potentials; not followed
                       by rhythmic efter effects.
                                                            AREA IV
                       Many conditioned slow potentials generalized or
                       or locelized at the vertex or on the occiput to S
                       of coupled SL.
                       Greater in the second session.
                       Absence of rhythmical conditioned potentiel.
                       Absence of conditioned evoked potential.
                       Absence of conditioned evoked potentiel on both
                       tracings after SL.
                       Large visual evoked potentials greater during the
                       first session.
    PSYCHOLOGICAL TRAITS
   Trusting,
   Tolerant to ambiguity.
   Absence of temper tantrums.
   Not easily impressed by authority.
   Not overly meticulous,
   Susceptible to past events.
-  Lawebiding,
-  Stable,

-  Slow,
-  Optimistic.

-  Not whimsical.

-  Thinking few people have saxuel
   preocupations.
-  Having tolerant pirents.
    Restless.
    Quick.
    Not optimistic.
    Tolerant to law.
    Thinking many people have sexual
    preoccupitions.
    Having intolerant parents.
 -  Distrustful.
 -  Intolerant to ambiguity.
 -  Presence of many temper tantrums.
 -  Easily impressed by authority.
 -  Overly meticulous.
 -  Not susceptible to past events.
 -  Whimsical.
                                                                      AXIS 1                                   •>•
  Fig. 4.  Factorial analysis of relationships. Each electrophysiological factor was situated in relation to psychological
  traits. Each psychological trait was situated in  relation to electrophysiological factors.  The two spaces were pro-
  jected on a two-dimensional space. A progression for electrophysiological factors as well as for psychological traits
  was seen along Axis 2.

-------
 368
                                   Bruneau et al.
This electrophysiological pattern implies conditioning
with early latent responses and complex morphology
in primary sensory cortex corresponding to the uncon-
ditioned stimulus (Lelord and Maho 1969). The elec-
trophysiological-psychological structure of Factor I is
reminiscent of the concept of "epicritic" sensitivity
proposed by Head (1920). The traits of emotional
between the psychological and electrophysiologjcal
domains. Laffont et al. (this volume) illustrate the use
of this  technique  in the study of developmental dis-
stability and confidence are consistent with  the pleas-
ant sensibility of the epicritic mode.
    Conversely, Factor IV reflected a different type
of conditioning characterized by the absence of rhy-
thmic or early evoked potentials, but the presence of
diffuse late slow waves at the vertex and occipital area.
This pattern represents global facilitation correspond-
ing to the "protopathic" mode of Head. The traits of
emotionality, instability, anxiety, and intolerance are
also consistent with the unpleasant sensations  attri-
buted to protopathic function.
    Although the present study is preliminary, the
results provide encouraging evidence of a systematic
relationship between personality trait structure and
characteristic  electrophysiological patterns. Further-
more, factor analytic techniques appear to provide an
effective method to elaborate the complex relationship
orders. These findings suggest that the search for elec-
trophysiological  correlates of psychopathology may
be more fruitful if investigators focus on discreet mea-
sures of personality traits rather than diffuse, ill-defin-
ed diagnostic categories.

 Summary

     The relationship between evoked potentials and
 personality traits was examined  in 95 psychiatric and
 41 control subjects during sensory conditioning. Two
 major  electrophysiological groups were distinguished
 by localized conditioned evoked potentials and general-
 ized slow  conditioned  potentials.  Epicritic evoked
 potential  patterns  were associated with emotional
 control,  while  protopathic slow  potential patterns
 were associated with emotional prevalence.

-------
CONTINGENT NEGATIVE  VARIATION IN
PATIENTS AFFECTED BY SPECIFIC PHOBIAS
 H. BARBAS, L. SOLYOM, AND B. DUBROVSKY

 Departments of Psychiatry and Physiology, McGill University, Montreal, Quebec, Canada
    The essential clinical feature of phobic neurosis is
 the anticipatory fear of certain objects and situations.
 The anticipatory component of the phobia seriously
 disrupts the patient's life (Marks 1969).  Research
 has focused on experimentally created anxiety rather
 than actual phobic states observed in patient popula-
 tions. Measures of experimentally induced fear  in-
 clude: skin conductance (Ladar 1967, Wilson  1967),
 forearm blood flow (Kelly  and Walter 1968), bio-
 chemical parameters (Marks  1969), and muscle activ-
 ity changes (Malmo and Davis 1951).

     Reduced CNV amplitude has been reported in
 anxiety  situations  (Knott and  Irwin 1968,  1973;
 McCallum and Walter  1968). Knott and Irwin (1968)
 examined  CNV  and galvanic skin response (GSR) in
 subjects divided into high and low anxiety groups on
 the basis of scores on the Bendig Scale of Emotional-
 ity. No differences were observed between  groups
 under normal conditions, but  smaller  CNV  ampli-
 tudes were observed in the high anxiety group than in
 the  low  anxiety group under stressful conditions.
 GSR did not vary between  groups; however, using a
 clinical  model,  where response  differences between
 patients and normal controls or between responses to
 different stimuli in the same patient were  evaluated,
 Wilson (1967) was able to show differences in GSR
 between phobic and nonphoblc subjects.


     The seeming discrepancy in results suggests that
 reactions of normal subjects to experimentally pro-
 duced fearful situations may not be comparable to
 clinical  conditions  such as phobic neuroses. These
 results also point to the need to distinguish between
 the emotional experience of anticipatory anxiety and
 thote lomatic, visceral, and cognitive reactions to fear
 that are moat likely  related to different  aspects of
 central nervous system activity.

     CNV  studies  employing prolonged foreperiodi
 (4 or more sec) have demonstrated that different pro-
 cesses  contribute  to  the development  of the CNV
 (Weerti and Lang 1973, Klorman and Bentaen 1975,
Loveless and Sanford 197S, Rohrbaugh et al. 1976).
Weerts and Lang (1973) suggested that anticipation
of the imperative stimulus is important in generation
of CNVs. Weinberg (1975)has argued that CNV devel-
opment corresponds  to  the expectancy  to receive
rather than respond to information.

    Anticipatory processes appear to play a central
role in the etiology of phobic neuroses as well as the
CNV. The CNV, therefore, should provide a sensitive
index of the electrocortical effects of phobic neuroses
as well as provide a prognostic index of therapeutic
efficacy. In  the present study, nondisturbing  and
phobogenic stimuli were used  to study the CNV in
phobic patients prior to and following desensitization
therapy.

Method

    Subjects were 14 female out-patients ranging in
age from 21 to SO yean and suffering from phobias
concerning dogs, spiders, bees, snakes, birds, frogs,
and dolls. Subjects had an average phobic anxiety
score of 2.8 and average avoidance score  of 3.0 on a
04 scale questionnaire (Solyom et al. 1973). Average
scores on the IP AT Anxiety Scale (Cattell 1963) and
 Fear  Survey Schedule (FSS) Wolpe and  Lang 1964)
 were  4.9 (normal  is between 5 and 7)  and  53.6
 (normal is approximately SO), respectively.

     All subject! participated in two recording sessions
 a week apart. The first allowed  them to become
 acquainted with the procedure and sequences of the
 test.   The second  provided the data to be used in
 analysis.

     The experimental paradigm uied to elicit the CNV
 coniiited of a warning click (SI) followed after l.S MC
 by a illde-projectod picture (82), which the lubject
 terminated by pressing a button. Three seriis of stim-
 ulus  pain were presented. A nondisturbing picture
 waa used ai the  imperative stimulus in the first and
 third series, and a picture of the feared object served
 as 52 In the middle section. ISI varied from 20 to 60

-------
 370
                                    Barbas et al.
 sec. Ten sweeps free of eye or skin potential contami-
 nation were summated for  each series. The ratio of
 sweeps accepted varied in the population studied from
 half to one quarter of the trials presented.

     Recording methods have been detailed elsewhere
 (Dubrovsky et al. 1973). Briefly, EEC was recorded
 at Cz using nonpolarizable Ag/AgCI electrodes referred
 to linked earlobes. The EEC preamplifier was modified
 to provide a 9.5-sec time constant with a 75-Hz upper
 frequency limit. Eye movements (EOG) were monitor-
 ed  with a supraorbital electrode referred to linked
 earlobes. Skin potentials were monitored by recording
 bipolarly between  the two mastoids. EEC, EOG, and
 skin potentials were summated on line with a Mneme-
 tron CAT 400A.

    CNV amplitude was measured as the largest neg-
 ative peak during the S1-S2 interval relative to a pre-Sl
 baseline. The baseline was set as the mean amplitude
 of a 1.5-sec epoch  preceding SI. Duration of postim-
 perative negative variation (PINV) was measured as
 the time between S2 and return of negativity to the
baseline (Dongier 1973b). T-tests for dependent sam-
ples were used for statistical comparisons.

Results

    Representative CNV patterns observed in patients
 in anticipation  of  disturbing and phobogenic stimuli


           NONDISTURBING
                                          CNV
        S1     S2
                                           EOG
                                           Itec
 are shown in Fig.  I and 2. RT was shorter [t (13) =
 2.92, p < .02] while CNV amplitude [ t (13) = 6.48,
 p < .01]  and PINV duration [ t (13) = 2.92, p < .02]'
 were greater for disturbing than nondisturbing stimuli
 (Table 1). No significant  differences  were noted
 between  series  one and  three when nondisturbing
 stimuli were used.

     Following the training and recording sessions, a
 group of the patients underwent deserialization thera-
 py until symptomatic improvement was observed. Six
 subjects, who successfully concluded desensitization
 treatment, were tested again with nondisturbing and
 with phobogenic stimuli. CNV averages were obtained
 again for  the two conditions: Fig. 2 (lower traces)
 shows CNVs obtained in the posttherapy session. In
 all six cases, no differences were observed in anticipa-
 tion of nondisturbing compared to previously phobo-
 genic stimuli. Two other subjects who had successfully
 completed their desensitization therapy were not avail-
 able for posttherapy testing.  Six subjects did  not
 complete therapy.

    Table 2 presents the pre- and  posttherapy scores
 for the subgroup of treated subjects. Statistical analy-
sis showed that following desensitization, CNV ampli-
tude and PINV duration for phobogenic stimuli were
reduced in comparison to pretherapy levels [ t (5) =
5.2, p < .01 for amplitude; t(5)=2.12,p <  .05 for
duration].
                   PHOBOGENIC
                                       I
10/uV
 sentation o
                                                                                             w™ Pre.

-------
CNV and Phobias
                                         371
                                        PRETHERAPY
         NONDISTURBING
            PHOBOGENIC
                                        CNV
        SI    82
                                          Isec
        SI     S2
                                    POSTTHERAPY
        SI    S2
                                                           S1    S2
 Fig. 2.  Comparison of CNV samples in response to nondisturbing and phobogenic stimuli obtained from patients
 affected by specific phobias before and after treatment.
            Table 1. Mean Value* of
         CNV Amplitude, Duration and
                Reaction Time

Nondisturbing
stimuli
Phobogenic
stimuli
Amplitude,
MV
11.3
20.4b
Duration of
negativity
after S2,
msec
1900
2300a
Reaction
time
after S3,
msec
907.28
711.0
•t-tert wultt: N-14,df-13,p < .02, two-tailed.
bp<.01.
    Table 2. Mean Values of CNV Amplitude
   Duration, and Reaction Time before and after
           Desensitization Therapy

Nondisturbing stimuli
Amplitude, juV
Duration, msec
Reaction time, msec
Phobogenic stimuli
Amplitude, y. V
Duration, msec
Reaction time, msec
Before
therapy

17
1966
10438

28a
2925b
735
After
therapy

16
1833
850

16
1858
977
at-tastresultt:N-6,df-5,p < .01, one-tailed.
bp<.06

-------
 372
                                    Barbas et al.
     Behavioral and psychological test results also indi-
 cated a normalization of scores following therapy. RT
 increased slightly to phobogenic stimuli while the rela-
 tively long pretherapy  RTs to nondisturbing stimuli
 were reduced [t (5) = 3.7, p < .01],  Statistical com-
 parison  of pre- and posttherapy psychological tests
 showed  scores were reduced to normal values after
 completion of therapy  [t (5) = 6.34, p < .01 for pho-
 bic anxiety; t (5)= 5.41, p < .01 for avoidance ;t (5) =
 2.5, p < .05 for PAT (STEN); and t (5) = 2.5,p< .05
 for FSS].

 Discussion

     The results demonstrate that CNVs from phobic
 patients are significantly larger and PINVs more pro-
 longed when imperative stimuli depict phobogenic
 rather than nondisturbing stimuli. These differences
 may  be  attributed to the presumed  importance of
 anticipatory processes in the etiology of phobic neuro-
 ses and the CNV.

    Sharply rising high-amplitude negative shifts and
 prolonged  PINVs  following the warning stimuli were
 observed in phobic patients (cf. Fig. 1) when the para-
 digm  contained phobogenic stimuli. According to
 Weerts and Lang  (1973) and Loveless and Sanford
 (1975), the early negative phase of the CNV is related
 to signal stimulus  orientation although the results of
 Rohrbaugh et al. (1976) are not consistent with this
 hypothesis. Loveless  and Sanford suggest that  the
 warning signal affects the gain of the threshold mech-
 anism rather than its  trigger level, i.e., the warning
 stimulus modulates the effective intensity of the  im-
 perative  stimulus.   These authors  also showed that
 warning signal  intensity  affects the response charac-
 teristics for at least 8 sec after imperative stimulus
 onset.  Perhaps  the  conditioned  emotional signifi-
 cance  of a stimulus modulates response threshold in
 a manner similar to physical  intensity, an hypothesis
 that could account for the  increased  amplitude and
 prolonged  duration  of PINV observed  in  patients
 when the imperative stimulus was phobogenic.

    The results may also be related to the compulsive
 search of phobic patients for the feared object (Marks
 1969). A signal warning of the temporal proximity of
 the feared  stimulus should produce a  large orienting
response and hence an orienting cerebral wave of large
amplitude  and  duration.  Similarly, the ittentlonal
proceii "which facilltatei the selection of relevant
stimuli from the environment" (Tecce 1972) should
 be enhanced while the patient searches for the feared
 object.  According to Tecce's hypothesis, attention is
 positively and monotonically related to CNV ampli-
 tude. Differential CNV amplitudes and reaction times
 in the phobogenic versus nondisturbing conditions are
 consistent with this hypothesis.
     Patients with specific phobias showed low levels
 of general anxiety consonant with theMaudsley study
 (Marks 1969). CNVs of this population were larger in
 amplitude  when phobogenic  stimuli  compared  to
 nondisturbing stimuli were used. These  results and
 other reports of low CNV amplitudes in high anxiety
 subjects reveal that  CNV amplitude can provide  an
 objective indication of specific behavioral states, such
 as fear and anxiety. Furthermore, these electrophysio-
 logical results support the clinical view that fear, anxi-
 ety, and  phobia encompass distinctive phenomena
 (Marks 1969).

    Finally,  CNV differences observed with phobo-
genic and nondisturbing stimuli disappeared in  six
patients after successful desensitization therapy, while
responses  to nondisturbing stimuli remained stable.
These results suggest that changes in brain processes
 associated with phobic neuroses are  reversible.  Avail-
 able data do not Indicate whether amplitude and dura-
 tion differences with phobogenic and nondisturbing
 stimuli involve  changes  in magnitude in the same
 CNV generators, or whether phobogenic stimuli acti-
 vate other central nervous system areas specifically
 related to the pathological phobic condition.
    In conclusion, we propose that sequential CNV
evaluation during and after treatment may be used at
an objective indicator of the treatment course in speci-
fic  phobic  neurosis. By comparing behavioral normi
of improvement with CNV recordings, a better predic-
tion of relapses may become possible, which may allow
for the  development  of appropriate prophylactic
measures, an important goal in psychiatric  manage-
ment.

Acknowledgments

    We  thank Mr. Joel Tepper for excellent technical
support and  Dr. Maurice  Dongjer for reading  the
mamucript. Thii work wu supported by a grant from
Le  programme F.C.A.C.,  Minister* de 1'education,
Gouvememont du Quebec.

-------
RELIABILITY OF CONTINGENT NEGATIVE
VARIATION IN PSYCHOPATHOLOGY
M. TIMSIT-BERTHIER, J. DELAUNOY, B. XHENSEVEL, AND M, TIMSIT

Laboratory of Clinical Neurophysiology, Department of Medical Psychology and Psychosomatic
Medicine, University of Liege, Liege, Belgium
    The reliability of a clinical testing procedure is an
important point  in determining its usefulness. With
the CNV, however, the measurement of reliability in
normal subjects  has  raised problems. While  Cohen
(1969) and Roth et al. (1975) obtained a high correla-
tion between two CNV recordings, Straumanis et al.
(1969) reported  a low correlation. These differences
might result from two important independent varia-
bles in test-retest studies: (1)  the number of trials
selected for averaging by different investigators (10 to
36) and (2) the time interval between successive record-
ing sessions (5 minutes to 8 days). In psychopathology,
a third variable is added: the possible change of the
mental state of the patients. The present study was
undertaken to determine the reliability of CNV record-
ing across two different test-retest intervals (1  month
and 1 year) and with two different-but complement-
ary-psychopathological conditions (with improve-
ment or aggravation of mental state).

Methods

    One-hundred-six male subjects were selected on
the basis of psychopathological history and psychiatric
interviews from  the 300 patients of the psychiatric
hospital at LaVoliere, Liege (mean age, 32 years;S.D.,
4.52). The distribution of patients according to classi-
cal diagnostic categories is shown in Table  1. The
frequency  distribution is broken down in terms of
patients who showed no change (N=76), who improved
(N=24), and who deteriorated (N=6) during the test-
retest period.

   Two selection criteria were used: (1) patient co-
operation in EEC  and CNV recorded  and  (2) the
presence of the behavorial  disorder for  at least  1
year. All patients received  psychotropic medicat-
ions.

    In order to study CNV reliability among patients
with stable behavior disorders, two recordings were
made for each of 76 chronic patients. The test-retest
interval was 1  to 2 months for half the patients and 1
to 2 years for the  other half. No important clinical
changes were reported in any of these subjects during
an observation period of  2  weeks before and after
recording sessions.

    In order to study CNV reliability among patients
who showed important changes in clinical state, CNVs
were recorded from 30 chronic patients using a test-
retest interval of 1 to 2 years: 24 displayed clear sym-
tomatic improvement; 6  showed an aggravation of
mental state.

    EEC was recorded at Cz  from Ag/AgCl electrodes
referred to linked  earlobes. EOG was recorded from
     Table 1. Frequency Distribution of Patients According to Classical Diagnostic Categories
                         and Clinical Change During Test-Retest Period

Hebephrenic schizophrenics
Paranoid schizophrenics
Paranoics
Manic-depressives
Personality disorders
Obsessional neurotics
Totals
No change
21
27
7
9
5
7
76
Improved
1
18
0
5
0
0
24
Deteriorated
3
0
0
2
0
1
6

-------
  374
     Table 2. Frequency Distribution of CNV
    Resolution Modes for Test-Retest Interval
         of 1-2 Years for 38 Patients with
              Stable Symptomology8
Second recording
Type I CNVs
Type II CNVs
Type III & IV or
flat CNVs
First recording
Type I
CNVs
6
2
2
Type II
CNVs
1
0
1
Type III & IV
or flat CNVs
2
0
24
  8Test-retest correlation (Goodman-Kruskal Index) - 0382

    Table 3. Frequency Distribution of CNV
    Resolution Modes for Test-Retest Interval
    of 1-2 Months for 38 Patients with Stable
 	Symptomology3	
                         First recording
Second recording
Type 1 CNV*
Type II CNVs
Type III & IV
or flat CNVs
Typel
CNVs
7
2
1
Type II
CNVs
0
2
0
Type III & IV
or flat CNVs
4
6
16
 aTeit-rete»t correlation (Goodman-Kruskal Index) =0.784.

 above  and below  the  right eye.  Both signals were
 recorded with a time constant of  11 sec. Calibration
 signals of 20 juV were  recorded in series  with CNV
 and EOG leads on each trial. The data were stored on
 analog tape  for off-line  summation. Subjects were
 required  to  keep eyes  closed during testing. Trials
 containing EOG artifact were rejected from averages.

    The paradigm used  for testing was a warning click
 (SI) followed after 1.5  sec by a series of light flashes
 (S2), which the subject terminated by pushing a but-
 ton. The CNV was obtained by summating 20 trials.


    Three CNV parameters were  measured: pre-S2
 amplitude,  post -S2  duration, and resolution mode.
 Amplitude was  defined  as the difference in  mean volt-
 age between  a 200-msec pre-S2  epoch and a 1-sec
pre-Sl  baseline epoch. Post-82 duration was defined
as the  delay,  in  seconds, between  S2 and  the return
to baseline of the CNV. In normal curves this delay is
about 0.4 sec; in some  patients, it ranges between 1
                          Timset-Berthier et al.

    Table 4. Frequency Distribution of CNV
    Resolution Modes of 30 Patients Showing
     Distinct Symptom Change During the
    1- to 2-Year Interval Between Recording
                    Sessions8
Second recording
Type | CNVs
Type II CNVs
Type III & IV
or flat CNVs
First recording
Type I
CNVs
1b
1d
3d
Type II
CNVs
2c
0
0
Type III & IV
or flat CNVs
15b
7"
1d
  T«tt-rete*t correlation (Goodman-Kru»kal Index) - -0,50
  Improved.
c One Improved; one deteriorated.
  Deteriorated.
 and 4 sec. Resolution mode is judged in relation to
 the four types of CNV identified previously (Timsit-
 Berthier et al. 1973): Type I-CNV returns to baseline
 immediately  after S2;  Type  H-baseline  return is
 delayed,  although  the  drop  begins  immediately
 after  S2;  Type III- baseline return  is delayed with
 no immediate drop after S2, i.e., the CNV remains at
 creases in  amplitude after S2, which  yeilds a charact-
 eristic dome-shaped curve.

     The  Bravais-Pearson Index (Kendall and Stuart
 1967) was used  to calculate test-retest correlations
 for amplitude and duration measures. The Goodman-
 Kruskal Index was used to correlate qualitative ordinal
 test-retest measures of CNV resolution.

 Results

    In the stable patient population, significant corre-
 lations (p<.05) for amplitude measures were obtained
 for test- retest intervals of 1  to 2 months (p=0.55).
 Test-retest correlations for the duration measure were
 significant only for the  1-to 2-year interval (p=0.44).
The highest test-retest correlations were obtained for
 resolution mode ratings. The Goodman-Kruskal Index
 was 0.88 for the 1- to 2-year interval  and 0.78 for the
 shorter interval., The frequency distributions of reso-
lution modes obtained  for the longer and shorter
intervals are shown in Tables 2 and 3, respectively.

    Table 4 shows the frequency distribution of reso-
 lution modes in 24 subjects exhibiting symptomatic
 improvements and in 6 subjects (2 manic-depressives
 3  acute schizophrenics, and 1 neurotic who had a
 hallucinatory  episode) who deteriorated. These two
 populations were  combined in order to test for the

-------
Reliability ofCNV
                                             375
effect of absolute change in mental state during the
Ho 2-year  test-retest interval. A significant positive
correlation (p=0.48) was obtained for the amplitude
measure; a significant negative correlation (e=-O.S6)
was obtained for resolution mode;  but the correl-
ation of  the duration measure  (p=0.15) was  not
significant.
Discussion

    The test-retest reliability of three CNV parameters
was examined in psychopathological  patients.  The
most informative and reliable of these  measures was
the qualitative rating of resolution mode. This measure
correlated positively for both short and long intervals
in patients with stable symptomatology, but correlated
negatively in patients who showed a distinct change
in mental  state  during the  1-to 2-year-test-retest
interval.  This parameter  appears to  be a sensitive
index of clinical change  and, therefore, may be of
prognostic value.

Acknowledgment

    This work  was supported  in part by  Grant
20397-FRSM. The authors  are grateful to  Mr. D.
Defays, Statistics Department, Psychology  Institute,
Liege University, for his assistance.

-------
  CNV  IN PATIENTS AFTER PSYCHOSURGERY
  AFFECTING THALAMOCORTICAL PATHWAYS
  TO PREFRONTAL CORTEX1
 R. ZAPPOLI, M. PAPINI, F. DENOTH, 2 A. PASQUINELLI, L. ROSSI, M. G. MARTINELLI
 AND S. GUERRI

 Service of Neurophysiopathology, Institute of Nervous and Mental Diseases, University of
 Florence, Florence, Italy
     This report examines the role of thalamofrontal
 pathways, including the anterior thalamic radiations,
 in the genesis of frontal CNV. The necessity of these
 connections for the production of prefrontal cortical
 negative slow potentials (SPs) elicited in cats during
 behavioral conditioning has been reported by Skinner
 and  Lindsley (1973). Zappoli et al.  (1973, 1976)
 observed normal frontal CNV in two patients treated
 with  bilateral  prefrontal  lobotomy (Freeman  and
 Watts 1939,  1947, 1950, 1966), which severed the
 thalamofrontal pathways. To shed further light on this
 problem, waveforms of frontal and vertex CNVs in
 patients subjected to various psychosurgical proced-
 ures  were  examined. These  tractotomy procedures
 interrupted pathways originating in the dorsomedial
 and anteromedial nuclei of the thalamus projecting to
 various areas of prefrontal cortex.

 Method

   Eight nonschizophrenic patients (seven of whom
were  male), aged 37-74, were selected on the basis of
 fair neuropsychiatric condition and precise knowledge
 of lesions. The location and extent of lesions were de-
 termined by neuroradiological examinations (pneumo-
 encephalography) and in  two cases by  computed
 tomography (Baker 1975, Baker et al. 1975). Psycho-
 surgical history  and present condition of each patient
 are summarized  in Table 1.

   An extensive bilateral  prefrontal lobotomy using
 the classic  "radical" plane  of sectioning was perform-
 ed on six patients (M47_, M50, M53, M55, M56, and
 F74), 14-24 yean ago (X=18 years). This plane passes
just rostral to the lateral ventricle anterior horn, cut-
 ting the greater part of the central white matter of
 the prefrontal areas, and interrupting intercommunica-
1 Supported in part by CNR Grant No. 73-00837-
 04/115-6978.
Jlnititute for Information Processing (IEI) of Na-
 tional Research Council (CNR), P1M, Italy.
 tions  between  thalamic dorsomedial  nuclei  and
 cortical areas 9, 10,  11, 13, 14, and 46 as mapped
 by Brodmann (1909, 1925). In these patients, in-
 sertion  of the leucotome was extended  with differ-
 ent techniques to the midline in an attempt to sever
 the more medial fibers connecting thalamic antero-
 medial  nuclei and  Brodmann's  areas 24 and 32.
 In five patients,  operation resulted in some improve-
 ment of phobias, obsessive thinking, compulsions,
 and  emotional changes.  One patient (F74) did not
 benefit  from the operation. In five Ss there was no
 evidence that psychosurgery impaired I.Q., memory,
 imagination, or  social attitudes, although distracti-
 bility increased noticeably in patients following sur-
 gery. One  patient (M50) exhibited an  intellectual
 deficit (Wechsler-Bellevue AIS I.Q. 85 and M.D.I. 60).

    The seventh  patient (M48), suffering from severe
 anxiety  neurosis  with episodes of intense  agitation
 and depression, was treated by two successive psycho-
 surgical  operations. The  first,  performed 23 years
 previously,  was  a left frontal  lobotomy with sure
 destruction  of the dorsomedial thalamo-prefrontal
 fibers. The incision was on a plane so posterior that it
 produced a  lesion of the lateral ventricle left frontal
 hom (documented with pneumoencephalogram). The
 patient's  psychopathological  condition  remained
 unaltered, and he was treated a year later by a right
 extensive gyrectomy, with removal of the whole area
 9 and upper half of areas 10  and 46 (Fig. 1). No
 neurological or intellectual deficits were apparent.

    The last patient (M37), afflicted with a serious
 painful left lower phantom limb, had been subjected
 2 years earlier to a right prefrontal chemical stereo-
 leucotomy (alcoholization). This procedure generally
 causes well-delimited  extensive  lesions in the white
 matter of the prefrontal lobe  with destruction  of
 fibers originating in the thalamic nuclei, particularly
 in the dorsomedial and anteromedial.  By means of
 computed tomography,  this  patient was found  to
have  a roundish lesion of about 2.5-cm  diameter a
little below and about  1 cm in front of the right later-
al ventricle anterior horn. In view  of its extent and
site, this lesion must certainly have caused interruption

-------
                  Table 1. Psychosurgical Patient History and Present Condition
Men*
Sac
M
M
M
M
M
F
M
M
A*
55
SO
47
56
S3
74
48
37
nm*tr**r*1n^
HMiiiUI condition
Chronic high anxiety,
ulaeaive compulsive
psychoneuroHS
Chrome neurotic
depression with
anxiety and obsessive
tendencies
Chronic high anxiety,
obwiiiiHf romnultivB
psychoneurosis
Chronic moderate
anxiety, obsesshrecomoulsive
psychoneurosis
Chronic high anxiety,
obsessive-compulsive
psychoneuroris
Chronic manic depressive
psychosis with obsessive
Chronic severe anxiety
psychoneurosis whh

Left lower phantom
limb pain
Psychosurgical
procedure
Freeman-Watts radical
bilateral prefrontal
lobotomy; lateral approach
FiiMim •» M/nllc rnrlirnl

Miaterai preTronui
lobotomy; lateral approach
Freeman-Watts radical
bilateral prefrontal
lobotomy; superior approach
Freeman-Watts radical
bilateral prefrontal
lobotomy; superior approach
Freeman-Watts radical
bilateral prefrontal
lobotomy; superior approach
Freeman-Watts radical
bilateral prefrontal
lobotomy; superior approach
1st left radical prefrontal
lobotomy
2nd right gyrectomy of areas
9. 10. and 46.
Right prefrontal chemical
stereoJeucotomy
Years since
operation
24
22
18
15
18
14
23
22
2
Present
condition
Little improvement;
no intellectual deficit
Little improvement;
moderate intellectual
deficit
Significant improvement;
no intellectual deficit
Little improvement;
no intellectual deficit
Little improvement;
no intellectual deficit
No improvement;
no intellectual deficit
Little improvement ;
no intellectual deficit
Transitory significant
improvement;
no intellectual deficit
Vesent anxiety
level - STEN
score"
9
6
7
5
9
8
8
7
•iPAT -wietv
R.B. Catt* «1963».

-------
378
                                  Zappoli et al.
of pathways originating in the dorsomedial nucleus
and projecting to right  frontal areas 9, 10, 11, 13,
14, and 46.

    All patients were capable of performing a typical
CNV task, which was in this case a click warning stim-
ulus (SI) followed after I or 1.5 sec by a tone imper-
ative stimulus (S2)  of moderate intensity. A button
press, which S was instructed to make as fast as possi-
ble, terminated the tone (operant response). Eighty
to 100 paired click-tone trials were presented, with a
variable intertrial interval (ITI) ranging  from 10  to
60 sec.

      B.F. J 48
    Ag/AgCl  recording electrodes were applied at Cz,
Fp2, Fpl,  F4, F3, F2, and  Fl;  the  last  two were
centered equidistantly between, respectively, Fp2, F4
and Fpl, F3. Two methods of reference were used in
a session:  (1) Cz to linked earlobes and  frontal  elec-
trodes to ipsilateral mastoid and (2) all electrodes to
linked  mastoids.  Also recorded were  vertical EOG,
finger  photoplethysmograrn,  the  EMC  of forearm
flexor  muscles  involved in  the  operant  response,
stimulus markers and  reaction times (RT) to S2.  EEC
and V-EOG were recorded using a 6-sec TC. Data  were
stored  by an FM  tape recorder for off-line analysis
                                       1B
                                                                      2A
                                                                       3B
 Fig. 1.  Computed tomography of patient M48. Quite visible lesion produced by left frontal radical lobotomy and
 the result of right pre frontal extensive gyrectomy (probably the whole of Brodmann's area 9 and upper half of
 areas 10 and 46).

-------
CNV after Psychosurgery
                                            379
by  a  general-purpose  computer. Relatively artifact-
free 5-sec EEC epochs were averaged with 8 or 16
trials  in  an  average. CNV amplitude was quantified
in  terms  of the excursion from  baseline at 200,
150, and 100 msec preceding S2. Baseline was defined
as the average amplitude of EEC activity occurring
800 msec preceding SI. Methods are described in
detail elsewhere (Zappoli et  al. 1973).

Results

    CNV, postimperative negative variation (P1NV),
and RT measures for  each  patient and for a control
group of 10 volunteer Ss  are reported in  Table 2.
CNVs recorded in patients  were slightly attenuated,
especially in Ss with high anxiety levels, but negative
deflections could consistently be elicited on the front-
al thalamo-disconnected areas.

    In one (M55) of the six patients treated by bilater-
al extensive  "radical" prefrontal  lobotoiny, no  1SI
slow potential shifts were observed in any derivations.
The patient  responded repeatedly only to S2,and the
response almost  always coincided with the start of an
ample PINV of about 1450-msec duration, present
symmetrically over  frontal areas  and at  the vertex.
When questioned afterward, the patient, who was very
tense, answered that he clearly noticed SI, but never
considered it as a warning signal.
    In the other patients of this group, fairly typical
CNVs were elicited over frontal areas and at the vertex
(Fig. 2). In three patients (M50, M53, and F74), slight
frontal  CNV asymmetries  were observed  in  some
averages (Fig. 2). These asymmetries were inconstant
as to  laterality, were without clinical or neuroradio-
logical explanation, and were not enhanced by use of
ipsilateral mastoid references—a method previously
shown to accentuate  asymmetries (Zappoli in  press).
In patients M50 and  F74, frontal and vertex CNVs
were often followed by an evident PINV (Fig. 2A, B),
especially when Ss exhibited excessive emotional ten-
sion and were requested to avoid eye and head move-
ments during trials.


    Patient  M48 had been treated by  left frontal
lobotomy  and  right extensive  frontal  gyrectorny
(Fig.  1). At  Cz and left frontal area, he presented
CNVs  of fairly normal features, but of low voltage,
sometimes followed  by  a PINV.  At  Fp2 and F2
no  CNV activity or PINV was  observed (Fig. 3A).
Only  at  F4, immediately behind  the gyrectomized
area,  did low voltage CNVs and PINVs become ob-
servable (Fig. 3B).
                 M50
                    S2
                                                           50;itV|    V.EOG|
                                                                                           [SOjuV
                                                                     FP2-M2
                                                                      -       -
      V-EOG
                                      Cz-2A
Fig. 2. CNVs showing moderate asymmetries.  (A) Normal CNV followed by an evident PINV of about 1200-msec
duration in prefrontal and vertex derivations. Mean RT about 280 msec.  (B) CNV with attenuated amplitude
and almost normal features, followed by an evident PINV of about 1300-msec duration, in prefrontal and vertex
derivations.  Mean RT about 410 msec.   /C) CNV of normal shape but low amplitude  in prefrontal and vertex
derivations.  Mean RT about 340 msec.  In A,  B,  and C, moderate asymmetries of  the prefrontal CNVs are
observable.  These asymmetries appeared, however, very irregularly as to side In different sets of eight trials and
in these patients were apparently without clinical and neuroradiological explanation. Pre-Sl has 800-msec base-
line, 5-sec total display time with negative up and TC of 6 sec.

-------
  Patient

 Sex Age



   M55




   M50




   M47



   M56




   M53



   F74



   M48



   M37

^^^=5=


Meant (SO)


 Controls

 10 Ss aged

 2449

Mean: 36.6
Table 2. Stow Potentials and Reaction Time Measures
^-^"—— ^^^
Fp2-2M
NoCNV
-10.1
—
-10.4
-7.2
-9.1
NoCNV
-7.2
B^^HBM
-8.8 (2.0)
===s

• ' -^•••B.
Fp1-2M
NoCNV
-10.1
—
-10.4
•6.2
-9.2
-9.2
-7.1
••••••••^
-8.7 (1.9)
=======

•^•^^^^^^^•^ta.
F2-2M
—
-11.2
-9.9
—
-8.6
-10.8
NoCNV
-8.3
••••^••H
-9.8 (1.5)
========
^^•— —^^^»*.^»
CNV amplitude3. MV
F1-2M
—
-12
-10
—
-8.8
-9.3
-11.2
-8.6
M^HB^^
-9.9 (1.5)
======sss

F4-2M
NoCNV
—
-11.6
-13.2
—
-10.6
-9.4
-7.2
-10.4 (2.3)
=====
-12"
(range:
-6 to -175)
F3-2M
NoCN
—
-11.6
-13.4
—
-10.6
-9.5
-7.2
-10.5 (2.3)
=====
-11.8d
(range:
-6.2to-17.5J
Cz-2M
NoCNV
-15.4
-15
-17.6
-11.2
-11.8
-12.6
-13.1
-13.8 (2.4)
=====
25.6d
range:
19.5 to -36)
Cz PI NV duration,15
msec
1450
1225
No PINV
No PINV
No PINV
1295
1255
No PINV
1308(119.3)


Reaction
timd C meat*
415
245
150
180
280
385
225
350
278 (136)
16 7.4 d

Range msec
365-515
135-380
105-250
110-275
155-430
210-460
140-295
21O680
==a^^B
^^^^^^^•MMM^V
88-235

      of all *ns of eight averaged triab; amplitude
                                                                                                                                          u>
                                                                                                                                          00
                                                                                                                                          o
 T«wn of all ms of eight awraged triab in which PINV

 CMMn of aU trials Mtactad for awwaging.
Ml 200 msec ppor to S2.


F plOfOJIt.
                                                                                            in the
                                                                                                    I
                                                                                                    a
                                                                                                    a

-------
CNV after Psychosurgery

                      S2
                                                                     381
H
RT
                                                      20
                                                     +20
F2-2M
F1-2M
                 AV.n = 8
                                                     •50

                                                     +50
                                                               S1
                                                                            S2
                                                                                                  5 sec
                                                                         AV=8
                                                                                                 V-EOG
/i1^. J  C7VV i/i 3f 48, the left frontal lobotomized and right frontal gyrectomized patient.  (A) with F2-linked
mastoids, CNV activity and PINY are absent.  With other derivations (F2-M2, Fp2-2M, Fp2-M2), this absence
was noted on the whole area of the right frontal extensive gyrectomy. Mean RT about 218 msec.  (B) Low-vol-
tage CNV followed by a PINV is present immediately behind the gyrectomized area (F4-2M).  These SPs were
easily elicited on all the other areas explored. Averages of eight artifact-free trials.
     In the case of M37, who had been subjected to
 right extensive prefrontal stereoleucotomy, CNVs of
 low voltage were easily elicited from all areas explored
 (Fig. 4). These CNVs always reached maximum ampli-
 tude after S2, often with  slow  return to baseline.

     In most  patients, RT to S2 was fairly long. The
 mean  RT of all trials  selected for  averaging in the
 entire group of patients was 278 msec (range: 105-
 680).  RTs for each patient and control Ss are given in
 Table 2.

 Discussion

     These results show that it is possible to elicit front-
 al  CNV even when the integrity of mediothalamic-
 frontocortical pathways is interrupted. These findings
 are inconsistent  with  observations  by Skinner and
 his colleagues (Skinner 1971, Skinner and Lindsley
 1973,  Skinner  and Yingling  1976) that cryogenic
 blockage in  cats of  the inferior  thalamic peduncle
 (ITP) abolished surface negative SPs in the ipsilateral
 frontal cortex. On  the basis of this evidence .Skinner
 has  hypothesized  that  the  mediothalamic-fronto-
 cortical  biderectional  system  plays an important
 role in  generating and  regulating  frontal  negative
                         SPs elicited by behavioral conditioning (cf. Skinner's
                         neurophysiological  model  for  the  regulation  of
                         sensory input to  cerebral cortex, this volume). The
                         presence of frontal CNV in lobotomized patients is
                         not easily explainable in terms of this model.

                             It is impossible on the basis of autoptic findings
                         in lobotomized patients  to sustain any  regeneration
                         of the sectioned mediothalamic prefrontal connections
                         (Freeman and Watts 1947,1966). The results reported
                         here show that formation and regulation  of prefrontal
                         CNVs  in humans may occur independently of the
                         anatomo-functional integrity of communicating path-
                         ways to and  from mediothalamic nuclei and anterior
                         frontal cortex. CNV patterns in lobotomized patients
                         suggest that  the prefrontal cortical areas continue to
                         receive and process information  relevant  to stimuli.

                             Possible  explanations of frontal CNVs in loboto-
                         mized patients include:

                              1. Scalp electrodes, especially when  referred to
                         linked mastoids,  may reflect  ERPs from much wider
                         cortical areas than those immediately underneath the
                         electrodes. This possibility, however, would seem to
                         be refuted by observations  of patient M48, treated by

-------
  382
                                      Zappoli et al.
    F4-2M
   F3-2M
   Cz-ZIM
   V-EOG
 Ffc,  4.  CNV in M 37.  Averaged potentials obtained with different electrode montages.  2M: linked mastoids:
 Ml,  M2: ipsilateral mastoids; and  2A: linked ears.  Low-voltage CNVs were  easily elicited from all explored
 areas, and negative shifts always reached maximum amplitude after 52, with slow return to baseline.  MeanRT
 about 308 msec.
 right frontal gyrectomy. No measurable CNV activity
 was observed over the ablated cortical area. Activity
 was observed immediately behind the gyrectomized
 area.
     2.  Transcortical  mechanisms,  i.e.,  tangential
 propagation  of activity from distant cortical centers,
 may be involved in generating and regulating CNV in
 frontal cortex anterior regions, although there are no
 EP or neurophysiological data to unequivocally sup-
 port this hypothesis. Ennever (1975), for instance,
 found no evidence for tangential transmission across
 the  cortex  in  rats of the short-latency surface  re-
 sponse evoked by weak somatic stimulation.

     3. Frontal CNV generation and regulation in an
 operant conditioning situation may, in humans, be
 mediated  by cortical and  subcortical connections
 forming  part of a diffuse,  probably extrathalamic,
 projection  system of which the exact source, dis-
 tribution, and  structure are  presently  unknown.
Since CNV-like  waveforms  have been recorded at
many locations along the neuraxis from brainstem to
cortex, contingent negative variation may simply be
an electrocortical sign of a general neuronal process
occurring  diffusely throughout  the  brain. This hy-
pothesis would account for the present observations
as well as the bilaterally symmetrical CNVs reported
by Gazzaniga and Hillyard (1972) in "splitbrain" pa-
tients cued unilaterally to respond.  Hillyard (1973),
in fact, suggested  that human  CNVs contain a bi-
  lateral component activated by a nonspecific sub-
  cortical mechanism. Thus, neither bilateral thalamo-
  frontal tractotomies nor transection of the forebrain
  commissures impairs the development or symmetry
  of premotor CNVs.

     The method of cryogenic lesions (Skinner 1971)
 that transitorily abolishes frontal SPs probably influ-
 ences cerebral structures other than the FTP and alters
 the state of consciousness, as manifested by behavior-
 al  changes.  With  respect  to  the  cryogenic  data
 McSherry (1973) stated  that "it is difficult to inter-
 pret these data in terms of an organism in a state of
 mind capable of generating a CNV."

     4. It  is  also conceivable that unknown compe-
 satory mechanisms could have produced new function-
 al connections (other than the regeneration of severed
 pathways) during the 20  years since lobotomies were
 performed in  these patients.
     Each of these hypotheses requires further assess-
 ment in human and animal models.
Acknowledgments

    The  authors wish to express their appreciation
for the technical assistance provided by A. Versari
andC.Nencioni.

-------
ABSENCE OF CNV REBOUND IN
PSYCHOSURGERY PATIENTS1
J.J. TECCE2
Laboratory of Neuropsychology, Boston State Hospital, Department of Psychiatry,
Tufts University School of Medicine, Boston, MA, U.S.A.

M.H. ORZACK AND A.F. MIRSKY3
Divison of Psychiatry,  Boston University School of Medicine, Boston, MA, U.S.A.
    Event-related slow brain potentials are useful as
neurophysiological indicators of information process-
ing.  For example,  the  development  of contingent
negative variation (CNV)(Walteret al. 1964) is disrupt-
ed and reaction time toS2 is lengthened (CNV distrac-
tion effect) when a unified attention set to S2 is dis-
turbed by the imposition of a short-term memory task
for letters presented in the auditory mode  (Tecce et
al.  1976) or for letters or numbers presented in the
visual mode (Roth et  al. 1975; Tecce et al., this sec-
tion). In addition, the CNV distraction effect has re-
cently  been demonstrated with a short-term memory
task in which trials including letters were randomly
interspersed with trials having no letters (Tecce, this
volume). An unexpected finding in the latter study
was the significant elevation of CNV amplitude during
the no-letter trials. This supra-normal magnitude of
CNV was interpreted as a rebound function and as a
possible index of plasticity in human brain function-
ing.  In the present study, an attempt was  made to
assess  the  CNV rebound  effect  in  psychosurgery
patients.

Methods

    Participants were  14 young normal volunteers
(6 males and 8 females) aged 19 to 26 (X = 21.43) and
 19  psychosurgery patients (7 males and 12 females)
aged 34 to 73 (X = 47.11). Patients had received a bi-
medial  prefrontal  leucotomy (Valenstein  1973, p.
280) within a 12-year period prior to  testing. In this
procedure,  medial fibers (white matter) connecting
prefrontal cortex to thalamus are destroyed. With this
technique, there is an expectation of a high degree of
consistency in surgical lesions. No individual was an
inpatient at the time of testing. Eighteen patients had
a history of electroshock therapy  and, at the tune of
testing,  all  patients were  on medication  including
psychotropic and analgesic substances.  Assessments
of clinical status were made on the basis of interviews
by two psychologists, a social worker, and a psychia-
trist  and by standardized  rating procedures such as
the  Psychiatric Status Scale and  Problem  Appraisal
Scale. Information on psychopathology will appear in
a subsequent report.

    The experimental task consisted of a constant-
foreperiod simple  reaction-time paradigm. The prepra-
tory stimulus (S1) was a 150-msec  Hash of a black "X"
on a circular patch of dim light 2.5 cm in diameter.
The second stimulus (S2) was a continuous 1000-Hz
tone of 70 dB (SPL) presented  through earphones
1.5  sec  after the  "X" and terminated by a telegraph
key press  (KP).  Intertrial intervals varied  randomly
from 8 to 14 sec  (X = 11) within  a rectangular distri-
bution of values 1 sec apart.

    The two experimental conditions were "control"
and "50% letters." The control condition consisted of
31 trials of the S1-S2-KP sequence and lasted 7 min.
A second control condition after "50% letters" was
omitted since previous work has shown no difference
in CNV amplitude values between control conditions
administered before  and after letters (Tecce et al.
 Supported by U.S. Public Health Service Research Grant AG-00515 (JJT)and Contract No. 278-76-0064 (ER)
 (MHO and AFM).
 2Research Scientist Development Awardee 1-K02-00016 from the National Institute of Mental Health.
 3Research Scientist Awardee 5-K5-14915 from the National Institute of Mental Health.

-------
 384
                                     Tccce et al.
  1976; Tecce,  this volume).  The   50%-letters task
 consisted of 31 trials divided randomly into two types:
 letter trials and no-letter trials. No-letter trials and
 control trials were identical. Letter trials were similar
 to no-letter trials except that three  successive letters
 were presented visually between flash and tone (within
 the S1-S2 interval) in a short-term memory task. Sub-
 jects were instructed to repeat aloud the three letters
 following key press. The letters were a random sample
 for each  trial from  the following: A, C, E, H, K, L,  N,
 P,  S, and U.  The first letter appeared 50 msec after
 termination  of SI; the interletter interval was 200
 msec. Each letter lasted 200  msec. In the 50%-letters
 task, the  first trial had noletters and was omitted from
 data analysis. The sequence of the remaining 15 letter
 trials and 15 no-letter trials was random and the same
 for all subjects.

     Electroencephalographic  (EEC)  recordings were
 made from frontal (Fz), central (Cz), and parietal (Pz)
 scalp areas with linked earlobes as reference. A vertical
 electrooculogram (EOG) was recorded from  above
 and below the right eye. Electrodes  were Beckman
Ag/AgCl standard (for EEC)  and miniature (for EOG
 and reference) types.  A Grass Model 7B polygraph
 was used  to record  EEC and EOG (time constant: 8
 sec). High frequency cutoff was 75Hz (50% amplitude
 reduction with 12 dB/octave  roll-off). EEC and EOG
 were recorded on magnetic tape to permit editing of
 artifact trials  during off-line averaging with  a CAT
 1000. Averaged CNVs were based on 6 to 12 trials, the
 number being constant for a given individual. Trials
with eye movement, includingeyeblinks, or pre-S2 key
presses were excluded from analysis.

    Mean  differences  in CNV  amplitude between
patient  and nonpatient  groups were  evaluated by
independent t tests  based on  31 df; units of analysis
were either raw scores  for control trials or difference
scores (letter trials minus control trials and no-letter
trials minus control  trials). Within-group comparisons
of both letter and no-letter trials with control trials
were  evaluated  by   correlated  t  tests based  on  18
df (patients) and 13 df (non patients).

Results and discussion

    Table 1 presents CNV amplitude values for letter,
no-letter,  and control trials of patient and nonpatient.
 groups. Fig. 1 shows difference scores for CNV ampli-
 tudes recorded at Fz,  Cz and Pz sites of patient and
 nonpatient groups.  Examples of CNV tracings for a
 patient and a nonpatient appear in Fig. 2.

     As shown  in Table 1 and Fig. la, CNV amplitude
 was significantly lower in letter trials than in control
 trials for patients at all recording sites and for non-
 patients at  F/ und Pz. This reduction of CNV by let-
 ters is relatively greater for patients than nonpatients
 at both  Fz  and Cz (see Fig. 1). For the psychosurgery
 patients then, CNV development appeared to be more
 disrupted by a short-term memory task  than for non-
 patient controls.

     As shown in Table 1 and Fig. 1 b, there is a pattern
 of  elevation of CNV amplitude  in  no-letter trials
 compared to control trials for nonpatients (p < .02 for
 Cz; p < .10 for P/.) but not for patients, who show a
 pattern  of  slight  CNV decrease in  no-letter trials.
 Group differences between patients and nonpatients,
 shown in Fig. 1 for no-letter minus control difference
 scores,are significant for Cz  and  approach significance
 for Fz  (p < .10). Tims, the pattern  of  a rebound in
 CNV amplitude seen on no-letter trials for nonpatients
 is missing in patients.

     In  conclusion, patients receiving psychosurgery
 show greater disruption in  CNV development defined
 in two ways greater amplitude reduction by a short-
 term  memory task for letters and a lack of recovery
 from this disruption as shown by absence of a CNV
 rebound function. The fact that the patients are older
 than nonpatients raises the question of possible organ-
 ic deterioration due to the normal aging process rather
 than surgery. Preliminary findings indicate that elderly
 individuals show a clear lack of CNV  rebound only at
 Fz. Since departures from  normal CNV reported for
 psychosurgery patients appeared at other than  the Fz
 location,  whatever organic factors   contributed to
 CNV  impairment  are likely to be outside the normal
 aging process. In  addition,  it is unlikely that surgery
 alone (intended for frontal  brain areas) could account
 for the  absence of CNV  rebound  in the  posterior
 (Pz) recording site. Aside  from  the lesions in the
 frontal white matter, other factors that  differentiate
 the  two groups include psychopathology,  previous
 history  of medication, and/or electroshock  therapy.
 Baseline  CNV amplitude values were comparable for
 patient and  nonpatient groups; rebound values were
 not. The possibility of the  influence  of these other
 variables that distinguish the two groups cannot be
 ruled  out. Whatever the underlying neurophysiolog-
 ical mechanisms responsible for these  changes, the
 CNV rebound effect appears to reflect a type of brain
 function not detectable in  the routine measurement
 of baseline CNV (in control trials).

Acknowledgments

    The  assistance  of Debra  Yrchik  and  Debbie
Meinbresse in carrying out the study and Tim Clifford
and Connie  Dessonville in data  analysis  is gratefully
acknowledged.

-------
Absence of CNV Rebound after Psychosurgery                                                385

Table 1. Means (and Standard Deviations) for Amplitude of Contingent Negative Variation a

Nonpatients
(n-14)
Patients
(n-19)
Control
Letters
No letters
Control
Letters
No letters
CNV amplitude, 
      u
6


4


2


I


2


4


6


3


N
                ».  LETTER TRIALS
                   MINUS CONTROL
                   TRIALS
                                       I
                  Fi         Cl          Pi

                     RECORDING SITES
 6


 4


 I


 ll


•1


 4


 6


 1


 M
b.  NO-LETTER TRIALS
   MINUS CONTROL
   TRIALS
                                                    Fi         Ct         Pi

                                                       RECORDING SITES
 Fig. 1. CNV amplitude difference scores in psychosurgery patients (n*19) and nonpatient controls (n*>14).  The
 downward direction of bars in a indicates reduction in CNV amplitude in letter trials relative to control trials.  The
 upward direction of bars in b indicates enhancement in CNV amplitude in no-letter trials relative to control trials.
 while the downward direction indicates reduction in CNV amplitude in no-letter trials relative to control trials.

-------
386
                            Tecce et al.
          81
                                                                  CONTROL
                                                                  LETTERS
                                                                  NO LETTERS
SI
                                                                                0.5 sec
Fig. 2. Examples of vertex (Cz) CNV traces of a psychosurgery patient (left side of figure) and a normal volunteer
subject (right side of figure) for control, letter, and no-letter trials. For the normal subject, there is an elevation in
CNV magnitude for no-letter trials compared to control trials (CNV rebound effect) that is missing in the psycho-
surgery patient. Negativity at the Cz recording site relative to linked earlobes is shown as upward.

-------
SOME METHODOLOGICAL AND THEORETICAL
ISSUES OF ERP IN PSYCHIATRIC POPULATIONS


B. DUBROVSKY AND M. DONGIER
Allan Memorial Institute, Department of Psychiatry, McGill University, Montreal, Quebec, Canada
    Results obtained by recording event-related slow
 potentials (ERP) from psychiatric  patients in Liege
 and Montreal have been reported in the proceedings
 of  the three previous ERP meetings (Dongier et al.
 1973, Dubrovsky and Dongier 1976, Timsit-Berthier
 et al.  1969). Since these findings have not been re-
 plicated in other laboratories, we think it important
 to comment on certain methodological aspects that
 may account in part at least, for some of the discrep-
 ancies. We should also like to raise some issues per-
 taining to the interpretation  of ERPs in  psychiatry.

    ERP recording techniques have not yet been stan-
 dardized.  Serious limitations in interpretation  are
 posed by the use of techniques that cancel out from
 scalp-recorded  activity potentials originating in extra-
 cerebral sources. Such is the case when potentials orig-
 inating  during  eye  movements recorded  bipolarly
 (Hillyard and Galambos 1970, McCallum and Walter
 1968) are substracted from brain potentials recorded
 monopolarly  as  conventionally  done  for  CNV and
 readiness potential (RP). As noted by Cooper (1959)
 bipolar derivations can seriously exaggerate the phase
 and time difference in records of activity from sources
 that are not radial, symmetrical, and stationary. Sub-
 stracting bipolarly derived EOG potentials from con-
 ventionally recorded CNV and RP is then a very hazard-
 ous procedure.

    Furthermore, since cerebral potentials can reach
 the electrodes for eye-movement  recording  (EOG),
 they  may be cancelled out when potentials  derived
 from the EOG are subtracted from potentials recorded
 from the vertex :(Papakostopoulos etal. 1973; Rosen,
 this volume). It is possible then, that cancellation of
 vertex-derived brain potentials by anomalous compen-
 sation from EOG potentials recorded in parallel may
 eliminate significant phenomena of cerebral origin.


    In our laboratories, we monitor the influence of
 eye movements on event-related brain potentials by
 recording in a parallel channel from a  supraorbital
 electrode referred to the linked earlobes electrode.
We use the same reference to record CNV and RP.
The amplification and time constant (9.5 sec) is also
similar for both the vertex and supraorbital channels
(Dubrovsky et al. 1973).

    Another aspect that deserves serious considera-
tion,  especially  in  psychiatric populations,  is the
instructional set given to the subjects. A graphic exam-
ple of the importance of prior instructions and their
influence on ERP characteristics is shown in  Fig.  1.
The CNV recorded from subjects suffering from spe-
cific phobias is significantly larger and more prolonged
when  phobogenic stimuli are used as S2 (Fig. IB)
(Barbas et al., this section), than when non-disturbing
stimuli are used (Fig. 1 A). In Fig. 1C, the patient was
expecting the phobogenic stimuli because of misinter-
preted instructions. Consequently, even though non-
disturbing stimuli were actually used as S2, the CNV
obtained had characteristics like  the ones obtained
when  phobogenic stimuli were employed (Fig. IB).
This result  indicates that  the expectation of phobo-
genic  stimuli has greater influence in determing the
CNV  amplitude and duration than the characteristics
of the stimuli used per se.
    The parameters generally evaluated in ERP (amp-
litude and duration) presuppose that the central ner-
vous system (CNS) is an energy transfer system. We
are searching, with the help of sophisticated com-
puter analyses, for changes in overall levels of electri-
cal activity that either accompany, or cause, mental
disease. We are still working within the framework of
the Freudian  dictum of looking at changes in quan-
tities of energy (physical or chemical) to account for
psychopathological changes in behavior. It appears to
us that a method for analysing ERP that allows search-
ing for qualitative changes in brain waves and for pre-
dominance of certain cerebral areas at different stages,
rather than for an evaluation of overall level of activity,
will be more in accordance with the present-day views
that consider the nervous system as an  information
transfer  system (Dubrovsky  and Melzack  1970,
Dubrovsky and Dongier 1976).

-------
 388
                         Dubrovsky and Dongier
           SI    82
                   i
GNV
      ISM
 Fig. 1.  Contingent negative variation and eye movement potential! for patient suffering from phobia to snakes:
 A.  when presented in S2 with a picture of a cat (nondiiturbtng stimulus), B. when presented in S2 with a picture
 of a snake (photogenic stimulus), and C. when presented In S2 with a picture of a cat, but was expecting the pic-
 ture of a snake because of misinterpretation of instructions.
    Trends in  this direction include  the analysis of
CNV form under different behavioral  requirements
during recording (Weinberg et al. 1976). In addition,
studies of the topographical distribution of ERPs have
 revealed significant differences in CNV form, recorded
 from different leads in normal subjects (Weinberg and
 Papakostopoulos 1975). At EPIC IV, Tecce also re-
 ported   that  abnormally prolonged postimperative
 negative variation  (PINV) (Dongier  1973 a) during a
 standard CNV paradigm can be observed in psychiatric
 patients only from certain leads.

     These observations reopen the issue of the possi-
 ble differential origin of the prolonged PINV, which
 may be cortical (Dubrovsky et al. 1976), as opposed
 to the CNV wave perse, which appears to be triggered
 subcortlcally  (Gazzaniga and Hillyaid  1972, McCal-
 lum et al.  1973). Work using prolonged reaction time
 foreperiods (4  sec or more), during the  classical CNV
 paradigm (Loveless and Sanford 1975, Rohfbaugh et
 al.  1976, Weerts and Lang 1973), clearly reveals that
different processes are involved in the generation and
resolution  of CNV phenomenon. These studies have
serious implications for the interpretation of ERPs in
psychiatry. In  recent studies  on  electrophyiiological
manifestations  of psychopatholojry  (Shagass  1975,
Callaway 1975), it  has been emphasized that consis-
tent alterations in later evoked activity recorded from
frontal, nonprimary receiving areas, reflect alterations
of the early evoked response. In  these studies, it is
assumed that late evoked activity in frontal areas is
elicited from projections originating  In primary sen-
sory cortex, after being partly processed there. This
assumption involves essentially a sequential type of
sensory information processing.

    As the bioelectrical activity that gives rise to CNV
and RP is preferentially  recorded from frontal brain
areas, it is important to remember that evoked electri-
cal activity in these areas can be elicited by activation
of parallel fiber systems, independent of projections
to primary receiving areas. Moreover, we have argued
(Dubrovsky and Garcia-Rill 1971) that these parallel
projection systems are involved with the processing
of different aspects of the original input  from the
primary projection systems. The existence of parallel
information processing in the CNS must then be taken
into account  when interpreting late 'frontal evoked
responses. On  account of the classical derivations for
ERP, it will be of great  importance  to elucidate the
question of whether the  frontal brain potentials are
mainly  dependent  on previously processed cortical
neural  activity or  are independently generated from
subcortical areas that do  not project to primary sen-
sory cortex (and are, therefore, part of a parallel sys-
tem for information processing).
    While we believe that the problems raised for dis-
cussion concern the general field of event-related slow
potentials and behavior,  we think that resolution of
these problems ii of utmost Importance for a better
definition and agreement of the uses (and abuses) of
event-related slow potentials in psychiatry.

-------
MORPHOLOGICAL ANALYSES OF  THE  CNV  IN
PSYCHIATRY: COMPARISON OF  RESOLUTION
MODE AND  CUMULATIVE CURVE METHODS1
M. TIMSIT-BERTHIER, J. DELAUNOY, AND A. GERONO

Laboratory of Clinical Neurophysiology, Department of Medical Psychology and
Psychosomatic Medicine, University of Liege,Liege, Belgium
   Several years  ago, with  a purely empirical  and
medical point of view, a morphological classification
system based on the "resolution mode" of the CNV
was developed at Liege (Timsit-Berthier et al. 1973).
Although this technique was of some value in distin-
guishing  slow potential  patterns of mental patients
from those of control subjects, morphological analy-
sis provided little help in the differential diagnosis of
psychotic and neurotic patients. In order to improve
the diagnostic sensitivity of the electrophysiological
test, the  recording session was extended and a cumu-
lative  method of assessing the temporal evolution of
SPs was  developed. In  this report, the cumulative
curve  method will be described and compared with
the earlier resolution mode method.


Method

Subjects

    This study included 55 control subjects (17 men;
38  women; mean  age, 29 years; S.D.-5.5)  and 240
patients  (166 men; 74 women; mean age, 33 yean;
S.D.=6.1). Ninety-four were recruited from the poly-
clinic of our department, 90 from the open psychiatric
service of the general hospital, and 56 from the closed
psychiatric service. The patient group included 85 psy-
chotics and 75 neurotics. Eighty patients were rejected
because of inadequate clinical data.

Apparatus

    EEC was recorded at Cz with Ag/AgCl electrodes
referred  to linked earlobes using amplifiers modified
to yield  an 1 l*ec time constant with an upper cutoff
of 50 Hz. EEC and EOG were recorded on FM tape
for off-line summation. Five to 35 percent of the trials
from individual subjects were rejected because of
EOG deviations, muscular artefacts, or erroneous
responses.

Experimental procedure

    The recording room was darkened and isolated
from outside sounds by white noise delivered over
loudspeakers at low intensity. Subjects were asked to
keep their eyes closed during testing.  In a 1-hour ses-
sion, subjects received a block of 100  stimuli pairs.
Each trial began with a warning click (SI) followed
after 1.5 sec by a series of clicks (S2), which the
subject  terminated by pushing a button with the right
hand. The interval between the motor  response and
the warning stimulus  for the next trial varied ran-
domly from 10 to 30 sec. Six to 8 successive aver-
ages (12 trials each) were calculated for each block
of 100 trials.

Resolution mode analysts

    The first 12 trials were averaged and analyzed for
resolution mode. Baseline was estimated visually as the
average voltage of the 1000-msec  pre-Sl epoch. CNV
amplitude was measured as the average voltage of the
200-msec pre-S2 epoch relative to baseline. The reso-
lution mode of the CNV for each subject was also
judged as described in the paper by Timsit-Berthier et
al. (this section).
    When CNV amplitude and resolution mode had
 been determined, all waveforms were classified accord-
 ing to statistical data as "normal" or "abnormal" as
 illustrated in Fig. 1. "Normal" curves included Types I
 and II, while "abnormal" curves included Types III
 and IV as well as flat CNVs (amplitude < 5/uV).
 1
 Work was supported in part by grant 20397 - FRSM.

-------
 390
                          Timsit-Berthier et al.
                              TYPE ICNV
                             TYPE II CNV
              NORMAL CURVES
                              FLAT CNV
            r
             XT^
                            TYPE III CNV
                            TYPE IV CNV
          ABNORMAL CURVES

Fig. 1 Normal and abnormal curves classified on the
basis of the resolution mode

Cumulative curve analysis

   A  method of cumulative analysis, illustrated in
Fig. 2, was devised to reduce the amount of data ob-
tained  from the six. to  eight sequential averages ob-
tained  from each subject  to a single positivity/neg-
ativity  (O'P/ON) ratio. First, negative and positive
measures were obtained for  each sequential average.
Negative values  (CNV)  were obtained  as  described
above.  Positive values were  calculated  as the mean
voltage  during a 300-  to  500-msec  post-82 interval
relative to the CNV value. Positive (0'P=P, + P2+...+Pn)
and negative (ON= Nt  + NJ+...+ lNn) values  of each
average in the sequence were then summed, and the
cumulative  values were expressed  as the O'P/ON
ratio.  Cumulative curves of the  positive  and nega-
 Fig. 2.  Cumulative curve analysis of slow potentials.
 A.  Principle of measurement: measures of pre-S2
 negativity  (n) and post-52 positivity (p) are obtained
 for sequential series of 12 trials.  B. Cumulative dis-
 play  and reduction of data:   cumulative curves are
 constructed  for  successive positive  and  negative
 values.   The cumulative  sum  of positive fO'PJ and
 negative (ON) values are then expressed as the cumu-
 lative ratio (R).


 live  measurements  were  also drawn  to illustrate
 graphically the temporal evolution of the phenomena
 (see Fig. IB).

    According  to  statistical data, cumulative  curves
 were then classified  as "normal" or "abnormal"  as
 shown  in Fig.  3. In brief,  normal cumulative curves
 included: (1) Type 1 or II patterns with relatively
                \
                \
                           vrfTTI
           NORMAL         ABNORMAL
Fig.  3. Normal and abnormal cumulative curves clas-
sified on the basis of the six successive measures of
pre-S2 negativity and post-82 positivity.

-------
Morphology of CNV
                                                               391
 invariant positive and negative amplitudes across time
 (2) initially flat  CNVs, which progressively increased
 in negativity across time; and (3) initially prolonged
 Type III or IV CVNs, which progressively decreased
 in duration across  time. Abnormal cumulative curves
 included: (1) Type III, IV,  or  flat CNVs, which re-
 mained  relatively  invarient across  time, or  (2)  a
 labile  temporal pattern of CNVs  (e.g., flat, then
 Type III, then Type I) with 0'P/ON<-0.1.
 Results and discussion

     Results of the resolution mode and cumulative
 curve analyses are shown in Fig. 4 and 5, respectively.
 Note  that  the cumulative  curve method classified
 practically all normal control and neurotic subject SP
 patterns as "normal," but classified psychotic patient
 SPs as "abnormal" 49% of the time. The discriminant
 power within the psychotic  population, therefore,
 was greater for the  earlier resolution mode than the
 cumulative curve method. Neither method,  in fact,
       87%
                       59%
           13%
                           41%
NORMAL

ABNORMAL
       64%
                                     36%
     CONTROLS    NEUROTICS    PSYCHOTICS
      n =  55             75              85
 Fig. 4. Results of the resolution mode classification of
 slow potential patterns observed in control, neurotic,
 and psychotic subjects.
1

18% Q,0/


1
2%
j«f /«



7%
H
D
•


CONTROLS NEUROTICS
n= 55 75
NORMAL
ABNORMAL
5

1%

49°,
1
0

PSYCHOTICS
85
Fig.  5. Results of the cumulative curve classification
of slow potential patterns observed in control, neurot-
ic, and psychotic subjects.

provides  an unequivocal measure for the differential
diagnosis of neurotic versus psychotic syndromes.

    However, results of this comparative evaluation
do  provide  some  optimistic  signs.  While  a  normal
cumulative curve  for  a psychotic  subject  has  no
particular meaning and does not in the least preclude
a diagnosis  of psychosis, an  abnormal cumulative
curve (0'P/)N) ratio -01) allows us  to affirm a diag-
nosis of psychosis within  a very  small margin of
error. It  is also apparent that the temporal analysis
of  evolving slow potential patterns provides a  dif-
ferent, and probably more sensitive, index  of mental
state than a single CNV provides.  It is worthwhile
to  point out that the cumulative   curve method
also includes a second phenomenon (usually called
the  P300 wave) in the  diagnostic equation. Further
refinement  of the electrophysiological criteria  for
differential  diagnosis is obviously  needed.  On  the
other hand, morphological analysis of slow potentials
has  provided  to  be a  useful adjunct to  traditional
diagnostic  procedures  at  Liege  for  many  years
(Timsit-Berthier et al. 1970, 1973, 1975).

-------
 SOMATOSENSORY EVOKED  POTENTIAL AS A
 MEASURE OF TOLERANCE TO ETHANOH


  D. M. SEALES, P. NAITOH, L. C. JOHNSON, AND M. A. SCHUCKIT 2
  Naval Health Research Center, San Diego, California, U.S.A.
     Ethanol, in amounts commonly imbibed at social
  functions, reduces the amplitude of late components
  of evoked potentials  recorded from central scalp
  regions. Gross et al. (1966) reported that 100 cm3 of
  90-proof whiskey substantially reduced the amplitude
  of components of the auditory  evoked  potential
  between 30 and 220 msec, recorded from a left cen-
  tral scalp region.  Lewis et al. (1970) found a similar
  influence of ethanol on components from approxi-
  mately 130 to 230 msec of the visual and somatosen-
  sory evoked potentials at blood alcohol concentrations
  (BACs)  ranging from  70  to  100  mg%. Significant
  changes occurred at C3 and C4, but not at 01 and 02.
  Salamy and Williams (1973) confirmed the effect of
  ethanol on the somatosensory evoked potential (SEP)
  at the vertex and provided evidence that amplitude
  reduction was not the result of gross changes in back-
 ground EEC or of increases in variability of peak laten-
 cy. They showed  also that speed of peripheral nerve
 conduction  was unchanged and that short-latency
 components of the SEP resulting from specific nervous
 input were resistant to the presence of ethanol, except
 at high concentrations (145 mg%). The observation
 that ethanol-induced decrement of evoked potential
 amplitude occurs primarily in  late components often
 associated with cognitive functions suggested that the
 amount of evoked potential decrement might relate
 to drinking history, since experienced drinkers evi-
 dence less cognitive disruption from a given amount
 of ethanol than do inexperienced drinkers. Objectives
 of the present work were to  replicate the effect of
 ethanol on  the vertex-recorded SEP and to determine
 the relation between SEP change and drinking history.

Methods

    Eleven males were chosen on the basis of a drink-
ing history  questionnaire,  modified from  that  of
 Cahalan and  Cisin  (1968). The subjects ranged in
 drinking history from one or  two beers a month to
 two cocktails or five beers nearly every day, plus
 moderate intake of other alcoholic beverages.

    Subjects were studied during 2 nonconsecutive
 days. The placebo day was first. Three drinks were
 served, each of 200 ml orange  juice with a sufficient
 quantity of 95% USP ethanol floated on top (2 ml per
 drink) to resemble the odor of an actual mixed drink.
 On the second day, subjects received 0.9 g/kg ethanol
 in three equal drinks of 200 ml of orange juice. Each
 drink was imbibed over a period of 10 min.

    Fig. 1  illustrates the daily schedule. Subjects had
 fasted for  at least 4 hours before the  start of the
 experiment, usually since the preceding evening meal.
 They were  given a light breakfast (toast)before record-
 ing and stimulating electrodes  were applied. Testing
 began at 0930 and continued  in 0.5-hour sessions.
 Drinks were given from 1030-1100, lunch was from
 1230-1300, and a relaxation period was held from
 1430-1500. Data from individual testing sessions were
grouped in the predrink, midday, and late afternoon
time periods (see Fig. 1). Because of rapid changes in
BAG, data from only the 1130-1200 testing session of
the morning period were used.

    Ag/AgCl cup electrodes were applied at vertex
 and mastoids. EEC was amplified and recorded (band-
 pass 0.02-30.0 Hz, 8-sec TC). Beckman Biominiature
 electrodes  were attached superior and lateral to one
 eye. SEPs were elicited by 500-/zsec pulses delivered
 through a stimulus isolation unit with a predetermined
pseudorandom sequence of interstimulus  intervals
 ranging from 2-6 sec in 0.5-sec increments. The cath-
ode was a flat  Ag  disk, applied  over the median nerve
2 cm proximal to  the flexion crease of the wrist. The
anode was  a large metal plate on the dorsal surface of
^This research was supported, in part, by Department of the Navy, Bureau of Medicine and Surgery, under Work
Unit MROOO.01.01-6006. The views presented in this paper are those of the authors. No endorsement by the
Department of the Navy has been given or should be inferred.

2Pre«ent Address: Alcohol and Drug Abuse Institute, University of Washington, Seattle, Washington.

-------
 Ethanol and Somatosensory EPs
                                            393
                   TIME OF DAY

Fig 1. Lower portion: Daily time schedule for both
placebo and drug days.  Upper portion: Average blood
alcohol concentrations for 11 subjects plotted in time
relationship to daily schedule (drug day: solid line).
the wrist. Subject ground was placed on the upper
portion of the stimulated arm. Current was set near
1 mA to produce a minimally observable thumb twitch.
Care was taken to maintain a constant effective cur-
rent by continually monitoring the extent of the
thumb  twitch and  by measuring impedance before
each series of stimulations, adjusting voltage appropri-
ately. Sixty to 72 electrical stimulations were given in
a recording session. Subjects were instructed to count
the number of stimulations to themselves, to visually
fixate a spot on the wall, and to refrain from blinking
in the 1 sec following an electrical stimulation. Breath-
alyzer readings (Smith and Wesson Model 1000) were
taken at the midpoint of each 0.5-hour session, follow-
ing SEP recording. Behavioral  tests and a mood scale
were administered, but these are not discussed in this
report.

    Evoked  potentials were  averaged off-line on  a
PDP-12 computer. Trials associated with eye or move-
ment artifact or  with  EEC signs of drowsiness were
excluded. An average consisted of the first 32 accept-
able trials in any one recording session. All statistical
comparisons  were two-tailed  and were evaluated at
the 0.05 level. Dunn-Bonferroni criteria (Dunn 1959)
are indicated in tables.

Results

BAC values

     Average BAG values for the 11 subjects are plotted
in the upper portion  of Fig.  1. On the placebo day
(dashed line), BAG remained  at or below the 10 mg%
noise level of the Breathalyzer, except in the morning
period when the 6 ml of ethanol used in the placebo
drinks produced a slight increase in BAG. On the drug
day (solid line), BAG rose rapidly, to peak at about
1145, and then  fell slowly. One subject achieved peak
BAG at 1115, three  at 1215, and one at 1315. Peak
BACs ranged from 77-121 mg% (mean 94).
SEP waveform

    Fig. 2 illustrates the components of averaged SEPs,
a positivity at 100 msec (PI), a negativity at 155 msec
(Ml), and a broadly peaked positivity from 220-400
msec (P2). The latter peak sometimes consisted of two
positive peaks at 230 and 300 msec, although not con-
sistently enough to allow analysis as separate peaks. No
components were detected consistently prior to about
70 msec,  and there was no late negativity at  about
350 msec as reported by Salamy and Williams (1973)
and by Lewis et al.  (1970). Peak-to-peak measures of
amplitude were used: P1N1 and N1P2. Predrink SEP
amplitudes were stable for each subject. Accordingly,
SEP amplitudes did  not differ significantly from one
predrink  testing  session (0930-1000)  to  the  next
(1000-1030) or from the placebo to the drug day.
Predrink  SEP  amplitudes,  however, were variable
among subjects. P1N1 ranged from 6 to 46 pV (mean
19); N1P2 ranged from 14 to 71 jxV (mean 35).
           PREDRINK
           N1
         MIDDAY
   MORNING
LATE AFTERNOON
            -PLACEBO DAY
             DRUG DAY
 Fig.  2.  Composite averaged somatosensory evoked
 potentials for 11 subjects during four time periods:
 predrink (704 trials),  morning (352  trials), midday
 (1056 trials), and late afternoon (1056 trials). Epoch
 length  was 500 msec.  Stimulus was  at trace onset.
 Reference level drawn was  through  averaged EEC
 50-msec poststimulus.


 Alcohol and evoked potential amplitude

     Alcohol  clearly reduced the amplitude of the
 vertex-recorded potentials (Table 1). Fluctuations are
 expressed in  Table  1 In terms of percentage change
 from predrink amplitude, usually percentage decre-
 ment. On the placebo day, both peak-to-peak measures

-------
  394
                                      Scales et al.
              Table 1. Percentage Decrements from  Predrink  Period of  SEP Amplitude
                               on the Placebo (P) and Drug (D) Days

(P) Mean
(P)S.D.
(P) t -test
(P)Sig.8
(D) Mean
(D)S.D.
(D)t-test
(D)Sig.a
P1N1
Morning
5
26
1.638

-------
Ethanol and Somatosensory EPs
                                            395
   Table 2. Drinking History and SEP Amplitude: Percentage Decrements from Predrink Period
                             on the Placebo (P) and Drug (D) Days

Mean (P) of light drinkers
S.D.
Mean (P) of heavy drinkers
S.D.
U
Sig.
rho
Associated _t -value
Sig.
Mean (D) of light drinkers
S.D.
Mean (D) of heavy drinkers
S.D.
U
Sig.
rho
Associated jt-value
Sig.
P1N1
Morning
9a
28
17
18
6
(ns)
-0.450
-1.512
(ns)
67
9
9
41
0
0.004
0.791
3.879
<0.01
Midday
8
30
20
14
14

-------
  396
                                     Scales et al.
                Table 3.  Controlled Variables in Comparing Light and Heavy Drinkers

Mtin of light drlnkin
3.0.
Mean of hiivy drinkiri
S.O.
U.
Slj.
rho
Anoclitid i-vilu*
81,.
Am,
yr
30
3
33
4
6
(ni)
0.553
1.991
(ni)
Weight,
kg
73
8
82
11
8
(ni)
0.379
1.225
(ni)
Main predrink SEP amplitude, [M
Placebo
P1N1
17
8
21
12
9
(ni)
0.524
1.848
(nil
N1P2
40
13
35
IB
10
(ni)
•0.192
•0.565
(ni)
Drug
P1N1
17
7
19
12
15
(m)
0.223
0.696
Ini)
N1P2
36
8
30
16
9
(nt)
•0.205
•0.628
(ni)
Peak
BAC,'
my%
90
18
96
8
7.5
(ni)
0.328
1.042
(nil
Time to
Peek
BAG,'
hr
1.41
0.28
1.52
0.28
11
Ini)
0.009
0.027
Ini)
BAC slope,8
mg%/hr
Ascending
58,7
S8.0
61.6
11.9
12
(ni)
0.309
0.975
(ni)
Descending
1S.4
3.7
16.0
2.6
15
(ni)
0.214
0.657
(ni)
Mean predrink
BAC, mg%
Placebo
5
3
6
4
7.5
(ns)
0.696
2.238
0.06
Drug
9
9
6
3
11.6
(ni)
•0.475
•1.619
(nil
 *0rug day, Ounn-6onl«rrono criterion: 0.08/12-0.004.
 In tolerance preceding alcholiun, even though the bio-
 chemical bases of tolerance are still under investiga-
 tion.

     The correlation between drinking  history and
 decrement  in amplitude of N1P2 during the morning
 period of  the placebo day is puzzling.  The gradual
 amplitude decline noted for all subjects on the placebo
 day appean to be a long-term habitation. This inter-
 pretation it supported by the full recovery of ampli-
 tude in the predrink period of the drug  day. Why
 would individuals with a history of moderately heavy
, drinking and, by Inference, with a relatively high toler-
 ince toethanol, consistently show greater habitation
 than light drinkers to 11 repetitions of a short series
 of electrical stimulations? Predrink BACi are an unlike-
 ly factor since, on  the average, they fell within the
 10 mg% noise range of the Breathalyzer (Table 3).
 The 6 ml of ethanol in placebo drinks may have had
 an Influence. The different rates of habitation may
 have occurred  by chance. Or, It may be that habitue-
 tlon and tolerance have features in common and that
 rate of habituttion cut be uted at • measure of toler-
 ance without the administration of ethanol.

     There  are  only a few objective measures of toler-
 ance that can  be obtained and quantified in a clinical
 setting (e.g., optokinetic nystagmus, Mlzoi et al. 1969).
 At a result, the process of  tolerance is poorly under-
 stood! and the relation of tolerance to addiction is
unknown.  Alcohol rehabilitation centers  could use
noninvasive, easily administered measures of tolerance
to determine the amount of LJbrium or other drug
necessary to sustain patients while at various stages of
withdrawal (SchucWt 1975). Measures of tolerance
could also  be used on general medical  and surgical
wards to determine whether  surgical candidates sus-
pected of alcoholism are denying the ailment and,
thereby, are placing themselves in danger of receiving
an insufficient dose of anesthesia. SEP measures, in
combination  with other  measures, might provide a
useful clinical index of chronic tolerance to ethanol.
Summary

    This study examined the relationship of somato-
sensory evoked potentials (SEP) and ethanol tolerance.
Eleven male subjects were  chosen  on the basis  of
drinking  histories  ranging from light  to moderately
heavy, Vertex SEPs to median nerve stimulation were
recorded on placebo and drug days, Ethanol (0,9 g/kg)
was  administered  orally to  obtain  an avenge peak
blood alcohol concentration of 94 mg% derived from
 Breathalyzer samples. The reduction  in SEP amplitude
during ethanol  intoxication compared to  placebo
measures was inversely related to drinking habit (I.e.,
less reduction in heavy drinkers). Results suggest that
the SEP may serve as a sensitive index of chronic toler-
ance to ethanol.

-------
STIMULANT AND DEPRESSANT EFFECTS
OF CIGARETTE SMOKING, NICOTINE,  AND
OTHER  DRUGS ON  THE CNV IN  MAN1
H. ASHTON, J.E. MILLMAN, M.D. RAWLINS, R. TELFORD, AND J.W. THOMPSON


Clinical Psychopharmacology Unit, Department of Pharmacological Sciences, University of New-
castle upon Tyne, England
     Although vast  numbers of people smoke ciga-
 rettes, no one, including the smokers, seems to know
 why.  Undoubtedly there is no simple explanation.
 Smokers often  comment  that they feel  either 're-
 laxed' or 'stimulated' after  a cigarette, although
 direct evidence of depressant or stimulant changes in
 the brain is  difficult to obtain. An obvious way to
 study this problem is to examine the  electroen-
 cephalographic   changes   that  follow   smoking.
 Murphree et al. (1967)  and Lambiase  and Serra
 (1957) found,  in  general, only stimulant effects.
 However, Armitage et  al. (1969)  reported both
 stimulant and depressant effects of cigarette smoke
 and  nicotine in animals,  both  on electrocortical
 activity and  on the release of acetylcholine from the
 cerebral cortex. The biphasic action of nicotine, i.e.,
 its ability to first stimulate and then depress nervous
 tissue, has been known for a long time.  This  fact,
 coupled  with statements of smokers concerning the
 stimulant and  relaxant effects of nicotine and the
 findings  of Armitage et  al (1969) suggested that a
 biphasic  action of nicotine in cigarette smoke ought
 to be detectable in  the human brain. The contingent
 negative  variation (CNV), first described by Walter
 et al. (1964),  seemed to provide an appropriate
 measure, and the hypothesis was made that stimulant
 effects of nicotine on the brain might increase the
 magnitude of the CNV, whereas depressant effects
 might decrease it.

     In order to study rigorously the action of nico-
 tine  in cigarette smoke, three series of CNV experi-
 ments were performed:  (1) a study of the effects of
 cigarette smoking,  (2) a study of the effects of drugs
 with known central stimulant and depressant actions,
 1 This work was supported by the Tobacco
 Research Council of England.
e.g., caffeine and nitrazepam, and  (3) a study of the
action of pure nicotine administered intravenously.

Cigarette smoking
    Cigarette smoking was associated with significant
changes in CNV magnitude which either increased or
decreased after smoking (Ashton et al. 1973). Twen-
ty-two  regular smokers were studied and CNV in-
creased in 7, decreased in 11, (and showed biphasic
changes in 4,  e.g. Fig. 1). If an increase or decrease
in CNV magnitude  reflects stimulant or depressant
actions on brain activity, this study indicates that cig-
arette smoking can exert both effects in man. These
findings are compatible with earlier studies on human
performance in a car simulator which showed that
reaction time  could be either increased or decreased
by smoking (Ashton et  al.  1972a,  1972b). There
was no correlation between the effects of cigarette
smoking on the CNV and changes in heart rate, blood
pressure, skin temperature, or blood carboxyhemo-
globin. Moreover, sham smoking of an unlit cigarette
had no effect  on the CNV, suggesting that changes in
CNV magnitude associated with smoking were in fact
due to central actions  of nicotine (Ashton et al.
1974).

    From experimental work in animals, it is known
that the effects of nicotine are dose-dependent, i.e.,
smaller doses  cause central stimulation whereas larger
doses cause depression. There is strong evidence that
human smokers smoke for  some optimum dose of
nicotine by appropriate  alterations in their smoking
behaviour, particularly  puffing  rate  (Ashton  and
Watson 1970). The dose of nicotine varies according
to circumstances: a smaller  dose was taken when
subjects were  carrying out a task,  whereas larger
amounts of nicotine were taken after the task. An
important  additional determinant  of nicotine dosage
is the personality of the smoker.  Of 16 smokers,

-------
 398
                                   Ashton et at
                  PRESMOKING
                SMOKING 2nd HALF      I
                                                   500msec
Fig. 1.  CNVs (average often trials) in one subject showing decrease after smoking first and second half of one
cigarette,  with postsmoking recovery.  CNV magnitude ffJ. V-sec) is indicated. SI = flash (0.3 joules, 100 msec).
S2 = tone (500 Hz). Calibration as shown. Negative up. (Reprinted from Ashton et al. 1978a by permission of the
publisher.)
the 8 more extraverted subjects has a lower rate of
nicotine  intake, which produced a stimulant effect
on the CNV, whereas the 8 more introverted subjects
had a higher rate of nicotine intake, which depressed
the CNV (Ashton  et  al.  1974).  From these results,
it  was concluded that smokers unconsciously select
different doses of  nicotine from a cigarette  in order
to achieve some optimum dose that meets their  own
requirements,   determined by  circumstances   and
personality. AsArmitage et al.(1968) stated, smokers
have  literal finger-tip control  of  nicotine  intake.

Caffeine and nitrazepam

    When  it  became evident  that nicotine could
either increase or decrease the magnitude of the CNV,
it was considered essential to validate these results by
examining  the effects of known central stimulant and
depressant  drugs on  the  CNV.    It has since  been
shown repeatedly that nitrazepam (2.5 mg) decreases
and caffeine citrate (300 mg) increases CNV magni-
tude (Ashton et al. 1974).

    These  small doses, of nitrazepam and  caffeine,
although producing significant changes in CNV  mag-
nitude, did not cause significant alterations in heart
rate  or fingertip  temperature  and had  virtually no
subjective effects.  It was  concluded that the CNV
was  a  sensitive and reliable indicator of drugs  with
central  stimulant and  depressant actions.   Further
experiments using other centrally acting drugs, such
as pemoline and  diazepam, have provided addition-
al supporting evidence  (Ashton et al. 1976  and un-
published results).

Pure nicotine administered intravenously

    Having  demonstrated  that  cigarette  smoking
produced both stimulant  and   depressant   effects
on brain activity in  man, measured in terms  of CNV
effects, the next step was to determine  whether the
effects of cigarette  smoking could be accounted for
by the  nicotine content of the smoke.

    Armitage  et al. (1969) had  shown that  intrave-
nous (iv) nicotine in animals can  mimic the effects of
cigarette smoking if the drug is given as  intermittent
'shots'  but not if it is  administered as a continuous
infusion.  Volunteers were  therefore given intermit-
tent  iv injections of nicotine and physiological saline
(as a  control) using  the  technique described by
Armitage et al. (1974). The dose of nicotine in each

-------
Smoking, Nicotine, Drugs and CNV

                                   si
              SALINE 1
              (1-ml "shot" each min) x 5
                                                                                                  399
               SALINE 2
               (1-ml "shot" each min) x 5
               SALINE 3
               (1-ml "shot" each min) x 5
                NICOTINE 1
                (150-pg"shot"eachmin)x5
                                                   500msec
                NICOTINE 2
                (150-M9 "shot" each min) x S
                 POSTNICOTINE SALINE 1
                 (1-ml "shot" each min) x S
                                                    500msec
Fig. 2.  CNV (average of ten trials) in one subject showing responses to six series of intravenous 'shots'(five per
series) of physiological saline and nicotine {5 x 150 jiG = 750 tig total). CNV magnitude (n V-sec) is indicated,
SI and S2 as for Fig. 1. Negative up. (Reprinted from Ashton etal. 1978b by permission of the publisher.)

-------
  400
                                   Ash ton et al.
 shot and the total amount of nicotine were compar-
 able to  that  obtained  by a cigarette smoker who
 inhales when  smoking a cigarette of a brand  with
 medium  nicotine yield  (1  to 2 mg) (Ashton et al.
 1975, 1978).

     After each series of shots of nicotine or physi-
 ological saline, the CNV was measured. Fig. 2 shows
 records  obtained  from   one  subject  who  received
 five shots of 150 ;ug nicotine  on two occasions after
 control saline  shots.  There  was a small reduction
 in  CNV magnitude after the first dose of nicotine
 and a 50%  reduction in CNV  magnitude after the
 second dose.  During the  post-drug saline  control,
 the CNV magnitude returned to normal.

     During the initial experiments with intravenous
 nicotine, a single dose was administered to each sub-
ject based upon his  smoking habits.  Under  these
 conditions,  nicotine  produced an  increase in  CNV
magnitude in some subjects  and  a decrease in others.
 It  then became important to determine whether the
response of each subject  depended on dose or wheth-
er it depended on  the individual and was largely or
completely  independent   of dose.   Further  experi-
ments  were  carried out  using  a  range of nicotine
doses, and Fig. 3 shows the dose-response relationship
in one subject; the  relationship was characteristic for
all subjects in the series.   The dose-response curve is
one of an  unusual form in that the magnitude of the
CNV at first  increases with  increasing doses of nico-
tine, but thereafter further increases led to increasing
reductions in the magnitude of the CNV.

    The results of the experiments using intravenous
nicotine indicated that (1)  the effects produced by
cigarette smoking and pure  nicotine are very similar
and  (2) nicotine  is capable  of causing a biphasic
response that is dose-related.   This biphasic  effect
of nicotine is not unique.  For  example, Tecce and
Cole (1974) reported that 10 mg of amphetamine can
produce behavioural  alertness and  increased  CNV
amplitude   in some subjects,  whereas in others it
produces initial drowsiness and reduces  the  CNV
               »      so      100     no
              TOTAL iv NICOTINE (LOG SCALE)

Fig.  3.  Typical dose-response curve showing effect of
intravenous nicotine (each dose given as five "shots ").
Note:   Highest dose given  in separate experiment.
(Reprinted from Ashton et  al. 1978b by permission
of the publisher.}

More recently, Pirch (this volume) has shown that
amphetamine causes  a biphasic dose-related effect on
cortical slow potentials recorded from unanesthetised
rats.  All these findings, taken with the present  re-
sults, emphacize  the  importance of  dosage  when
studying the effects of drugs on event-related slow
potentials.

Conclusions

    In summary, it was concluded that changes in the
magnitude  of  the CNV produced  by cigarette smok-
ing are principally due  to nicotine absorbed from the
inhaled  smoke and  that  nicotine can exert  either
stimulant or depressant effects on the  human  brain.
From the  combined results of the  series  of experi-
ments,  it  was  also  concluded that the CNV is a
sensitive, reproducible, and quantitative test for the
central effects  of drugs; that it has been the means of
throwing light  on  the effects of cigarette smoking on
the human brain; and  that it  has given considerable
insight into the reasons why people smoke.

-------
PSYCHOPATHOLOGY REFERENCES
Abraham,  P., McCallum, W.C. and Gourlay, I. The
   CNV and its relation to specific psychiatric syn-
   dromes. In W.C. McCallum and J.R. Knott (Eds.),
   The responsive brain. Wright, Bristol, 1976:144-
   149.

Armitage,  A.K.,  Hall, G.H.  and  Morrison,  C.P.
   Pharmacological basis for the smoking habit. Na-
   ture (London), 1968, 217:331-334.

Armitage,  A.K., Hall, G.H. and Sellers, C.M. Effects
   of nicotine on electrocortical activity and acetyl-
   choline release from cat  cerebral cortex. Brit. J.
   Pharmac., 1969,35:152-160.

Armitage,  A.K., Dollery, C.T., George, C.F., House-
   man, T.H., Lewis, P.J.  and Turner, D.M. Absorp-
   tion and  metabolism of  nicotine by man during
   cigarette  smoking.  Brit.  J. Clin.  Pharm.,  1974,
   1:180P-181P.

Ashton, H. and Watson, D.W. Puffing frequency and
   nicotine intake in cigarette smokers. Brit. med. J.,
   1970,3:679-681.

Athlon, H.,  Savage, R.D., Thompson, J.W. and Wat-
   son, D.W. A method for measuring human behav-
   ioural  and physiological  responses at  different
   stress levels  in a driving  simulator. Brit. J. Phar-
   mac., 1972a, 45:532-545,

Ashton, H.,  Savage,  R.D., Telford, R., Thompson,
   J.W. and Watson, E.W. The  effects of  cigarette
   smoking on the response  to stress in a driving sim-
   ulator.  Brit.  J.  Pharmec.,  1972b, 45:546-556.

Ashton, H.,  Millman, J.E., Telford, R. and Thompson,
   J.W. Stimulant and depressant effects of cigarette
   smoking  on  brain activity  In man.  Brit. J.  Phar-
   mac., 1973,48:715-717.

Ashton, H., Millman, J.B., Telford, R. and Thompson,
   J.W. The effect of caffeine, nltraupam and cigar-
   ette smoking on the contingent negative  variation
   in man. Electroenceph, din, Neurophysiol,, 1974,
   37:59-71.

Ashton, H.,  Millman, J.E., Telford, R. and Thompson,
   J.W, The contingent negative variation: an ob-
   jective  electroencephalographlc method for meas-
   uring the effects of centrally acting drugs on brain
   activity in man,  Brit, J. Clin. Pharmac,, 1975,
   2:177-178.

Ashton, H., Millman, J.B., Telford, R. and Thompson,
   J,W, A comparison of some physiological and psy-
   chological effects of propranolol and dlaispam in
   normal subjects. Brit. J. Clin, Pharmac,,  1976,
   3:551-539,
Ashton, H., Marsh, V.R., Millman, J.E., Rawlins,
   M.D., Telford, R. and Thompson, J.W. The use of
   event-related slow potentials of the brain  as a
   means to analyse the effects of cigarette smoking
   and nicotine in humans. In R£. Thornton (Ed.),
   Smoking  Behavior.  Churchill Livingstone, Edin-
   burg, 1978a.

Ashton, H., Millman, I.E. Rawlins, MD. Telford, R.
   and Thompson, J.W. The use of event-related slow
   potentials of the brain in the analysis of effects of
   cigarette smoking and nicotine in humans. In K.
   Battig (Ed.), Behavioral Effects of Nicotine. Int.
   Workshop Behav. Effects  of Nicotine.  Karger,
   Basel, 1978b:26-37.

Baker, H.L., Jr. The impact of computed tomography
   on  neuroradiologic  practice.  Radiology,  1975,
   116:637-640.

Baker, H.L., Jr., Houser, O.W., Campbell, J.K. Reese,
   D.F. and Holman, C.B. Computerized tomography,
   of the head. J.A.M.A., 1975, 233: 1304-1308.

Begleiter, H. and Platz, A. Evoked potentials: modifi-
   cations by  classical conditioning.  Science, 1969,
   166:769-771.

Braff,  D., Callaway, E. and Naylor, H. Very short-
   term memory dysfunction in schizophrenia. Arch.
   gen. Psychiat., 1977,34:25-30.

Brodmann, K. Vergleichende  Lokalisationslehre der
   Grosshirnrinde In ihren Prtnziplen dargestellt auf
   Grund des Zellenbaues. F.A. Berth, Leipzig, 1909
   and 1925.

Buchsbaum, M. and Pfefferbaum, A. Individual differ-
   ences in stimulus intensity response. Psychophys-
   iol., 1971,8:600-611.

Cahalan,  D. and Cisln, I.H. American drinking prac-
   tices: Summary of findings from a national proba-
   bility sample. Q. J.  Stud. Alcohol., 1968,29:130-
   151.
Callaway, E. Brain electrical potentials and Individual
   psychological differences,  Orune  and  Stratton,
   New York,  1975.

Cattell, R.B. IPAT, Self Analysis Form. Institute for
   Personality  and  Ability  Testing,  Champaign,
   Illinois,  1963:1-4.

Cohen, J, Very slow brain potentials relating to ex-
   pectancy;  The  CNV.  In B.  Donehin and  D.B.
   Llndiley  (Eds.),  Averaged  evoked  potentials:
   Methods, results and evaluations. National Aero-
   nautics  and  Space Administration, Washington,
   D.C..NASASP-191, 1969:143-198.

-------
 402
                   Psychopathology References
 Cooper,  R. An  ambiguity  of bipolar recordings.
    Electroenceph. din. Neurophysiol., 1959, 11:819-
    820.

 Donchin, E. Average evoked potentials and uncertain-
    ty resolution. Psychon. Sci. 1968, 12:103.

 Dongier,  M. Clinical application of the CNV:  a re-
    view.  Electroenceph.  clin. Neurophysiol., Suppl.
    33, 1973a:309-315.

 Dongier,  M. Event related slow potential changes in
    psychiatry.  In  S. Bogoch  (Ed.), Biological  di-
    agnosis of brain disorders. Spectrum,  Flushing,
    New York, 1973b:47-59.

 Dongier,  M., Timsit-Berthier M., Koninckx,  N. and
    Delaunoy,  J. Compared  clinical significance of
    CNV and other slow potential  changes  in psy-
    chiatry.  In W.C. McCallum and J. Knott (Eds.),
    Event  related slow potentials of the brain: Their
    relation  to behavior.  Elsevier, Amsterdam, 1973:
    321-327.

 Dubrovsky, B.  and Dongier,  M. Evaluation of event-
    related slow potentials in selected groups of psychi-
    atric patients. In W.C. McCallum and J.R. Knott
    (Eds.),  The  responsive  brain.  Wright, Bristol,
    1976:150-153.

 Dubrovsky, B.  and Garcia-Rill, E.,  Convergence of
    tectal and visual input in cat motorsensory  cortex.
    Expl. Neurol., 1971, 33:475484.

 Dubrovsky, B,  and Melzack,  R. Prolonged changes
    in single cell  activity in lateral geniculate nucleus.
    Expl. Neurol., 1970, 28:123-131.

 Dubrovsky, B., Garcia-Rill, E.,Tepper,J.and Vanagas,
   J. Some aspects of the methodology  of recording
   slow potential changes. In S. Bogoch (Ed.), Biologi-
   cal diagnosis  of  brain disorders. Spectrum, Flush-
   ing, New York, 1973:300-309.

 Dubrovsky, B.,  Simkus, R.,  Garcia-Rill,  E.  and
   Dongier, M.  Effects of changes in carbon dioxide
   tension in abnormally prolonged  contingent nega-
   tive variation. Bio. Psychiat. 1976, 11:535-541.

 Dunn,  O.J. Confidence  intervals for the means  of
   dependent,  normally  distributed variables.  Am.
   Statistn. Assn., 1959, 54:613-621.

 Ennever, J.A.  Intracortical transmission of specific
   evoked activity across  the cerebral cortex. Electro-
   enceph. clin.  Neurophysiol.,  1975,   38:281-293.

Freeman,  W. and Watts, J.W. Interpretation of func-
   tion of frontal lobes  based  upon observation  in
   48 cases of  prefrontal lobotomy. Yale J. Biol.
   Med., 1939, 11:527-539.

Freeman, W. and Watts, J.W. Retrograde degeneration
   of the  thalamus following prefrontal lobotomy.
   J. Comp. Neurol., 1947, 86:65-93.
 Freeman W. and  Watts, J.W. Psychosurgery in the
    treatment of mental disorders and intractable pain
    (2nd ed.). Charles C. Thomas, Springfield,  111.,
    1950.

 Freeman. W.  and Watts,  J.W. The thalamic projec-
    tion  to  the  frontal lobe.  In Association for re-
    search in nervous and mental disease: The frontal
    lobes, Hafner, New York, 1966:200-209.

 Gauther, P. and Gottesmann, C. Etude de la Varia-
    tion  Contingente   Negative  et  de 1'onde post-
    imperative  en  presence  d'inteferences.  Electro-
    enceph.  clin.  Neurophysiol., 1976, 40:143-152.

 Gazzaniga,  M.S. and  Hillyard, S.A. Attention mech-
    anisms  following  brain  bisections. In  S. Korn-
    blum (Ed.), Attention and performance, Vol. IV.
    Academic Press, New York, 1972:221-238.

 Gross,  M.M., Begleiter,  H.,  Tobin, M. and Kissin,
    B. Changes in auditory evoked response induced
    by alcohol. J. nerv. ment. Dis., 1966,143:152-156.

 Head,  H. Studies in neurology.  Oxford Univ. Press,
    London, 1920.

 Hillyard, S.A. The  CNV and human behavior: A re-
   view. Electroenceph.  clin.  Neurophysiol., 1973
   Suppl. 33:161-171.

 Hillyard, S.A. and Galambos, R. Eye movement arte-
    fact  in  the  CNV.  Electroenceph. clin.  Neuro-
    physiol., 1970,28:173-182.

 Jewett,  D.L.,  Romano,  M.N.  and Williston,  J.S.
   Human auditory evoked potentials: Possible brain
   stem  components detected  on the scalp.  Science
   1970,167:1517-1518.

 Jones,  B.M. Memory  impairment on  the  ascending
   and descending limbs of the blood alcohol curve.
   J. abnorm. soc. Psychol., 1973, 82:24-32.

 Jones,  B.M. and  Vega,  A. Cognitive performance
   measured on  the ascending and  descending limb
   of the blood alcohol curve.  Psychopharmacologia,
   1972,23:99-114.

 Kalant, H., LeBlanc, A.E. and Gibbins, R.J. Tolerance
   to, and dependences on, some non-opiate psycho-
   tropic drugs.  Pharmac. Rev., 1971, 23:135-191.

 Kelly,  D.M.W.  and Walter,  C.J.S. The  relationship
   between clinical diagnosis amd anxiety assessed by
   forearm   blood   flow  and  other measurements.
   Brit. J. Psychait., 1968,  114:611-626.

Kendall,  M.G. and Stuart, A.E. The advanced theory
   of statistics, vol. 2. Griffin, London, 1967:690.

Klorman, R. and Bentsen, E.  Effects of warning sig-
   nal duration on the early and late components of

-------
Psychopathology References
                                            403
   the  contingent negative variation. Biol. Psychol.,
   1975,3:263-275.
Knott, J.R. and  Irwin, D.A. Anxiety, stress and the
   contingent negative variation. Electroenceph, clin.
   Neurophysiol., 1967, 22:188.
Malmo, R.B. and Davis. J.F. Electromyographic stud-
   ies  of muscular tension in psychiatric patients
   under  stress. J. clin.  exp. Psychopath.,  1951,
   12:45-66.

Marks, I.M. Fears and phobias. Heineman Medical,
   London, 1969.
 Knott J.R. and  Irwin, D.A. Anxiety, stress and the
   contingent negative  variation  (CNV). Electro-
   enceph. clin.  Neurophysiol., 1968,  24:286-287.


 Knott, J.R.  and Irwin, D.A. Anxiety, stress and the
   contingent negative variation. Arch. gen. Psychiat.,
    1973,29:538-541.

 Lader,  M.H. Palmar skin conductance  measures on
   anxiety and  phobic states. J.  Psychosom.  Res.,
   1967, 11:271-281.

 Lader,  M.H.  Psychophysiological aspects of anxiety.
   In M.H. Lader (Ed.), Studies of anxiety. Headley
   Brothers Ltd., Kent, 1969:53-61.


 Laffont, F., Jusseaume, Ph., Bruneau, N. and Lelord,
   G. Aspects differents de conditionnement des acti-
   vities evoquees par couplage du son et delalumiere
   chez 1'Homme. J. Physiol., Paris, 1972, 65:434-
   435.

 Lambiase, M. and Sena, C. Fume e sistema nervoso.
   1. Modificazioni dell1 attivita electrica corticate da
   fumo.  Acta Neurol. (Napoli), 1957,  12:475493.

 Lelord  G.  and Maho, C.  Modifications  des activites
   evoquees  corticales et  thalamiques au cours d'un
   conditionnement    sensoriel:   Localisation des
   responses  et  variations avec la  vigilance. Electro-
   enceph. clin.  Neurophysiol., 1969,  27:258-262.

 Lelord. G.,  Calvet, J.  Fourment, A.  and Scherrer, J.
   Le conditionnement de la  reponse evoquees elec-
   trocorticale chez 1'Homme. C.R. Soc. Biol., 1958,
   152:1097-2000.

 Lelord G., Laffont, F. and Jusseaume, Ph. Condition-
   ing  of evoked potentials  in children of differing
   intelligence. Psychophysiol., 1976, 13:81-85.

 Lewis,  E.G., Dustman, R.E.  and Beck,  B.C.  The ef-
   fects of  alcohol  on  visual  and  somato-sensory
   evoked  responses.  Electroenceph.  clin.  Neuro-
   physiol., 1970,28:202-205.

 Loveless, N.F. and Sanford, A.J. The impact of warn-
   ing  signal intensity on reaction time and  compo-
   nents of  the contingent negative variation. Biol.
   Psychol.,  1975,2:217-226.

 Low, M.D. and  Stoilen, L. CNV and EEC in children
   maturational  characteristics  and   findings  in the
    MCD  syndrome.  Electroenceph.  clin.  Neuro-
    physiol.,  1973, Suppl. 33:139-143.
McAdam,  D.W.  Changes in somato-sensory  evoked
   potentials during the contingent negative variation
   (CNV). Electroenceph. clin. Neurophysiol., 1968,
   24:286.

McAdam,  D.W. Increases in CNS excitability during
   negative cortical slow potentials in man. Electro-
   enceph. clin. Neurophysiol.,  1969, 26:216-219.

McAdam,  D.W., Knott, J.R. and Rebert, C.S. Cortical
   slow potential changes in man related to interstim-
   ulus interval  and to  pre-trial  prediction of inter-
   stimulus  interval.  Psychophysiol., 1969, 5:349-
   358.

McCallum, W.C. and Walter, W.G.The effects of atten-
   tion and distraction on the contingent  negative
   variation in normal and  neurotic subjects. Electro-
   enceph. clin. Neurophysiol.,  1968,  25:319-329.

McCaUum, W.C., Papakostopoulous,  D., Grombi, 1.,
   Winter. A.L., Copper, R. and  Griffith, H.B. Event
   related slow potential  changes in human  brain
   stem. Nature, 1973, 242:465-467.

McSherry, J.W. Physiological mechanisms of the CNV:
   Origins -  A  review. Electroenceph.  clin. Neuro-
   physiol., 1973, Suppl. 33:53-61.

Mendelson, J.H. Biochemical mechanisms of alcohol
   addiction. In B. Kissin and H.  Begleiter (Eds.). The
   biology  of  alcoholism,  Vol.  I, Biochemistry.
   Plenum, New York, 1971:513-544.

Miller, L.H., van Veen, W.,  Sandman, C.A. and Knott,
   J.R. Intermodal  and intramodal  distraction and
   CNV.  Electroenceph. clin. Neurophysiol.,  1973,
   Suppl., 33:209-211.

Mizoi, Y., Hishida, S.  and Maeba,  Y.  Diagnosis of
   alcohol intoxication by the optokinetic test. Q.J.
   Stud. Alcohol, 1969, 30:1-14.

Murphree, H.B., Pfeiffer, C.C. and Price, L.M. Electro-
   encephalographic changes in man  following smok-
   ing. Ann.  N.Y.  Acad.  Sci.,  1967, 142:245-260.

 Papakostopoulos, D., Winter, A., and Newton, P. New
   techniques for the control of eye potential artefacts
   in multiple-channel CNV recording. Electroenceph.
   clin. Neurophysiol., 1973, 34:651-653.

 Peters, J.F., Knott, J.R., Miller, L.H., van Veen, W.J.
   and Cohen, S.I. Response variables and magnitude
   of  the  contingent  negative variation. Electro-
   enceph.  clin. Neurophysiol., 1970, 29:608-611.

-------
404
                  Psychopathology References
Picton, T.N. and Hillyard, S.A. Cephalic skin poten-
   tials   in  electroencephalograph.  Electroenceph.
   clin. Neurophysiol., 1972, 33:419-424.

Raskin, D.C., Kotses.H. andBever, J. Autonomic indi-
   cators of orienting and defensive reflexes. J. exp.
   Psychol., 1969, 80:423-433.

Rohrbaugh, J.W., Syndulko,  K. and Lindsley, D.B.
   Brain  wave  components of the contingent negative
   variation in humans.  Science,  1976, 191:1055-
   1057.

Roth, W.T. Late event-related potentials and psycho-
   pathology.  Schizophrenia Bull., 1977, 3:105-120.

Roth, W.T., Kopell,  B.S., Tinklenberg, J.R., Darley,
   C.F., Sikora, R. and Vesecky, T.B. The contingent
   negative variation during a memory retrieval task.
   Electroenceph. clin. Neurophysiol., 1975,38:171-
   174.
Salamy, A. and Williams, H.L. The effects of alcohol
   on sensory evoked and spontaneous cerebral poten-
   tials in  man. Electroenceph. clin. Neurophysiol.,
   1973,35:3-11.

Schucklt, M.A. Drugs in combination with other ther-
   apies for alcoholism. In M. Grenblatt (Ed.), Drugs
   in combination with other  therapies. Grune and
   Stratton, New York, 1975:119-134.

Shagass,  C, Evoked brain  potentials  in psychiatry.
   Plenum, New York,  1972.

Shagasi, C. An  electrophysiological view of schizo-
   phrenia. Presidential address, Soc. Biol. Phychiat.,
   New York, June 197 5.

Shagass, C. An  electrophysiological view of schizo-
   phrenia. Biol. Piychiat,, 1976,11:3-30.

Skinner, J.E. Abolition of conditioned, surface nega-
   tive, cortical  potential during cryogenic blockade
   of non-specific thaltmocorticel system. Electroen-
   ceph. clin. Neurophysiol., 1971,31:197-209.

Skinner,  J.E. and  Lindsley, D.B.  The  nonspecific
   medtothalamlc frontocortlcal system: iti influence
   on  electrocortical activity and behavior. In  K.H.
   Pribram  ind A. R. Luria (Edi.), Piychophyslology
   of the frontal lobes, Academic  Press, New York,
   1973:183-234.

Skinner, J.E. and YingHng, C.D. Regulation of slow
   potential ihifti in  nuoleui reticularii thalami by
   the meiencephilic  reticular formation and the
   frontal granular cortex, Electroenceph. din, Neuro-
   phyalol., 1976,40:288-296.

Small, J.G.  and Small,  l.F. Contingent negative varia-
   tion (CNV) correlations with psychiatric diagnosis,
   Arch, gen, Piychiat,, 1971,25:530-554,
Solyom, L., Heseltine, G.F.D., McClure, D.J., Solyom,
   C., Ledwidge, B. and Steinberg, G. Behavior ther-
   apy versus drug therapy in the treatment of phobic
   neurosis. Can. psychiat. Ass. J.,  1973, 18:25-32.

Straumanis,  J.J., Shagass, C. and Schwartz, M. Prob-
   lems associated with application of the contingent
   negative variation to  psychiatric research. J. nerv.
   ment. Dis., 1969, 148:170-179.

Tecce, J.J. Contingent negative variation and individ-
   ual differences: A new approach in brain research.
   Arch. gen. Psychiat., 1971, 24:1-16.

Tecce, J.J. Contingent negative variation (CNV) and
   psychological processes in man.  Psychol. Bull.,
   1972,77:73-108.

Tecce, J.J.  Additional  electrophysiological  studies:
   Electrical brain activity (contingent negative varia-
   tion) and related neuropsycholical functions.  In
   Report  and  recommendations:  Psychosurgery.
   Appendix: Psychosurgery. The National Commis-
   sion for the Protection of Human Subjects of Bio-
   medical and Behavioral Research. DHEW  Publica-
   tion No.  (OS) 77-0001. Washington, D.C., 1977:
   11-44 to 11-64.

Tecce, J.J. and  Cole, J.O. Amphetamine effects in
   man: Paradoxical drowsiness and lowered electri-
   cal brain  activity (CNV). Science, 1974, 185:451-
   453.

Tecce, J.J. and Cole, J.O. The distraction-arousal hy-
   pothesis,  CNV, and schizophrenia. In D.I. Moatof-
   sky (Ed.), Behavior  control and modification of
   physiological  activity.  Prentice-Hall, Englewood
   Cliffs, N.J., 1976:162-219.

Tecce, J.J. and Hamilton, B.T. CNV reduction by sus-
   tained  cognitive  activity (distraction).  Electro-
   enceph. clin. Neurophyaiol., 1973, Suppl.  33:229-
   237.

Tecce, J.J., Savignano-Bowman, J. and Melnbresse, D,
   Contingent negative variation and the distraction-
   arousal  hypothesis,  Electroenceph. clin. Neuro-
   physlol.,1976,41:277-286.

Tecce, J,J,, Cole, J.O., Mayer, J. and Lewis, D. Barbit-
   urate effects on  brain functioning  (CNV) and at-
   tention  performance In  normal  men.  Psycho-
   phirmacol. Bull,, 1977,13:64-66.

Tecce, J.J,, Savignano-Bowman,  J. and  Cole,  J.O.
   Drug effects on contingent negative variation  and
   eyeblinki: the diitraction-arouial  hypothesis In
   M.A, Lipton, A, DiMascio and  K.P. KiUam (Edi.),
   Piychopharmacology-A  generation of progress.
   Raven,  New York, 1978:745-758,

Tecce, J.J, Cole, J.O., Mayer, J. and Lewis, D. Math-
   odone effects on human brain  functioning (CNV)
   and attention behavior. Payohopharmacol, Bull,, in
   preai.

-------
 Psychopathology References
                                             405
Timsit-Berthier, M. La variation contingente negative
   et  les maladies mentales. These  de doctoral  es
   sciences naturelles. Paris, 1975.

Timsit-Berthier, M. and Hausman, J. Etude de la VCN
   et du phenomene de preparation motrice chez des
   enfants de  5  a  15  ans. Rev. Electroenceph.  clin.
   Neurophysiol., 1972, 2:141-146.
Timsit-Berthier M., Koninckx, N., Dargent, J., Fon-
   taine, O. and Dongier, M. Etude de la duree de la
   VCN chez un groupe de sujets normaux, un groupe
   de nevroses et  un groupe de psychotiques. In J.
   Dargent and  M. Dongier (Eds.), Variations  con-
   tingentes negatives. Congres et  Colloques de 1'Uni-
   versite de Liege, 1969: 206-214.
Timsit-Berthier, M.,  Koninckx, N., Dargent, J., Fon-
   taine, O. and Dongier, M. Variations contingentes
   negatives en psychiatric. Electroenceph. clin. Neu-
   rophysiol., 1970, 28:41-47.
 Timsit-Berthier, M., Koninckx, N., Delaunoy, J. and
   Rousseau, J.C. Nouvelles etudes statistiques de la
   Variation Contingente  Negative  en  Psychiatric.
   Rev.  Electroenceph.  clin.  Neurophysiol., 1972,
   2:333-339.
 Timsit-Berthier, M., Delaunoy, J., Koninckx, N. and
   Rousseau, J.C. Slow Potential Changes in Psychia-
   try.  1. Contingent  Negative Variation.  Electro-
   enceph. clin. Neurophysiol., 1973, 35:335-361.
 Valenstein, E.S. Brain control. Wiley and Sons, New
   York, 1973.

 Van Veen, W.J., Knott, J.R., Peters, J.F., Miller, L.H.
   and Cohen, S.I. CNV shape differences  and  per-
   ceptual mode.  In W.C. McCallum and J.R. Knott
   (Eds.),  Event-related slow potentials of the brain:
   Their relations to behavior. Electroenceph. clin.
   Neurophysiol., 1973, Suppl. 33:327-328.
 Walter, G. Electrophysiologic contributions to psychi-
    atric therapy. Curr. psychiat. Ther., 1966: 6:13-25.

 Walter, W.G., Cooper, R., Aldridge,  V.J., McCallum
    W.C. and Winter, A.L. Contingent negative varia-
    tion: An electric sign of sensorimotor association
    and expectancy in the human brain. Nature, 1964,
    203:380-384.
Waszak, M. and Obrist, W.D. Relationship  of slow
   potential changes to response speed  and motiva-
   tion  in  man. Electroenceph.  clin. Neurophysiol.,
   1969,27:113-120.

Weerts,  R.C. and Lang, P.C. The effects of eye fixa-
   tion  and stimulus response location  on the con-
   tingent  negative variation (CNV). Biol.  Psychol.,
   1973,1:1-19.

Weinberg,  H.  The contingent negative variation:  Its
   relation  to feedback  and expectant  attention.
   Neuropsychol., 1972, 10:299-306.

Weinberg,  M. The contingent negative variation:  Its
   clinical  past and future. Am. J.E.E.G. Technol.,
   1975,15:51-67.

Weinberg,  H. and Papakostopoulos, D. The frontal
   CNV:   its dissimilarity to  CNV's recorded from
   other sites. Electroenceph.   clin.  Neurophysiol.,
   1975,39:21-28.

Weinberg,  H. and Michelewski, H., and Koopman, R.
   The  influence  of discrimination on  the  focus of
   the  contingent  negative  variation.  Neuropsy-
   chologia, 1976, 14:87-95.

Wilson, G.D.  GSR responses to fear related stimuli.
   Percept. Mot. Skills, 1967, 24:401-402.

Wing, J.K., Birley, J.L.T., Cooper, I.E., Graham, P.
   and  Isaacs, A.D. Reliability  of a procedure  for
   measuring and  classifying   "present psychiatric
   state." Brit. J. Psychiat., 1967, 113:399-515.

Witkin, H.R., Dyk, R.B., Faterson.H.F., Goodenough,
   D.R. and Harp. S.A. Psychological differentiation.
   Wiley, New York, 1962.

Wolpe, J.  and Lang, P.J. A  fear survey schedule for
   use  in behavior therapy. Behav. Res. Ther., 1964,
   2:27-30.

 Zappoli,   R.  Gli,  "Event related slow potentials"
   (ERSPs) in  neurologia.  Riv. Neurol. (in press).

 Zappoli,  R., Papini, M., Briani,  S., Benvenuti, P.  and
   Pasquinelli, A. L' "onda  d'aspettativa" (CNV) in
   sogetti con lesioni accertate  dei lobi frontali.  Riv.
    Pat nerv. ment., 1973, 94:341-357.

 Zappoli,  R., Papini, M., Briani,  S., Benvenuti, P.  and
    Pasquinelli, A. CNV in patients with frontal-lobe
    lesions and mental disturbances. In W.C. McCallum
    and  I.R. Knott  (Eds.),  The  responsive brain.
    Wright, Bristol, 1976:158-163.

-------
VII.  ENVIRONMENTAL NEUROTOXICOLOGY
      Section Editors:

      David A. Otto and Lawrence W. Reiter
      Neurotoxicology Division
      Health Effects Research Laboratory
      U.S. Environmental Protection Agency
      Research Triangle Park, North Carolina, U.S.A.

-------
NEUROBEHAVIORAL ASSESSMENT
OF ENVIRONMENTAL  INSULT
D. OTTO AND L. REITER
U. S. Environmental Protection Agency, Research Triangle Park, NC, U.S.A.
     An  extensive  correspondence  among environ-
 mental toxicology  panelists was undertaken prior to
 EPIC  IV to identify critical issues for discussion.
 These questions are reviewed here1 and elaborated in
 the  papers that follow. The issues are primarily meth-
 odological ones dealing with the comparative sensitiv-
 ity  and  utility of neurobehavioral  measures  in the
 assessment of environmental insult. The papers also re-
 port neurobehavioral results of human and  animal
 exposures to adhesive solvents, carbon monoxide (CO),
 noise, ozone, and pesticides.
     Two  EPIC IV sessions were  devoted to the dis-
 cussion of environmental Issues, the first dealing with
 clinical toxicology research and the second with ex-
 perimental toxicology research in animal models. In
 view of the substantial overlap of issues considered by
 the  clinical and  experimental panels,  papers from
 both have been combined in this section. The  broad
 spectrum  of issues and insults  discussed  in  these
 papers  defies  any  simple  organization. Therefore,
 they appear In alphabetical order by senior author,
 Panelists included:
     1,  D.  Otto (Clinical  Panel Chairman),  U.  S.
        Environmental  Protection Agency, Univer-
        sity of North Carolina at Chapel Hill.
     2.  L. Reiter (Experimental Panel Chairman), U.
        S,  Environmental  Protection  Agency,  Re-
        search Triangle Park, North Carolina.
     3,  R,  Dyer, The  Johns  Hopkins University,
        Baltimore, Maryland,
     4.  N.  Grandstaff, Stanford  University, Stan-
        ford, California,
     S.  L, Grant, Unlvertity of North Carolina at
        Chapel Hill.
   6.   E. Groll-Knapp, Institute for Umvelthygiene,
       Vienna, Austria.
   7.   A.  Loizzo, Institute Superiore  di  Sanita,
       Rome .Italy.
   8.   N.  Loveless, The University, Dundee, Scot-
       land.
   9.   J. Pirch,  Texas  Tech University, Lubbock,
       Texas.1
  10.   M. Rudnev, Kiev Research Institute of Gen-
       eral and Community Hygiene, Kiev, USSR.
  11.   A.  Seppalainen,  Institute of Occupational
       Health, Helsinki, Finland.
  12.   R.  Traystman, The Johns Hopkins  Univer-
       sity, Baltimore, Maryland.
  13.   H.  Weinberg, Simon Fraser University, Burn-
       aby, British Columbia, Canada.
  14.   G.  Winneke,  University  of  Dusseldorf,
       Germany.
  15.   D. Woolley,  University  of California  at
       Davis.
  16.   R. Zappoll, University of Florence, Italy,


    Neurotoxiclty is not a new  problem; chemicals
adversely affecting the nervous system have troubled
mankind since antiquity. For  example,  one theory
(Hammond 1969)  postulates  that  the  fall  of  the
Roman Empire was due to lead toxicosis in the ruling
class,  stemming from the use of lead pipes In  the
water  system. What is new, however,  is the systematic
study  of unwanted  effects of toxic agents on nervous
system function, i.e., neurotoxlcology (cf. Zenlck and
Reiter, In press). Recent Interest In neurotoxlcology is
largely due to the growing concern over the number of
neurotoxlcants to which we are exposed In our highly
industrialized society, Industrial use of chemicals, the
  1 Uneited quotations or references to panelists in the
  following text refer to premoetlng eonespondinee,
2Dr,  Pirch's  paper appears
Section.
in  the Eloctrogenesii

-------
410
                                Otto and Reiter
 increased number of synthetic materials being pro-
 duced, as well as the large amounts of industrial waste
 products being introduced into the environment have
 all added significantly  to the problem of  neuro-
 toxicology.

    The  risk of intoxication  now  extends beyond
 accidental ingestion and includes occupational and
 environmental  exposure. Examples  of the  diversity
 and magnitude of this problem range from the sub-
 clinical behavioral manifestations  produced by lead
 intoxication  (De la  Burde and Choate  1972)  to the
 gross neurotoxicological  manifestations recently re-
 ported   in  workers  manufacturing  the pesticide
 Kepone  (Martinez et al.  1977). Firemen may be at
 even  greater risk from the toxic fumes produced in
 the combustion of synthetic materials, such as vinyl
 flooring (Dressier et al. 1975).

    The   basic objective  of both  toxicology  panels
 was to evaluate neurobehavioral techniques, particu-
 larly  ERP measures, as indicators  of the adverse ef-
 fects  of  environmental insult. Very little direct evi-
 dence exists  concerning the effects of environmental
 stressors  on  ERPs. Few  environmental toxicologists
 are familiar with ERP techniques, and few ERP re-
 searchers have addressed  themselves to issues of toxi-
 cology. We,  therefore, face a twofold  challenge-to
 demonstrate  the usefulness of ERP measures to envi-
 ronmental   toxicologists, while  introducing  envi-
 ronmental toxicology concerns to other ERP investi-
 gators. General methodological factors that affect the
 choice and  utility of  neurobehavioral  indicators of
 environmental insult (including behavior, neurochem-
 istry, neuropathology, and other bioelectric phenom-
 ena) will  be considered before discussing ERPs.

 Methodological considerations

 What factors should be considered in selecting
appropriate tests  to  determine  the neurobe-
 havioral effects and assess the risks of environ-
mental toxicant exposure?
    One of the first considerations is the known bio-
logical effect of the test agent. Does the agent affect
the nervous system? It is not uncommon for the toxi-
cologist to encounter an agent with undefined neuro-
toxic  properties.  A new substance may be untested
and, therefore, may have unknown  biological prop-
erties, or the toxicity may be established, but with an
undefined CNS component. In these instances,  it is
desirable to  conduct  a battery  of  neurobehavioral
tests that screen for CNS toxicity. Such "apical" tests
would  require  utilization  and integration of several
neural systems so that a chemically induced change in
any one system would produce a change in the meas-
ured endpoint.  Obviously, caution must be exercised
in interpreting such data since screening tests are de-
signed to detect a change rather than to identify a site
of action.

    Grant  (this section) elaborates  theoretical  and
methodological  issues in the design  of  innovative
screening procedures to detect neurotoxicological ef-
fects early in development. Neuropathological and be-
havioral indicators of "functional brain capacity" are
important elements in this approach.

    A distinction  must  be  drawn between a direct
effect on  the nervous system and an indirect change
in nervous system function resulting  from peripheral
effects of a toxicant. The complex role of the nervous
system in maintaining the organism's milieu interieur
makes it possible to  detect early  toxic manifestations
in an organism by utilizing  neurobehavioral indices,
even when the changes may be secondary to a periph-
eral effect. Since there are few toxicants that act on a
single target organ, the nervous  system may provide
for early detection of the impending toxicity.

    Our greatest interest, however, is in the study of
direct toxicant interactions with  the  nervous system.
We will limit  our definition of  neurotoxicants to
those agents acting  directly on  the nervous system,
bearing in mind that it is sometimes difficult to estab-
lish this relationship and sometimes advantageous to
study indirect effects of toxicants  on nervous system
activity.

    Once an agent has been shown to be neurotoxic,
several  determinations are required,  and each deter-
mination is likely  to place certain  restrictions on the
testing methodology. The objectives include:
     1.   Establishment  of  dose-response  relation-
        ships.
    2.  Determination of the extent of CNS involve-
        ment.
    3,  Determination of the mechanisms  responsi-
        ble for the neurotoxicity.

    The primary objective of neurotoxicology  is to
create a data base  for  the  evaluation of potential
health hazards associated with exposure to environ-
mental  pollutants. The above determinations, there-
fore,  are critical. The  better we understand the bio-
logical effects of a toxicant, including its mechanism
of action, the better we can define the risks involved
in exposure. Such information will then permit the
establishment of realistic environmental and occupa-
tional exposure standards. The standard-setting  proc-
ess requires the determination of dose-response  data,
especially at low levels of exposure. However, when
we attempt to determine threshold levels, we must,
by definition, work at the  level of  biological noise
such that variability from sources other than the ex-
perimental manipulation begin to mask any treatment

-------
Neurobehavioral Assessment of Environmental Insult
                                             411
effects (Burt 1975). It is important, then, to develop
and utilize testing procedures with maximum sensitiv-
ity. Obviously, an  understanding  of  a toxicant's
mechanism of action facilitates the choice of methods
for threshold determinations.

     Problems associated with the administration and
duration of toxicant exposure  also affect the choice
and utility of a  particular  neurobehavioral index.
Consideration must  be given to the treatment,  dura-
tion, age at treatment, and time of testing relative to
treatment. These aspects of experimental design  place
considerable restrictions  on testing  methodologies.
Perinatal exposure to an environmental pollutant, for
example, precludes  the establishment of a pretreat-
ment baseline and thus places certain demands on the
response variability. This exposure regimen would dis-
qualify testing methods that do not  permit intergroup
comparisons.  Similarly, a response that is unstable
over time would be  an unlikely candidate for chronic
toxicity studies, although it may be quite useful for
testing acute intoxications.

     These  factors  may  seem  elementary and are
mentioned only to  provide a framework for discus-
sion. Neurotoxicology  is  still in its infancy and the
data available on neurobiological effects of toxicants
have had little impact  on governmental policy  deci-
sions related  to exposure standards.  Greater impact
will come only with a better understanding of both
the methods employed and the  results obtained.

     Eastern European countries have, for some  time,
utilized neurobehavioral data in determining environ-
mental  and  occupational standards.  This fact  may
account, to some extent, for the generally lower ex-
posure  standards in  these countries. Rudnev et al.
(this section) describe neurobehavioral methods used
by the  Kiev Institute of General and  Communal Hy-
giene to evaluate the harmful  consequences of envi-
ronmental agents. The Kiev Institute employs  both
classical behavioral and evoked potential methods to
characterize the effects of toxicants.
 What are the comparative sensitivity and util-
 ity  of  behavioral,  biochemical, neuropatho-
 logical,  and neuroelectric measures of environ-
 mental  insult?
     Sensory deficit can be reliably assessed in many
cases by simple tests of sensory threshold (e.g., audio-
met ry and eye chart tests). The effect of environmen-
tal toxicants on CNS function, in fact, has been infer-
red from behavioral measures such as visual threshold
(McFarland et al. 1944) or psychophyslcal discrimin-
ation of time intervals (Beard and Wertheim 1967).
Neuroelectric measures offer more direct, noninvasive
measures of CNS  function. Furthermore, the effec-
tiveness of behavioral measures may be compromised
by language or motor deficits in some instances (e.g.,
preverbal children  or injured patients), necessitating
neuroelectric assessment. The relative sensitivity of
behavioral  and neuroelectric measures is thus an im-
portant  issue  in  environmental  toxicology  and
medicine.

    The functional significance  of ERP changes can
be inferred from  concomitant alterations in behav-
ioral measures. What if ERP changes are observed in
the absence of behavioral effects? Winneke et al. (this
section) describe an  interesting dissociation of behav-
ior and the auditory evoked potential (AEP) in stud-
ies of  alcohol and trichloroethylene (TCE).  TCE ex-
posure decreased P2 amplitude of the AEP, but failed
to produce any vigilance decrement. Alcohol, on the
other hand, impaired vigilance without any effect on
the AEP. Winneke et al. discuss the problems of es-
tablishing the reliability and validity of these neuro-
behavioral  measures.

    Grandstaff and  Beard  (this  section) describe an-
other  intriguing dissociation  of behavioral and elec-
trophysiological  measures:  0.4  ppm ozone exposure
produced no observable change in visual evoked po-
tentials, but this exposure  did produce a significant
improvement in vigilance performance. This finding
contrasts with a substantial  body of evidence that
ozone  impairs pulmonary  function  (cf. Stokinger
1954).  Orthogonal  data  derived from  behavioral
measures can provide useful information concerning
the functional significance of observed changes in
other  physiological  measures.  In  this  case, for in-
stance, improved  vigilance performance might have
been due to an irritant  effect of ozone, which coun-
teracted the usual decrease in  arousal across time.
There  is less reason  to  expect  changes in ERP para-
meters, however, when  the CNS is not the  primary
target organ of the test substance.

     Biochemical  assays of  toxicant  levels  in  the
 blood, urine, or various tissues of the body are neces-
 sary to determine the ultimate distribution and meta-
 bolic  fate of a test  substance. Clinical decisions con-
 cerning therapeutic  intervention in humans  are often
 based  on biochemical measures-e.g.,  chelating agents
 may be administered if blood lead  levels  exceed a
 certain criterion  level. Most biochemical measures,
 however, are  invasive and  some (e.g., brain  tissue as-
 says) may require  the sacrifice of test animals. Animal
 models are thus necessary to study the mechanisms
 underlying the  effects of neurotoxicants.  Although
 behavioral and  neuroelectric  measures may provide
 rapid,  simple, reliable, and noninvasive indices  of
 neurotoxic effects, these  measures  must be coordi-
 nated  with biochemical  measures  in  order  to

-------
 412
 determine the underlying mechanisms. Furthermore,
 electrogenesis of ERPs reduces ultimately to a neuro-
 chemical process (of. reviews by Libet, Marczynski,
 and Somjen, this volume).  As Dr. Upton noted in
 his opening remarks, the "plumbers and electricians"
 of the brain must work together to resolve many of
 the critical questions in neurotoxicology.
     The work of Woolley and Reiter (this  section)
 provides a good example of the  coordinated assess-
 ment of electrophysiological and biochemical para-
 meters. These investigators compared the effects of
 the  insecticide parathion on visual evoked potentials
 (VEPs) and acetylcholinesterase (AChE) inhibition in
 rats and monkeys. The latency of VEP components
 increased markedly 2 to 4 hr after parathion admin-
 istration in the visual cortex of rats (4 to 8 hr in
 monkeys) and returned rapidly to pretreatment levels
 within  8 hr (24 to 48 hr in monkeys). Complete re-
 covery of brain AChE activity, however, required 2 to
 4 weeks in rats  and extended periods in monkeys.
 The authors discuss  possible mechanisms of func-
 tional tolerance to AChE inhibition suggested by the
 dissociation of electrophysiological and biochemical
 measures.
                                Otto and Reiter

    There are, of course, no simple answers. The util-
ity  of any  given measure depends on complex varia-
bles, including the extent of nervous system function
affected, the question of whether  the toxicant exerts
a specific or nonspecific effect on CNS function, and
the dose and duration of toxicant exposure. That is,
the relative utility of different neuroelectric measures
must  be determined empirically for each environmen-
tal  toxicant. Unless the specific target system in the
nervous system is known, multiple measures should
be  used to screen for effects in different  brain sys-
tems and functions.

    The  value of the comparative approach is exem-
plified in CO research. Dyer and Annau (this section)
studied  the effect of CO on visual evoked  potentials
recorded in the superior colliculus of rats.  No effect
was observed below 20% carboxyhemoglobin levels, a
point at  which humans begin  to perceive physiolog-
ical symptoms (cf. Stewart 1975). Results of Groll-
Knapp et al. also point to  the  insensltivity of the
visual system,  but suggest  marked effects in the
somatosensory system. Assessment of a single sensory
modality may thus provide an incomplete and mis-
leading picture of altered  CNS function. We will re-
turn to this point later.
     A number of papers in this section provide data
 on the comparative utility of different neuroelectric
 measures.  Seppalainen  found  the  somatosensory
 evoked potential  (SEP) slightly more sensitive than
 peripheral nerve conduction velocity in assessing the
 effects of occupational lead exposure. Groll-Knapp et
 al. reported that the SEP was  sensitive, the AEP less
 sensitive,  and  the VEP insensitive  to CO anoxia.
 Winneke  et al. found the AEP insensitive to the ef-
 fects of CO exposure, although Winneke and Kastka
 (in press) previously  reported a decrease  in the P2
 component  of the AEP  following  TCE  exposure.
 Zappoli et al., however, did not find the CNV or SEP
 as useful as routine electroneuromyographic and EEC
 examinations for diagnostic signs  of subclinical neuro-
 pathy in  workers exposed to  industrial solvents.
 Zapponi  et  al. illustrate the utility  of yet another
 EEC measure, spectrum analysis, in the assessment of
 methyl-parathion poisoning.
    The comparative utility question has been posed
from an experimental perspective. In the environmen-
tal medicine, this question takes a different form, as
discussed below.
Which neuroelectric measures provide the best
diagnostic and prognostic signs in the clinical
assessment  of  acute  and  chronic toxicant
effects?
 What are the comparative effects on ERPs of
 environmental  toxicants,  stimulants,  and
 depressants?
    Although there is a paucity of evidence concern-
ing the effects of environmental toxicants on ERPs,
the effects of many stimulant and depressant  drugs
on ERPs  have been studied (cf. Ashton et al. and
Thompson et al., this volume). Drug effects can thus
be used as models for comparison with toxicant ef-
fects. Winneke et al. (this section) compare the effect
of alcohol, a well-known CNS depressant, with TCE.
This approach provides a useful frame of reference
for calibrating human  performance, as well as  ERP
changes, during exposure to stressors.
How well do subjective ratings correlate with
behavioral and ERP measures during exposure
to environmental stressors?
    Subjective perception of discomfort during toxi-
cant exposure may produce significant, although non-
specific,  effects  on performance and CNS  function
(e.g., irritation of the mucosa by ozone may increase
arousal).  Several  panelists  (GrolUKnapp, Otto,
Winneke) routinely use subjective rating scales in an
effort to detect  and quantify such effects.  Loveless

-------
Neurobehavioral Assessment of Environmental Insult
                                            413
suggests that "annoyance" related to the "perceived
loudness" of noise bursts may be a critical  variable
determining performance decrement in noise studies.
The transitory effects of noise bursts on performance
could parallel a decrease in subjective loudness and
annoyance.  Loveless  (this section) speculates  that
ERPs might  provide an objective correlate of "per-
ceived noisiness."
    Many of the papers in this section deal with the
neurobehavioral  effects  of specific environmental
insults rather than methodological issues. The remain-
der of this summary is devoted to questions concern-
ing the  neurobehavioral effects  of two  common
stressors, noise and carbon monoxide.
Neurobehavioral effects of noise
    Noise is an environmental stressor that can dis-
rupt cognitive function under certain conditions. Fur-
thermore, noise is omnipresent in urban and industri-
al environments, where other toxicants are likely  to
be encountered. Thus, the neurotoxicological effects
of noise per se, as well as the  synergistic (or con-
founding) effects of noise in combination with other
environmental agents, merit careful study.
What are the neurotoxicological consequences
of intermittent and continuous noise?
    Psychological research on noise has concentrated
on effects of continuous high-intensity noise on tasks
requiring sustained attention (Broadbent 1971). Con-
tinuous white noise at intensities above 90 dB  pro-
duces a consistent decrement in vigilance perform-
ance, which Broadbent attributes to overarousal, al-
though the exact mechanism is poorly understood.


    Loveless (this section) notes that  the focus on
continuous noise has diverted attention from inter-
mittent noise, which  might well produce neurobehav-
ioral deficits at lower intensities.  He proposes that
the effects of intermittent noise can be conceptual-
ized in terms of orienting and defensive reactions
"operating through mechanisms  of perceptual selec-
tion".  Noise bursts  of moderate  intensity,  for in-
stance, should evoke a transitory orienting response
that rapidly habituates.. In tense bursts, on the other
hand, ought to produce a nonhabituating defense re-
sponse, characterized by reduced  sensory intake or
reduced sensitivity. The defense response might entail
peripheral or central gating of sensory input.  Predic-
tions can be made and results interpreted within the
framework of Sokolovian  theory and the growing
body of ERP evidence concerning selective attention
and information processing (cf. Tueting, this volume).

    Theoretical interest in  noise bursts and  the ori-
enting response stems from recent work at Dundee.
In studies of simple reaction time performance, Love-
less and Stanford (1974a,b; 1975) differentiated two
components, an  early orienting response (O-wave)
and  a later anticipatory  response (E-wave), which
may summate in the CNV. These results raise meth-
odological  questions with regard to short interstim-
ulus intervals.  Loveless warns that "CNV research
seems to have fallen into the trap that classical condi-
tioning was in, before it was realized that the use of a
short 'optimal' interval confounds orienting responses
to  the conditional  stimulus  with  anticipatory re-
sponse. . .my own work, like that of Weerts and Lang
(1973), strongly  suggests  that slow potentials simi-
larly show orienting and anticipatory phases, which
are clearly distinct when the ISI is long, but summate
when it is short."

    Loveless reviews noise  research from behavioral
and  ERP perspectives  and relates the  data to basic
issues in the broader field of stress research. He also
describes preliminary evidence of the effectr of inter-
mittent noise  on ERPs and performance during sim-
ple and selective  RT tasks. Otto and  Benignus (this
section) report increased errors and  decreased ampli-
tude of  the N110 component in  a numeric moni-
toring  task during low frequency (11.5 to 350 Hz at
80 dB) noise. Weinberg et  al. (this section) failed to
observe any detrimental effect of either a "real-life"
noise (telephone  ring) or  artificial (white) noise on
the CNV.
Neurobehavioral effects of low-level
CO exposure
    CO is a colorless, odorless asphyxiant gas that
binds tenaciously to hemoglobin in red blood cells to
form  carboxyhemoglobin  (COHb).  CO produces
anoxic stress by reducing the oxygen-carrying capac-
ity of the blood and impairing the release of oxygen
to body tissues.
What is  the threshold level at which CO pro-
duces an observable decrement in behavioral
function?
    U. S. air quality standards for CO are 35 ppm (1
hr) and 8.7 ppm (8 hr) in nonindustrial environment
(U.S. EPA 1971). These stringent standards are based,
in part, on behavioral deficits reported to occur at 2
to  5%  COHb (Beard and Wertheim  1967). Three

-------
 414
                                Otto and Reiter
 other laboratories (O'Donnellet al. 1971a; Stewart et
 al. 1973; Otto et al., in press), however, have been
 unable  to replicate the  impairment in temporal dis-
 crimination reported by Beard and Wertheim. The ef-
 fect of CO on vigilance performance is likewise equiv-
 ocal: Horvath et  al. (1971), Fodor and  Winneke
 (1972), and Groll-Knapp et al. (1972) reported vigi-
 lance decrements below 5% COHb, but subsequent
 efforts  by  each of these laboratories to replicate the
 earlier  experiments have  failed  (Christensen et al.
 1977; Winneke et al. and Groll-Knapp et al., this sec-
 tion). CO effects observed by Otto et al. (this section)
 in a continuous performance vigilance task were mar-
 ginal. Stewart (1975) has concluded that observations
 of behavioral decrements  at COHb levels below  5%
 "must  be  considered suspect" until independently
 replicated.
Is general arousal level a critical variable medi-
ating the effect of CO on ERPs and behavior?
     Beard  and Grandstaff (1975) noted that perfor-
mance decrements associated with low-level CO ex-
posure have only been observed in very monotonous
tasks under low arousal-conditions.   Arousal  level,
moreover, has not been adequately controlled in stud-
ies  reporting negative results. Grandstaff emphasized
that general arousal level was probably  "the  most
important variable to be considered in designing or
evaluating any study with CO." She proposed a sim-
ple  test of the CO-arousal hypothesis: sample back-
ground EEG throughout the  exposure period to de-
termine if  activation levels change as  a  function of
COHb concentration.  Fast-Fourier  analysis  tech-
niques can then be used to quantify the frequency
content of  EEG samples. Haider et al.  (1976) re-
ported a  slight decrease in alpha accompanied by in-
creased theta and beta activity following 4 hr expo-
sure to 200 ppm CO. This finding is not  impressive,
considering that COHb levels  should have reached at
least  15%.  Other EEG studies  of CO reviewed by
Dinman  (1969) reported inconsistent  or  negative
results.

    Measurement of motor activity level  constitutes
an alternative or concomitant method of testing the
CO-arousal  hypothesis.  In view of the inconsistent
findings of EEG and behavioral studies in the past,
both indices of CO effect should be examined concur-
rently in future studies.


    Winneke et al. (this section)  directly tested the
CO-arousal  hypothesis.  "Monotony"  was varied by
changing  target probability (0.03 vs. 0.1) and by pro-
viding rest breaks with performance feedback during
the   "less monotonous" condition. Random  clicks
were  superimposed on  the vigilance task to  obtain
AEP measures. Each subject repeated the experiment
under 0, 100, and 200/150 ppm CO, conditions that
yielded terminal COHb values of 7.5% and 11.3% for
low and moderate CO conditions. They did not find
any effect of CO exposure on vigilance performance
or AEPs.  In order to  rule out  the  possibility that
random clicks enhanced arousal, Winneke et al. ran
20 additional subjects  without clicks. Results were
the same  and thus do  not support  the  CO-arousal
hypothesis.

    Groll-Knapp et al. (this section) used a classical,
straightforward method  to minimize arousal: a sleep
study was undertaken  to avoid the uncontrollable
variability caused  by voluntary shifts of attention
during waking. Exposure to 100 ppm during 7 hr of
sleep increased the amount of stage 3 and 4 sleep, but
decreased  rapid eye movement (REM) sleep. All com-
ponents of  the  AEP shifted  in  a positive direction
during stage  4 sleep. Groll-Knapp et al.'emphasize
that CO effects are more consistent during sleep than
waking and recommend  this approach as an effective
method of controlling arousal in toxicant research.
Are the neurobehavioral effects of CO mod-
ality-specific?

    Xintaras  et  al. (1966a,b)  described  changes in
flash evoked  potentials in the  superior colliculus of
rats following exposure to 50 ppm  CO. Dyer  and
Annau (this section) did  not observe any significant
changes below 20% COHb in an attempted replication
of the Xintaras experiment. Reasons for the discrep-
ant results  are unclear, although several strands of
evidence suggest  that visual function, at least in the
photopic range, is not particularly susceptible to CO
impairment.

    Parameters of visual function reported to be im-
paired at  3 to 20% COHb levels include brightness
thresholds  (MePariand et al.  1944;  Halperin et al.
1947,  1959; Beard and Grandstaff 1970), visual  acu-
ity (3 to  10% COHb:  Beard and Grandstaff 1970),
and dark  adaptation (17% COHb: McFarland et al.
1972,  McFarland 1973).  On the other hand, COHb
levels below 20% do not appear to affect brightness
discrimination (Ramsey 1972,  1973; Weir and Rock-
well 1973), depth perception (McFarland et al. 1972;
McFarland  1973; Ramsey 1972, 1973; Wright et al.
1973), glare recovery (McFarland et al. 1972,McFar-
land 1973; Wright et al. 1973), or critical flicker fu-
sion (CCF) (Fodor  and Winneke  1972, Guest et al.
1970,  Johnson et al. 1974,  O'Donnell et al.  197la,
Ramsey 1973).  Beard  and Grandstaff (1970) did a
marginal decrease in CCF at 3 to 10% COHb levels.

-------
Neurobehavioral Assessment of Environmental Insult
                                            415
    Groll-Knapp  et  al.  (this  section) compared
the effects of low-level  CO on auditory, visual, and
somatosensory  EPs. As noted previously, the VEP
was insensitive to CO effects. Other evidence also sug-
gests  differential sensitivity of auditory and visual
function to CO impairment.  Beard and Wertheim
(1967) demonstrated a CO dose-related decrement in
the estimation of tone duration, an effect that Beard
and Grandstaff (1970) were unable to reproduce in
the visual modality.

    Results of other  experiments employing  visual
stimuli  are  also inconsistent.  Horvath et al. (1971)
observed  a  significant performance decrement in a
visual vigilance  task  following 2-1/4-hr exposure to
111 ppm  CO. Winneke et al. (this section) and Chris-
tensen et  al. (1977) both failed to replicate the Hor-
vath et al.  study. Otto  et al. (this section), on the
other hand, observed  CO-related changes in CNV and
N1P1 components  during  a  visual monitoring task.
Further comparative  studies of CO effects on  dif-
ferent parameters of auditory,  somatosensory,  and
visual function are needed to resolve the inconsisten-
cies of  existing data and to identify the  mechanisms
by which  CO impairs sensory function.

    Luminance level  is  an important parameter of
visual function that could possibly account for dis-
crepant results in CO studies employing visual stimuli.
At high  levels of luminance, photopic  vision  is
mediated  by a dense population of cones in the fovea,
while at low levels, scotopic vision is mediated by a
much less dense population of rods distributed more
uniformly throughout the retinal Held. McFarland et
al. (1944)  showed that  hypoxic hypoxia  affected
visual acuity much more at scotopic than photopic
levels of  illumination. Merigan and Blick (1978) and
Evans and Carman (1978) have argued,  on the basis
of relative  neuronal  density  and redundancy,  that
scotopic  vision is  probably  more susceptible than
photopic  vision to neurotoxicant effects. Winneke et
al. (this section) point out that they used  a  higher
luminance level than  Horvath  et al. (1971) used, and
suggest that this factor could possibly  explain  the
lack of CO  effects. Further study is needed to clarify
the relationship of CO exposure, luminance level, and
visual function.
 What are  the short-term  compensatory and
 long-term adaptive mechanisms of response in
 man to low level CO stress?
    The behavioral muddle of CO  research suggests
that the critical variable controlling CO effects has
not yet been identified. Activation of a physiological
compensatory mechanism was proposed as a possible
explanation  of  negative  findings  by  Fodor  and
Winneke (1972). The paradoxical effects observed in
some studies of behavioral decrements early in  the
uptake  cycle,  followed  by a return to normal per-
formance at higher  COHb concentrations,  support
this  speculation. Beard  and Grandstaff  (1975),  for
example, observed "a definite trend for CO to have
its greatest  effect  during  the  initial portion of  the
uptake period  shortly after it was introduced (within
30 min.). Therefore, if the impairment  in perform-
ance of a given task is slight, or occurs primarily dur-
ing  the  uptake  stage, this  effect  would easily  be
obscured."

    What is known about the physiological reaction
of the organism to CO? The  cardiovascular system
responds to  anoxic stress by increasing cardiac output
or selectively increasing blood flow to specific organs.
Patients with impaired oxygen-uptake capacity (e.g.,
coronary artery disease) are thus uniquely susceptible
to CO  poisoning (Ayres et al. 1969, Aronow and
Isbell 1973, Anderson et al. 1973).  Paulson et  al.
(1973)  reported that cerebral blood flow increased
at   moderate   (20%)   COHb  levels  in  humans.
Traystman and Carlson (1974) extended this work in
rats  to show that  cerebral vasodilation results from
both CO hypoxia and hypoxic hypoxia at comparable
arterial  02  levels.  This evidence indicates that  the
brain can effectively compensate for the  oxygen de-
ficiency produced by CO anoxia at levels up to about
20% COHb. The conditions  that control the onset
threshold of this mechanism, however, have not been
established.

     Determination of threshold conditions could pro-
vide a key  to the contradictory behavioral findings
obtained during low-level CO exposures. If cerebral
vasodilation does not occur until COHb  levels reach
5%,  subtle behavioral measures may show effects at 2
to 5% COHb. If the compensatory mechanism triggers
at 5% COHb, then performance should rapidly return
to control levels. Fodor and Winneke (1972), Beard
and  Grandstaff (1975), and Otto et al. (this section)
have observed such paradoxical findings. Research on
cerebral hemodynamics  (Traystman, this section),
however, provides no  direct evidence of any lower
threshold triggering of the  cerebral compensatory
mechanism.

     Grandstaff  suggested that behavioral as well as
physiological  compensatory mechanisms may coun-
teract the effect of CO.  Behavioral reactions might
take  the form  of head-shaking,  eye-blinking,  or
squirming around  to  restore  alertness as the subject
perceives  himself  becoming drowsy.   Grandstaff
further pointed out that activity levels of subjects
ought to be measured along with other electrophysio-
logical and  behavioral measures. EOG, EMG, or other
kinds of electromechanical transducers could be used

-------
 416

 for this purpose. Activity levels constitute a standard
 measure of toxicant effect in animal studies. Measure-
 ment  of  activity  levels  in  humans could provide a
 useful bridge for extrapolation from animal to human
 research.

     Grandstaff also raised the question of adaptation
 in  relation  to  the  compensatory response of the
 organism  to CO.  The time course of adaptation  to
 low levels  of CO has not been systematically investi-
 gated in humans, particularly  in relation to perform-
 ance on specific tasks. (Smokers are generally rejected
 from CO  studies on the assumption that adaptation
 to  elevated COHb levels does occur.) Adaptation  to
 the  hypoxic conditions  of living  at high altitude is
 well documented.  Adaptation to CO is an important
 question for further study.
Conclusions
    Neurotoxicology is an emerging discipline devot-
ed  to  the  study of the adverse effects of environ-
mental insult, both physical and chemical, on nervous
system function. The issues addressed in this  review
and the  papers that follow are primarily methodo-
logical questions dealing with the comparative utility
of  available  neurobehavioral measures of  toxicity.
The application of ERP techniques  in the assessment
of  environmental  insult has  been limited, although
the results of several studies in this section suggest
that certain ERP parameters are sensitive indicators
of neurotoxicant effects.  Interpretation of the func-
tional significance  of observed ERP changes presently
requires the concomitant  recording of behavioral or
other  neurobiological  criterion measures.  As  our
understanding of the neurophysiologica] and neuro-
chemical  substrates  of  ERPs grows, these measures
                                Otto and Reiter

should provide an increasingly important index of the
neurobehavioral  consequences of exposure  to envi-
ronmental stressors.

     H.  E.  Stokinger,  Chairman of the  Threshold
Limits  Committee  at  the National  Institute of
Occupational  Safety and  Health,  has stated  that
"changes in the visual evoked response (VER) are pre-
sently not  sufficiently  well understood  in overall
physiological  terms  to serve as basic criteria for air
standards" (1974, p20).  Although the neuroanatomi-
cal substrates and functional significance of many
ERP components  are  presently unknown, ERPs do
offer a convenient,  noninvasive window  on central
nervous system  activity.  If appropriate precautions
are taken to avoid extracerebral artifacts,  changes in
ERP latency or amplitude induced by environmental
agents should  provide a legitimate and useful index of
the effect of toxicant exposure on brain function (cf.
Otto, in press, for extended discussion of  this  issue).
The fact that the  precise mechanisms  underlying
ERPs may not be completely understood is essential-
ly irrelevant to the use  of these bioelectric events as
measures of neurotoxicant effects.

     This volume  and others (Donchin and  Lindsley
1969; Regan 1972;McCallum andKnott 1973, 1976;
Desmedt 1977) document the rapid growth of know-
ledge in the ERP field. The clinical utility  of ERPs
has  been  clearly established in audiometry (Davis
1976), neurology (Starr and Achor  1975),  ophthal-
mology (Sokol 1976), psychiatry (Shagass  1972), and
psychopharmacology (Tecce et  al.  1978).  The re-
markable progress in clinical applications in reviewed
extensively  in Callaway  et  al. (in press).  Thus, the
view that ERPs may be  mere epiphenomena of brain
function (Uttal 196S) and the reluctance to use ERP
evidence  as criteria  for exposure  standards are no
longer defensible.

-------
CARBON MONOXIDE
AND SUPERIOR  COLLICULUS  EVOKED  POTENTIALS

 R. DYER AND Z. ANNAU
Department of Environmental Medicine, The Johns Hopkins University, Baltimore-, MD, U.S.A.
    Attempts to determine the level of carbon mono-
xide  (CO)  exposure  that  produces observable
behavioral changes have produced conflicting results.
Since evoked potentials (EPs) provide information
about the general functional integrity of the sensory
system studied, they  might be useful in determining
the level  of exposure  to CO that results in changes in
neural  function. In turn, since changes in behavior
must be preceded or accompanied by changes in neu-
ronal activity, the EP technique may also be useful in
predicting the CO level that can be expected to pro-
duce behavioral changes. To this end, Xintaras et al.
(1966) recorded flash evoked potentials from  both
the visual cortex and superior colliculus of unanesthe-
tized rats and reported  that the superior colliculus
evoked potential (SCEP) was  uniquely sensitive to
low concentrations of CO. Changes in the SCEP were
observed  at levels of 50 ppm CO. Procedural difficul-
ties make generalization from  this study hazardous.
The results were presented in descriptive terms based
on  a few animals and no effort to control for periph-
eral variables such as pupil diameter was described.
Since pupil  diameter can be  expected to vary with
respect to arousal level and hypoxic condition, and
since variations  in pupil diameter can be expected to
produce  variations in both amplitude and latency of
the evoked potential, failure  to control for it repre-
sents a serious  shortcoming.  The present paper ac-
commodates these  shortcomings and reports results
of SCEP experiments using short-term (2- to 2.5-hr)
exposures to CO in chronically implanted unanesthe-
tlzed rats.
 Methods


     Eight adult male Long Evans rats were used in
 the  experiments. Animals  were anesthetized with
 Equithesin, and  bipolar 0.25-mm twisted nichrome
 'This  work  was  supported  by  Fight  for  Sight
 grant-in-aid, G-SS3, HL 05453 and EHS 00454.
wires were lowered into the superior colliculus under
stereotaxic  guidance. Tips of the electrodes  were
separated from each other in the vertical plane by 1
mm, and the electrodes were lowered into the SC 5.5
mm posterior to bregma, 1.5 mm lateral and 3.8 mm
below the  cortical surface, according to  the atlas of
Skinner (1971).  Electrodes  were cemented in place
with dental acrylic and connected to an Amphenol
receptacle. A 0-80 stainless steel screw was inserted
into the skull over the frontal sinus and connected by
an insulated nichrome wire to the receptacle for pur-
poses  of grounding  the animal. At  least 1  week of
recovery was allowed before  any  recordings were
made.
    After pupils were dilated with atropine sulfate
and  the  animals were  connected  to  the recording
apparatus via an Amphenol plug and Microdot mini-
noise shielded cable, they were placed in a chamber 8
cm wide, 20 cm long,  and 38 cm high, which had
mirrors on three walls, ceiling, and  floor. The fourth
wall was  clear Plexiglas, and had a Grass PS-2 photo-
stimulator lamp mounted  flush against it. Either air
or  a mixture  of  air and CO was blown into  the
chamber  at 6 liters/min through a 0.6-cm hole 5 cm
from the floor.  Throughout  the  experiment  the
chamber  was illuminated  by overhead lights in  the
laboratory, yielding  a  luminance  of  about   100
millilamberts.
    Recordings were  made  with  high- and  low-
frequency filters set at 10 kHz and 0.2 Hz, respective-
ly.  Amplified signals were led to an oscilloscope for
monitoring, and to a PDP-8 or POP-12 computer for
averaging. The poststimulus analysis epoch was 240
msec,  each  millisecond  representing one bin  of a
240-point plot  display.  Voltage was sampled every
333 Msec, and  the three samples  taken  each  milli-
second were averaged to produce the value deposited
in a given bin. A signal from the computer triggered a
Grass  S44 stimulator,  which in  turn triggered the
photostimulator. The photostimulator was set  at its

-------
 418

 highest intensity, which produced a 10-^sec flash of
 about  1,500,000 candle power.

     Averaged responses were displayed on an oscillo-
 scope,  and a cursor controlled  by  the  teletype and
 one analog channel printed the latency to the nearest
 millisecond and  amplitude to the nearest microvolt
 for any bin requested. The displayed signal could be
 photographed with a Polaroid camera or printed by a
 Centronics printer. The printouts were dot plots with
 numeric descriptors of each dot.

     Desired concentrations of CO were  mixed in a
 large drum, continuously monitored with a Beckman
 infrared analyzer, and  pumped  into the recording
 chamber.

     The experimental paradigm consisted of record-
 ing the  response to 500 flashes presented at 0.5 Hz
 while air was blown  into the  chamber.  The air was
 then replaced with the desired concentration of CO,
 the animal allowed 2 hr to equilibrate, and the series
 of flashes repeated.  No systematic attempts were
 made to measure activity during this period, but the
 animals  were observed to be generally quiescent.
 Latencies of peaks PI, N3, P3, N4, N5, and P5 and
 peak-to-peak amplitudes  of N1P1, P3N4,  N4P4,
 P4N5, and N5P5 were then expressed as a percentage
 of the  pre-exposure values,  a procedure that mini-
 mizes the interanimal differences known to occur in
 bipolar  recordings from  the  SC (Dyer  and  Annau
 1977). These experiments were performed with 1000,
 500, 250,  150, and 0 ppm CO.  All conditions were
 presented to each animal in  random order  with at
 least 1  week between sessions. A repeated measures
 analysis  of variance was  performed separately  for
 each component  analyzed, and  individual compari-
 sons were made using the t-test technique.

    Carboxyhemoglobin (COHb) saturations  at the
 different exposure levels were  determined in a sepa-
 rate series of rats from the same strain with chronical-
ly  implanted venous catheters  using the methods of
Weinstein and Annau (1967) and Small et al. (1971).
Results

    COHb  saturations reached  equilibrium before
120 min of exposure in sedentary rats. Values of 13,
22, 38, and 55% were found to result from exposures
to 150,250, 500, and 1000 ppm.

    Fig. 1  shows a typical SCEP with the different
peaks identified.  Low levels of exposure to CO in-
creased the amplitude of late components, but did
not affect  early components. Severe exposure (1000
                                                                                      Dyer and Annau

                                                       ppm) depressed amplitudes of late  components and
                                                       increased all latencies.
                                                           Fig. 1. Superior colliculus evoked potential.

                                                           The only significant departures from compressed
                                                       air values  occurred  at  exposures to  500  and 1000
                                                       ppm. At 500 ppm, there were significant increases in
                                                       P3N4 and N4P4 amplitude. At 1000 ppm, there was a
                                                       significant  depression in both P3N4 and N5P5 ampli-
                                                       tudes, and a significant  increase in latency of PI, N3,
                                                       P3,  N4, and P4  peaks. At all concentrations, there
                                                       was marked interanimal variability, not only in mag-
                                                       nitude but in some cases even in  direction of effect.
                                                                     10'    20   30    40   80    80
                                                                              COHb, K
                                                                        ISO  280     800       1000
                                                                              CO, ppm

                                                      Fig. 2.  Effects of CO on superior colliculus evoked
                                                      potential amplitudes.


                                                          Fig. 2 and 3 show the effects of varying concen-
                                                      tration  of CO upon amplitude and latency of dif-
                                                      ferent peaks. These  figures indicate that amplitude is
                                                      more sensitive than  latency as a measure of CO toxi-
                                                      city.

-------
Carbon Monoxide and Superior Colliculus Evoked Potentials
                                                                                               419
      200
2
X
u
UI
£
   o
       180
       160
       140
       120
       too
            •—•  PI
            0—0  N4
            A—A  NS
   £  ao
               10    20   30   40   80    «0
                       COHb, %
                  IBO  Z50     600
                      CO, ppm
                                     1000
Fig. 3.  Effects of CO on superior colliculus evoked
potential latencies.
 Discussion
    The present results show that 150 and 250 ppm
concentrations  of CO, which produced COHb levels
of 13 and 22%, do not significantly change the SCEP.
These results are contrary to those of Xintaras et al.
(1966) who reported EP changes at 50 and 100 ppm.
A number of explanations  for the discrepancy are
possible. Dyer and Annau (1977) have shown that it
takes at least 2 hr for dark adaptation to occur when
rats are placed in a recording chamber darker than the
home cage. During this time, repeated flashing elicits
EPs of greater and greater amplitude. Further obser-
vations from our lab suggest that even when no dark
adaptation occurs  in  the  course of an experiment,
there is  an increase in amplitude of several compo-
nents from the  first block of 50  to the second block
of 50 trials. Thus, the increased amplitudes observed
by Xintaras et al. (1966) during exposure to low CO
levels may have reflected adaptation phenomena. The
interanimal variability observed in the present experi-
ment indicates, furthermore, that descriptive findings
from a small number of animals may be very mislead-
ing.

    The finding that a concentration of 500 ppm CO
significantly increases EP amplitudes is of interest. It
is  well  known that the electrical activity of the
nervous system during the course of anoxia progresses
through a  stage of  relative increased  activity. EP
amplitudes  increase, spontaneous  cell firing increases,
and the EEC is characterized by low-amplitude, high-
frequency waves (Baumgartner et al. 1961). At least
three mechanisms may be postulated  to account for
this phase of activation: greater sensitivity of inhibi-
tory neurons to hypoxia (Gelfan and Tarlov 1955),
activation of carotid body  chemoreceptors by low
arterial 02 pressure  with  subsequent reticular and
cortical activation  (Dell and Bonvallet  1956), and
partial  depolarization induced by electrolyte shifts
known to  occur  during hypoxia (Michael 1973).
Finding the activation phase with exposures to 500
ppm CO  appears to rule out the chemoreceptor hy-
pothesis since the chemoreceptors are presumably not
activated at this CO concentration. Electrolyte shifts
cannot account for the activation either, since they
are also presumed not to occur during CO exposures
(DeValois and Schade 1967). Thus, the most plausi-
ble explanation appears to  be a  release from inhibi-
tion produced by  a greater sensitivity of inhibitory
synapses to hypoxia.

    Traystman (this section) has  shown that cerebral
blood flow in dogs increases markedly to compensate
for reduced blood oxygen under conditions of both
CO hypoxia and hypoxic hypoxia. Moreover, Trayst-
man's data indicate that brain oxygen  consumption
remains relatively stable up  to 30% COHb, but begins
to  decrease at an  undefined point between 30 and
51% COHb.  If one can generalize from dog to rat,
then the clear depression  of SCEP components ob-
served  at  1000 ppm (55% COHb) may be attributed
to decreased oxygen consumption of the brain at high
COHb  levels. The  increase  of SCEP components ob-
served  at 500 ppm CO (38% COHb), however, cannot
be  attributed to a decrease in oxygen  consumption
and may, in fact, reflect activity  of the presently un-
known  compensatory  mechanism  that   mediates
cerebral  blood flow. Further study is required to
elaborate  the relationship  of EP changes, cerebral
blood  flow and oxygen consumption, and  the com-
pensatory response of the brain to CO hypoxia.

-------
 EFFECTS  OF OZONE
 ON HUMAN CENTRAL NERVOUS SYSTEM FUNCTION


 N. GRANDSTAFF and R. BEARD
 Stanford University School of Medicine, Stanford, CA, U.S.A.
     The purpose of this study was to assess the ef-
 fects of exposure to low levels of ozone on CNS
 function  In humans.  Twenty young, adult, paid
 volunteers were assigned to experimental or control
 groups  of 10 subjects each. All subjects completed
 four 4-hour testing sessions on successive days as fol-
 lows: day 1: orientation; day 2: pro-exposure base-
 line;  day 3:  ozone exposure  (experimental group
 only); and day 4: post-exposure baseline. Subjects In
 the experimental group were exposed doubleblind to
 0.4 ppm (800 Mg/m3)  ozone for 4 hours In a small
 audiometric booth on day 3.

    The following parameters were measured during
 each  testing session: errors and response latency in
 detecting target lights  in a peripheral vision testing
 device (Grandstaff 1974), respiration rate, EEC, and
CNV. Clinical measures of peripheral (Goldman Pe-
rimeter) and central (Tangent Screen and/or Goldman
Perimeter) fields  and  dark  adaptation  (Goldman
Adaptometer) were  also obtained before and after
each testing session.

    Subjects  exposed  to  ozone  made  significantly
fewer errors and responded  more rapidly to target
lights on day 3 than subjects not exposed to ozone.
This effect was generalized throughout the peripheral
field. Ozone exposure did not produce any significant
change in clinical measures of peripheral or central
fields, visual thresholds, or In respiration  rates. The
EEG/CNV  analysis  was inconclusive. The observed
behavioral change may  reflect a general Increase in
arousal level from the stress induced by  ozone expo-
sure.
'This  study  was  supported  by  Contract  No.
68-02-1762 from the U.S. Environmental Protection
Agency.

-------
NEUROBEHAVIORAL ASSESSMENT OF EFFECTS OF
ENVIRONMENTAL INSULTS EARLY IN  DEVELOPMENT
 L. GRANT

 Departments of Psychiatry and Anatomy, Biological Sciences Research Center,
 University of North Carolina, Chapel Hill, NC, U.S.A.
    This paper focusses on neurobehavioral assess-
 ment  of  toxicity  following  environmental  insults
 early in  development. More specifically, theoretical
 and methodological considerations providing a ration-
 ale for development  of a neurotoxicity screening
 approach for detection of early developmental insult
 effects are discussed, with primary emphasis on the
 use of  neuropathological  and  behavioral  testing
 methods.  Within that context, the question of the
 relative sensitivity  of such assessments for detecting
 neurotoxicity is considered and recent efforts aimed
 at  improving neuropathological screening procedures
 are noted.  It is  hoped that raising the issue of relative
 sensitivity  of neuropathology and behavioral screen-
 ing approaches  will stimulate discussion on  (1) how
 electrophysiological methods might be employed to
 augment  neuropathological  and  behavioral assess-
 ments of  neurotoxicity following insults  early in
 development and (2) what special problems are apt to
 be encountered in trying  to employ electrophysio-
 logical techniques in that manner.

     Assessment of  changes  in   "functional brain
 capacity" is a key  theoretical issue implicitly address-
 ed by most studies of neurotoxicity. Functional brain
 capacity, reflecting the number of functionally intact
 neurons in the central nervous system (CNS), number
 of functionally intact synaptic connections, and the
 integrity of neurochemical processes, determines the
 maximum  complexity of  neurobehavioral  responses
 mediated by the CNS of an organism. As indicated in
 Fig.  1, functional brain capacity is  thought to in-
 crease prenatally and neonatally with the rapid prolif-
 eration of neural  tissue during early development. It
 likely asymptotes  sometime after puberty, reflecting
 offsetting  effects  of normal neuronal cell losses and
 gains  in processing efficiency with experience. In old
 age functional capacity gradually decreases as the
 effects of cell losses predominate and lead to the
 manifestation of neurobehavioral deficits defined as
 "senility" or "senescence". Even  very small excess
 rates  of neuronal cell  losses of less than 0.1%, if
induced by  environmental insult early  in develop-
ment, could lead to substantial reductions in rate of
maturation and ultimate maximum level of functional
capacity  attained.  Or the effects of early  "mild"
damage might not be manifested until  late in life,
when "premature senescence"  theoretically could oc-
cur, even 10 to 20 years sooner than would otherwise
be expected.

    For assessment of toxic effects of exposures to
environmental agents early  in  development, such
exposures can be restricted to either prenatal or early
postnatal periods, or they can be of a more chronic
nature, extending throughout  these periods and pos-
sibly  throughout the entire life  of the animal. With
such exposures, assessments of toxicity can  only be
undertaken after the initiation of the exposure, thus
not allowing for accumulation of pre-exposure base-
line  data against  which  to  compare postexposure
effects. This necessitates  use of between-subject  ex-
perimental designs, with  longitudinal study of sub-
jects demanded by the nature of the problem in order
to detect possible immediate effects of exposures on
early development, as well as possible delayed effects
not  appearing until  adult life or old age. Toxicity
assessments appropriate at different  life  stages start-
ing during prenatal exposures include: (1)  prenatal
evaluations of unborn fetuses for embryotoxicity and
teratogenic  effects, including neuropathology evalua-
tions of brain damage, (2) periodic sampling of sub-
jects at various  life stages  postnatally for neuro-
pathology  screening, and (3)  postnatal development
assessments and adult behavioral testing on remaining
living subjects.

     With high exposure levels for many agents, neural
damage is  severe  enough to be detected by existing
routine neuropathology screening procedures. These
procedures  mainly consist of: (1) gross  external ex-
amination  of the  brain for obvious alterations in size,
shape, or surface  features, for signs of hemorrhage or
abscess, or  for other indications of abnormality, and

-------
422
                                                                                Grant
  RELATIVE FUNCTIONAL
     BRAIN CAPACITY
                                                               Maturity
 LIFE PERIODS
  AND EVENTS

 TOXICITY
  TESTING
  POINTS

 TYPES OF
 EXPOSURE
 PERIODS
     Conception   Birth         Puberty
         Prenatal    Postnatal
         Teratology   Developmental
         Screening     Testing
                                                                         Death
 Adult
Behavioral
 Testing
Late Adult
(Geriatric)
 Testing
Pregestation  Prenatal   Neonatal
                                  0   Q
                                   Acute
      Subacute Chronicdntermedlate, Whole-life, etc.)
                                                   Provocative Drug Tests ?
    Fig. 1.  Schematic representation of conceptual framework upon which are based suggestions for longitudi-
    nal designs for behavioral teratology studies. Normal changes in relative levels of functional brain capacity,
    determining the  maximum complexity of neuro behavioral responses and the neural reserve of an organism,
    are plotted (top)  in  relation  to periods and events occurring during the  life span of most mammalian
    species. Various  types of experimental exposure periods are depicted in relation to different points in life
    when  they might be administered to reflect actual types of exposure problems. Toxiclty testing points (*),
    when  different types  of anatomical and behavioral evaluations are appropriately conducted, are noted in
    relation to the exposure periods.  (From Grant 1976.)
 (2) rough serial sectioning to allow scanning for read-
 ily observable  internal lesions,  enlarged ventricular
 spaces, etc. It has been estimated that at least a 30%
 loss of neurons in a particular brain area would have
 to occur before existing standard  neuropathology
 screening procedures  could  detect such  a defect.  It
 thus appears unlikely  that such procedures would be
 adequate to  detect   small  reductions  in  neuronal
 density or other subtle changes in brain morphology
 resulting  from low-level exposures to various environ-
 mental agents,  which are  becoming  of increasing
 concern  to  environmental  lexicologists and  regu-
 latory agencies.

    Efforts  aimed  at improving  neuropathology
 screening procedures  include  work  by  Dr.  Martin
 Krigman  at  the  University  of North Carolina  in
 Chapel Hill. His screening  procedures attempt  to
 quantify  damage to  the  CNS  at  both the light-
 microscope  and  ultrastructural  level.  Light-
 microscope analyses include measurements of cortical
 mantle thickness and determinations  of neuronal
 density by meticulous counting of neurons in cortical
 or other brain  areas on sections subdivided  by grid
 coordinates  for  evaluation.  On  an ultrastructural
level,  analyses include  estimates of density of synap-
 tic contact points in  neuropil of cortex or neostria-
tum by  counting of  synaptic profiles in  electron
micrographs from sampled  brain regions. These quan-
tification  techniques  are estimated  to  be  sensitive
enough to reliably detect changes of less than 10% in
neuronal or synaptic density.
                                        Another morphological approach holding some
                                    promise for improved neuropathology screening is the
                                    use  of histochemical methods for the demonstration
                                    of  specific  catecholamine, indoleamine, or  other
                                    neurotransmitter-containlng neural pathways in the
                                    CNS. By  employing such methods in my laboratory
                                    (Nemeroff et al.  1977), it was possible to demon-
                                    strate  a marked prolonged  reduction  in numbers of
                                    dopamine  (DA)  neurons   present  in  the  arcuate
                                    nucleus of the hypothalamus in rats neonatally treat-
                                    ed with monosodium-1-glutamate  (MSG), a widely
                                    used food additive. The reduction in DA neurons was
                                    shown to still exist at time  points in the adult life of
                                    MSG-treated animals  when  an early lesion in the
                                    arcuate nucleus was no  longer readily detected with
                                    standard neurohistological procedures.

                                        Even with  improved  neuropathology screening
                                    procedures of the type outlined above, certain types
                                    of neurotoxic effects may not be detectable morpho-
                                    logically.  For example,  subtle  alterations in neuro-
                                    transmitter release effects on postsynaptlc transmitter
                                    receptor sites would not be picked up with improved
                                    neuropathology screening  and are even difficult to
                                    demonstrate biochemically. Presumably, however,
                                    such effects or neuronal  damage below neuropatholo-
                                    gy detection limits may  be reflected by alterations in
                                    behavioral responses  mediated by affected neural
                                    pathways or biochemical processes. Thus, in addition
                                    to neuropathology screening, assessments of toxic ef-
                                    fects on behavior would  seem to be advisable. In ad-
                                    dition,  it is here that electrophysiological techniques

-------
Neurobehavioral Assessment of Environmental Insults in Development
                                             423
 might also  provide  an increased level  of sensitivity
 over neuropathology screening methods for detecting
 subtle alterations in neural function.


     As for  postnatal developmental assessments and
 later adult behavioral  testing, such relatively expen-
 sive and time-consuming procedures should probably
 be reserved for exposure levels below those producing
 significant effects detected by prenatal teratology or
 neuropathology  screening  procedures. Evaluations
 designed to assess the  progress of maturation early in
 development might  profitably include not  only
 behavioral  tests,  but  also measures of growth and
 physical development. This  allows  for estimates of
 whether behavioral changes likely indicative  of  alter-
 ed neural  function  occur at exposure  levels below
 those producing general effects  on growth or matura-
 tion of other organ  systems. Thus, ages at which cer-
 tain  physical development  landmarks  appear  (e.g.,
 incisor eruption, eye  opening, vaginal  opening  or
 testes  descent)  should be  recorded.   In addition,
 behavioral assessments would focus early in  develop-
 ment on the maturation of certain  reflexes  (e.g.,
 righting,  auditory  startle,  and  visual  placing  re-
 sponses), as well as indexing the shift of the immature
 pattern of ambulation called "pivoting" to the  adult
 pattern of  straight-line forward locomotion  and the
 appearance  of  other  behavior  patterns (sitting up,
 grooming, climbing).  Kimmel et al. (1978)  describe
 an  approach employed  collaboratively by several
 laboratories for assessment of chronic low-level lead
 exposure effects on early postnatal development, us-
 ing many different  types of behavioral and  morpho-
 logical assessments of toxic effects.
     Adult behavioral testing might then proceed in a
 sequence  of increasingly more sophisticated and diffi-
 cult assessment procedures.  The  first, and grossest,
 level of assessment would consist of continuous moni-
 toring for overt signs of toxicity. This would include,
 in addition  to changes in body weight, deterioration
 of fur condition, observations on abnormal postural
 or  gait characteristics,  increased  sensitivity  to  han-
 dling, the occurrence of tremor  or repetitive stereo-
 typed motor  patterns, and  loss of reflexes.  The
 second level  of testing would involve  attempts at
 quantifying changes  in spontaneous  behavioral  re-
 sponses such  as locomotor activity, food and water
 consumption,  and  sex  behavior.  The third level of
assessment would involve analyses of behavior in
much  more  controlled  circumstances  and would
include operant conditoning paradigms to assess an
agent's  impact on  learning and  memory  functions
and/or shifts in sensory processes.  Since delayed ef-
fects of early insults may not be manifect until late in
adult life, one or more components of the above pro-
cedure may need to be repeated, especially as subjects
approach  old  age in  order  to  detect  premature
senescence.

    The  above testing approach  would permit  a
broad  and detailed description of neurobehavioral ef-
fects induced by a given agent. It would, however, be
very costly and time-consuming, with final results not
available  possibly for several  years if testing during
the geriatric period was necessitated by the lack of
detected  effects earlier  in life.  Alternatively, for
screening purposes in which any  sign of alteration in
neural function would perhaps suffice to warrant the
raising of "caution" signals for a given  agent, certain
heuristics might be adopted, such as (1) following up
findings of developmental delays,  however small, in
maturation of reflexes to see  if they are predictive of
later,  more  severe  behavior deficits;  (2) use of
 "apical"  behavioral  tests  in which paradigms are
employed that require a number  of  behavioral re-
sponse systems to be intact  for successful perform-
 ance  of the  tasks,  e.g., operant conditioning para-
 digms where food or water motivation, motor func-
 tions, and learning or memory processes must all be
 intact to  permit performance of the conditioned re-
 sponse^) at  normal levels; and (3) employment of
 "provocative" drug tests to look for increased vulner-
 ability of performance in exposed animals to deleteri-
 ous or facilitating effects of neurally active pharmaco-
 logical agents.

     Where electrophysiological techniques might ap-
 propriately be employed in  the above framework as
 possible  screening procedures for neurotoxicity is a
 subject for further  discussion.  Electrophysiological
 techniques  have  been applied successfully  to the
 assessment of neurotoxic effects in adult animals ex-
 posued to environmental agents (cf. other reports in
 this section).  However, there are  inherent methodo-
 logical difficulties in assessing the effects on subse-
 quent development of prenatal and neonatal toxicant
 exposure. The utility  of electrophysiological  tech-
 niques as neurotoxicity screening tools in such cases
 remains to be demonstrated.

-------
 NEURO-  AND  PSYCHOPHYSIOLOGICAL  EFFECTS
 OF  MODERATE  CARBON  MONOXIDE EXPOSURE


 E. GROLL-KNAPP, M. HAIDER, H. HOELLER, H. JENKNER, AND H. G. STIDL

 Institute of Environmental Hygiene, University of Vienna, Austria
    Numerous experiments have been conducted on
 the effects of CO exposure on human behavior (cf.
 reviews by Permutt and Farhi 1969, Coburn  1970,
 USDHEW (1970), and Beard and Grandstaff 1975).
 Probably due to differences in experimental design
 and parameter measurement, the effects of CO seem
 contradictory. Some aspects of human behavior are
 affected by low COHb levels while others show no
 impairment until the COHb level is higher. The capac-
 ity for continuous observation of weak or in frequent
 stimuli, as in a vigilance task, can be a sensitive indica-
 tor  of behavioral  impairment while complex psycho-
 motor tasks show no reliable impairment. Beard and
 Grandstaff (1975) have suggested that failure to de-
 monstrate impairment  on a complex task may be due
 to the intrinsic activating characteristics of the task
 that possibly facilitate  the mobilization of compen-
 satory  mechanisms. In the presence  of significant
 changes in behavior, some investigators infer an im-
 pairment of CNS function due to increased CO bur-
 den. In an effort to determine the mode of CNS im-
 pairment, spontaneous EEC activity, EPs, and CNV
have been examined.

    EEC has yielded  no  reliable  differences during
acute  exposure to low levels of CO.  Zorn (1964),
Shul'ga (1964), Stewart et al. (1970, 1973), Hosko
 (1970), and Groll-Knapp et al. (1972) found no al-
 teration in  spontaneous EEC with COHb levels as
 high as 33%, although Grudzinska (1963) observed
 an augmented occurrence of slow wave components
 in the EEC of workers chronically exposed to 100
 ppm CO with COHb  levels of less than  1%.  Ad-
 ditionally, an  increased  occurrence  of  flat  EEC
 accompanied by  irregular alpha rhythm was noted.
 In contrast to the absence  of EEC changes at high
 COHb  levels.   O'Donnell et al.  (197la) found al-
 terations in EEG sleep  patterns during  exposure
 to 75 and 150 ppm for 7 hours of nocturnal sleep.
 A nonsignificant  trend toward increased duration
 of deep  sleep  (stages  3 and 4) was noted.   CO-
 induced changes were more pronounced but not
significant during  the early  phases of sleep when
COHb levels were  lower.  It  was hypothesized  that
early  adaptation  mechanisms were stimulated  by
CO exposure during the initial 3 hr.

    The  analysis  of evoked and slow  potentials
permits investigation  of processes that are more spe-
cific  than those reflected in  the spontaneous EEG.
EPs as an indicator  of CO-induced functional CNS
deterioration in man were first investigated by Hosko
(1970). No significant changes in peak latencies or
amplitudes of VEPs were found below  15% COHb.
With COHb levels between 20  and 22%, an increase In
amplitude of the late potential component (70 msec)
and a negative-going shift after PI20 were  described.
Due  to  the  small  number of subjects  in the high
COHb group, these results must be regarded with cau-
tion.

    AEPs in CO experiments were studied by Groll
-Knapp in 1971 (unpublished). Experiments using 14
subjects at 0, 15,100 and 150 ppm exposure for 2 hr
showed a consistent, but nonsignificant  reduction in
amplitudes and a lengthening of peak latencies of late
components. The extent of AEP decrement appeared
to be proportional to the level of CO exposure.


    Slow potentials, which are considered to be CNS
correlates of attention, anticipation, and motor readi-
ness, were first used as indicators of CO-induced CNS
disturbance by Groll-Knapp et al. (1972). A signifi-
cant diminution of CNV amplitude during moderate
(50,100, 150 ppm) exposure was found.


    EP experiments with  animals have yielded con-
tradictory results.  XJntaras et al. (1966b) described
an increase in superior colliculus (SC) EP in rats  ex-
posed to CO levels  of 50 ppm. In contrast, Dyer and
Annau (this  volume) found  significant  changes in
SCEPs only during exposure of rats to 500 and 1000
ppm CO. During exposure to  500 ppm SCEPs were
 1 With financial support of the Osterreichlschen Fonds zur F8rderung der wlssenschaftlichen Forschung  and the
 commission of the European Communities.

-------
Effects of Carbon Monoxide Exposure

augmented, but during exposure to 1000 ppm, SCEPs
were decreased in amplitude.

    No reliable neurobehavioral effect has been de-
monstrated for  moderate COH6 levels. Furthermore,
several Investigators (cf. Winneke et al., this volume)
have been unable to replicate their own findings. This
predicament suggests either considerable variability in
the sensitivity of different neurobehavioral measures
to the hypoxlc effects of CO, or that slight changes In
experimental design or arousal level can mask the sub-
tle effects of moderate COHb levels. Studies of the
comparative sensitivity to CO exposure of different
neurobehavioral and electrophysiological  measures
and exploration of more effective methods for .con-
trolling the arousal  level  of  subjects are needed to
clarify these issues.

    Two experiments pertinent to these questions are
reported here. The first study examines the compara-
tive  sensitivity  of several behavioral (aiiditory vigi-
lance, simple addition, short-term memory) and neu-
roelectrlc (CNV; auditory, somatosensory,  and  visual
evoked potentials;  EEC power spectrum) measures
during exposure to 200 ppm CO. The second explores
nocturnal sleep as  a method to control voluntary mo-
dulation of attention and arousal.
 Experiment 1


 Methods
     Twenty  nonsmoking  unpaid  volunteers  (11
 male), aged 18 to 24 yr, served as subjects. Each sub-
 ject was exposed double-blind  to two conditions (0
 and 200 ppm CO) in a pneumatic chamber with ses-
 sions separated by at least 4 days. Each session lasted
 210 min and  resulted  in a maximum COHb level of
 12%. COHb levels are determined spectrophotometri-
 cally from blood samples obtained after each session.
 Observed values agreed well with values calculated by
 means of the Coburn et al. (1965) formula. Subjects
 completed the tasks twice during each session. For
 the 200 ppm condition, tasks were administered at a
 low COHb level (below 6% COHb) and again at a
 higher level (above 8%).

     EEC was  recorded on FM tape from Ag/AgCl
 electrodes at the  vertex  referenced  to  linked mas-
 toids. EOG was also recorded  for offline editing of
 eye movement artifact. A  CAT 1000 and Sunai Fre-
 quency  Analyzer were used  for off-line  data proces-
 sing. Power spectra (at 4 and 10% COHb) were com-
 puted in 10 sec epochs for frequencies of 5 to 15 Hz
 with subjects  at rest. Average power spectra were
 evaluated for  10 min periods.
                                            425

    An auditory vigilance task was used for CNV as-
sessment.  Pairs of 1  msec clicks were presented via
earphones with an intersignal interval of 1 sec and an
average interpair Interval of 4.17 sec. The intensity of
both stimuli was equal (96 dBA) for irrelevant pain,
while the Intensity of the second stimulus was slightly
less (93 dBA)  for relevant pairs to  which subjects
were required to respond. The vigilance task consisted
of three  continuous  series, each containing 41  rele-
vant and  200 nonrelevant  pairs. CNVs were selec-
tively  averaged for hits, misses,  and  irrelevant pain
using a 4 sec sweep rate and N-100. CNV amplitude
was measured as shown In Fig. la.
                        im
                                                              Al Al
                                                                                                1NC
                        TIMt,
 Fig. 1,  Measurement procedures for CNV amplitude
 (a) and peak-to-peak components of auditory (b), vis-
 ual  (c),  and somatosensory (d)  evoked  potentials.
 CNV amplitude was measured at S2 onset relative to
 the maximum positive peak 300 to 400 msec after SI
 onset. Negativity upwards.

      Auditory (A), visual (V), and somatosensory (S)
 evoked potentials  were assessed twice  beginning at
 times 10 and 120 min. Blocks of 100 stimuli in each

-------
 426

 modality were presented in the order A-V-S. A fixed
 ISI of 4.17 sec was used in all modalities. AEPs were
 elicited by 1 msec, 96 dBA clicks. A special preheated
 fluorescent tube 40 cm in front of the subject's eyes
 was used  for visual stimuli.  Intensity was adjusted
 during a pretest to obtain a clearly defined VEP with
 eyes closed. SEPs were  elicited by dc stimulation of
 the median nerve of the left arm 10 cm  proximal to
 the wrist. The intensity  was adjusted for  each subject
 to obtain a clear sensation just below the pain thresh-
 old. A 1-sec sweep rate and N=100 were used for EP
 summations.  Peak-to-peak component measures  for
 each modality are illustrated in Fig. Ib-d.

     Heart rate was monitored throughout the experi-
 ment. Average heart rate/min was determined during
 each subtest.

     Behavioral measurements included auditory vigil-
 ance performance (between  6 and 8.4% COHb), an
 addition task from the Horn Intelligence Test, and a
 memory test (10-11% COHb). The  memory test con-
 sisted of 11 nonsense syllables presented visually and
 orally, each  repeated 10 times for 2.5 sec. Subjects
 were tested  immediately after presentation and 17
 min later (Butollo  1969). A subjective  mood  scale
 was completed  at the beginning and end of the ex-
 periment. Nonparametric tests were used to evaluate
 significance.
 Results
     Most variables failed to show clear changes in
 response to CO exposure, although selected measures
 showed significant effects. No difference in the EEC
 power spectrum  could  be demonstrated for COHb
 levels of about 10%, although a general trend in dimi-
 nution in power  density values was seen at levels of
 4% COHb for all frequency bands. During CO expo-
 sure, AEPs  were  slightly  reduced  in  amplitude  for
 components later than 50 msec.  No changes in peak
 latencies could be seen.  Amplitude reduction in rela-
 tion to control was seen during the  first measurement
 (< 4.5% COHb), but not the second (> 8.6% COHb).
 For some subjects this difference was very clear (Fig.
 2), but for the whole  group the differences were not
 significant  (Table 1). No changes  in amplitude  of
 VEPs attributable to CO were observed. Longer laten-
 cies  under  CO for the  late  components were more
 pronounced during the first measurement series (4.9%
COHb) than the second  (9.1% COHb, Table 1). After
median nerve stimulation, amplitude reductions of all
late  EPs could clearly  be seen  under CO, but  no
 changes in latencies were observed. The CO effect was
 more pronounced during the first (5.5% COHb) than
 the second measurement (9.7% COHb, Table 1).
         200 msac
  6//V
                                GrolI-Knapp et al.
CO
CONTROL
                                     4.5% COHb
                                      6% COHb
 Fig. 2.  Example  of AEPs  recorded in one subject
 during control and 200 ppm  CO conditions at two
 different points on the COHb uptake curve.
                                                            20JUV
                                                            10 JiV

1


i 	 1
C
2
0
(X
NTRO
Jppm
L

n p n p n p
HITS MISSES NRS
 Fig, 3.  Mean CNV amplitude for hit, miss, and non-
 relevant (NRS) trials of the vigilance test during expo-
 sure and control conditions,

     CNV  amplitude  was measured  in relation to
 P330. The only significant difference in CNV ampli-
 tude occurred between CO and control conditions for
 relevant signals that  were  not correctly  detected
 ("misses"). In other categories of stimuli (hits, nonre-
 levant  pairs),  no CO  effect was observed  on CNV
 amplitude (Fig. 3). Pulse rates during exposures be-
 low 6% COHb showed a slight depression. In contrast,
 COHb  levels greater  than 8%  showed a relatively
 higher pulse rate than control (Fig. 4).

    No significant difference in missed signals or false
alarms in the vigilance task (Table 2), or in the addi-
tion test was observed between CO and control condi-
tions. Not quite as many nonsense  syllables were re-
called in the CO as in the control condition.

-------
Effects of Carbon Monoxide Exposure

             Table 1.  Peak-to-Peak Amplitudes (juV) of Evoked Potential Components
                       after Auditory, Visual, and Somatosensory Stimulation
                                                                                              427




AEP A1
A2
A3
A4
VEP A1
A2
A3
A4
SEP A2
A3
A4
0 ppm'
Series
1
7
16
29
24
3
8
13
8
8
20
18
Series
2
4
10
18
14
3
7
17
12
8
22
21
Diff
1-2
*
ft
*
*
NS
NS
*
*
NS
NS
NS
200 ppm
Series
1
5
13
26
24
3
7
13
9
5
13
11
Series
2
4
9
18
14
3
8
16
11
6
17
15
Diff
1-2
NS
*
*
*
NS
NS
NS
NS
NS
NS
NS
Diff
0-200
Series
1
NS
NS
NS
NS
NS
NS
NS
NS
NS
*
»
ppm
Series
2
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
*
•p < .05 (Wilcoxon-matched-pair tigned-rank companions); N - 20 for all comparisons. NS-nonsignificant.
    There were no significant differences In mood
ratings,  although  subjects tended to feel  less ex-
hausted, less overburdened and  less  tense after CO
exposure.

Experiment 2


Method


    Ten  nonsmoking, unpaid student volunteers (5
male) aged 20 to 25 and having normal EEGs, partici-

         PULSE RATE/min

            75''
         •|  74
         •0
         1  73
          5  72
          oc
                                                      pated. Subjects were screened for absence of sleep,
                                                      cardiac, or pulmonary disturbances. The single-blind
                                                      experiment included three sessions, separated by at
                                                      least  14  days. The first session was used for adapta-
                                                      tion,  and the second and third sessions were used for
                                                      either control (0 ppm CO) or experimental (100 ppm
                                                      CO)  purposes. Subjects  slept in  an experimental
                                                      chamber for about 7 hr each night (10 p.m.  to 6
                                                      a.m.). EEC  was  continuously recorded  during the
                                                      night from  Cz referenced to a mastoid (Ml).  EOG
                                                      was also  recorded for rejection of eye- movement-arti-
                                                      fact  trials  during offline sleep stage scoring. Sleep
                                                      stages were determined by visual EEC inspection, on
             70
                                         CONTROL
                                         200 ppm CO
                r_
                    -»	h
                                   »        1         2         1       tt     X
                                   X                                   "     S
                                                 VIGILANCE            5     £


                    Fi£. 4.  A verage heart rate during successive phases of the experiment.

-------
 428
                                                             Groll-Knapp et al.
  Measurement
                                   Table 2.   Vigilance Performance
   Condition
   1-20
                                                                   Series (Minutes)
  21-40
41-60
Mean
% Misses

% False alarms

Control
200 ppm CO
Control
200 ppm CO
31.4
32.1
7.1
6.7
43.5
48.3
4.9
4.6
55.3
50.9
2.7
4.3
43.4
43.7
4.9
5.2
 the basis of 1-min intervals according to Rechtschaf-
 fen and  Kales  (1973) criteria. For  EP evaluation,
 clicks (70 dBA, 3 kHz, 8.9 msec duration) were pro-
 duced at a  random  interval averaging 22.6 sec by a
 loudspeaker placed 50  cm above the subject's head.
 AEPs were analyzed selectively  for  different  sleep
 stages. Each AEP was a summation of 30 successive
 trials. For statistical comparisons of CO and control
 situations,  an  average  potential  was calculated for
 each sleep stage. Amplitude of negative and positive
 peaks was measured  in relation to a prestimulus base-
 line. A subjective mood scale was completed before
 and after each session. Additionally, the memory sub-
 test of the  I-S-T intelligence  test (Amthauer  1953)
 was used to assess recall. A 3-min learning period was
 followed by testing 6 min later and again the follow-
 ing morning. COHb level was calculated by means of
 the  Cobum et al.  (1965) formula. Nonparametric
 tests were used to evaluate the  data (p <  .05).
Results


     Exposure to 100 ppm CO during sleep, calcu-
lated to raise COHb levels to 10-12%, caused a signifi-
cant change in sleep-stage distribution. During CO ex-
posure, the percentage of sleep stages 3 and 4 in-
creased, and stage 2 and REM sleep decreased (Table
3). Compared to the control  condition, REM sleep
decreased by  18% and deep sleep  (stages 3  and 4)
Increased by about 29%.

    AEP  latencies  analyzed  during  sleep stage  2
showed no differences between control and CO con-
                               ditions (Table 4). All positive components clearly in-
                               creased during CO exposure. The P200 difference was
                               significant, while the P900 and P2000 differences fell
                               just  short  of significance. Negative  components did
                               not differ between  control and CO conditions. The
                               amplitude  change observed during sleep  stage 3 re-
                               sulted exclusively from  enhancement of the positive
                               peaks.

                                   In sleep stage 4 (Table 4), the peak latency of the
                               P200 component was significantly lengthened during
                               CO exposure.  Amplitude measures also showed very
                               clear changes (Fig. 5, Table 4). All positive compo-
                               nents were significantly larger  under CO, and negative
                               components showed a nonsignificant decrease in com-
                               parison to  the control condition.  While  the AEP of
                               sleep stage 4 showed no change in absolute size due
                               to CO, there was a significant shift of the whole wave-
                               form in the positive direction.

                                   Significantly  more words from  the memory test
                               were remembered after the  control night than the
                               experimental night.  No significant  differences were
                               found on the subjective mood scale, but there  was a
                               tendency towards negative responses in the CO condi-
                               tion. Similar results have been reported by Bender
                               (1971).
                               Discussion
                                   The results indicate that the SEP is the most sen-
                               sitive  to CO influence of the EPs examined.  Reduc-
                               tions  of all  late SEP components under CO were
                               clearly apparent (Table 1). That  all sense modality
                             Table 3.  Percent Time in Each Sleep Stage
Condition
Awake
Stag* 1
Stag* 2
Stag* 3
   REM
Control
100 ppm CO
12.2
13.0
6.3
5.5
58.5
57.8
10.4
13.4«
12.6
10.3«
 *p < ,OB, WllcoKon m«tohtd'|Mir ilgned-nnki compirlioni v». control valu«i

-------
Effects of Carbon Monoxide Exposure
                                         429
                        Table 4.  Latencies (L) and Amplitudes (A) of AEP
                                    during Different Sleep Stages
Steep Measure-
Stage ment Condition
Control
L.msec
100 ppm CO
Control
A,/LiV
1 00 ppm CO

P1
206

206
+15

+19a

N1
330

335
-16

-18

P2
424

412
-2

-5

N2
602

577
-29

-37

P3
1024

1017
+24

+31

Control
L.msec
1 00 ppm CO
Control
100 ppm CO
198

204
+10
+15a
418

441
-43
-41
922

919
+28
+34
1572

1600
-20
-20
2036

2152
+7
+13

Control
L.msec
100 ppm CO
Control
100 ppm CO
186

209a
+4
+138
445

464
-43
-29
963

920
+30
+37a
1493

1524
-21
-19
2038

1986
+1
+108
  •p < .06 (compared to control value); Wllcoxon; N • 10.
                                  — 0 ppm CO
                                   ~" tOO ppm CO
        l 4
   Fig, 5.  Effect  of CO exposure  on the auditory
   evoked potential during ttage 4 sleep.

   decrements in amplitude or changes in latency were
   greater at lower (4.5 to  5.5%) than at higher (8.5 to
   9.7%)  COHb levels Is noteworthy. Wlnneke et  al. (this
   volume)  examined AEPs in a CO itudy using a vigi-
   lance task with two levels of monotony. The  authors
   summarize their results by stating that no obvious
effects due to CO or to monotony were apparent.
However, analysis was  done separately for negative
and positive peaks and examination of peak-to-peak
amplitudes  suggests that a reduction in mean ampli-
tude occurred in the monotonous condition. Reduc-
tions of 13, 14, and 23% were seen at COHb levels of
6.5,7.4, and 11.3%, respectively.
    Since conscious mediation of cortical function is
suspended during sleep, impairments in CNS function
due to exogenous or endogenous factors may be ob-
served more clearly.  Colmant (1972)  found a CO-in-
duced  decrease in  REM  sleep  in animals  and
O'Donnell et al. (197la) reported trends toward alter-
ations in  human sleep patterns. The data reported
here revealed a significant alteration of sleep patterns
under CO. Ail EP components in sleep stage 4 were
significantly enhanced, whereas in sleep stage 2, only
P100 was significantly larger in amplitude. Subjects
appear more  susceptible to CO effects in deep sleep
than  in  shallow sleep. The  positive  translation
under CO should be  considered in relation to general
dc shifts, facilitation, and the availability of neurons
during different sleep-stage arousal levels.

-------
430

    EP  methods  can  demonstrate  CO-lnduced
changes of CNS functioning at low COHb levels when
compensating mechanisms are not in effect-and at
high COHb levels when compensating mechanisms are
insufficient. During sleep stage 4, EP component am-
plitudes were significantly augmented, but no differ-
ences in AEP amplitudes were found in subjects a-
roused after 7 hr sleep exposure to 100 ppm CO. One
may hypothesize that awakening activated a compen-
satory mechanism absent during sleep.

    No significant impairment of vigilance perform-
ance was observed during CO exposure. Winneke et
al. (this volume) reported similar results. Groll-Knapp
et al. (1972) found a significant  decrement in  vigi-
lance performance during CO exposure (50, 100,150
ppm) for 2 hr.  Failure to  replicate these results may
be due  to  variations  in experimental design.. Heart
rate was higher during CO exposure than the control
condition, and one could infer that the increased rate
resulted from compensation  for the additional CO
burden.

    During the acoustic vigilance task, CNV measure-
ments  were obtained, but CNV differences were ob-
served during CO and control conditions only when
trials  were  selectively   averaged  for  "hits"  and
"misses." Groll-Knapp et al. (1972) reported a signifi-
cant decrease in  CNV amplitude  as the level of CO
exposure increased. Failure to replicate these results
may again be due  to alterations  in  design or other
uncontrolled variables.

    CO effect on subjective feelings was different af-
ter short-term and sleep exposure. Following short-
term exposure, subjects felt less exhausted, less over-
burdened, and less tense than after the control condi-
tion,  although  these. changes were  not  significant.
Weber et al. (1975) reported similar, but significant,
results.  After sleep exposure, however, subjects felt
more negative.
                              Groll-Knapp et al.

    Different studies of behavioral and neurophysio-
logical parameters under moderate CO exposure have
shown contradictory  results.  Neurophysiological
measures appear to be more sensitive than behavioral
measures to the effects of low COHb concentrations.
These  effects are more consistent during sleep ex-
posure when the inter- and intraindividual variability
is lower. One reason that effects are difficult to re-
plicate in awake subjects may be that healthy young
adults  are able to mobilize compensatory mechanisms
during short-term exposure. Differences between ex-
periments using low-level CO exposures could result
from  a disassociation between  the  onset of CO-in-
duced changes in CNS function and the onset of com-
pensatory mechanisms.
Conclusions
    The public is exposed to environmental toxicants
without choice, knowledge, or defense. Healthy per-
sons as well as  others with increased risk due to func-
tional  disturbances or disease are exposed  indiscri-
minantly. The  adverse effects of simultaneous expo-
sure to multiple environmental  stressors, moreover,
may be synergistic. Results  of  the present experi-
ments  suggest (1) that appropriate neurophysiological
measures reflect functional changes during moderate
CO exposure;  (2)  that  measures of CNS function
should be included in standard test batteries used to
assess  the  adverse  effects  of  environmental  sub-
stances; and (3) that neurophysiological evidence of
this type should be considered in establishing expo-
sure standards.  Sleep studies may be helpful in a-
voiding certain theoretical and practical problems in
the study of influences of environmental factors on
physiological and neurophysiological mechanisms.

-------
APPLICATION  OF ERP  TECHNIQUES
IN NOISE RESEARCH
N. LOVELESS
Department of Psychology, The University, Dundee, Scotland
    Unlike other agents in clinical toxicology that are
subject to neurobehavioral research techniques, noise
is not a toxicant that produces central nervous system
(CNS) poisoning, and there is no conclusive evidence
that it has any effect on health other than the loss of
hearing  that  may  result from prolonged exposure.
However, few would deny that noise is an environ-
mental insult simply  because of its psychological ef-
fects; and since  it may be important to determine
how far other agents have specific effects that may be
differentiated from general reactions such as distrac-
tion,  irritation, or fatigue, noise  is certainly among
the problems that  should be considered  in stress  re-
search.

    The potential  contribution of ERP techniques to
noise research may best be examined by reviewing
what behavioral research has, and has not, already
achieved. A striking feature of research in this field is
that psychologists have had great difficulty in demon-
strating any noise effects of practical importance, and
those they have found have proved in  some  ways
rather puzzling (Broadbent 1957,  1971). Noise can be
a nuisance and the reasons why hardly seem mysteri-
ous. Noise distracts us, and may alarm or startle us as
well. So where is the problem?

     Investigations  of distraction have tested perform-
ance  on a wide variety of tasks, ranging from simple
reaction time (RT)  to complex intellectual skills,
under the most distracting noises imaginable. The uni-
form finding has been a transitory disturbance of per-
formance when  the  noise Is turned on and another
when it is  turned off, the amount of disturbance
 being roughly proportional to the change in noise lev-
 el, up or down  (Teichner 1963). These features are
 reminiscent  of an orienting response, in which case
 habituation  would  be expected, and it is indeed
 found  that after  the  initial disturbance the subject
 continues    with    his   performance   apparently
 unimpaired.
    "Apparently" is used advisedly. Persistent  dis-
tracton effects can be found if performance is closely
examined. Fisher (1972), using a choice reaction task
with bursts of noise at random intervals, found that
response was slowed-but only on those trials where
the noise  occurred less than 200 msec  before  re-
sponse. It  appeared to affect response initiation, as if
deciding to ignore the noise interfered with deciding
which response to make. This suggests  that distrac-
tion effects may be so subtle that they will not be
detected by gross average measures but only by such
microanalysis, or by increasing the sensitivity of per-
formance  measurement in some other way (Poulton
1965). Such an approach runs the risk of obtaining
effects that are of theoretical interest but little prac-
tical importance (Chapanis 1967). Common observa-
tion indicates that  people  can  perform many tasks
surprisingly well in noise. Nevertheless, they are apt
to complain of the strain it places  upon them, and
this complaint is  supported by evidence of increased
muscle tension and elevated heart rate (Davies  1968).


     The possibility that environmental stressors may
 have effects that are not apparent in performance is a
 major reason for considering whether ERPs may pro-
 vide more sensitive measures. In  the case of distrac-
 tion, there is an obvious lead: in his review of psycho-
 logical research  on the CNV, Tecce (1972) remarks
 that "distraction is one of the most powerful  vari-
 ables that can disrupt CNV development." Pioneer
 investigations at  the  Burden Institute (Dargent and
 Dongier 1969) showed that  the CNV was diminished
 by distractions, such as conversation or music, that
 continued throughout the trial. The effect seemed to
 be most marked  on the "early CNV," and to depend
 on how effective the distraction was for the individ-
 ual subject. A series of experiments by Tecce and his
 collaborators (1972, and this volume) have attempted
 to ensure the effectiveness  of distraction by  using a
 secondary task. Again it was mainly the  early  CNV
 that decremented, and RT increased as well.

-------
432
                                         Loveless
   QUIET
   NOISE 1
   NOISE 2
            ZOjuV
Fig.  1.  CNVs from three sets of eight  consecutive
trials, one set performed in quiet followed by two in
continuous white noise at 100 dB.


     None of these investigations involved what is nor-
mally referred to  as noise. The present writer there-
fore carried out some informal observations, using a
paradigm in which both  SI and S2  were visual. Con-
tinuous white noise  at  100 dB produced a slight ef-
fect on the early CNV on the first few trials; but
  QUIET
  CLICKS 1
  CLICKS 2
            20
Fig. 2.  CNVi from  thrtt ten of tifht contecuttvt
trials,  OM »*t ptrformtd in quttt followed by two
with loud cltckt pruenttd at ihort random  intervals.
 averaged over the first set of eight, it was barely per-
 ceptible, and the effect was not apparent on the sec-
 ond set (Fig, 1).  RT was, if anything, a little faster,
 with some tendency toward premature response.

     This  result is perhaps not surprising. Compared
 with conversation, white noise is meaningless and, by
                                                         QUIET
                                                        SOFT TONES
 LOUD TONES
A
            20juV
                           1stc
Fig.  3.  CNVs from three sets of eight  consecutive
trlalt, one performed in quiet, one with an unpredict-
able  series of soft (60-dB) tones preceding SI, and
one with a similar series of loud (100-dB) tones,


definition, featureless. It is difficult, therefore, to see
white noise as distracting. Loud clicks and tones, pre-
sented at  irregular intervals, certainly sound more dis-
tracting. Clicks clearly depress the early CNV during
the first eight trials, but there is substantial recovery
on the  second set of trials (Pig. 2). Even loud  tones
are soon gated out.

    All these effects come from stimulation that con-
tinues throughout the  trial. What happens with more
localized stimuli? Tecce (1972)  reports that distrac-
tion after SI  also decreases CNV and increases RT,
Distraction before SI  decreases  CNV, but RT is not
affected.  McCallum and Walter  (1968) report little
effect on  CNV when the distractor  is in a different
modality  from SI. My observations  suggest that in-
tensity is  important, since compared with soft tones,
loud tones  have  a marked  effect across modalities
(Fig.  3).  Since the early CNV  again seemed to be
affected, the foreperiod was increased to 3 tec. (Fig.
4). The whole CNV appeared to be flattened, except
for a very local readiness potential. RT was unaffect-
ed.

-------
Application of ERP Techniques in Noise Research
                                             433
    Summarizing, stimulation presented throughout
the trial, or during the foreperiod, or before the warn-
ing signal, all depress the CNV, especially the early
CNV; but the effect Is transitory, and RT Is affected
only when distraction is present during the foreperi-
od. Stimulation  before  SI  does not affect  RT, even
though flattening of the early CNV suggests that the
alerting effect of the  warning signal has been largely
eliminated. The explanation is not clear. Tecce (1972)
pointed out that speech (or any continuous stimulus)
causes a baseline shift. Does this mean that the alert-
ing effect of the warning signal has been preempted?
Or are these ERP effects epiphenomena without func-
tional significance?
  QUIET
  LOUD TONES
 Fig. 4.  CNVt from  two sets of eight  consecutive
 trtalt, one performed in quiet, the other  with an un-
 predictable  series  of loud (100-dB) tones preceding
 SI. The foreperiod has been lengthened to 3 sec.

      The relevance of this question to the present dis-
  cussion lies in the apparent dissociation of ERP and
  behavioral effects. The effect on RT seems to be lo-
  calized; distraction must occur close in  time to the
  response (Fisher 1972). The ERP effect  seems much
  less localized; loud stimuli before  SI have an effect
  luting several seconds. Can this be a case of an ERP
  providing a more sensitive measure than behavior?
  This seems arguable. There would be no difficulty If
  stimulation before SI had a greater effect on the ERP
  with increasing intensity (which teems to be true) and
  only began to  affect RT at high Intensity; but the
  limited  observations  presented hare suggest that  it
  does not affect RT at all.

      This result raises two Issues. First: to use an ERP
  effect  as an index of environmental insult requires
  validation against some practical criterion, such as dis-
  ease, discomfort, or Inefficiency; so what constitutes
  practical validation? Second: it is not enough to get
  an effect; It is equally Important, for practical as well
  as theoretical reasons, to know what kind of effect It
  Is. Tecce (1972) proposed  that to demonstrate dis-
  traction, It is necessary to show both that the alleged
dlstractor is actually processed, and that there is per-
formance decrement on  the  central task. According
to this criterion, the effect of stimulation before SI
cannot be described as distraction. The interpretation
of dual task performance itself poses problems that
should not  be underestimated (Brown  1964,  Rolfe
1969). It seems clear  that the nature of effects ob-
tained with the CNV requires further Investigation.

     The problem of identifying distraction can be il-
lustrated further by turning to  the  one case  where
noise has been shown  to have more than a transitory
effect. Persistent  noise  effects have  been  shown on
so-called "continuous performance  tasks,"  such as
vigilance, serial  reaction, and  tracking (Broadbent
 1971). The main feature of these tasks seems to be
that it is difficult to compensate for failing perform-
ance by increasing effort. Compensatory effort is so
endemic a problem in stress research that we should
seriously consider whether ERP techniques could pro-
vide an Indicator of such effort.

     The effect of noise on continuous performance
 tasks was initially  attributed to distraction (Broad-
 bent 19S3). It was argued that as subjects continued
 to  work at extremely  boring  tasks, their attention
 tended  to  wander, and was Increasingly likely to be
 captured by noise. The effect on performance might
 well be described as  a  lapse of attention; but there
 was never any direct  evidence that attention was cap-
 tured by noise,  and there  are  a number of  reasons
 why this seems implausible. As  noted above, it is not
 clear how continuous  broadband noise, commonly
 used in these experiments, can be described as distract-
 ing. This doubt was reinforced when it was discovered
  that tasks sensitive to  noise were alto sensitive to a
  wide range of other stressors, including some, such as
  sleep deprivation, that are  even more difficult to see
  as distracting. This raised the question whether differ-
  ent stressors affected the tasks in the same way; and
  this question was answered by  investigating the inter-
  action of stressors.

      Interaction  effects are of  considerable practical
  Importance. Stressors  that have little  effect  in Isola-
  tion may have marked effects in combination (Chap-
  anlt 1967). Noisy environments commonly contain
  other ttressors - one reason why it has been difficult
  to determine whether  noise exposure is implicated in
  circulatory disease (Broadbent 1961). Interactions are
  also Important theoretically, as a way of determining
  whether stressors affect the same or different me-
  chanisms, If  different mechanisms,  their  effects
  should be Independent and additive; if the tame, they
  should Interact.

       Experiments on combinations of stresses (Wilkin-
   son 1969) have shown that noise Is antagonistic to
   sleep  deprivation but  synergistlc with Incentive, thus
   placing it on  a continuum from drowsy,  slow but

-------
434
                                         Loveless
 accurate performance to over-excited, fast but erratic
 performance - a dimension that has been identified
 as "arousal." The arousal theory of stress (Broadbent
 1971) supposes that arousal is related to performance
 by an inverted-U function. The main effect of arousal
 is upon selective attention of the  "stimulus set" or
 filtering" type, which operates upon  the intake of
 sensory information. In states of high arousal,  atten-
 tion  becomes  selective to the point  of  "tunnel vi-
 sion," when performance deteriorates because there is
 insufficient sampling of low probability sources. Con-
 versely, low arousal reduces the bias towards the pro-
 bable,  attention is insufficiently selective, and per-
 formance deteriorates  because  there is  insufficient
 sampling of task stimuli (Hockey 1973).

     This leads to a predicted ERP effect, since Picton
 and Hillyard (1974; see  also Tueting, this volume)
 have argued plausibly that filtering is reflected  in the
 amplitude of the Nl component of  the  auditory e-
 voked potential and analogous components in  other
 modalities. These ERPs are  supposed to reflect the
 activation of a mechanism that selectively passes sen-
 sory  information for further processing on the basis
 of primitive feature analysis, and in accordance with
 attentional demands  that  have  usually been set  by
 instruction. Loveless (1977) presents evidence that
 when subjects  anticipate noisebursts  loud enough to
 be  disturbing,  attention is biased away from a  visual
 task, and that  the  behavioral decrement is accompa-
 nied by enhancement of the N120 component  of the
 visual evoked response to the imperative stimulus.

     A similar effect should be obtained when a con-
 tinuous performance task is performed in  the pres-
 ence of intense broadband noise, which is believed to
 heighten arousal, since Nl is increased by  incentive
 and decreased in drowsiness. Otto and Benignus (this
 volume) report the novel  finding that when perform-
 ing  a similar  task  in low-frequency noise, subjects
 complain of drowsiness rather than irritation. There is
 an  increase in the number of missed signals,  which
 with a  constant false-alarm rate would indicate  a dec-
 rement in the d'parameter of signal detection theory,
 which is the result obtained under sleep deprivation
 (Wilkinson 1969). While their ERP measurements do
 not entirely  dissociate evoked response components
 from slow potentials, they strongly  suggest that the
 amplitude of the Nil 0 component of the visual evok-
 ed response decreases. The data therefore  consistently
 indicate a decrease in selectivity with lowered arousal.
 It  is not clear, however, why low-frequency noise
 should have this effect.

     The point here is that arousal  theory specifies
 only the relationship between arousal level and per-
 formance; the relationship between stressor level and
 arousal level remains to be determined, and experi-
 ence with a number of stresson suggests that  it is
 unlikely to be simple. A good example is the case of
thermal stress, which also illustrates more generally
the problems  that arise in applied research when in-
vestigators aim directly at practical results with little
concern for theory (cf.  review  by Griffiths 1970).
Much early  effort attempted to  link subjective com-
fort directly with environmental variables, resulting in
a number  of stress indices, such  as the  effective tem-
perature scale, none of which is really satisfactory-
Attempts  were then made to relate  performance to
effective temperature. This led to a mass of inconclu-
sive  and   inconsistent results, which left  it unclear
even what kind of tasks were  affected by thermal
stress, let alone why or how.
     It was then suggested that performance might be
 related to body temperature rather than ambient tem-
 perature, but this approach was not entirely success-
 ful  either. A particularly  puzzling  feature was the
 temporal pattern of performance: there seemed to be
 an  initial disturbance,  followed by  recovery,  but
 sometimes there was a later deterioration. The expla-
 nation appears to be that  the initial impact of heat
 raises the level of arousal, but arousal subsides as soon
 as  the disturbance of thermostasis  is countered by
 physiological compensation. As long as the compensa-
 tory reaction is succeeding, arousal remains normal;
 but if compensation shows signs of failing, arousal
 rises again. The critical variable is neither skin temper-
 ature  nor core  temperature, but  the  relationship
 between  them;  performance  deteriorates when  this
 relationship  changes in such a way  as to threaten a
 displacement  of  core  temperature.  The  same
 relationship  appears to underlie subjective discom-
 fort (Provins et al.  1973). Thus,  with an adequate
 account  of  the  stress  reaction, physiological,  per-
 formance, and  subjective  effects all fall into place.
 This seems to be a crucial issue  in stress research. In
 the case of  hypoxia it may be equally important to
 understand  the  compensatory  mechanisms and the
 factors that control their threshold.
      It may be precisely because such an analysis has
  been lacking that the psychological effects of noise
  stress have been puzzling. It has never been clear why
  effects have been found only at intensities above 90
  dB; and again  the temporal pattern has been odd. The
  effects do not appear until the subject has been work-
  ing in noise for at least IS min, and when the noise is
  turned off, performance in quiet continues to suffer
  for about the  same length of time (Hartley 1973). A
  time-lag  of this  order strongly suggests a humoral
  mechanism, and  recent  evidence  (Simpson  et al.
  1974) strongly implicates blood-sugar level. This idea,
  of course,  is not new - Selye (1950) saw it as part of
  his "general alarm reaction" - but the precise details
  are important. It appears that  an initial neural re-
  sponse, sustained by  the secretion of glucagon and
  epinephrine, produces a peak in blood-sugar in about

-------
Application of ERP Techniques in Noise Research
                                             435
10 min. If the stress is not too severe, this is counter-
ed by the release of insulin (to prevent loss of glyco-
gen from the muscles) so that the  trend is reversed
towards  hypoglycemia.  This time-course correlates
very well with the performance change, and the rela-
tionship  appears to be causal because the effect can
be prevented by preloading with glucose.

    If the reaction to intense noise is viewed as  part
of a general  stress response, it may be easier to formu-
late a comprehensive view of noise effects. The dis-
covery of the  persistent effect  of continuous noise
has  tended  to  divert  attention from intermittent
noise, which is perhaps more common in practice.
While sounds of moderate intensity tend to produce
only a transient disturbance, a series of experiments
by Woodhead (see Broadbent 1971) has shown that
intense bursts of noise produce a more prolonged and
persistent effect that seems to involve an impairment
of sensory  intake. It can hardly be coincidence that
the threshold for this effect, Uke that of continuous
noise, is in  the region of 90 dB,  nor that it is at about
this level that the defence reaction replaces the ori-
enting reaction.

     Attempting to interpret the distraction and stress
 effects  of  noise in terms  of orienting and defence
 reactions is not just a translation into jargon; it allows
 us  to draw  upon  what  is known about these re-
 sponses. For example,  there should be differences in
 threshold between  meaningful   and meaningless
 noises, because Sokolov (1963) has shown that  when
 a  sound has signal value,  it  elicits the orienting re-
 action over a wider range of intensities.  Distraction
 may be produced by meaningful stimuli near auditory
 threshold,  and a sound that may be stressful  when
 task-irrelevant  may be acceptable  as a signal. Stimuli
 that are biologically important, either because  of in-
 tensity  or because of  acquired significance, do not
 habituate  readily. This is  of practical importance be-
 cause it has proved difficult  to recommend a critical
 level for intermittent  noise precisely because of un-
 certainty about habituation.

      An adequate account of noise should also com-
 prehend  subjective  reactions.  Psychologists   have
 shown  some tendency to shy away from the problem
 of annoyance, perhaps sharing  the view of Broadbent
 (1957) that  it is  "largely  unpredictable by scientific
 methods." While  some people are more  sensitive to
 noise than others, and annoyance is to some extent
 idiosyncratic, treating  it as strictly subjective provides
 an easy escape for noise makers, and perhaps diverts
  scientists  from problems  that merit study. The mini-
  mal condition for annoyance seems to be distraction,
  in that it  is precisely the "attention-getting" features
  of sounds that are universally  found  to be annoying.
  These  features correspond strikingly with standard
  determinants of the orienting  reaction, such as inten-
  sity, novelty, uncertainty, and complexity. These are
also  features that have linked with late components
of the auditory evoked potential. Nl, for example,
though  clearly  related to physical features such as
intensity, is  not  simply  related to them. It reflects
change in intensity, unexpectedness, and the extent
to which a stimulus demands further evaluation of its
significance and so diverts processing capacity from
current  activity.  If annoyance is subjective in the
sense that some people are more sensitive than others,
the same  has been found in evoked potentials  — for
example,  Buchsbaum and Silverman (1968)  report
that  Nl  amplitude  varies  along the  personality
dimension of augmenting-reducing - and the idiosyn-
cratic aspect of  annoyance  seems to relate to the
acquired significance of sounds, which may be reflect-
ed in the evoked potential. In short, ERPs might serve
as indices of noisiness  that  would be less open to
attack as  being  subjective  in the pejorative  sense.
Insofar  as  there is reason  to distinguish orienting,
defense, and startle reactions to noise, more sophisti-
cated multidimensional  measures  of subjective re-
sponse need to be developed.

     In  applying ERP techniques to noise research,  it
may be useful to recognize that there are two relative-
ly  distinct arousal systems, and that the phasic de-
fense reaction  may be sustained  by a humoral mech-
anism.  A number of ERPs have been  tentatively link-
ed with  the orienting  reaction (Klinke et al. 1968;
Tueting, this volume; Loveless 1976), but there has
been little attempt to work out a model of its mech-
 anisms   and the relation  of ERPs to its presumed
 function of increasing perceptual sensitivity (Loveless
 1975).  A  few attempts have been  made to relate
 ERPs with traditional  autonomic measures of ori-
 enting  (Roth et al., this volume), but unfortunately
 not those such as the forehead vasomotor response  01
 heart rate (Graham and Slaby 1973) that differentiate
 the orienting and defense reactions.
      Furthermore, there has been no  exploration of
  relationships  with EEC  desynchronisation. For  in-
  stance, the effect of hyperventilation on the EEC is
  potentiated by hypoglycemia, a fact  which may be
  relevant  when considering  possible  similarities  be-
  tween noise and hypoxia. Although this relationship
  between  the  EEC  and brain metabolism is well-
  known, very little is yet known about the relationship
  between either and ERPs.

           This review  has  emphasized  the  critical
  importance of theory in dealing with practical issues.
  Among the many difficulties in  applying laboratory
  research  to  practical situations  is  what  Chapanis
  (1967) calls  "variable naming."  It  is  tempting to
  apply the results from a continuous performance test
  to some real-life task, such as industrial inspection, on
  the  grounds  that both  involve "vigilance." Clearly,
  this sort of extrapolation is justified only to the

-------
436
                                         Loveless
extent that the two tasks do in fact involve the same
functions, which is by no means guaranteed by the
naming. This consideration is particularly important
in stress research, because researchers have tended to
limit themselves to a handful of stress-sensitive tasks.
There is  an  obvious  strategic advantage in doing so,
but it then becomes very difficult to apply the results
to a wide variety of real-life situations. We do not
have  an  adequate taxonomy of  tasks  (Wilkinson
1969), and it  is  difficult to see how one will be
achieved  without analyzing  tasks  into  component
functions. The  usefulness of ERP techniques in meet-
ing this  theoretical  challenge might be  their most
important contribution to stress research.

-------
LOW-FREQUENCY  NOISE, SELECTIVE  ATTENTION,
AND EVENT-RELATED POTENTIALS
D. OTTO AND V. BENIGNUS

U. S. Environmental Protection Agency, University of North Carolina at
Chapel Hill, NC, U.S.A.
    Low-frequency noise (LFN) generated by major
appliances such  as heating and  refrigeration equip-
ment is a pervasive component of domestic and indus-
trial environments, but little  is known about the ef-
fect of LFN on the central nervous system. Noises
below 100 Hz are seldom perceived as irritating, al-
though people sometimes complain that the continu-
ous "hum" of motors makes them drowsy. This study
explored the neurobehavioral effects of LFN in the
11.5- to 350-Hz range. Methods and behavioral results
have been detailed elsewhere (Benignus et al. 1975).
This paper describes  the analysis of vertex event-
related potentials (ERPs).


Methods
    Twenty-seven male subjects aged 18 to 31 were
 paid to perform a 2.5-hour numeric monitoring task
 in an acoustic chamber.  Single-digit  numerals were
 displayed  for O.OS sec at the rate of I/sec using a
 light-emitting diode.  Subjects viewed nine pseudo-
 random series of digits, with instructions to press a
 hand switch whenever three consecutive even or odd
 numbers appeared. One-digit, two-digit,  and three-
 digit sequences were equiprobable. Each series lasted
 12.5  min and contained approximately  111 target
 sequences.

    The output of a  random-noise generator was fil-
 tered to obtain an upper band of 91 to 3 SO Hz and a
 lower  band of  11.5  to 44  Hz. Noise was radiated
 free-field at a moderate sound-pressure level of 80 dB
 during series 2,4,6, and 8. No noise was radiated dur-
 ing series 1,3,5,7, and 9.

     EOG and vertex EEC referred to linked ears were
 recorded using amplifiers with 2.2-sec time constants.
 Signal averages were computed separately for hit (H),
 correct rejection  (CR), and  miss (M) trials using a
PDF-12  computer.  Trials  containing EOG artifact
were rejected from averages. ERP data from six sub-
jects were not analyzed because of excessive  eye
movements.
    ERP measurements are illustrated in Fig. 1. CNV
was measured as the mean amplitude of a 128-msec
epoch preceding S3 relative  to a 128-msec baseline
preceding S2. The  maximum negative and  positive
peaks following S3 were measured relative to the
same  baseline. The  mean latencies  of these peaks,
averaged across subjects and conditions, were 110
msec (Nl 10) and 456 msec  (P456), respectively, re-
lative to S3 onset. In order to minimize the  possi-
bility that N110 measures were confounded by the
slow negative shift preceding S3, CNV values were
subtracted from N110 values prior to statistical test-
ing.
                                 CONTROL
                                 imftBAND
                                 LOWER UNO
 Fig.  1.  Summary averages of hit trials during con-
 trol,  upper-, and  lower-noise band conditions for 17
 subjects. Time base: 1 sec from S2 onset to S3 onset.

-------
 438
                              Otto and Benignus
     Vertex ERPs were averaged selectively for H, CR,
 and M trials. Twenty-one subjects were used in  the
 CR analysis. Four low-performance subjects were ex-
 cluded from the H analysis due to insufficient hits in
 some  conditions. Only seven subjects missed enough
 trials for inclusion in the M analysis.

 Results
    Subjects missed more target strings during noise
than control conditions, although M rates did not dif-
fer significantly between  the  two noise conditions.
Table 1 shows the average number of misses for noise
and  control series  of high-, medium-  and  low-
performance subjects. False alarm rates were uniform-
ly low in all conditions.
 [F(l,16) = 8.53, p <  .01J, and P456 amplitude was
 not affected by LFN.
     None of the  ERP amplitude measures discrimi-
 nated between  noise  and  control conditions  in CR
 trials. No behavioral criterion was available, however,
 to indicate whether or not subjects were attending to
 stimuli during CR trials (i.e., subjects were not re-
 quired to respona following two-digit strings).  Selec-
 tive averages of CR and M trials in seven subjects (Fig.
 2) illustrate  the problem of interpreting CR averages
 without  a behavioral  criterion. Note the  attenuation
 of CNV  and Nl 10 and the absence of P456 in the M
 average.  The extent to which CR averages included
 and were  distorted by M  trials  could not be  deter-
 mined.
                          Table 1.  Average Number of Misses during Control
                                            and Noise Series


level
High (N=7)
Medium (N=7)
Low (N=7)
± 1.96xS.E.b
Average misses/series

Control
2.2
7.0
26.9
±2.69
Noise8
3.0
8.4
38.8
±4.68
                 aUpper and lower bands combined.

                 b95% confidence Interval • mean ± 1.96 x standard error of mean.
     Fig. 1 depicts the summary averages of H trials
during control, upper-, and lower-band noise condi-
tions for 17  subjects.  The subtle effects of noise
exposure on ERP amplitudes are shown in Table 2.
Control means were subtracted from upper and lower
band means for each subject in order  to reduce the
dimensionality of the data  for statistical testing. The
null hypothesis  of no noise effect was rejected by
means of a single-factor multivariate test [F(6,ll) =
6.40, p < .004]. Univariate tests indicated that Nl 10
amplitude decreased  during upper-  [F(l,16) = 10.93,
p<  .004] and lower-band noise [F(l,16) = 4.35,p<
.053], CNV decreased during lower-band noise only
    Latencies of the largest positive component fol-
lowing each stimulus in CR trials were also measured.
Latencies increased successively following SI (155.9
±  37.4  msec), S2 (184.1  ± 46.1), and S3 (455.6 ±
99.6). The 28-msec difference between SI and S2 was
significant (Wilcoxon:  T=10, N«21, p < ,01).  The
discontinuity between  S2  and S3, however, suggests
that PI84 and P456 do not represent the same neuro-
electric  process. Fig. 2  illustrates this point. Two dis-
tinct positive components  are evident following S3 in
the CR  average. The earlier peak corresponds to the
positive  component following SI and S2. As noted
above,  the  second positive  component  (P456) is
absent hi M trials.
                      Table 2.   Hit Trials: ERP Amplitude fcV) as a Function
                                          of Noise Exposure
Noise exposure
Control
Upper band
Lower band
CNV
-3.9
-3.8
-3.1a
N110
-5.7
-4.?a
-4.4a
P456
13.1
12.0
12.7
•Significantly different from control values as determined by pott-hoc unlvariate twtt dee text).

-------
Low-Frequency Noise, Selective Attention, and Event-Related Potentials
                                            439
                                             SjiV
                        S3
Fig.  2.  Summary averages of miss trials  superim-
posed on  correct rejection  (CR) trials of seven low-
performance  subfects. The  P456 component follow-
ing S3 in hit and CR trials was absent in miss trials.
to sleep. (3) If LFN induced drowsiness, a decrease in
beta and alpha frequencies accompanied by an in-
crease in theta and delta frequencies would be expect-
ed. No noise-related changes in EEC spectra were ob-
served (Benignus et al., in preparation).

    An alternative hypothesis is that selective atten-
tion, rather than general  arousal,  was impaired by
LFN. That  is, noise may have interfered with the abil-
ity of  subjects to "tune in" or concentrate on the
numeric display.  Schwent  and  Hillyard  (1975) pro-
vide evidence that the auditory Nl component, anala-
gous to N110, reflects selective attentional processes.
They observed an 82% intramodal enhancement of
Nl in   attend compared  to  ignore conditions  of a
complex pitch discrimination-spatial localization task.
Related findings have also  been reported in visual and
somatic modalities  (Schecter and Buchsbaum  1973,
Velasco et al. 1973).
Discussion

    Slight, but  significant, decrements  in  vigilance
performance  and  N110 amplitude  were  observed
during exposure to low-frequency continuous noise
of moderate intensity.   The experimental  situation
may be compared to the monotony and social isola-
tion of an individual driving a long distance.  Engine
hum, tires, and air-conditioning could generate noise
of comparable frequency and intensity.  The poten-
tial  deleterious  consequence of  impaired  vigilance
during driving is obvious.

     Do the convergent decrements in behavioral and
ERP measures reflect an impairment in  selective  or
general attention? Available  evidence does not sup-
port a general attentional hypothesis. (1) A  parallel
change in P456  amplitude  would be expected if the
effect were nonspecific, (2) There is no evidence that
components analagous to Nil0 vary as a function of
arousal level. For instance, Picton et al. (1974) did
not  observe any change in  auditory components ear-
lier than 250 msec during the transition from waking
    The  latency difference between the prominent
positive component following the first and second
digits is worthy of comment.  This difference was not
associated with any change in amplitude. Presumably
SI  provided little information, while S2 alerted sub-
jects  to  prepare for possible response. Thus,  the a-
mount of information conveyed by the stimulus ap-
pears to have been encoded in the latency rather than
the amplitude parameter. This evidence suggests that
latency and amplitude parameters may reflect orthog-
onal dimensions of information processing. Friedman
et al. (1975) have reported similar findings in a lin-
guistic task.  The present paradigm could provide a
useful vehicle to further elaborate  the functional sig-
nificance  of the expanding  family of late positive
ERP components.

     In conclusion, LFN  of  moderate intensity was
found to impair performance of a numeric-monitor-
ing task and  to reduce the amplitude of a negative
evoked potential component peaking at 110 msec.
 Results suggest that LFN interferes with selective at-
 tention rather than general arousal.

-------
  PARADOXICAL EFFECTS OF CARBON MONOXIDE  ON
  VIGILANCE PERFORMANCE AND EVENT-RELATED
  POTENTIALS
  D. OTTO, V. BENIGNUS, J. PRAH, AND B. CONVERSE

  U.S. Environmental Protection Agency and  University of North Carolina, Chapel Hill, NC, U.S.A.
     Behavioral impairments have  been observed in
  several vigilance experiments during low-level carbon
  monoxide exposure (Beard and  Grandstaff 1975,
  Fodor and Winneke 1972, Groll-Knapp et al. 1972,
  Horvath et al, 1971).  Two recent negative reports,
  however, demand a cautious reappraisal of the rela-
  tionship between CO exposure and vigilance perform-
  ance.  Groll-Knapp  et al. (this section) failed  to find
  any CO-related decrement in a complex visual vigi-
  lance  task, and Winneke et al. (this section) were un-
  able to replicate the results of Horvath et al. (1971).

     The present experiments were designed to study
  the effect of 0,  100, and 200 ppm CO on vigilance
  and event-related potentials (ERPs). A  continuous
  performance  task that is sensitive  to psychoactive
  drug effects (Mirsky and Rosvold 1960) and environ-
  mental stress  (Benignus et al. 1975) was chosen.
  McCallum (1976) has also demonstrated  the  useful-
  ness of this paradigm in assessing ERPs.
 al. 1971). Mean pre/post COHb levels were 0.06/0 12
 0.36/4.64, and 0.41/11.99% for 0,100, and 200 ppm
 groups, respectively.

     BOG and EEC referred to linked ears were re-
 corded at Cz, P3 and P4  by means of amplifiers with
 8-sec time constants. Analog data were recorded on
 magnetic  tape and computer-averaged off-line. Trials
 containing EOG artifact  or blocking were automati-
 cally rejected.

     Stimuli consisted of single-digit numerals display-
 ed on  a light-emitting diode for 0.05 sec  at 1.5-sec
 intervals. Subjects viewed six pseudorandom series of
 digits  with instructions to  press a button whenever
 three consecutive  even or odd  digits appeared. Each
 run of 667 stimuli contained 27  target strings. Rest
 breaks (3.3 min) alternated with monitoring periods
 (16.7 min) during the 2-hr test.
 Experiment I

Methods I
    Twenty-eight male volunteers (18 to 35 yr old)
were  paid  $5/hr to  participate.  Subjects  were
screened  by  MMPI,  Duke Medical Inventory, and
physical examinations to obtain a normal, healthy,
nonsmoker sample.

    Subjects  were  exposed  double-blind to   0
(N»8), 100 (N»10), or 200 (N-10) ppm CO for 2 ru-
in a »mall dimly lit booth (1.7 x 1.7 x 2.0m), venti-
lated at 100 fr /min. CO concentration was continu-
ously  monitored with a Beckman Infrared Analyzer.
Blood »ample« were taken before and after expoiure.
Carboxyhemoglobin (COHb) levels  were determined
by the Radford jpectrophotometric method (Small et
Results I

    Percent misses across runs are shown in  Fig. 1.
The 100-ppm group missed slightly fewer targets than
the control  or 200-ppm groups,  although  group
                                                       40 I—
    10
                                    XIN'10>
                                    , (N • 101
                         4

                        RUN
Fig. 1.  Percent misses across  time as a function of
CO exposure level during Experiment I.

-------
CO, Vigilance, and ERPs
                                                                 441
                                         J
              0.8
                     VO
                           1.8
                                  2.0
                                         2.8
                       TIME.m
 Fig. 2. ERP measurements included CNV amplitude
 relative  to  pre-S2  baseline and  three peak-to-peak
 measures: NIPl folio wing S2 and S3 and N2P2.

 differences were not significant. No  vigilance decre-
 ment was apparent across time.

     Vertex averages digitized at 128  Hz were con-
 structed for target strings, including both hit and miss
 trials. Four ERF measures were made as shown in Fig.
 2. Four subjects were rejected because of excessive
 eye movement. Two additional subjects who, failed to
 show  any consistent CNV were eliminated from the
 analysis. Run 1 was considered a training run. Meas-
 urements were averaged across runs 2 to 6.

     ERP results are summarized  in Table 1. The ef-
 fects of CO exposure  on CNV amplitude were mar-
 ginal,  with a small increase in the  100-ppm group and
 a small  decrease in the 200-ppm group  relative to
 control subjects. Summary averages of runs 2 to 6 for
 subjects In the three groups are shown in Fig. 3. CNV
 amplitude was  significantly larger for the 100-ppm
 than 200-ppm group, but neither exposure group dif-
 fered significantly from controls.

     NIPl amplitude increased with CO exposure fol-
 lowing S2 and  S3, although the effect  was significant
 following S3  only. Subsequent analysis, using the
 CNV  epoch as  baseline, indicated that this effect was
                                                                                                 10 nV
                                 O.B
                                        1.0    1.6

                                         TIME. MO
                                                     2.0
                                                            2M
                     Fig. 3. Summary ERP averages of runs 2 to 6 from
                     0-, 100-, and 200-ppm CO subjects.

                     limited to the PI component with a mean latency of
                     161 msec. The N2P2 measure did not vary with CO
                     exposure.
                         Fig. 4 illustrates temporal trends in CNV and PI
                     amplitude during the session. In contrast to the con-
                     trol group, CNV amplitude tended to  decrease as
                     COHb  levels  increased  for both exposure groups.
                     Larger samples are needed to evaluate the significance
                     of this trend.
                         Results of this experiment are difficult to inter-
                     pret because of small sample size, lack of clear behav-
                     ioral differences, and paradoxical ERP findings. The
                     large  intersubject  variability  in  miss  rate within
                     groups masked any  possible group differences. Data
                     can be normalized for individual baseline differences
                     if an appropriate  baseline measure is available.  This
                     experiment did not include a pre-exposure baseline
                     measure. A second experiment  was undertaken to
                     correct this problem.
               Table 1.  Neurobehavioral Results Averaged Across Runs 2-6 for Control
                                             and CO Groups
CO, ppm
0 (N-6)
100 (N-8)
200 (N-8)
95% Cl
F
Mines
6.3
4.2
8.2
1.36
0.94
CNV, uV
8.1
10.5
8.0
1.45
3.29'
N1P1,^V
13.9
17.4
19.3
1.44
4.18b
  •p",088 (100 x 200 ppm compirlion:
  bp  £.08  (Ox 100ppm compirlion:
           (0 x 200 ppm compirlion:
t»2.4S,dM4,p<,05).

t-2,0fl,df-12,p<,10).
t-2.98,df-11,p<,02).

-------
442
                                                             Otto et al.
                               CNV  i-n
                                                                            PI
           16
                                         CO, ppm
                    1.2
3,4
5,6
1.2
                                                                            3.4
6,6
                                                     RUNS
           Fig. 4.  Temporal trends in CNV and PI amplitudes across time as a function of CO exposure
           level Data were pooled for runs 1 +2, 3 + 4, and 5 + 6.
 Experiment II

 Methods II

     The behavioral task used was identical to Experi-
 ment I. To obtain a pre-exposure baseline, the testing
 period was lengthened to  10 runs. Run 1 was used to
 train subjects and run 2 served as a baseline. Subjects
 were given a  10-min  rest break  outside the  chamber
 after run 2, CO exposure commenced with run 3.

     Fifty-two nonsmoker male  subjects, screened as
 before, were assigned double-blind to 0 (N=17), 100
 (N=16), or 200 (N=19) ppm CO groups. EEC was not
 recorded. Blood samples were  drawn before and after
 testing. Mean pre/post  COHb  levels  were 0.02/0.14,
 0.20/4.95, and 0.40/12.04% for 0,100, and 200 ppm
 CO groups, respectively.

     Fig. 5  summarizes the  results of Experiment II.
 Differential miss scores were  calculated by subtrac-
 ting baseline scores (run 2) from exposure run scores.
 Differential scores were then averaged for each quar-
 ter of the exposure period. Fig. 5 indicates that the
 only striking difference between groups occurred dur-
 ing the initial  40 min of exposure.  Control  and
 200-ppm  groups showed a slight improvement in per-
 formance (relative to baseline) during the initial quar-
 ter of exposure, while the 100-ppm group showed an
 impairment in performance. A one-way ANOVA was
 computed for quarter I scores only [F(2,49) = 3.845,
 p < .028]. Mann-Whitney comparisons indicated that
 differential miss scores of 100-ppm  subjects were
                                                       higher than control (U = 77, p <  .05) or 200-ppm (U
                                                       = 80.5,p< .02) subjects.
                       Discussion

                      Electrophysiological data  (Experiment I)

                          A paradoxical increase in amplitude of a positive
                      component  peaking at about 160 msec was observed
                      during CO exposure sufficient to raise COHb levels to
                       12%.  Dyer and Annau (this section) also reported in-
                      creased amplitude of visual evoked potentials (EPs) in
                      the superior colliculus of rats,  but at much higher
                      COHb levels (40%) under radically different testing
                      conditions.  On the other hand, Groll-Knapp et  al.
                      (this section) reported decreased amplitude of soma-
                      tosensory and auditory  EPs, but no change in visual
                      EP amplitude, at COHb levels up to 10%. Moreover,
                      Winneke et  al. (this section) failed to find any change
                      in auditory  EPs at COHb levels comparable to the
                      present study.
                           The lack  of behavioral effects or an appropriate
                       pre-exposure baseline confounds the interpretation of
                       ERP results in the present study. The lack of consis-
                       tency in testing  and analysis procedures employed in
                       other CO studies reported in this section further lim-
                       its the  generalization of results.   The  collective re-
                       sults do suggest, however, that certain ERP param-
                       eters are sensitive to CO effects under certain con-
                       ditions.  Further research is needed to determine the
                       validity and functional significance of these findings.

-------
CO, Vigilance, and ERPs
                                            443
                                o co IN-1?)
                        	100 COIN-19)
                        	200 COIN-19)
                                   I
      BASE      >         »       'I'       IV
      LINE   QUARTERS OF EXPOSURE PERIOD

Fig. 5.  Differential misses for Q-, 100-, and 200-ppm
CO subjects during Experiment II. The number of
misses made by each subject during pre~exposure run
2 was subtracted from misses made during each suc-
cessive exposure run to obtain differentia] scores. The
transient increase in  misses during  the initial 40 min
of exposure (Quarter 1) of the 100-ppm  group was
the only significant finding.

 Behavioral data

     CO exposure had minimal effect on performance
 in  tlus study. No significant behavioral impairment
 was observed  in Experiment I. The transient increase
 in  miss rate  observed during the  initial 40 min of
 100-ppm CO  exposure in Experiment II may be dis-
 missed conservatively as a sampling error (Benignus et
 al. 1977). Similar findings in other studies, however,
 suggest another possibility. Beard  and Grandstaff
(1975) state that behavioral decrements tend to be
maximal  during the initial 30  min of CO uptake.
Fodor and Winneke (1972) reported a temporary vig-
ilance decrement during the initial observation period
at COHb levels of 2.3 to 3.1%. These authors suggest
that a  physiological compensatory mechanism could
account for their results.

     What evidence is available  concerning the com-
pensatory mechanism of  response  to CO hypoxia?
Traystman (this section) observed increased cerebral
blood  flow and cerebral vasodilation in the rat brain
during CO  exposure. An analagous  compensatory
mechanism could  enable  humans  to sustain normal
cognitive  function at COHb  levels up to 20% and
could  explain the numerous  negative  reports of be-
havioral effects at 5 to 20% COHb levels (cf. review
by Stewart 1975).

     The more intriguing  question concerns  positive
reports of behavioral impairments at COHb levels be-
low 5% (Beard  and  Wertheim  1967,  Fodor and
Winneke 1972, Groll-Knapp et al.  1972, Horvath et
al. 1971).  The conflicting evidence suggests that the
behavioral effects of low-level  CO exposure are, at
best, extremely fragile. The key to this enigma may
lie in determining the threshold conditions that medi-
ate  the activation of the  physiological compensatory
response to CO hypoxia.

-------
  THE USE  OF EVOKED POTENTIAL AND  BEHAVIORAL
  MEASURES IN THE ASSESSMENT  OF ENVIRONMENTAL
  INSULT

  M. RUDNEV, A. BOKINA, N. EKSLER, AND M. NAVAKATIKYAN

 A.N. Sysin Institute of General and Community Hygiene, Academy of Medical Sciences,
 Moscow, USSR
     Neurophysiologlsts have accumulated extensive
 data on evoked potentials (EPs) and behavioral re-
 sponses in human suffering from a variety of organic
 disorders. Neurobehavioral  assessment   techniques,
 however,  have not been widely used In hygienic re-
 search to evaluate the possible harmful effects of
 chemical and physical factors of the environment. E-
 voked potential methods have been used In the relat-
 ed field  of toxicological research (Xintaras  et  al.
 1966, Khachaturyan and Mitarevskaya 1969, Bokina
 and Eksler 1973, Lehotzky and Meszaros 1974).

 Theoretical Basis for the Interpretation of
 Evoked Potentials
     In accordance with current theory,  the primary
 EP complex is considered to be a signal-releasing pro-
 cess because  EP parameters correlate well with the
 intensity and  frequency of external signals (Gershuni
 1963, Peymer and Modin 1963, Shevelev 1971). Sec-
 ondary or late  EP components are associated with
 later  processes of complex  analysis and  integration.
 The amplitude of late EP components depends on the
 information content of the stimulus and its meaning
 to the organism (Peymer 1971, Batuyev 1971, Shumi-
 lina 1971).

    There is  disagreement whether or not primary
 components of  the EP  reflect shifts in  the general
 functional state of the CNS. While Khananashvili et
 al. (1971) answer affirmatively,  Kratin et al. (1971)
 disagree because they  observed a decrease in ampli-
 tude of early components against  a background of
 desynchronized EEC (excitation) and during the  de-
 velopment of extinction (inhibition) of the condition-
 ed reflex.

    Distinct changes in primary EP components  ap-
 pear only under extreme conditions. High 02 concen-
 trations (96%) cause clear secondary alteration of the
initial positive  EP component in visual cortex on the
second day, followed  by the disappearance of this
 component  in  3  to  4  days  (Agadzhanyan and
 Kalyuzhnyi 1969).

     In a conditioned reflex  study, Shugalev (1970)
 distinguished three stages of amplitude change in the
 primary negative EP component: decrease, increase,
 and decrease. The first stage in the development of
 the  conditioned reflex is  characterized by  reduced
 amplitude of  late  responses  (Kondrat'yeva  et al.
 1970). During reinforcement of the conditioned re-
 flex, primary EP components diminish and secondary
 components increase (Shilyagina 1971); late EP com-
 ponents disappear during extinction of the condition-
 ed reflex (Shilyagina 1971).

     Some authors note distinct changes in the config-
 uration of slow potentials (SP) when amplitude, dur-
 ation, and latency correlate with the animal's condi-
 tioned behavioral response. This  finding emphasizes
 the significance of the late component of the visual
 evoked potential (VEP) in evaluating both the mean-
 ing of the conditioned stimulus  and the  functional
 state of various structures of the brain during a specif-
 ic activity (Shumilina 1971).

    The effective  application of evoked potential
techniques in the analysis  of normal brain function
requires the theoretical formulation of the origin and
functional  significance of individual components of
evoked potentials.  According  to  generally accepted
views in the Soviet Union (e.g., Gershuni 1963, fieri-
tov 1948,  Roytbak 1955, Kogan 1956), the initial
positive component reflects the postsynaptic process,
i.e., excitation of pyramidal cells in cortical layers III
and IV. The initial negative component is associated
with excitation of superficial cortical layers.  Most
authors consider that this component  reflects the
postsynaptic potential  of apical dendrites  (e.g.,
Roytbak 1964, Anokhin 1964).

    Studies of unit  activity from cells in the visual
cortex in response to a specific stimulus indicate that

-------
EP and Behavioral Measures of Environmental Insult
                                             445
the initial discharge corresponds to the primary posi-
tive component of the VEP. The inhibitory phase,
which follows discharge,  coincides in part with the
primary negative component of the VEP (Krondrat'-
yeva 1967, Polyanskiy 1965). Intercellular recordings
are consistent with this interpretation; i.e., the  sur-
face negative  component of the  primary VEP is asso-
ciated with positive hyperpolarization of pyramidal
cells in the visual cortex. This component, therefore,
reflects  true  inhibitory  postsynaptic   potentials.
(Skrebitskiy and Voronin  1966, Fuster et al. 1965).

    Amplitude of primary EP components, however,
is not an absolute index of the level  of cortical exci-
tatory-inhibitory processes because these components
do not  vary  consistently  with successive  changes in
state. The  secondary  late negative  wave  provides a
supplementary index  of the functional state of the
cerebral  cortex.  This late wave indicates both the
state and location of cortical inhibition.

Evoked Potential Methods in Environmental
Research
    On the basis of the theoretical  framework sum-
marized  above,   investigators   in   Moscow   have
studied the effects of chronic exposure to neurotrop-
ic  substances including  formaldehyde,  ozone,  and
carbon bisulfide on the visual evoked potential in rab-
bits. Primary and secondary EP components were re-
corded monopolarly from electrodes implanted in the
visual cortex. Data were analyzed by means of Daw-
son's method of superposition and signal averaging
with an ATAS-201 computer.

    Baseline  recordings were  obtained 5 to 7  days
after electrode implantation and again 1  week later.
Animals were then  placed in  special chambers for
treatment  and were exposed continuously to a neuro-
tropic substance for 45 days. Concentrations of form-
aldehyde (0.1 mg/m3), ozone (0.05  mg/m3) and car-
bon bisulfide (2.0 mg/m3) were tested. A control
group of  animals  were exposed to clean air under
 otherwise identical conditions. During the test period,
VEPs elicited by arrhythmic light stimuli were record-
ed at 2-week intervals in special chambers, isolated
 from sound.

     Measurements  included  the peak  latency and
 amplitude of primary  positive  and negative  VEP
 components, as well as the amplitude and duration of
 the secondary late negative wave (Fig. 1).  The reliabil-
 ity of  EP differences between  baseline and exposure
 measures was assessed statistically. The latency of pri-
 mary  components remained very  stable during ex-
 posures to the test substances, but the amplitude of
 primary components decreased differentially across
 time for the three substances. Dissimilar changes were
 observed  in  the secondary negative wave, although a
 decrease in  amplitude of this component was com-
Fig.  1. (1)   Typical  VEP recorded from the visual
cortex of a  conscious rabbit in response to a single
flash  of  light, (2)  Superposition of five VEPs,  (3)
Schematic representation of VEP showing the meth-
od  of computing amplitude-time  characteristics of
 VEP components: a, latency of positive phase of ini-
tial response (IR); b, latency of negative IR phase; c,
IR  amplitude; A, slow negative wave (SNW) maxi-
mum; 'E, duration ofSNW half-wave; B, duration of
1R-SNW  complex;    C,  SNW amplitude, Negative
upwards.


mon  to all substances (Fig. 2). In brief, we found that
prolonged low-level exposure to  neurotropic toxi-
cants significantly altered the temporal pattern of pri-
mary  and secondary  VEP  component amplitudes.
 Neurophysiological  analysis indicated, furthermore,
 that  the  changes occurred  in the cortical inhibitory
 system.

 Behavioral Methods in Environmental
 Research
     Behavioral methods are used in many countries
 to  assess the possible harmful effects of chemical and
 physical  factors in  the environment. Computer  tech-
 nology is employed in some laboratories to  conduct
 simultaneous tests on a large  number of animals
 (Weiss and Laties 1975, Guy and Chou  1975). Other
 simpler,  but informative, methods that do not require
 complex  computers  are also used  to  measure the
 general response of  an organism to environmental
 insult (Spyker 1975). Investigators at Kiev have also

-------
446
                                                                                            Rudnev et al.
1T~2
                                 0246
               IR                   SNW
                       TIME, weeks
 Fig. 2,  Variation in IR and SNW amplitudes during
 6-week inhalation of formaldehyde, ozone, and car-
 bon disulfide in concentrations determined in the air
 of industrial cities. At the end of the exposure period,
 the  amplitude of both parameters was definitely re-
 duced. Control observations  conducted in  clean air
 for  the same periods reveal the homeostatic nature of
 the  oscillations of these IR parameters.
measured a number of behavioral parameters, includ-
ing motor  activity,  aggressiveness,  efficiency under
dynamic and static  load, unconditioned alimentary
responses, and sensitivity to electric shock.

    Although the inhabitants  of technologically ad-
vanced societies are exposed extensively to electro-
magnetic radiation  emitted  by  communication net-
works, very little is known about  the physiological
effects of this pervasive phenomenon. We, therefore,
felt it  advisable to  study  the  behavioral  effects of a
superhigh  frequency  (SHF)  field of  nonthermal
intensity.

    Fifty male white rats were used; half were  ex-
posed to radiation and half served as controls. The
source of  radiation was a SHF-radiation generator
with a wave length of 12.6 cm. Animals were radiated
for 1 month at 500 ^iw/cm^ for 7 hours a day.

    Activity was measured by counting the squares in
an open field that were intersected by the rat over a
3-minute test (3  times/min).  Activity  was measured
twice at each log point (on the 9th and 10th days, on
the 19th and 20th days, and so on). On the first  day
the activity was considered to be exploratory; on the
second, motor. Aggressiveness was scored as the out-
come of combat  between a test and control rat  fol-
lowing electric shock. Endurance under dynamic load
was defined as the length of time  the rat maintained
its balance on a cylindrical rotating treadmill; endur-
ance under static load was measured as the time the
rat maintained its balance on  an inclined bar at a
fixed height from the ground.

    The magnitude of the unconditioned alimentary
response was measured  as the amount of food con-
sumed during a 20 min. period after 23 hrs. of food
deprivation.  Sensitivity to electrodermal shock  was
measured as  the  voltage of  a  100 Hz  square wave
electrical stimulus needed  to elicit withdrawal of the
paws from the metal rods in the cage floor.  Measure-
ments were  made several times  prior to radiation,
on days  10, 20 and 30 of exposure, and  on days 15,
30,45,60, 75 and 90  after exposure.

     Results of the study indicated a reduction in ali-
mentary response and static  and  dynamic efficiency
by the  10th day of radiation exposure (Fig. 3). By
the  20th day of exposure, dynamic efficiency and
exploratory activity were  decreased and sensitivity  to
electrodermal shock was increased (Fig. 4 and 5). By
15
c
°E
g 10
cc
LU
£ ,
LU a
n

T


_L-


1





—

LU"
a
z
C9
s
0
g
UJ
CC
                                                                        £  1
                                                                                 T
                                                                                                 1
                                                     CONTROL   TEST         CONTROL  TEST

                                                                     SUBJECT

                                               Fig.  3.  Effects  of SHF  radiation  on alimentary  re-
                                               sponse in animals on 10th day of exposure.

-------
EP and Behavioral Measures of Environmental Insult
                                                   447
    +60
       0  10  20 30  0      30

                       TIME, days
90
 Fig.  4.  Effect of SHF radiation on exploratory activ-
 ity of white rats. Relative deviations of mean values
 of the test group from the  control group and devia-
 tion errors are presented.
0  10  20  30  0      30
                TIME, days
                                                                                             60
        Fig.  5.  Effect of SHF radiation on  sensitivity to
        electrodermal shock.
 the 30th day, exploratory and motor activity, static
 efficiency,  and sensitivity to electric  shock were re-
 duced. In the test group, some time after cessation of
 exposure, signs of inhibition of the nervous system
 were observed: on the 15th and 30th days, the latent
 period of alimentary response was increased, and on
 the 15th day dynamic efficiency was reduced. Some
 variations,  however,  indicated excitement of the cen-
 tral  nervous system.  Exploratory activity was in-
 creased over the entire  postexposure period (30th,
 60th, 90th day),  motor activity was increased on the
 90th  day,  and sensitivity  to electric  shock was re-
 duced  on the 30th and 60th days. All the differences
 cited were statistically significant.
           Thus, SHF radiation at 500 /uw/cm^ appears to
       produce inhibition of the central nervous system dur-
       ing radiation exposure and excitation after exposure.
       These general  behavioral disturbances are attributed
       to the effect of SHF radiation on the central nervous
       system.

           In conclusion,  we  found that evoked potential
       and behavioral assessment methods provide useful in-
       tegral measures of the effect of chemical and physical
       environmental insults on  the organism.  These tech-
       niques can be employed to demonstrate  subtle func-
       tional changes that are  imperceptible or inaccessible
       by other methods of measurement.

-------
 DIAGNOSTIC  UTILITY OF NEUROELECTRIC MEASURES
 IN ENVIRONMENTAL AND OCCUPATIONAL  MEDICINE
 A. M. SEPPA'LA'INEN
 Institute of Occupational Health, Helsinki, Finland
     Formerly,  occupational health  was concerned
 with the prevention of manifest disease, while mod-
 ern thinking cannot accept a work environment that
 causes even subclinical disorders. This development
 has prompted a need for sensitive and valid measure-
 ments of subclinical effects. Liver toxicity or dys-
 function of the hematopoetic system can be demon-
 strated  by  traditional, well-documented laboratory
 tests, but in many cases first signs do not appear in
 the  liver or blood cells. This makes  those measures
 unsuitable  in setting threshold limit values (TLVs),
 which aim to protect workers in occupational set-
 tings, or any individual when ambient air concentra-
 tions are concerned. In addition, many commonly
 used chemicals are neurotoxic at  low levels in long-
 term exposure, and this neurotoxicity has  not been
 taken into consideration adequately in the setting  of
 TLVs.


     In Finland, there is a definite trend toward devel-
 opment and application of more sensitive diagnostic
 techniques  (SeppalHinen 1975). Neurobehavioral in-
 dices are considered valuable in the diagnosis of occu-
 pational diseases  in cases that receive compensation
 from insurance.  For example,  a  person who com-
 plains  of subjective symptoms that  decrease  work
 capacity and who exhibits neurophysiological signs of
 neuropathy and/or encephalopathy and psychological
 impairment is considered to have an occupational dis-
 ease if it can be shown that he or she has been expos-
 ed to neurotoxic chemicals.
    Neurobehavioral tests have been employed sever-
al times to aid in the setting of standards and In the
evaluation of working places (Seppalainen and Tolo-
nen 1974, Seppalainen and HSrkonen 1976, Seppalai-
nen et al. 1975). Relatively large groups of workers
with documented exposure data, as well as controls,
were studied to establish exposure-response relation-
ships as well  as noneffect levels. In this manner, neu-
 ropathy and coronary death were shown to be risks in
 long-term occupational exposure to 20 to 30 ppm of
 carbon disulfide (Tolonen et al. 1975), and  the Fin-
 nish TLV was consequently lowered to 10 ppm. Vari-
 ous  neurophysiological  methods have also been ap-
 plied in the study of other solvents.
     Recently, several types of harmful effects on the
 nervous system have been reported after exposure to
 relatively  low levels of lead. Elevated  blood  lead
 (PbB) levels (PbB 40 to 60 jug/100 ml) were found in
 young  children who showed hyperactivity, impair-
 ment of  fine motor functions,  and psychological
 changes (David et al. 1972, Albert et al. 1974, Pue-
 schel 1974, Landrigan et al. 1975).  An association
 between mental retardation and lead in drinking wa-
 ter has also been observed (Beattie et  al. 1975). Lead
 poisoning can cause clinical or subclinical neuropathy
 in both adults and  children, with a  slowing of the
 maximal motor conduction velocity (MCV) or other
 impairment of nerve conduction (Catton et al. 1970,
 Behse et al. 1972,  Seppalainen and Hernberg 1972,
 Vasilescu 1973,Feldmanetal. 1973).
    Using electrophysiological methods, investigators
at the Institute of Occupational  Health, Helsinki
(Seppalainen et al. 1975), detected subclinical neuro-
pathy in 26 workers exposed to lead for 1 to 17 years
in a storage battery factory. According to factory re-
ports  and a careful  check of exposure history, the
workers' PbB never exceeded 70 jug/100 ml. PbB de-
terminations were evaluated in an international com-
parison program and were  found to be  valid. The
main findings were slowing of the MCV of the median
and ulnar nerves,  particularly of the slower motor
fibers (CVSF) of the ulnar nerve. Electromyographic
abnormalities included fibrillations, reduction in the
number  of  motor unit potentials  during maximal
muscle contraction, and abnormally long duration of

-------
Diagnostic Utility of ERPs
                                            449
motor unit potentials. Motoi unit potentials are elec-
trical  phenomena recorded with needle electrodes
during voluntary muscle contraction. A motor unit
potential Is usually a bl- or triphaslc wave generated
by muscle fibers that are innervated by the same ante-
rior horn cell. The  characteristics of a motor unit
potential undergo different changes In myogenic and
neurogenlc diseases.  All of these findings in lead-ex-
posed workers were  compatible with peripheral neu-
rogenic lesion. More pronounced abnormalities have
been observed in heavily exposed workers (SeppHlMi-
nen and Kernberg 1972).
     The previous studies suggested  a dose-response
 relationship between  the level of lead exposure and
 degree of neuropathy. They  also showed that slight
 damage within the nervous  system  is  produced by
 lead  during exposures hitherto  regarded as safe. A
 more extensive study has been  Initiated to confirm
 earlier results and to determine the threshold level for
 subclinical neuropathy from lead exposure.
 Methods
     Sixty-four workers  from a storage battery fac-
 tory were studied (18 female). The exposure level of
 workers has been reliably monitored throughout their
 employment, and  none ever showed overt clinical
 symptoms  of lead poisoning. Subjects were divided
 into four exposure categories:
tion velocities (SCV) of the median, ulnar, and sura}
nerves. Nerve conduction velocities were measured
using skin electrodes for stimulation and recording.
MCVs were determined by standard methods (Smorto
and  Basmajian  1972,  SeppalHinen  and  Hernberg
1972) with a Disa electromyograph and stimulator. A
Nokia Pulse Analyser (LP4840) was used for averag-
ing SCV responses (Sepp&lalnen et al. 1975).
    For somatosensory evoked potentials (SEPs), the
left median nerve was stimulated at the wrist with
0.2-msec square-wave pulses sufficient to elicit a clear
movement of the thenar muscles. EEC activity of the
right  Rolandic(C4) and parietal (P4) areas, with the
Ipsllateral  ear as  reference, was  amplified with  an
Elema Mingograph.  The time constant was 0.3 sec
and the upper bandpass limit was set to 70 Hz. Re-
sponses to SO signals were averaged and further ana-
lyzed with a Wang computer.
                                                       Results
     Conduction velocity data are summarized in Fig.
 1. In some cases, slightly abnormal nerve conduction
 velocities were detected among  the lead-exposed
 workers. Mean SCV values of the arm nerves tended
 to be lower among the lead-exposed groups than con-
 trols,  although the  difference was not  statistically
 significant.  MCV of the median nerve was lower in
 1. Group A:  38 workers who had just  begun to
              work with lead and who had no prev-
              ious history  of lead  exposure. Mean
              age was 25.9 (SD 8.1) years. Group A
              served as controls.
 2. Group Bl: 9 workers with occupational lead ex-
              posure  from 2 to 11 years with PbB
              values  that  had never  exceeded 40
              Mg/100 ml. Mean age was 33.6 (SD
              9.7) years.
 3. Group B2: 9 workers with lead exposure from 2
              to 13 years, the highest PbB  having
              been between  40 and 50 jug/100 ml.
              Mean age was 33.3 (SD 8.9) years.
 4. Group B3: 8 workers with lead exposure from 2
              to 7 years with the  highest PbB be-
              tween 50 and 70 jug/100 ml. (PbB ex-
              ceeded  60 jug/100  ml  in  only one
              case.) Mean age was 36.5 (SD 7.6)
              years.

     Several  neurophysiological parameters  were
 measured including: MCV of the median, ulnar, deep
 peroneal, and  posterior tibial nerves; CVSF  of the
 ulnar and deep  peroneal nerves; and sensory conduc-
  m/Mc
  66
  60
  58
  50 L.
    nt_
MEDIAN N SCV

ULNARNSCV



MEDIAN N MCV
                                PERONEAL N MCV
     CONTROLS 40     50     70
               BLOOD LEAD, MB/100 ml

 Fig. 1.  Mean nerve conduction velocities at different
 levels of exposure among  26 lead-exposed workers
 and 38  controls. SCV = sensory conduction velocity;
 MCV - maximal motor conduction velocity. Blood
 lead values indicated refer  to the highest blood lead
 recorded.

-------
 450
                                  Seppaftinen
Fig. 2.  SEPs of a lead-exposed subject (blood lead
65 w/100 ml) and a control (blood lead  12 fig/100
ml) recorded at Rolandlc area (C4). The measurement
of peak-to-peak amplitude of N18-P22 and N40-P7Q
Is indicated.


Group B3 than in Group A (t = 2.42, p  <  0.5). No
slowing of nerve conduction  velocities was  observed
in leg nerves.
    Typical  SEPs observed in a control and lead-ex-
posed subject are shown in Fig. 2. Certain group dif-
ferences in SEPs were noted.  Peak-to-peak amplitude
measured between N18 (latency about 18 msec) and
P22 (latency about 22 msec) increased directly with
lead  exposure  level  (Fig.  3). On the  other hand, a
later  component (N40-P70,   which commenced
around 40 to 55 msec and reached peak around 70 to
80 msec)  tended  to decrease in amplitude with in-
                                                          - SEP
                                                            AMPLITUDE OF THE FIRST
                                                            NEGATIVE PEAK
                                                        2 -
                                                        1  -
                                                                                        p<.01
      CONTROLS  40      SO      70
               BLOOD LEAD,/K|/100 ml

Fig.  3. Peak-to-peak amplitude of N18-P22 of the
somatosensory evoked potential (SEP) on the Rolan-
dic area (C4) among subjects with different maximal
blood lead  levels.  The difference between the expo-
sure  groups and the controls tested with Student's t.


creasing lead exposure. Table 1 presents group means
for amplitudes of both components at both recording
sites. These  amplitudes showed great intersubject vari-
ability. Latencies of N18 remained within a narrow
range, those of the  N40 component varied more, but
neither of the latencies correlated with exposure.
Discussion
    The occupational lead  exposure of workers was
relatively low and the number of subjects was small In
the present study. Although some individuals showed
Table 1.  Peak-to-peak Amplitudes (juV) of N18-P22 and N40-P70 on the Rolandic (C4) and
Parietal (P4) Areas among Controls and among Workers Exposed to Different Levels of Lead


Group
A (controls)
B1
B2
B3
N18-P22

Na
31
8
7
6
Rolandic
Mean
1.76
1.41
2.57b
3.12C

SD
0.92
1.62
1.12
1.37
Parietal
Mean
2.86
3.06
4.52
3.55
SD
1.66
3.13
2.95
0.84
N40-P70
Rolandic
Mean
7.90
5.04
5.34
6.54
SD
5.85
2.74
2.80
3.34
Parietal
Mean
5.68
6.15
3.77
4.04
SD
4.24
3.08
2.39
1.41
   •N • number of lubjecti.
   bB2 >A , t  "  2.03, p  »  .05.
   CB3>A , t  • 3.07, p <  .01.

-------
Diagnostic Utility of ERPs
                                             451
signs of subclinical neuropathy, nerve conduction ve-
locity measures were not sensitive enough to differen-
tiate between exposed and unexposed  workers until
blood lead levels exceeded SO Mg/100 ml. Only work-
ers in Group 63, with maximal PbB between SO and
70 Aig/100 ml, exhibited significant slowing of MCV
of the median nerve.  This measure was previously
found  to be  sensitive to lead effects (Seppalainen
et al. 1975).
pathological release of inhibitory control in the brain;
masses of neurons activate simultaneously, resulting
in the diffuse spread  of excessive neuronal discharge
(cf. Gastaut and  Tassinari  1975). Low-level lead ex-
posure is not  likely to induce a "grand mal seizure"
but may well exert a cumulative disinhibitory effect
on CNS function reflected in increased amplitude of
early  SEP components.
    SEPs were more sensitive than nerve conduction
velocities in differentiating  between subjects  with
lead exposure  and controls.  Workers in  groups B2
as well as B3 showed an increase in amplitude of the
N18-P22 measure. This increase cannot be explained
as an early sign of neuropathy, since latencies of early
peaks did not increase with increasing lead exposure.
Moreover,  impairment  of  peripheral nerve function
would hardly increase amplitude of early evoked phe-
nomena; in neuropathy one would expect an opposite
change,  i.e.  decrease  in  amplitude. Thus changes
noted  in  SEPs  after  occupational lead exposure
should be  interpreted as direct central nervous ef-
fects.
     Can the mechanism underlying the observed ef-
 fect of lead exposure on SEPs be inferred from exist-
 ing data? Several strands of evidence suggest that the
 increase in N18-P22 amplitude could reflect the disin-
 hibition of central inhibitory mechanisms. Stewart et
 al. (1972) reported increased peak-to-peak amplitude
 in early components  of the vertex evoked potential
 during  human exposure to methylene chloride. This
 chemical solvent is metabolised to CO, which is distri-
 buted throughout the organism in the form  of car-
 boxyhemoglobin (COHb). The  hypoxic effects of CO
 on cerebral circulation (Traystman) and neurobehav-
 ioral function (Dyer  and Annau, Groll-Knapp et al.,
 Otto et al., Winneke et al.) are  described elsewhere in
 this volume. Germane to the present argument  are the
 observations of increased amplitude of the  vertex e-
 voked potential in humans (Otto et al.) and the super-
 ior  colliculus evoked  potential  in rats  (Dyer and
 Annau) following CO exposure. Dyer and Annau at-
 tributed this  finding  to a "release from inhibition"
 produced by increased sensitivity of inhibitory synap-
 ses  to hypoxia. The effect of lead on evoked poten-
 tials may be similar to  the effect of CO and methyl-
 ene chloride.
     More direct evidence can be derived from classi-
 cal  observations of lead poisoning and epilepsy. En-
 cephalopathy  typically  develops in  severe cases of
 lead poisoning, and epileptic seizures are common in
 adult  encephalopathy (Fazullah  and  Ramamurthi
 1965; Whitfleld et al. 1972; Segal et al. 1974). Epilep-
 sy,  moreover, provides  a dramatic example  of the
    Other  evidence suggests that the hypothesized
disinhibitory  effect  could  be  mediated  in  the
thalamus. Skinner (this  volume) proposes a neuro-
physiological model in which thalamic nuclei func-
tion as an  "inhibitory gate" in  sensory pathways to
the cortex.  N20 (probably  the  same as N18 in this
report) has been assumed to be a subcortical response
(Allison et al. 1974). Augmentation of the N18-P22
component could be caused by "release from inhibi-
tion" in the thalamic reticular nuclei, although there
is no known neuropathologjcal evidence  that would
implicate this thalamic area as  the  site  of cerebral
damage in  lead encephalopathy. Neuropathological
changes are  usually diffuse  when changes can  be
found. In fatal subacute cases in  humans, conspicious
increase in brain volume associated with activation of
intracerebral capillaries has been found  (Pentschew
1965).  Dilatation of the capillaries with swelling of
the endothelial cells occurs at the earliest stage. Pent-
schew suggested a deficit in energy metabolism result-
ing in chronic metabolic hypoxidosis. Thus hypoxia
could be a feature common to CO and lead exposure.


    Changes in SEPs suggest an effect of long-term
lead exposure on brain function. Longitudinal follow-
up studies  on occupationally exposed subjects are
needed to  confirm these  results. One should start
with  pre-exposure studies so that each could act as
his own control in long-term studies.
     Several preliminary ERP  studies of  acute  ex-
 posure to neurotoxic chemicals have been described
 in  this section. ERP changes during acute exposure
 may provide a useful predictor of long-term hazards
 to  the nervous system, although long-term compara-
 tive studies are required to evaluate this possibility.
 Acute  experiments, moreover, may not reveal  the
 insidious cumulative effects of long-term low-level ex-
 posure. This is especially true concerning toxic neuro-
 pathy, which develops slowly at low exposure levels.
 Therefore, short-term and long-term studies of toxi-
 cant exposure should be undertaken  in parallel.


     Early detection  of occupational hazards to the
 nervous system is vitally important to prevent irreversi-
 ble damage. ERPs offer the promise of a sensitive,

-------
452                                                                                      SeppaKinen

noninvasive  predictor of chemical  insult to CNS      changes following acute exposure and clinical signs of
function. The functional significance of ERP changes      encephalopathy or neuropathy following chronic ex-
produced by chemical exposure Is difficult to evalu-      posure, ERPs will become a valuable diagnostic tool
ate, however, without more knowledge of the specific      in occupational medicine. Similarly, ERP techniques
neuroanatomic  substrate  of event-related potentials.      would also contribute Important evidence in the set-
If a clear relationship can be established between ERP      ting of exposure standards.

-------
EFFECT  OF  CARBON  MONOXIDE HYPOXIA
AND HYPOXIC  HYPOXIA  ON  CEREBRAL CIRCULATION1


R. J. TRAYSTMAN

Department of Environmental Health Sciences, The Johns Hopkins University, Baltimore,
MD, U.S.A.
    Several studies have shown behavioral and elec-
trophysiological abnormalities with various levels of
CO exposure (Xintaras et  al. 1966a, Beard and Wer-
theim 1967, Fodor and Winneke 1972, Horvath et al.
1971), and it is conceivable that these effects could
result from abnormalities in cerebral blood flow. The
responses of various vascular beds to arterial hypoxia
induced by decreasing arterial blood oxygen tension
(Pa02) (hypoxic hypoxia) has been well documented;
however, circulatory responses to hypoxia induced by
the administration of carbon monoxide (CO hypoxia)
is less well known. The literature concerning the ef-
fects of hypoxic hypoxia  on cerebral blood flow, in
general, indicates varying degrees of vasodilation.
Studies  of CO effects on cerebral blood flow also
tend to show cerebral vasodilation (Sjostrand 1948,
Haggendal and Norback 1966, Paulson et al. 1973).
However, many difficulties have been encountered in
these experiments, such as extracranial  contamina-
tion, surgical trauma to the cerebral vasculature, inad-
equate control of blood gas tensions, and failure to
measure carboxyhemoglobin (COHb) concentrations.
    A comparison of the  equivalent effects of both
types of hypoxia on cerebral blood flow has not been
done. The concept of equivalent effects of both types
of hypoxia has been  described by Permutt and Farhi
(1969) and involves the comparison of  physiologic
effects of COHb and low oxygen at equal reductions
in hemoglobin,  arterial  oxygen content, arterial or
venous oxygen tension,  or blood flow. The purpose
of the work described in this paper was to study and
compare  the effects of these two types of hypoxia on
the cerebral vasculature and to define any possible
differences between  them. In addition, the cerebral
hemodynamic response to the administration of low
levels of CO was studied.
 'Supported by ES-00454 and HL-10342.
Methods

    The autoperfused cerebral vasculature of 23 sodi-
um pentobarbital (30 mg/kg) anesthetized  dogs (18
to 26  kg) were studied. The comparative cerebral
hemodynamic responses to CO and hypoxic hypoxia
were observed in 13 animals, while the effects of low
levels of CO were studied in 10 animals. Measurement
of cerebral venous blood flow was made following the
technique of Rapela and Green (1964). Blood  flow
from the confluence of the sinuses was measured af-
ter extracerebral communications had been prevented
by occlusion  of the lateral sinuses  with  bone wax
(Fig. 1). Thus, intracranial blood flow from the sagit-
tal and straight sinuses was measured. From the con-
fluence of the sinuses, blood then passed through a
previously  calibrated electromagnetic flow probe, be-
fore  returning to  the  animal  via the  femoral  vein.
With this technique, approximately 50 to 70% of the
mass of the brain is drained at the confluence of the
sagittal and straight sinuses. Brain perfusion pressure
was  estimated as  systemic arterial  pressure (mean)
minus cerebral venous  outflow pressure. Intracranial
vascular resistance was calculated by dividing brain
perfusion pressure by cerebral venous outflow. Cere-
bral  venous  outflow  pressure and  femoral arterial
pressure were measured with Statham pressure trans-
ducers. Animals were paralyzed with succinylcholine
and ventilated with a positive pressure respirator con-
nected to a tracheostomy tube. Tidal volume and res-
piratory rate  were adjusted to give an alveolar (end-
expiratory) carbon dioxide of 4.0% as monitored by a
C02 gas analyzer.

    In the 13 animals in  which the hemodynamic
effects of hypoxic hypoxia  and CO hypoxia  were
compared, arterial 02 content was lowered by either
of two methods: (1) by inhalation of various 03 mix-
tures in nitrogen  at constant ventilation; this is  re-
ferred  to  as hypoxic hypoxia. (2) by inhalation of
various levels of CO, also at constant ventilation, to
produce equivalent reductions in arterial 02 content;

-------
 454
                                      Traystman
         FEMORAL ARTERY
            PRESSURE
        CEREBRAL VENOUSI
        aUTFLOW PKMURfl	
        MRCDRM. VENOUSI
        I    OUTFLOW   I	
     FEMORAL VEIN
                                       FLOWMETER
                                         LEVEL
                                       DETECTOR
 Fig. 1.   Procedure used to measure cerebral venous
 outflow.  Blood from the confluence of the sinuses
 (sagittal, straight, and lateral) was diverted through an
 electromagnetic flowmeter into a 20-ml reservoir and
 returned to the animal through a  femoral vein by a
 system  composed  of an electronic level detector, a
 switch,  and  a pump  that maintained the  level of
 blood in the reservoir constant.  Cerebral venous out-
 flow pressure was measured upstream  to the flow-
 meter.   Collateral communications between intra-
 and extra-cranial venous circulations were effectively
 occluded by  injecting  bone wax  into both lateral
 sinuses (marked with X). P. T. = pressure transducer.

 hypoxia produced in this manner is referred  to as CO
 hypoxia. 02 content for both  types of hypoxia was
 reduced from control (17.5 vol%)  to approximately
 16.0, 14.0,8.0, and 4.0 vol%. One important point to
 note is that  although  arterial 02 content is reduced
 with both types of hypoxia, with CO hypoxia there is
 no reduction in the arterial 02 tension.

     Animals  were maintained  at  a given  level of
 hypoxic hypoxia  for IS to 20 min and  CO hypoxia
 for 35 to 40 min to allow equilibration of ventilatory
 and blood gases before final gas samples were taken,
 and to allow time  for hemodynamic responses to oc-
 cur and to be maintained. Arterial and  cerebral  ve-
 nous blood samples  were taken,  respectively, from
 the femoral artery and cerebral venous outflow can-
 nulae. The experimental protocol was such that each
 animal acted as its own control for several  different
 levels of both  hypoxic hypoxia and CO hypoxia. In
 the 10 animals in which  the effects of low  levels of
 CO were studied, COHb levels were raised to  approxi-
 mately 2.5, 5.5, 8.0,  12.0,16.0, 22,30, and  50% and
 maintained at each level for 30 min.


    Oxygen tension, C02 tension, and pH at 37°C
were measured immediately after the samples were
obtained using IL  113  electrodes and analyzer. 02
and COHb saturation and hemoglobin were also meas-
ured immediately after samples were taken with an IL
CO-oximeter. Calibration of the electrodes was done
before and after each set of samples were taken.

Results

    The effects of CO hypoxia and hypoxic hypoxia
on cerebral blood flow are shown in  Fig. 2. (All data
in Figs. 2 to 5 were  analyzed for significance using
analysis of variance, and a point-by-point analysis for
significance was done using a paired t-test. Each point
represents the mean ± SE of 13 animal preparations.)
With hypoxic hypoxia, as arterial 02 content was re-
duced, cerebral blood flow increased to 108,  130,
164, and 271% of control.  With CO hypoxia, blood
flow increased to 110, 154, 196, and 232% of con-
trol. While  these increases in blood flow are signifi-
cantly different from  control for both types of  hyp-
oxia,  there is no significant difference between the
two groups. Note that while the arterial 02 content
scale  refers  to both CO hypoxia and hypoxic  hyp-
oxia,  an additional scale showing percent COHb is
also indicated.

    Fig. 3 shows the effect of both types of hypoxia
on  systemic (mean) blood pressure. Blood pressure
with  hypoxic hypoxia increased  to  104, 111,  114,
and 108% of control as 02 content was lowered.  With
CO hypoxia, blood pressure tended to decrease  only
slightly at the higher 02 contents, but was significant-
ly reduced at 4.1  vol%. Since cerebral blood  flow
increased equally with both types of hypoxia (Fig. 2)
and  considering these changes in systemic arterial
blood pressure,  it  is  obvious  that cerebral vascular
resistance must decrease more  with CO hypoxia than
with hypoxic hypoxia. This is illustrated in Fig, 4.
.£  60

I  50

O  40
it

§  X

£  20

<10
            <232)p\J<19e>
                   (164)
           _«   HYPOXIC HYPOXIA
           — o   CO HYPOXIA
             ()   X OF CONTROL
                                   Moai
            2   4   6   8   10  12   14   16  18
  0           ARTERIAL 02 CONTENT, Vol %
                rfc      Si            tS   ii  b
                      CARMXVHIMOOLOIIN,«

Fig. 2.  Effect of CO hypoxia and hypoxic hypoxia
on  cerebral blood flow. Each  point represents  the
mean ± SE of 13 animal preparations. Note that in
some  cases one half of the SE bar is omitted so as not
to overlap other points in the figure.

-------
 Effect of CO Hypoxia and Hypoxic Hypoxia on Cerebral Circulation
                                             455
     160


     140
y*~
t
I
I
f

8
2
2
i

120
100

80


•0
40
20
I ...i-f---^^
t|l 	 W) r(IOOI
¥'
I (70)


i « HYPOXIC HYPOXIA
0---0 CO HYPOXIA
( )V. of CONTROL


                4   I    8    10   IZ   14
                  ARTERIAL O2 CONTENT, Vol %
               T»
                       SI            50   II   0
                     CARBOXVHEMOOLOBIN. %
Fig.  3.  Effect of CO hypoxia and hypoxic hypoxia
on arterial blood pressure.

     Hypoxic hypoxia  resulted  in a decrease in cere-
 bral vascular  resistance to 91,  85, 69, and 39% of
 control, while  CO hypoxia produced an even greater
 fall  in resistance to 86, 63, 47, and 30% of control.
 Changes from control for both curves are significant,
 the  two curves are significantly different from each
 other  at  13.5  and 7.6 vol%, and the difference ap-
 proaches significance at 4.1 vol%. Thus, in the face
 of a decreased arterial 03 content  and resulting tis-
 sue  hypoxia caused by both hypoxic hypoxia and CO
 hypoxia, the brain appears to increase blood flow in
 order to maintain the required  O2 delivery, as shown
 in Fig, 5,  02 consumption of the brain initially in-
 creased and subsequently  decreased for both types of
 hypoxia. The  difference  between the curves  is not
 significant,  although the initial increase is significant
 with hypoxic  hypoxia, but  not with CO hypoxia.
            1391
               1(30)
                             •—• HYPOXIC HYPOXIA
                             0---0 CO HYPOXIA
                             (  )* of CONTROL
              4    •    (    10   It

                ARTIRIAL Oj CONTENT, Vol *
              75      61            30 11  0
                 CARBOXYHEMOQLOBIN, X
Fig. 4.  Effect of CO hypoxia and hypoxic hypoxia
on cerebral vascular resistance.
This increase probably results from the greater release
of catecholamines with hypoxic hypoxia, which has
been shown by other investigators to increase 02 con-
sumption (King et al. 1952, Sokoloff 1959). The de-
of catecholamines with hypoxic hypoxia, which has
been shown by other investigators to increase 0^ con-
sumption (King et al. 1952,  Sokoloff 1959). The de-
crease  in 02 consumption with hypoxic hypoxia is
significant  only at 4.1  vol%; with CO hypoxia it is
significant  at 7.6 and 4.1 vol%. These  data clearly
indicate that the brain can maintain oxygen consump-
tion until rather severe levels  of hypoxia.

    In the second series of experiments  involving 10
animals, the effects on cerebral blood flow of increas-
ing COHb to 50% with particular emphasis on COHb
levels below 20% were studied.  Fig.  6 shows that a
COHb  level as low  as 2.5%  resulted  in  a small, but
significant, increase in cerebral blood flow to 102% of
control. Each point in Fig.  6 represents the mean  ±
SE of 10  animals,  and the data were  analyzed for
significance by a paired t-test. With reduction  in 02
carrying  capacity of 10, 20, and 30% (COHb = 10,
20, and  30%), cerebral  blood flow increased to ap-
proximately 110, 120, and  130% of control, respec-
tively.  At each of these  levels, cerebral ©2 consump-
tion remained unchanged. At COHb levels above 30%,
cerebral  blood flow increased out of proportion to
the decrease in 02 carrying capacity, but  the  brain
could  no longer maintain constant oxygen consump-
tion. At COHb levels of 30  and 50%, cerebral  blood
flow increased to about 130 and 200% of control,
results comparable to the first series  of  experiments.
 Discussion

    These data support the conclusion that the brain
increases blood  flow in response to 02 needs with
both hypoxic hypoxia and CO hypoxia. Although the
means of producing tissue  hypoxia with hypoxic
hypoxia and  CO hypoxia are different, the results, as
     3.0
  K*
  *
  3 1.0

     1.5

     1.0

     Oft
                                                                         HYPOXIC HYPOXIA
                                                                     0---0 CO HYPOXIA
                                                                     ( } % Of CONTROL
                4   (    S   10    12   14

                  ARTERIAL 0| CONTENT I Vol %)
               7S
                       91             90
                        CARBOXYHEMOOLOBIN
Fig. 5.   Effect of CO hypoxia and hypoxic hypoxia
on cerebral O2 consumption.

-------
456
                                      Traystman
        200
        ISO
        140
        ItO
        100
                 10     tO     SO     40

                    CARBOXVHCMOGLOBIN, %
                                             eo
 Fig. 6.  Effect of increasing carboxyhemoglobin lev-
 els on cerebral blood flow, with special reference  to
 low-level administration  (below 20% COHb). Each
 point represents the mean ± SE of 10 animal prepara-
 tions.

 far as the brain's hemodynamic response, are essen-
 tially  the same. The difference in the blood pressure
 response is accounted for by  the  fact that hypoxic
 hypoxia stimulates  the chemoreceptors with its low-
 ered PaC>2 whereas CO hypoxia does not. These pres-
 sure changes have no significant  effect  on cerebral
 blood flow  since the brain  autoregulates and main-
 tains constant blood flow. The fact  that  cerebral  re-
 sponses to  both types of hypoxia  are similar implies
 that the mechanism of action of both types of hyp-
 oxia may also be similar.

     The mechanism  by  which  C>2  affects cerebral
 blood vessels is  unclear.  However, there  is evidence
 that oxygen can act directly on isolated perfused ves-
 sels, low p02 in the perfusate  causing vasodilation
 and' high p02 vasoconstriction (Guyton  et al. 1964,
 Detar and Bohr 1968). Hypoxia has also been report-
 ed to  result in a lactacidosis and a reduction in pH of
 the extracellular fluid of the brain, which  could result
 in cerebral vasodilation (Betz  1972). However, with
 hyperoxia, which does result in a slight vasoconstric-
 tion no alkalosis of the extracellular fluid has been
 demonstratea (Betz  1972). Sokoloff (1959) pointed
 out that neurogenic mechanisms  in the control  of
 cerebral blood flow with hypoxia might be involved.
 The mechanism of the low oxygen effect on  respira-
 tion is a reflex one via the chemoreceptors, and it is
 possible that the cerebral vasodilation with hypoxia is
 similarly mediated.  Indeed, it has  been  postulated
 that  the  carotid  chemoreceptors  and  neurogenic
 mechanisms are responsible for virtually all of the
 cerebral vasodilation  in response to hypoxia (Ponte
 and Purves 1974). However, it  has been recently dem-
onstrated that the carotid chemoreceptors are not re-
sponsible  for  the  cerebral vasodilator  response  to
hypoxia (Traystman, et al. 1978, Heistad et al. 1976,
Bates and  Sundt  1976)  to  CO  (Traystman  et  al.
1978), or  to cyanide hypoxia (Pitt et al. in press). It
remains possible, however, that the cerebral vasodila-
tor responses to hypoxia are mediated through higher
brain regulatory centers, including central chemore-
ceptors  that  may  be located  in cerebral arteries or
veins, or in the brain mass itself. This matter present-
ly remains unresolved.


    The mechanism of CO-induced cerebral vasodila-
tion differs  from  that  of hypoxic hypoxia  since
arterial p02 with CO remains at control levels.  How-
ever, cerebral venous p©2 decreases with both hyp-
oxic hypoxia and CO hypoxia and could be  involved
in a possible cerebral control mechanism. Unless an
O2 sensor that can respond to CO exists somewhere
in the circulatory system, the most  reasonable expla-
nation is that brain tissue itself controls its own blood
flow depending upon what the tissue requires.  Since
increases in cerebral  blood flow  with hypoxic hyp-
oxia and CO hypoxia were essentially  identical,  the
idea that the brain itself controls  its own blood flow
depending upon metabolic tissue  requirements  could
hold for both types of hypoxia. The precise mechan-
ism, however, by which tissue need results in cerebral
vasodilation is unknown.
     An important finding regarding the behavioral
 and electrophysiological  consequences  of  CO
 exposure is that cerebral blood flow increases progres-
 sively with increasing COHb levels (Fig. 2 and 6). In
 addition,  02  consumption  is  maintained  constant
 even at COHb levels  of 30% or  more (Fig. 5). This is
 consistent with the  finding that superior colliculus
 evoked potential latencies are not affected by COHb
 levels up to 40% (Dyer and Annau 1977)..At levels
 above  this, the brain cannot increase blood flow e-
 nough  to compensate for decreased tissue ©2 deliv-
 ery. At these levels, then, behavioral and neurophysi-
 ological abnormalities should be quite evident.


     The   idea  of a threshold level  below  which
 changes in COHb would not invoke increases in cere-
 bral blood flow (Otto  and Reiter, this section) was
 not substantiated by these studies. A threshold level
 such as this would have nicely  accounted for behav-
 ioral and electrophysiological decrements observed by
 some investigators at COHb levels less than 5% (Xin-
 taras et al. 1966a, Beard  and Wertheim 1967).  Our
 findings show that cerebral blood flow is elevated at
 COHb  levels  as low  as 2.5%. There appears to be an
 almost perfect compensation of blood flow as the 02
 carrying capacity is reduced, at  least up to about 30%
 COHb.

-------
Effect of CO Hypoxia and Hypoxic Hypoxia on Cerebral Circulation
                                           457
Summary

    Cerebral hemodynamic responses  to  arterial
hypoxia were studied in 13 anesthetized dogs. Arteri-
al 02 content was lowered from control (17.5 vol%)
to 16, 14, 8, and 4 vol% by two methods, decreasing
arterial Cb tension (hypoxic hypoxia) and increasing
COHb saturation (CO hypoxia)  at normal pO?. Cere-
bral venous  blood How (CBF) was measured  at the
confluence of the sagittal, straight, and lateral sinuses,
with the lateral sinuses occluded. Both hypoxic hyp-
oxia and CO hypoxia at each of the lowered arterial
02  contents resulted in  progressive, significant  in-
creases in CBF (108, 130, 164, and 271%, and 110,
154, 196, and 232% of control, respectively). Cere-
bral 02 consumption was unchanged at moderate hy-
poxic levels, but decreased with  severe hypoxia (8
and4vol%).

    The  effects of low COHb levels (below 20%)
were  examined in 10  animals.  It was found that
COHb levels as low as  2.5% resulted in a small but
significant rise in CBF. The evidence does not support
the concept of a threshold level below which changes
in COHb do not evoke compensatory  increases in
CBF.

-------
CONTINGENT NEGATIVE  VARIATION AS
AN INDEX  OF  ENVIRONMENTAL  DISTRACTION1


 H. WEINBERG, S. CURRY, AND J. PETERS
 Brain Behavior Laboratory, Department of Psychology, Simon Fraser
 University, Vancouver, British Columbia
     The effects of environmental noise on perform-
 ance have been generally  attributed to changes  in
 attention  or arousal. When noise facilitates perform-
 ance, the effect is presumed to result from an increase
 in attention or arousal that is not disruptive. When
 decrements occur, noise is  presumed, conversely,  to
 produce disruptive (i.e., excessive) increases in arousal
 or  performance. Hockey (1970) and  others have
 shown that  the facilitory or decremental effect  of
 noise on performance depends, in part, on the partic-
 ular demands of the task.

     Distraction is another concept employed to ex-
 plain decremental effects of input with respect to per-
 formance  of a primary task. Distraction is said  to
 occur because of excessive increases in arousal and/or
 the  necessity of dividing attention between two  or
 more tasks. Tecce et al. (1969, 1973) have pointed
 out that distraction may be evaluated by the subject's
 performance with  respect to the distracting task,  as
 well as the primary task. In effect, this view assumes
 that distracting information, irrelevant to the primary
 task, is being processed and is responsible for decre-
 ments in performance of the primary task.

    Electrophysiological measures of stimulus regis-
 tration or  information processing have also been used
 to as»ess  the effects of distraction. Contingent nega-
 tive variation (CNV) is presumed to reflect the degree
 of attention directed toward a task (Tecce and Scheff
 1969, Tecce and Hamilton 1973) and to be sensitive
 to changes in general  arousal (Tecce, 1972). McCal-
 him  and  Walter (1968) have also attributed the ef-
 1 This work was supported by grants from the Medical
 Research Council of Canada; the report of these re-
 sults at EPIC IV was facilitated by funds from the
 U.S.  Environmental Protection Agency;  and tele-
 phones used In .this study were generously provided
 by British Columbia Telephone Company, Educa-
 tional Division.
fects of intertrial punctate tones on CNV amplitude
to distraction.

    In the present experiment, the CNV was employ-
ed as a measure of the distractability of a common
environmental noise, the telephone ring, during a vis-
ual discrimination task. The degree of distraction was
hypothesized to vary as a function of the task rele-
vance  of the  distracting noise. It was hypothesized
that the ringing telephone would  result in reduced
CNV amplitude, increased reaction time (RT), and
more errors. Distractability of the telephone was also
compared with distractability  of random tone bursts
similar to those used by McCallum and Walter (1968).

Methods

Subjects

    The Subjects were four male and four female paid
volunteers between the ages of 18 and 39. They were
paid $2.50/hr plus an incentive for each correct re-
sponse. No subject had previous experience in a CNV
experiment or any known neurological deficit.

Procedure
     Electrical activity from the scalp and eye move-
 ments were recorded on an eight-channel Mingograph
 with time constants set  at 5.0 sec. High-frequency
 filters were set  at 30  Hz. Grass Ag/AgCl disc elec-
 trodes were affixed to the scalp at Fz, Cz, Pz, T5, and
 T6 with  collodian-soaked gauze patches. Scalp elec-
 trode impedances were less than 2 kft. Pairs of similar
 electrodes attached to the left (Al, A3) and right
 (A2, A4) mastoid processes served as references. Mid-
 line  electrodes  were  referenced to Linked  mastoids
 (Al  and  A2). Electrodes  T5 and T6 were referenced
 to the contralateral mastoid (T5-A4 and T6-A3).

-------
CNV as an Index of Environmental Distraction
                                             459
     Beckman biopotential electrodes were placed at
 the nasion, outer can thus,  and infraorbital ridge  of
 the right eye and were used for bipolar recordings of
 EOG  between nasion and outer canthus, and between
 outer canthus and infraorbital ridge.

     EEC and EOG activity were digitized on-line by
 an Hewlett-Packard  2116B  computer and  only arti-
 fact-free trials were  stored.  The sampling epoch was
 4.096 sec digitized into 1024 samples.

     Subjects were positioned  comfortably on  a bed
 within  an  electrically  shielded, sound-attenuated
 chamber. A television monitor was located in a com-
 fortable viewing position approximately 4 meters in
 front  of the subject, immediately  above the monitor
 were a red and green light to indicate the relevance or
 irrelevance of the concurrent auditory stimulation.

     Stimulus presentation was  under control of a
 Grason-Stadler (series 1200) solid-state programming
 system.  Both SI and S2 were visual stimuli presented
 tachistoscopically on the television  monitor.  A fix-
 ation  point was located in the center of the monitor
 screen. SI was a three- to six-letter word that  could
 be used either as a noun only or alternatively as both
 a noun  and a verb.  Stimulus duration of SI was  50
 msec. S2  consisted of a question  mark (?) presented
 1.5 sec after SI. The subjects response terminated S2,
 Intertrial intervals varied randomly from 5 to 50 sec.

 Conditions
     First Standard:  All subjects were given a suffi-
 cient  number of trials such that  eight  artifact-free
 trials  were  collected.  In this condition, all subjects
 were  required to make the discrimination described
 above, but without distraction.


     Distraction:  Following the  first standard condi-
  tion, 32 artifact-free distraction trials were collected:
  trials in each of four distraction conditions were pre-
  sented according to  a  table of random variations that
  ensured that an equal number (8) of trials occurred in
  each  condition. The four conditions were  designated
  Relevant Tone, Relevant Telephone, Irrelevant Tone,
  and Irrelevant Telephone. In the tone conditions, ex-
  traneous tone bursts (delivered  by loud speakers be-
  low the subject's bed) began within the intertrial in-
  terval and continued through the interstimulus inter-
  val.  The tone  bursts commenced at variable intervals
  within the intertrial interval and varied  randomly in
  duration between 100 and  1000 msec and in frequen-
  cy  between 640 and 2000  Hz. In the telephone con-
  ditions, a standard telephone, located near the sub-
  jects  bed, began ringing in the intertrial interval and
  continued through the interstimulus interval. The ring
mimicked the standard Bell telephone pattern of 1
sec on and 4 sec off. Both the telephone rings and the
tone bursts were approximately 70 dB sound pressure
level (SPL), measured at the subject's head. The rele-
vance or irrelevance of tones or telephone rings was
indicated to the subject by means of concurrent illu-
mination of one of the two lights located  above  the
television  monitor. A green light  indicated that  the
subsequent discrimination trial would be followed by
delivery of  relevant feedback, either  over the tele-
phone in the  Relevant Telephone condition, or over
an intercom system in the Relevant Tone condition.
A red light indicated that there would be no feedback
on the next trial.
    Second Standard:  After the distraction trials,
eight artifact-free  trials of a second standard condi-
tion were collected. The Second Standard condition
was identical to the first in paradigm.


Instructions
    The subjects were instructed to respond as quick-
ly as possible to S2 with the left hand if the word
appearing as SI  was a noun, and  with the right hand
if the  word  could be either a noun or a verb. They
were also told that they would receive $0.25 for each
correct response, defined as a correct choice with a
latency of less that 250  msec.
     In the relevant  conditions, the subjects were in-
structed  to respond to the telephone rings or to the
tone bursts by picking up the telephone receiver or, if
tones were occurring, by  pressing an intercom call
button, but only after responding to S2. After lifting
the telephone receiver or pressing the intercom but-
ton, subjects were informed as to the correctness of
their discrimination, the speed of their response (fast
or not fast enough), and their total accumulated earn-
ings. In the irrelevant conditions, the subjects were
given the  choice of terminating the tone or the  tele-
phone rings by  responding  in a similar fashion. If the
subjects did  not terminate the sounds,  they  were
automatically terminated within 30 sec after S2. Re-
gardless of how the noise was terminated, no informa-
tion was presented to the  subjects during the irrele-
vant conditions.

 Data Reduction
     The EEC from single trials, for each of the six
 conditions, were averaged off-line and reaction times
 were quantified. Baselines were defined as the aver-
 aged EEC 1.0 sec preceding SI. Averaged CNVs were
 quantified as mean amplitudes relative to the baseline

-------
                                                                                                                        o\
                                                                                                                        o
       EYE
    Cz
                                                              Y
T5
T6
FIRST STANDARD
(RT = 658)
-'- inj l/Srf-^'^V*^

RELEVANT TELEPHONE
(RT=351)
                         .
^
IRRELEVANT TELEPHONE
(RT = 351)
RELEVANT TONE
(RT = 350)
 IRRELEVANT TONE
 (RT = 373)
 SECOND STANDARD
 (RT =295)
                  N = 8
         SI     S2
                                                                                                                        3
                                                                                                                        sr
 «. 1. Summary averages across subjects for each standard and distraction condition. Mean reaction-time for each condition is indicated in the left column.

-------
CNV as an Index of Environmental Distraction
 within  the following intervals: Measure  1, 500 msec
 following SI; Measure 2, 500 to 1000 msec following
 SI; Measure  3, 1000 to 1500 msec following SI; Mea-
 sure 4,  200 msec  preceding S2.  These values were
 converted  to microvolts by comparison to a 50-fJiV
 calibration pulse that had been averaged in an analo-
 gous manner.
 Results
     Summary averages across all subjects at each elec-
 trode location in each condition are shown in Fig. 1.
 Mean RTs for  each condition are also shown in the
 left  column of Fig. 1. A stepwise discriminant analy-
 sis indicated that ERP Measures 3 and 4 at T5, T6,
 and  Fz were the variables that  contributed  most to
 correct classification of individual waveforms by con-
 dition.
     In order to assess the effects of the four distrac-
 tion conditions, repeated measures t-tests were run
 between conditions on the  four amplitude measures
 for Cz, T5, and T6. The only consistent significant
 (p< .05) differences seen at Cz were between the First
 Standard and three distraction conditions for Measure
 2 (the exception being the Irrelevant Tone condition)
 and between the Second Standard and  all other con-
 ditions for Measure 3. In these cases,  Cz  amplitude
 was larger in both standard  conditions than in the
 distraction conditions, and maximal in the Second
 Standard  condition.  The differences  between the
 standard and  the distracting conditions for TS were
 not as consistent as those for Cz; i.e., maximal nega-
 tivity was not invariably associated with the standard
 condition. Analysis of T6 amplitudes revealed signifi-
 cant differences between the First Standard and Irrel-
 evant Tone conditions.

     In view of the large contribution of T5 and T6 to
 the discriminant analysis, the  apparent differential re-
 sponsivity of T5 and T6 to distraction,  and visual
 inspection of the waveforms,  differences between TS
 and T6 data were analyzed with a matched t-test de-
 sign. Significant differences (p<.05) were found be-
 tween all  four  measures in each  condition; T5 was
 consistently more negative than T6.
     Matched t-tests were also used to assess the ef-
 fects of distraction on reaction times. Significant dif-
 ferences (p< .05) were found between the First Stand-
 ard  and  all other conditions, between the Second
 Standard  and the First Standard conditions, and be-
 tween the Second Standard and  the Irrelevant Noise
 conditions. RTs were longer in  the  First  Standard
 than in any of the other conditions. RTs in the Sec-
 ond Standard were shorter than  in any of the irrele-
                                             461
vant conditions and shorter than in the First Standard
condition.  No significant  differences in RT  were
found  between the four distraction conditions. The
number of errors in discrimination was not subjected
to statistical analysis since there  were almost no  er-
rors in the irrelevant or Second Standard conditions.
Discussion

     One of the most interesting results of this study
was the  consistently  larger amplitudes (greater nega-
tivity) observed at T5 compared to T6 in all condi-
tions. Since each average contained an approximately
equal number  of left- and right-hand responses, and
since there were no consistent differences in electrode
impedances or amplifier  gains, the asymmetry can
presumably be  attributed to the nature  of  the dis-
crimination. The laterality effect implies that the two
hemispheres  function differentially with respect  to
the processing  of language information presented vis-
ually.

     Distraction conditions did not yield decrements
in CNV  amplitudes of the consistency or magnitude
expected.  Since  reaction times  were  significantly
longer in  all  distraction  conditions, relative to the
Second Standard,  it appears that tones and rings were
somewhat effective in disrupting discrimination be-
havior. This conclusion is, however, not supported  by
the minimal number of discrimination errors observed
in all  conditions.  In  general, the  distractors did not
severely  disrupt discrimination behavior, a  finding
that is consistent  with  the  minimal  decrements  in
CNV amplitudes observed during distraction.


     Although  CNV decrements attributable  to noise
were minimal, it  is notable  that  the differences be-
tween the First Standard and the distraction condi-
tions, and also the differences between  the  Second
Standard  and   the distraction conditions, were pri-
marily in  amplitude measures of different  parts of
the waveform.  The First  Standard differed  from
other  conditions  primarily with respect to  Measure
2,  the mean amplitude in the interval 500  to  1000
msec  after SI.  The Second Standard  differed  on
Measure 4, which is  the interval 200 msec preceding
S2.  These observations suggest  a change of wave-
form  shape  from the First  to the Second Standard
conditions.  This  is probably attributable to the fact
that the First  Standard  waveforms reflected primar-
ily  task acquisition  processes, whereas  the  Second
Standard waveforms reflected primarily performance
processes of an already learned task.


     Tecce (1972) has argued that an important meas-
ure  of distraction is  the  degree to which distracting
information is processed  in addition to the effect of

-------
462

distraction on the primary task. This view conceptual-
izes distraction as information that must be processed
in parallel  with the primary task. This experiment did
not contain any measure of the extent to which the
telephone  rings or tones were processed. However,
the paradigm  is typical of experiments in which the
measure of distraction is the degree to which extrane-
ous stimuli influence primary task performance. The
observed effect was  small  even with  distractors pre-
viously shown to produce large CNV decrements, i.e.,
random  intertrial punctate  tones (McCallum and
                                  Weinberg et aL

Walter 1968). Failure to observe a behavioral or elec-
trophysiological distraction  effect may have resulted
from  the use of different modalities for distractors
and primary task.
    Is the telephone distracting? If the CNV is a relia-
ble index of distraction, the answer appears to be
"no," at least in the limited case of a visual semantic
task.

-------
CARBON MONOXIDE, TRICHLOROETHYLENE,
AND ALCOHOL:  RELIABILITY AND VALIDITY
OF NEUROBEHAVIORAL EFFECTS1


G. WINNEKE, G. FODOR, AND H. SCHLIPKOTER

Medical Institute for Air Hygiene and Silicosis Research,
The University, Dusseldorf,  Federal Republic of Germany
    Environmental toxicology is concerned with the
 establishment of dose-effect relations of noxious sub-
 stances that may be found in human habitats. A basic
 objective  is to  provide data for  setting  exposure
 standards  that will protect the health and welfare of
 people in residential and industrial environments.  Ex-
 posure standards aie based on the threshold at which
 significant impairment in function can be demonstrat-
 ed. Environmental lexicologists are thus faced  with
 the difficult task of defining empirical thresholds for
 a wide spectrum of noxious substances. Nor is it suffi-
 cient merely  to show an  effect; convincing evidence
 must be marshalled to establish that  an observed ef-
 fect is deleterious to health.
     "Evidence  used in setting standards must satisfy
 minimal requirements of reliability and validity, basic
 concepts in  psychological test  theory  (Cronbach
 1961). Reliability refers  to the  consistency of test
 scores in repeated measurements, whereas validity re-
 fers to the meaning of a test  acore. The question of
 reliability assumes the form: Can observed threshold
 effects be replicated within or between laboratories?
 The question  of validity underlies the need to demon-
 strate the functional significance of an observed im-
 pairment or effect. Reliability and validity pose criti-
 cal problems  for neurobehavioral research in environ-
 mental toxicology. These problems will be illustrated
 with experiments conducted in our laboratory during
  the past 3 years.
  'The experiments reported here were supported by
  "Kommission Europaischer  G erne ins chaften-
  Direction Gesundheitsschutz" (CO), and by "Deut-
  sche Forschungsgemeinschaft
Reliability of Neurobehavioral Measures
    The study of threshold effects requires extremely
sensitive measures, capable of detecting small changes
in  the  spontaneous  variability  or  physiological
"noiae" of the behaving organism. Many environmen-
tal chemicals such as lead, carbon monoxide, and sol-
vents are assumed or known to affect the central ner-
vous system (CNS). Signal averaging techniques have
been widely used to increase the signal-to-noise ratio
in studies of CNS function. Therefore, average evoked
potentials should  provide the sensitive kind of CNS
measure needed in environmental toxicology.
     The functional state of the brain Is influenced by
 numerous factors, Including amount of sleep, inges-
 tion of stimulants such as coffee or depressants such
 as alcohol, and motivational variables. These factors
 must be carefully controlled in order to  distinguish
 threshold toxicant effects.   When these  factors are
 inadequately  controlled, inconsistent  or  negative
 effects are likely to be observed.
     Neurobehavioral CO research is a case in point
 where there is frequent nonreplication of effects be-
 tween and within laboratories (Wlnneke 1974). Beard
 and Grandstaff (1975) argued  that inconsistent out-
 comes could be attributed to variable levels of activa-
 tion or arousal induced by different experimental
 procedures.  The following  experiment was designed
 (1) to  replicate the results of  Horvath et al. (1971)
 and (2) to test the "arousal hypothesis" proposed by
 Beard and Grandstaff.

-------
464
                                                                                        Winneke et al.
    Twenty nonsmoking students participated in Ex-
periment I. A visual vigilance task described by  Hor-
vath et al.  (1971) and  auditory evoked  potentials
(AEPs) were  used as neurobehavioral  measures. The
vigilance task consisted of a series of 1-sec light pulses
appearing  every 3 sec as "nonsignals" (dimmer) or
"signals"  (brighter); subjects  indicated  choice by
pressing an appropriate button. Signal  brightness was
preset  individually at 0.9 probability of  detection.
Hits and false alarms (FA) were recorded. The test
was conducted in two versions:  a monotonous (M)
and a less monotonous (LM) version. Both lasted 60
min  and  were  preceded  by a 5-min  pretest, which
provided a measure of  detection efficiency under
"alert" conditions. M was identical to the task used by
Horvath et al. (1971), with 10 signals and 290 nonsig-
nals within each of four consecutive 15 min observa-
tion  periods.  LM differed with respect to signal pro-
bability (30 signals, 270 nonsignals), interruptions (3
min  every 15  min), and verbal feedback (subjects
were informed of their performance every 15 min).
    Vertex EEC, with mastoid reference, was record-
ed before and  during the vigilance task (time con-
stant: 0.3 sec). Clicks  of 1000 Hz, 70 dB(A), and
0.01-sec duration  were  presented at random ISIs of
10, 18, 26, or 34 sec via earphones. AEPs were meas-
ured under "ignore instructions" before and during
consecutive  quarter  hours  of the test.  Subjective
states of arousal were assessed by means of 17 bipolar
           rating scales given before and after  the experimental
           session. Subjects were tested single-blind either in the
           morning or afternoon in counterbalanced order under
           each  of the  following conditions: Control, 2.5 hr of
           exposure  to 100 ppm  CO, and  1 hr of exposure to
           200 ppm  subsequently lowered to 150 ppm. CO level
           was monitored by nondispersive infrared analysis, and
           carboxyhemoglobin (COHb) was  determined by
           means of gas chromatography.

              Under conditions of low CO-exposure, neurobe-
           havioral measures were taken between average COHb
           values of about 3.6 and 7.5% and, under moderate
           conditions,  between 6.5 and  11.3%.  No  significant
           CO-effect was found for hits under condition M  or
           LM (Fig. 1).
               There  was a significant performance decrement
           over time, however (F = 14.6; p < .001). Overall false
           alarm rate was low. Meaningful comparisons could be
           made only  if computations were carried out across
           time. Even then FAs were absent for more than half
           the subjects under condition M in at least one CO-
           condition, whereas, across exposures, there was a sig-
           nificant increase under condition LM (U = 18,5; p <
           .01). For this condition, parameters d" and {? of signal
           detection theory (Green and Swets 1974) were deter-
           mined from Freeman's tables (1964). These are given
           in Table 1. No significant CO effect could be shown
           for these more-refined performance measures.
                   VIGILANCE
              MONOTONOUS CONDITION
  100


* 90

o

S3 80
IU
o
< 70
C9
" 60


   SO
                                        N-10
      CONTROL
b—A 100ppm
A--A 200/1 SO ppm
              100


              90


              80


              70


              80


              60
                                                                  VIGILANCE
                                                           LESS MONOTONOUS CONDITION
                                                                                       N-10
                      15       30      45       60             0
                                                    TIME, min
                           15
                                                                          30
45
60
        Fig. 1.  Performance (% hits) in a Horvath-type visual vigilance task under two conditions of monotony
        and three conditions of CO exposure.

-------
                                                                                                  465
Reliability and Validity of Neurobehavioral Effects

Table 1.  Meant  and Standard Deviation! of SDT Parameter! d' (Sensitivity Index) and 3 (Cri-
terion) Taken from Freeman's Tables (1964). Values are for Condition LM of Experiment I.

d'
0
CO exposure
Control
3.4
± 0.6
32.4
±43.6
100 ppm
3.4
± 0.9
26.4
± 17.8
200/1 50 ppm
3.4
± 0.6
44.4
± 39.6
    AEPs were measured before and during the vigi-
lance  test (Fig. 2). The first upward (negative) peak,
with a latency of about 100 msec, is  designated Nl,
and the following downward one  Is P2. Amplitudes
were  measured relative to a 200-msec  prestimulus
baseline.  Numbers 0 through IV  correspond to the
respective vigilance period.

    For the group as a whole, no  amplitude changes
foreither Nl or P2could be demonstrated under CO,
and no significant main  or Interaction effects were
noted (Fig. 3). There was, however, a significant am-
plitude decrement for Nl (F a 3.8; p <  .01) and  P2
                           Nl
                                     1000 Hz
                                       70 dB
                                       10mste
  IV
                                     100 msec
 Fig. 2.  Typical auditory evoked potentials (male sub-
 ject, 21 years old) to 30 irregular clicks (1000 Hz, 70
 dB) presented before (0) and during (1-IV) vigilance
 performance.
(F • 11.5; p < .001) over time. An analysis of the
bipolar rating scales revealed that the different con-
ditions of experimental monotony did indeed Induce
different degrees  of subjective arousal. The interac-
tion "task duration x monotony" was of borderline
significance (p <  .1) for 9 out of the 17 items used.

    These findings exemplify the lack of test-retest
reliability  of CD-induced neurobehavioral impair-
ment. We were unable to reproduce the  findings of
Horvath et al. (1971), although the experimental pro-
cedure was similar in both studies.  In order to rule
out the possibility that the presentation of clicks
might have Increased  general arousal level, a parallel
experiment without AEP measurement was conduct-
ed with 20 additional subjects. The results were essen-
tially the same. It must be mentioned, however, that
sensory restriction was more pronounced in the Hor-
vath et al. study, Whereas Horvath's subjects sat in a
dark, sound-attenuated booth (personal communica-
tion),  the present experiment was  run in a 15-rrr
laboratory lit by two 65-W neon bulbs with the am-
bient noise level- although attenuated by earphones-
at about 67 dB(A). Therefore, the possibility cannot
be ruled out that the Horvath et al. subjects were less
aroused than condition M subjects, a circumstance
that may have contributed to CO-induced hypoxia.
For the  conditions of the study presented here, how-
ever, arousal cannot be considered an important mod-
erator of CO-related vigilance decrement. AEP meas-
ures also did not reveal any CO-influence. This  find-
ing is  consistent with  Hosko  (1970), who  did not
observe  any effect on visual evoked potentials below
20% COHb. Dyer (this volume) also failed to find any
CO effect on  superior colliculus evoked potentials
below 20% COHb.
                                                       Validity of Neurobehavioral Measures


                                                           Establishment of the validity of neurobehavioral
                                                       measures, assuming that the measures are in  fact reli-
                                                       able, is a fundamental problem in environmental toxi-
                                                       cology. A  statistically  significant change in evoked

-------
466
                                                                                          Winneke et al.
                         CONTROL
  A—A  100 ppm
                                      A - A  200/150 ppm
40-

30-

20-
*
D 10-
AEP-N1
A-^'"-A^~~==-«- -o

""•^ A
oc
UJ
<
MONOTONOUS CONDITION
	 T 	 1
40- AEP-P2
° """"•»..
30 A~ *$V-
*" A ••V'--.
^-A_ *%*O---:::IA ~--—,. A
20- £ """---A
UJ — — ..

10- MONOTONOUS CONDITION
     1
0
I
40-

30-

20-


10-




^
oc
UJ
<
15 30 45 60

A-


A — —""""" ^ft»»""— """



' LESS MONOTONOUS CONDITION
0 15 30 45
40-

30-

20


10-


A^
^Ss>»»«.
£ X^v?"%% ,^~.x
UJ ^^^^" "™^^— *. A 	 *? A V>*
< "^^9^"""^ **'**'«
LESS MONOTONOUS CONDITION
60






r-*4

                     15
30
45
60
45
60
                                                     TIME, min
   Fig. 3. Decline of amplitudes ofAEP components Nl and P2 over time for two conditions of monotony and
   three conditions of CO exposure. Means are based on 10 subjects each.
potential amplitude ot latency produced by toxicant
exposure is not a sufficient index of validity, unless
the biological or behavioral significance of the electro-
physiological effect can be demonstrated. The signifi-
cance  of  behavioral impairments  is generally  self-
evident, although the predictive validity of behavioral
measures is rarely tested. For example, a vigilance
decrement  experimentally induced by exposure to a
noxious aerosol can be extrapolated to real life. The
validity of evoked and slow brain potential measures,
however,  is  difficult to  demonstrate because  the
physiological significance of these measures  is still
obscure. The meaning of observed changes in brain
potentials  must be inferred  from concurrent behav-
ioral measures  or changes induced by chemical com-
pounds of known pharmacological properties such as
alcohol. This procedure, known as convergent valida-
tion,  will  be  exemplified   by two  closely  related
experiments.


    Experiments II and III deal with the neurobehav-
ioral effects of trichloroethylene  (€2^013), an in-
dustrial solvent with narcotic properties, and alcohol.
In Experiment II (Winneke  and Kastka 1974), 18
volunteers were exposed for  about 31/2 hours to 50
ppm C^^Cl^- Vigilance performance was measured
with a bisensory task designed to produce a high false
alarm rate (FAR). In two sessions of 50-min duration,
                        subjects had to monitor auditory and visual events in
                        order to detect "signals" that were shorter than other
                        events. Neutral  events were  500-msec  duration;
                        critical ones were 200 (Experiment II) or 150 msec
                        (Experiment  III) shorter.  This  test started  77 min
                        after the beginning of the experiment and lasted 100
                        min.
                            Auditory  evoked potentials  (AEP) were  com-
                        puted  from  vertex  EEC, with  mastoid  reference
                        recorded just  before and  after the vigilance test. A
                        regular sequence (ISl  = 2 sec) of fifty 1000-Hz tone
                        bursts  of 30-msec duration was presented  via ear-
                        phones. Subjects were instructed to respond by but-
                        ton press to signals that were slightly weaker than the
                        majority of events. Two signals were randomly dis-
                        tributed among 50 bursts. Baseline-to-peak measures
                        were taken for components Nl  and  P2 as described
                        above.
                            The  methodology  of Experiment  III (Winneke
                        and Kastka  1975) was  identical  to  Experiment II.
                        Twelve subjects were tested in each of the following
                        conditions: Control, 0.6 g alcohol/kg, 3 1/2 hours of
                        exposure  to  50  ppm  02^013 and  alcohol  plus
                        C2H3C13. Ethyl  alcohol  was given  in orange juice
                        just before the vigilance task. Blood'alcohol level was

-------
Reliability and Validity of Neurobehavioral Effects
                                            467
24 mg% on the average at the end of the experiment,
and, with an hourly decrease of IS mg% assumed, is
estimated to have peaked at 50-60 mg%.
    A review of vigilance performance indicates that
the experiments differed in task difficulty (Table  2).
The signal-to-noise ratio was 300/500 msec for Exper-
iment II, whereas it was 350/500 msec in Experiment
III. Hits and FAR differed in the expected duration.
This was also true for the sensitivity index d', which
was clearly lower in Experiment III. As for vigilance
performance,  there  apparently was no C2H3C13
effect on hits, but there  was a marked effect if hits
and  FARs were combined to yield d'. Interestingly
enough, there was a significant increase in sensitivity
under the influence of C2HjCl3 as compared to con-
trol values in Experiment II but a slight, insignificant
decrease in Experiment HI. This unexpected reversal
suggests a possible interaction between task difficulty
and prenarcotic action of C2H3C13; for easy detec-
tion conditions, subjects may have been able to com-
pensate  or  even overcompensate for the C2H3Cl3-
induced handicap, whereas under more threshold-like
conditions (Experiment HI), compensation may have
broken down.
    This interpretation rests mainly on the results of
AEP measurements shown in Table 3. There was little
change in amplitude from beginning to end of the
experiment for control conditions, whereas there was
a marked amplitude decrement for component P2 af-
ter  3  1/2 hours of C2H3C13  exposure. There was a
slight increase of P2 amplitude from first to second
testing under control conditions, but a marked reduc-
tion after C2H3C13 exposure. The resulting inter-
action  "time  x  condition" was  significant  for
C2H3C13 in both experiments (p < .05), but of only
borderline significance (p < .12) for the alcohol con-
dition  of Experiment III.  There was  no consistent
effect on Nl amplitude of C2H3C13 or alcohol.
    While the results of Experiments  II and III de-
monstrated  test-retest  reliability of the C2H3C13-
induced decrement in P2 amplitude, the electrophysi-
ological and behavioral  data were inconsistent across
experiments. C2H3C13  exposure was associated with
increased sensitivity (d) in Experiment II, but with
no  significant change of d' in  Experiment  III. The
effect  of alcohol  on AEPs and behavior was  also
equivocal:  alcohol produced a  pronounced vigilance
                    Table 2.  Vigilance Performance for Experiments II and

Hits, %
FAR,%
d'
&
Experiment lla
(N=18)
Control
71.3
1.2
3.0
11.3
C2H3C13
73.1
1.0
3.2C
9.1
Experiment III3

-------
 468
                                                                                           Winneke et al.
decrement, but only a borderline P2 decrement (Ta-
ble 3). Thus,  the validity of observed electrophysio-
logical effects could not be demonstrated  with re-
spect to available behavioral measures. That is, we are
confronted with a reliable electrophysiological effect,
the meaning of which remains obscure.
    These findings illustrate a dissociation of behav-
ioral and electrophysiological changes under the influ-
ence of C2H3C13 and alcohol. Similar dissociations
have been  reported  by  others (Clark et al.  1969,
1970), although  these have been criticized on the
grounds that behavioral and neurophysiological meas-
ures were taken In different sets of trials, under dif-
ferent  instructions, and  with  inadequate  control of
attentional variables (Donchin  andSutton  1970, Paul
and Sutton 1973). The data reported here are open to
similar criticisms  because vigilance performance and
AEPs were, indeed, measured in different parts of the
experiment. Attentional variables, however, were con-
trolled by essentially  the same instructions for vigi-
lance and AEP measurements.
    Implicit in this discussion is the assumption that
there ought to be a direct relationship between the
parameters of CNS function and behavior chosen for
comparison. The observed dissociations raise doubts
about the validity of this assumption, particularly un-
der conditions of toxicant exposure.
 Conclusions
     Basic problems in assessing the reliability and va-
lidity of neurobehavioral measures have been examin-
ed in the context of data gathered during human ex-
posure to environmental toxicants. The extensive, but
contradictory, literature concerning the neurobehav-
ioral effect of low concentrations of carbon monox-
ide was  used to illustrate the problem of reliability.
Our failure to replicate the findings of Horvath et al.
(1971) adds another increment of uncertainty to the
CO story, although these results  indicate that varia-
tion of arousal level within the normal waking range
does not moderate the neurobehavioral effect of CO-
induced  hypoxia  as  hypothesized  by  Beard  and
Grandstaff (1975). The possibility remains, however,
that detection of subtle CO effects may be enhanced
during extreme states of hypo-arousal that occur dur-
ing deep  sleep (Groll-Knapp et al., this volume), pro-
longed vigilance performance, or sensory deprivation.
     Large differences  in individual susceptibility to
CO poisoning may also contribute to the confusion.
Most investigators,  including  the  present  authors,
computed means across subjects, thereby obscuring
possible individual differences in CO responsivity. It
is possible, for instance, that the normal adult popula-
tion contains a bimodal distribution of CO responders
who show decrements in performance  and nonre-
sponders who show no decrements. The reliability of
individual differences in CO susceptibility needs to be
assessed, and, if such differences are  established, the
distinctive features of CO-responders need to be iden-
tified.
    Another criticism concerns the subject  popula-
tion used in most human neurobehavioral CO studies.
Young healthy  adults are probably the least likely
segment  of the population to show deleterious effects
during low-levd exposure. The ability of humans and
animals to compensate for CO-induced hypoxia up to
COHb levels of 20% is well known (Dyer and Annau,
this volume). Data from Traystman (this volume) and
others suggest that increased cerebral bloodflow com-
pensates  for the reduced  supply of oxyhemoglobln.
Populations  with reduced oxygen uptake-e.g., eld-
erly  persons with  presumed  slight  impairment  of
cerebral circulation-are likely to be far more suscep-
tible to CO effects than young, healthy adults.
     Establishment of the validity of ERP measures in
environmental  toxicology and other clinical applica-
tions will probably remain a thorny problem for some
time to  come. The process of validation  entails  the
correlation of a known criterion measure with an un-
known test variable. In the present discussion, we
have assumed that behavioral measures (i.e., vigilance
performance) constitute meaningful criteria  for  the
validation  of  ERP  measures. While  a remarkable
correspondence between performance and  certain
ERP components such as the P300 has been demon-
strated (see review by Tueting, this volume), the cor-
respondence is relatively poor for other measures
such as  reaction time  and  CNV  amplitude (Robert
and Tecce 1973). The assumption that  vigilance per-
formance constitutes a meaningful criterion  for  the
validation of P2 changes observed during toxicant ex-
posure is therefore questionable. If the behavioral and
electrophysiological measures show parallel changes, a
straightforward  interpretation  of the  data can be
made, but, if the two measures are dissociated,  the
validity of the ERP measure  cannot be evaluated.
    What conclusions may be drawn from the data in
view of this dilemma? Although the evidence of disso-
ciation is equivocal, the results  suggest caution in ex-
trapolating from control to exposure  conditions. Al-
cohol, anesthetic agents, or other environmental toxi-
cants may decouple the relationship between  ERP

-------
Reliability and Validity of Neurobehavioral Effects                                              469

measures  and performance observed under  normal       dence in evaluating the health effects of environmen-
conditions. This  problem is not  unique to environ-       tal toxicants. Validation of ERP measures, as describ-
mental  toxicology. Normal brain-behavior relations       ed here,  will  require a systematic effort  to  study
may be disturbed by a variety of clinical disorders or       brain-behavior relations  in  a variety of normal and
pharmacological interventions, and the  effects  must       abnormal conditions in order to establish an extensive
be examined carefully in each case.                        network of empirical  associations. This network will
                                                        eventually provide the frame of reference necessary
    Event-related potentials are not understood well       to understand the significance of ERP changes observ-
enough at present to be accepted as independent evi-       ed without concurrent behavioral measures.

-------
 DISSOCIATION  BETWEEN TIME COURSE
 OF ACETYLCHOLINESTERASE INHIBITION AND
 NEUROPHYSIOLOGICAL EFFECTS  OF PARATHION
 IN  RAT  AND  MONKEY1
D. WOOLLEY AND L. REITER
Departments of Animal Physiology and Environmental Toxicology,
University of California, Davis, CA, U.S.A.
    Parathion (0,0-diethyl 0-p-nitrophenyl phospho-
 rothionate) is an insecticide with widespread current
 use  and  high  mammalian  toxicity. Cholinergic
 symptoms  and death due to parathion poisoning have
 been attributed to acetylcholinesterase (AChE) Inhi-
 bition by  paraoxon, the  active toxic metabolite of
 parathion  (O'Brien  1967). In this study, the time
 course of inhibition of AChE activity in the central
 nervous  system (CMS) and blood after parathion ad-
 ministration was compared with the time course of
 neurophysiological effects, as measured by  visual e-
 voked potentials (VEPs) in rats and monkeys bearing
 chronically  implanted electrodes and by changes in
 the maximal electroshock seizure (MES) pattern in
 rats. The results suggest either that adaptation or tol-
 erance to continued AChE inhibition occurs or that
 the cholinergic symptoms of parathion poisoning are
 due to an  effect other  than AChE  inhibition. Other
 possibilities foi mechanisms  of action are suggested.
recorded  from  side-by-side bipolar electrodes with
1-mm vertical tip separation. Each VEP included fifty
responses averaged with a Northern Scientific com-
puter. Usually six averaged VEPs were recorded from
each brain area at each time period. Computer read-
out was with a Moseley X-Y recorder.
    The effects of 3 mg/kg parathion sc on the aver-
aged VEP  in the visual cortex (Fig. 1) were to in-
crease the latency and decrease the amplitude of the
first major wave. Effects were maximal 2 and 4 hr
after administration; 8 hr after administration the re-
sponse had fully recovered. Responses recorded in the
superior colliculus under the same conditions showed
even more dramatic effects of parathion administra-
tion (Fig. 2). ColHcular VEP was almost completely
eliminated  45 mln and 2 hr after 3 mg/kg sc para-
thion. Recovery  was rapid, and by 4 hr after dosage
the response was back to normal.
Effects of parathion administration on VEP,
MES, and AChE inhibition in rats
    Effects of a single dose of parathion on the VEP
were determined in eight rats with electrodes chroni-
cally implanted in cortical and subcortical brain areas.
Parathion was dissolved in polyethylene glycol (Car-
bowax 300) and administered subcutaneously (sc).
Rats were recorded awake and unrestrained in their
home cages. Potentials were elicited by light flashes at
I/sec from  a Grass photostimulator. Responses were
1 Supported by NIH grant ES-00163.
    The effects of another cholinergic agent, tremor-
ine, on VEPs in the  rat were determined and com-
pared  with the effects of parathion. Tremorine is con-
verted to oxotremorine, which acts primarily to stim-
ulate central cholinergic receptors.  The effects of a
single 5- or 10-mg/kg dose of tremorine injected intra*
peritoneally (ip) were to increase the latency and de-
crease the amplitude of averaged VEPs recorded from
the primary visual cortex. These effects were  nearly
identical  to those produced by  parathion. This sug-
gests that these effects may reasonably be attributed
to central or peripheral (or both) cholinergic actions.
Histological evidence for involvement of central cho-
linergic synapses in VEP generation has been provided
by  Shute and  Lewis  (1967), who have described

-------
Dissociation Between Time Course of AChE Inhibition
                                             471
                        TIME AFTER PARATHION
                                     13 nig/kg sc)
                                          8hr
                                         20 hr
                                          1 wk
             (RAT 81 10/8/68)
Fig.  1.  Effects of parathion on  VEPs in the visual
cortex of an unanesthetized,  unrestrained rat with
chronically Implanted brain electrodes. Latency was
increased and amplitude was  decreased 2 and 4 hr
after parathion injection.  Recovery  was complete
after 8 hr.

AChE-containing fibers rising from the brainstem and
passing via the  dorsal tegmental pathway directly and
indirectly  to  the  superior colliculus, lateral  geni-
culate body, and pretectal area.

     The puzzle of these observations is not so much
the marked effects of the toxin on the VEP but the
rapid recovery of the response. Parathion administra-
tion Is believed to  result in irreversible or only slowly
reversible  inhibition of AChE activity  in  the CNS
(O'Brien 1967). However, some controversy exists re-
                        TIME AFTER PARATHION
                                     13 mg/kg tc)
              (RAT B2 10/18/68)
Fig. 2.  Averaged  VEPs In the superior colliculus of
the rat.  The response was nearly abolished 45 mln
and 2 hr after injection of parathion, but recovery
was complete at 4 hr.
garding the rate of recovery of brain AChE activity
after parathion administration in the rat. Although an
early report described recovery of brain AChE activi-
ty 4 hr after administration of parathion (DuBois et
al.  1949), another observed  that  several  days  were
required for 60% recovery from inhibition (Davison
1953),  and  a more recent report  (Giachetti et  al.
1966) stated that 1 month was required for complete
recovery after parathion  administered on alternate
days for 4 months.

    We therefore undertook an investigation of the
time-course  of AChE inhibition and recovery  after
parathion administration in the rat.  AChE activity
was measured  using a colorimetric method (Ellman et
al. 1961). A single dose of 2.0 mg/kg parathion was
administered  ip  or sc  in  separate experiments with
essentially similar results (Fig. 3). Inhibition of AChE
activity in the caudate nucleus, cerebral cortex, brain-
stem, spinal cord, and blood, and inhibition of plasma
pseudocholinesterase (ChE) activity occurred rapidly;
peak inhibition was reached 4 to  9 hr after adminis-
tration. AChE inhibition occurred more  rapidly and
was greater in blood than  in CNS. Recovery of blood
AChE  and of plasma  ChE activities was rapid and
complete during  the 1 week of this study, whereas
AChE  activity in the  CNS showed no evidence of
recovery during this time. In  an additional study, it
was found that 2 to 4 weeks were required for com-
plete recovery of AChE activity in the CNS after ap-
proximately 40% inhibition following injection of
parathion.
    Comparison of Figs. 1  to 3 reveals that recovery
of VEPs in the brain occurred long before recovery of
brain AChE activity. VEP changes after parathion ad-
ministration occurred only  during the falling phase of
AChE activity. By the time brain AChE activity had
reached peak inhibition and stabilized at this inhib-
ited level, VEPs had recovered.
    To determine if another measure  of CNS  func-
tion recovered as rapidly after parathion administra-
tion as the VEP, the effects of three  dose  levels of
parathion on the durations of the various phases of
the maximal electroshock seizure (MES) were investi-
gated. To produce MES, silver disc electrodes with a
small amount of electrode paste were  placed on the
scalp on top of the head just In front of the ears, and
a sine wave (60 Hz) stimulus of 250 mA was delivered
for 0.2 sec by a constant current stimulator, as de-
scribed previously (Woolley 1970). Administration of
l.S, 2.0, or 2.5 mg/kg parathion ip in a single dose
significantly altered the  MES pattern 1 hr later, but
the changes were greater at 4  hr (Fig. 4). Recovery
had occurred by 24 hr. The principal effects of para-
thion administration were to increase the duration of

-------
  472
                                                                       Woolley and Reiter
     C
     E
     o
     •o
     0)
     *•>
     re
 o
 a.
>'
K
>
U
LU
O
o
         50
         45
         40
         35
         30
         25
                                                      Cortex AChE
                                                                   1111
                                                                               J
    234567       01
                            7.5
Brain Stem AChE
                                                          234567
                                                      Spinal  Cord AChE
    234567-01234567
                          0.20 |
                                      Plasma ChE
                   Blood AChE
                   1234567*01234567
                     TIME AFTER PARATHION (2.0 mg/kg) ADMINISTRATION, days
Fig. 3. AChH and ChE activities after parathion administration in the rat. Enzyme activities in blood and CNS
were rapidly inhibited after administration.  Recovery was fairly fast in blood and returned to control levels by 1
week, whereas AChE activity in the CNS showed no recovery during this same time.

-------
 Dissociation Between Time Course of AChE Inhibition
                                                      473
         20
        15
        10
                          CZ3  PARTIAL EXTENSION
                          ES3  FULL EXTENSION
                                FLEXION
                     0  1.5 2.0  2.5
                         thr
       0  1.5 2.0  2.5

DOSE OF PARATHION, mg/kg ip

             4 tar

  TIME AFTER PARATHION
0  1.5  2.0 2.5
                                                                                        24 hr
 Fig. 4.  Durations of phases of the maximal electroshock seizure (MES) were altered 1 and 4 hr after parathion
 administration in the rat but had returned to normal at 24 hr.
the initial tonic flexor phase and to shorten the dura-
tion of the  following full extensor phase. There was
no  effect on the total duration of the seizure. Other
work in  our laboratory (unpublished) indicated that
tremorine administration  produced changes  in  dura-
tion of MES  phases similar to  those produced by
parathion administration;  thus, these MES  changes
may be typical of at least some cholinergic agents.


    Other studies (Woolley  1976) showed  that ad-
ministration  of scopolamine, but not  of methylsco-
polamine, blocked the  effects of parathion  on the
MES. Because methylscopolamine has marked periph-
eral  anticholinergic  effects but little central effects,
whereas scopolamine has both central and peripheral
anticholinergic effects (Innes and Nickerson 1975),
these drug-parathion interaction studies demonstrate
that the effects of parathion on the MES are due to
an action on  the CNS.
              The results of the studies of the effects of para-
          thion on the MES emphasize again that the effects of
          parathion  administration on  some aspects  of CNS
          function may  disappear before CNS AChE activity
          recovers.
          Effects on VEP, visual discrimination, and
          cortex AChE activity in monkeys


              Study of the effects of parathion  may have sig-
          nificance for  human health.  However, for that pur-
          pose, use of the monkey as a model would appear to
          be more relevant than use of the rat. For this reason,
          studies were undertaken to determine the time course
          of effects of parathion  on brain electrical activity, on
          blood and brain AChE activities, and on performance
          of a visual discrimination task.  Effects on brain elec-
          trical  activity were determined  in  three  bonnet

-------
 474
                                                                Woolley and Reiter
 (Macaco radiata) and two  rhesus (M. mulatto} mon-
 keys, each with 16 chronically implanted monopolar
 cortical surface electrodes. The indifferent electrode
 was located in bone over the frontal sinus.  The ani-
 mals were placed in primate restraining chairs during
 the  recording  sessions. VEPs were elicited by flash as
 in the study with rats, except that evoked potentials
 from six  electrode positions were  averaged  simul-
 taneously by  a Data General, Nova 1200, program-
 mable  computer. In some  cases, the averaged poten-
 tial, plus  and  minus the standard errors of alternate
 mean values (Fig. 5), were also calculated.  The be-
 havior of the  animal during a  recording  session was
 monitored via closed-circuit television.

     The effects of parathion were essentially similar
 for VEPs  recorded  in various locations in the primary
 and  association visual cortical areas. The major effect
 was  to increase latency, whereas effects on amplitude
 were variable. When 4 mg/kg parathion was  adminis-
 tered  orally  (po)  in fruit, significant  increases in
 latency were evident 3 or 4 hr later, were maximal at
 6 to 7 hr, usually showed some improvement at  10 to
 12 hr, and had returned to  pretreatment values by 24
 to 48 hr after administration (Fig. 5).
     Simultaneous  measurements of blood AChE ac-
 tivity and  VEP (e.g.,  Fig. 5) showed that,  after  4
 mg/kg paiathion, AChE activity in blood fell  rapidly
 during the first  4  hr, reached maximal inhibition by
 about 7 hr, and showed some recovery of activity by

      HOURS AFTER PARATHION
             (4 mg/kg P.O.)
 •VALUES FOR POTENTIALS ARE X ±S-E.   BLOOD AChE
                                       ^mol/ml/hr

                                           3.48
                                         (100%)
48
                               JSOjuV
                             SO msec
                   24 4/W^
                       4 V   .  ^
                                          1.81
E12  MEDIAL, POSTERIOR
                                          2.08
                                         (61 K)
                           E7 MEDIAL, ANTERIOR
                                MMU 7133 10/1/75
Fig,  5.  Averaged  VEPs with ± S.E, recorded from
the visual cortex of a monkey. Effects of parathion
were most marked 7 hr after administration. Some
recovery was evident at 24 hr, and responses were
back to normal at 48 hr.
                                    24 hr. However,  at  48 hr, when VEPs had usually
                                    recovered and the animals were obviously feeling bet-
                                    ter, blood AChE  activity had only  recovered about
                                    50% over the peak inhibited levels.
                                        An attempt was made to determine the degree of
                                    inhibition and recovery of brain AChE activity in
                                    three monkeys after administration of parathion. Bi-
                                    opsy samples weighing about 30 mg were taken suc-
                                    cessively  from adjacent areas  of parietal  cortex  2
                                    weeks before, 2  days after, and 8 or 14 days after
                                    administration of a single dose  of  3.0, 4.0, or 5.0
                                    mg/kg parathion  sc (Fig. 6). Blood samples for AChE
                                    activity were taken more frequently. Peak inhibition
                                    of blood  AChE activity occurred about 2 days after
                                    injection, which was considerably later than the time
                                    of peak inhibition after  po administration. At the
                                    times of  peak effect, blood AChE activities were 8,
                                    26, and 32% of control activities after 5.0,4.0, or 3.0
                                    mg/kg parathion, respectively. On the other hand,
                                    cortex AChE activities in the same animals were 30,
                                    64, and 102% of control values 2 days after parathion
                                    administration. The animal showing the greatest in-
                                    hibition  died at 8  days,  and  brain AChE activity
                                    showed essentially no recovery at this time. (Autopsy
                                    revealed that death was due to intestinal blockage.) In
                                    the remaining two animals, brain AChE activity show-
                                    ed no change at 14 days compared with values 2 days
                                    after parathion,  whereas blood  AChE activity  had
                                    shown significant recovery.
                                        When these studies in the monkey are compared,
                                    it is evident  that, as in the rat, the effects  of para-
                                    thion on VEPs were of relatively brief duration com-
                                    pared with effects on brain AChE activity.


                                        In previous work (Reiter et al. 1975), administra-
                                    tion of 2.0,  1.5, and 1.0 mg/kg parathion abolished
                                    performance of learned visual discrimination perform-
                                    ance in three rhesus monkeys 5 hr later and for as
                                    long as 3  to  7 days. When performance of the tasks
                                    returned after the 2.0  mg/kg dose,  the  level of per-
                                    formance remained below pretreatment values for up
                                    to 3 weeks. A dose of 0.5 mg/kg did not affect per-
                                    formance. Thus, effects of parathion on behavior last-
                                   ed considerably longer than did effects on VEPs.
                                   Possible explanations for the dissociation
                                   between changes in VEP and AChE
                                   inhibition

                                       The present findings show that after parathion
                                   administration changes in VEPs occur during the fall-
                                   ing phase of AChE activity, but that despite contin-
                                   ued inhibition of AChE activity the CNS effects dis-
                                   appear. This may be interpreted to mean  that devel-
                                   opment of adaptation or tolerance to depression of

-------
           i Between Time Course of AChE Inhibition
                                              475

        20
        10
                                                                  BLOOD  CORTEX       "~~'°
                                                      3.0 mg/kg S.C. —•	O MRA 526
                                                      4.0 mg/kg S.C. 	A	-A MRA 631
                                                      5.0 nig/kg S.C.       •	-Q MRA 97
                                                6
                    10   H    12   13
14
                               TIME AFTER PARATHION ADMINISTRATION, days
Fig.  6.  Blood and brain AChE activity after parathion
own control. AChE activity in blood was inhibited to
cortex in each monkey. Blood AChE activity showed
activity did not.
AChE  activity in the  CNS occurred  after parathion
administration in rats and monkeys. Development of
tolerance to a number of AChE inhibitors has been
reviewed recently (Bignami et al.  1975). Usually, an
AChE  inhibitor is administered and cholinergic
symptoms are immediately evident, but despite con-
tinued AChE inhibition, the symptoms soon disap-
pear.  The major present hypotheses regarding the
mechanism of onset of tolerance to AChE inhibition
(Bignami et  al. 1975), Woolley 1976) are as follows:
(1) acetylcholine (ACh) levels, which become elevat-
ed immediately after AChE inhibition, become lower
with time despite the persisting depression of AChE
activity; (2) prolonged exposure to elevated ACh lev-
els brings about a reduction in receptor sensitivity,
thus counteracting the effects of AChE inhibition and
allowing a  recovery of function; and (3) the activity
of other neurophysiological systems, either agonistic
or antagonistic, alters to counteract the overstimula-
tion of cholinergic systems.

     Some  evidence may be presented for or against
 each of these hypotheses, and still other possibilities
 remain to  be explored. AChE and the ACh receptor
 must  share some structural similarities because both
 combine with  ACh.  Paraoxon  also combines with
 AChE at the active esteratic site and probably also
 has  the structural requirements to react with the ACh
administration in three monkeys.  Each animal was its
a much greater extent than it was in biopsies of parietal
some recovery during the study, whereas cortex AChE

  receptor. It is possible that paraoxon first reacts with
  AChE to inhibit  it and permit a  buildup  of ACh,
  resulting in cholinergic symptoms. After some delay,
  paraoxon next may combine with the ACh receptor
  and in this way reduce receptor sensitivity to elevated
  ACh levels and permit a return to more normal func-
  tioning, even  though AChE is still  inhibited. Recent
  progress in the isolation of the ACh receptor may
  soon make it  possible to determine if paraoxon does
  indeed react with it.
      Still  another  explanation has been provided re-
  cently  by  the  observation that synaptosomes  pre-
  pared from rats shortly after injection of a single dose
  of paraoxon showed significantly increased release
  and synthesis of ACh. However, when a slightly small-
  er dose  of paraoxon was injected once  daily  for 5
  days for an even greater total injection, the rate of
  synaptosomai synthesis and release of ACh did not
  differ  from controls even though AChE activity was
  strongly inhibited in the treated animals (Speth et al.
   1975). The  interpretation appears to  be that  para-
  oxon  initially acts on synaptosomes to increase ACh
  release  and  synthesis,  thus  causing  cholinergic
  symptoms, but that the synaptosomes adapt to this
  effect. It remains to be seen how this possible expla-
  nation for the  development of tolerance  is  related to
   other  hypotheses presented above.

-------
 CNV  AND SEP IN  SHOE INDUSTRY  WORKERS WITH
 NEUROPATHY  RESULTING  FROM TOXIC EFFECT
 OF ADHESIVE SOLVENTS1
  R. ZAPPOLI,  G, GIULIANO, L.ROSSI, M. PAPINI, O. RONCHI, A. RAGAZZONI,
  AND A. AMANTINI

  University of Florence, Italy3
     The diagnostic utility of event-related potentials
 (ERPs) in assessing the effects of exposure to environ-
 mental toxicants  is  a question of practical interest
 (Seppaittinen, this volume). A closely related question
 is  the comparative  sensitivity of  ERPs and other
 measures  of central and peripheral nervous  system
 function uied In the diagnosii of nervous disorders.
 Whereas   previous  studies  (Gluliano et  al.  1965,
 Zappoli et al. 1966) have shown that routine EEC
 was of little help for an early  diagnosis  of inorganic
 mercury  poisoning, EEC, EMG, and conduction ve-
 locity abnormalities were frequently observed in shoe
 industry  workers  exposed  to adhesive solvents
 (Giuliano  et al. 1967,1974,1975). The present study
 was undertaken to evaluate the sensitivity of the CNV
 and somatosensory  evoked  potential (SEP)  to the
 neurotoxic effects of adhesive solvents in workers ex-
 hibiting clear electroneuromyographic signs of poly-
 neuropathy and EEC signs of diffuse brain damage.
 Methods
    Sixteen  patients (15 females; age  14-51, mean
30.9) with toxic polyneuropathy from long-term oc-
cupational exposure to  industrial  adhesive solvents
were studied. The  chemical composition of the sol-
vent  mixture  was:  n-hexane, 46.30%;  3-methyl-
pentane,  38.12%;  heptane,  8.44%;  methyl  cyclo-
1 This study was supported in part by CNR Grant No.
73.00837.04/115.6978. The authors wish to express
their appreciation  for the technical assistance pro-
vided by A. Vessari and C. Nencioni.
JG. Giuliano is associated with the Institute of Oc-
cupational Medicine,  the other authors with the
Institute of Nervous and Mental Diseases.
 pentane,  2.25%;  2-methylpentane, 2.24%; toluene,
 1.25%; benzene,  1.25% (gas chromatographic deter-
 mination).
    CNV was recorded  in six  patients, SEP in six
 patients,  and  both  EPs  in four additional patients.
 CNVi and SEPs  were also examined  in 10 normal
 volunteers (students and researchers: 7 females; age
 16-46, mean 28.7).
    All  patients exhibited hyporeflexia;  asthenia,
often associated with a clear motor deficit; more or
less severe paresthesias and hypoesthejias;headaches;
dizziness; and moderate psychic disturbances. Four
subjects  had  moderate  anemia  and liver  function
impairment. All showed EEC abnormalities indicative
of diffuse brain damage (prevalent slowing of EEC
rhythms). Electroneuromyographic examination in all
patients  indicated signs  of denervation associated
with decreased nerve conduction  velocity. The maxi-
mal motor conduction velocity (MMCV) was consid-
erably reduced in all patients. The maximal sensory
conduction velocity (MSCV) was  in the lower normal
range or borderline in 10 patients; in 6, a more severe
decrement was detected. EEC and electroneuromyo-
graphic data were within normal  limits in all control
subjects.

    A standard SI (click>S2 (50-dB intermittent
tone burst terminated by button pressing) paradigm
was used to elicit the CNV. Eighty to 100 trials with
fixed  1.0- or  1.5-sec interstimulus interval (ISI) and
variable  10- to 60-sec intertrial  interval (ITI) were
presented. EEC was recorded at F3, F4, C3, C4,and
Cz with a 6-sec time constant. In all subjects, electrodes
were referred to ipsilateral mastoids (Cz to linked ear-
lobes) during the first part of the experiment, and to

-------
CNV, SEP and Adhesive Solvents
                                                                                                        477
        vvfiZf^^*^^
        ,i^?!*^v^^
        frftf&sfi*^^
         e  V-EOG
         7 AV.CH. 6
                                              SI     S2   RT
         8 AV.CH.  3
                                                  AV.n-8
         9 AV.CH. 4
         10  AV.CH.  5
         F& A Typical data records for patient.  Female, aged 32, with severe industrial-solvent polyneuropatky, diffuse
         muscular atrophy and clear EEC and electroneuromyographlc abnormalities: excessive diffuse theta waves; more
         marked MMCV (MN 39.3 ml sec; normal 57.65 ± 6.72 - LPN practically inexcltable) and lest severe MSCV (MN
         52.2 m/sec; normal 63.39 ± 4.04) decrement. Patient unable to walk when examined. From top: (1) fingertip
         photoplethysmogram.  Signal and time marker: SI (click} - 1.5 sec - S2 (imperative tone); (2) electromyogram
         of acting muscles; (3),  (4), and (5) EEC recordings; (6} left V-EOG; (7), (8), (9) and (10) on-line averages of
         eight  consecutive artifact-free  trials of channels as indicated. 800-msec presttmulta analysis. TC * 6 sec. At
         19th trial, RT to S2 c. 150 msec; mean RT of eight trials: 214 msec. Normal vasomotor response to trial.

-------
478
linked mastoids during the second part. The first type
of montage has  proved  useful for  stable demonstra-
tion  of  possible  hemispheric  CNV  asymmetries
(Zappoli  1978).  Vertical  EOG,  finger  photo-
plethysmogram,  EMG of responding muscles, stimuli,
and reaction time (RT) were also  recorded.
    Averaging of EEC  and V-EOG  over  trials was
obtained by  a signal analyzer with a four-channel
input.  For  monitoring purposes, this signal analyzer
was modified  by  connecting it to  four of the poly-
graph  channels to provide on-line averaging on pri-
mary   records.  Additionally,  polygraph  data were
stored  on  analog  FM tapes  for off-line  computer
analysis.
    Sets of 8 or 16 artifact-free trials were averaged
in S-sec epochs.  CNV  was measured as  the  mean
amplitude of a 100- or 200-msec epoch preceding S2
relative  to an 800-msec pre-Sl baseline.
                                  Zappoli et al.

    The bandpass of EMG amplifiers was 15 Hz to 1
kHz or 10 kHz. Output  was summated by a Neuro-
averager (model 1172) triggered by the  stimulator.
The Neuroaverager offers the possibility of selecting
the epochs to be averaged and rejecting the ones that
are contaminated by artifacts. A single epoch is dis-
played about 3 sec, during which the operator may
decide whether to load it into the summation mem-
ory. If not, the epoch will be automatically erased.
An analysis time of 200 msec was used.
    The SEP data were stored on analog FM tapes for
further  off-line analysis. SEPs were averaged over 50
or  100 artifact-free responses.  Peak latencies and
peak-to-peak amplitudes  of successive components
were measured.

    The significance of differences in CNV, SEP, and
RT measures between patients and controls was eval-
uated by the Student's t-test.
    SEPs were elicited by rectangular electrical shock
of 100-^sec duration, delivered transcutaneously over
the median nerve (MN) at  the  wrist for upper limb
and over the lateral popliteal nerve (LPN) at the head
of the fibula for  lower  limb. Stimulating electrodes
consisted of silver discs with the cathode placed 3 cm
proximal to the anode. Stimulus intensity was liminal
for a weak muscular twitch.  ISIs were irregular, 4 to 6
sec. SEPs were recorded through needle electrodes in
the somatosensory  area contralateral to stimulation
and referred to linked earlobes.
Results
    In all 10 patients, it was very easy to elicit CNVs
over all areas explored with what may be considered
normal characteristics (Fig. 1) or characteristics very
similar to those elicited from control subjects of com-
parable age  range. For patients, CNV latency with
respect to SI (start of the negative deflection relative
to baseline) ranged from 362 to 407 msec, mean 378;
for control subjects, the range was from 358 to 413,
                 Table 1. Latencies and Amplitudes (±SD) of SEP Components for
                 Contralateral Stimulation of Median and Lateral Popliteal Nerves
                                           (MN and LPN)


Peak
latency,
msec


Amplitude
(peak to
peak),^V

Peak or
component
IPP
N1
P1
N3
P3
IPP-N1
N1-P1
N3-P3
Normal subjects (N=10)
MN
14.66(±1.66)a
19.5 (±1.35)
24.4 (±3.23)
57.2 (±3.48)
77.4 (±6.98)
2.96 {±1.1 1)a
4.4 (±1.36)
7.05 (±1.62)
LPN
25.0(±4.18)a
30.9(±3.21)
36.8(±3.55)
64.4(±5.14)
86.8 (±11. 36)
2.04{±0.56)a
2.57 (±0.82)
7.97(±3.9)
Patients (N=10)
MN
14.55(±1.59)8
19.9 (±1.35)
25.4 (±2.95)
56.4 (±4.55)
80.6 (±5.97)
3.33 <±1.56)8
5.0 (±2.92)
9.68 (±3.03)
LPN
26.0(±2.76)b
31.0(±2.0 )
36.8(±3.64)
64.2(±4.24)
86.8(±6.34)
2.81(±0.74)b
3.83(±2.23)
9.2 (±5.33)
            nine wbjectt; IPP absent In one.
 t>M«aiurM for ilx p.tl«ntt; IPP abient In four.

-------
CNV, SEP and Adhesive Solvents

 mean 372; the difference was not statistically signifi-
 cant. Amplitude, form,  total duration of ONVs, and
 CNV resolution were  normal,  allowing for typical
 individual  variability (the difference  in CNV ampli-
 tude and duration between patients and controls was
 not significant). These CNVs were without clear and
 constant asymmetries in the various  derivations uti-
 lized.  In many of these patients, we observed surpris-
 ingly low  RT to S2 (Fig. 1). The mean RT for all
 operant responses of the 10  patients was 212.4 msec
 (range 106 to 315); the  corresponding value for con-
 trols was 198.7 msec (range  118 to 348). The differ-
 ences between patients and controls were not signifi-
 cant. In two cases with rather severe polyneuropathy,
                                           479
    IPP PI       W
  20msec
 »^
 20msec  '
                                         LPN
Fig.  2.  Typical SEP in normal subjects to contr-
alateral stimulation of MN and LPN (average of 100
artifact-free responses). Only consistent early and late
deflections of the SEP are  labeled. Positivity down-
ward. A or Pj, needle electrode fixed over the post-
central MN focus contralateral to stimulation: about
7 cm from the midline and  2 cm behind the coronal
interaural plane. Pi  or  P^, LPN contralateral scalp
focus: approximately J to 10 mm from midline and 2
cm posterior to the vertex.
                P3
IPP PI ,
  I/-U/
   *w^
                                       f
                                   NM
    _
   20 msec

                                                                     A
                                                               II I
                                                                      "3
                                                        -- j^   ~ -
                                                        20msec

                                                          a FEMALE. AGE 32
                                  LPN  Pj-2A
                                                          i	i
                                                         20msec
                                                          b.  FEMALE, AGE 14
                                            LPN
Fig. 3.  SEP in patients with most severe toxic poly-
neuropathies, a.  Female, age 32 (same patient as Fig.
I).  SEP to MN contralateral stimulation, peak laten-
cies in msec:  IPP, 15.1; Nl, 21.2 (at upper limits of
normal range); PI, 24.2; N3, 52.1 ;P3, 70.2 (both at
lower limits of normal range).  SEP to LPN stimula-
tion:  IPP, 25.9; Nl, 34.2 (at upper limits of normal
range); PI, 45.2 (borderline for a little longer laten-
cy); N3, 64.8; P3, 83.2.  All amplitudes of the three
components measured in normal range values,  b.  Fe-
male, age 14.  Worked at  home with adhesive sol-
vents without precautions.  Severe polyneuropathy.
Excessive diffuse slowing of EEC rhythms. MMCV
at the Mn 46.9 m/sec (normal 57.65 ± 6.72) and at
the LPN 32.7 (normal 53.9 ±  4.68).  MSCV at the
MN  impossible to  measure as NAP at elbow  was
practically indistinguishable from noise. SEP to MN
stimulation:  IPP, 17.3; Nl, 21.4; PI, 29.9 (all at
upper  limits  of normal range); N3,  53.1; P3, 74.9
(both at lower limits of normal range).  SEP to LPN
stimulation:  IPP, 22.3;Nl, 32.4;PI, 35.7;N3, 60.8;
P3, 92.1.  Amplitudes of SEP components were all
within normal range.

-------
 480

 RT ranging from 275 to 315 msec was observed. Fair-
 ly  normal vasomotor responses to trials were present
 in  all  patients (Fig. 1). These  results were  obtained
 also in the cases that showed the greatest EEC abnor-
 malities (marked slowing of EEC).
     Table 1 shows latencies and amplitudes for early
 and late components of SEPs to MN and LPN contra-
 lateral  stimulation in normal subjects and patients.
 Fig. 2  shows typical SEPs obtained  in normal sub-
 jects. A small initial positive potential (IPP), extreme-
 ly variable in amplitude, was found in nine subjects to
 precede the  Nl  component both to MN  and LPN
 stimulation with  a  mean latency,  respectively,  of
 14.66 ± 1.66 and 25.0 ± 4.18. In order of appearance,
 the  more consistent early and late deflections ob-
 served with MN and LPN stimulation were: Nl, PI,
 N3.P3.
     IPPs were absent in one patient with MN and in
 four with LPN contralateral stimulations. Relative to
 normal subjects, none of the patients showed  clear
 differences in latency or amplitude of components
 measured (Table 1). None of the small latency and
 amplitude differences observed between patients and
 controls reached significance (p > .2)  Patients with
 the most severe toxic polyneuropathies and the great-
 est decrement of both MMCV and MSCV at the elec-
 troneuromyographic examinations  (Fig.  3)   were
 found  to have,  at the maximum, latencies of some
 components at the  upper limits  of the normal range
 of values measured in control subjects.
Discussion
    CNV and SEP patterns observed in patients with
toxic poly neuropathy and EEC signs of brain damage
resulting from industrial adhesive solvents were with-
in normal limits  relative to control subjects. These
results,  therefore, lead us to express a negative judg-
ment concerning  the clinical usefulness of CNV and
                                   Zappoli et al.

 SEP measures for early diagnosis, especially at a sub-
 clinical stage, of toxic effects of these  solvents  on
 central and peripheral nervous system function.

     When the  significance generally  attributed  to
 CNV activity is  considered it may be assumed on the
 basis of CNV data that these patients were not affect-
 ed  by important  disturbances of arousal-vigilance
 mechanisms, attentional capacities, or information
 processing and cognitive functions. No  other specific
 diagnostic tools were  used to  disprove the possible
 existence  of minimal disturbances of such complex
 functions. During clinical  evaluation,  however, we
 never  observed  a  real impairment of  these psychic
 parameters.
     As expected, neither SEP measure supplied any
 useful data for diagnostic purposes. It is well known
 that  clear  changes  in  SEP components (i.e., longer
 latencies and decreased amplitude) are observable in
 different types of  polyneuropathy that  are accom-
 panied by severe impairment of the peripheral senso-
 ry pathways, marked  MSCV decrement, damage  of
 the dorsal column system (Giblin 1964,Bergaminiet
 al.  1965, Halliday  1967). On the other hand, the
 toxic polyneuropathies resulting  from adhesive sol-
 vents are generally  characterized by a marked decre-
 ment of the MMCV and a much smaller decrease  of
 the MSCV, without clear neurological signs of dorsal
 column damage. We also observed cases with marked
 decrement of the MMCV and with MSCV within the
 normal  range or borderline. In these  forms of poly-
 neuropathy, therefore,  for early diagnosis (at times
 even at  a subclinical stage), evaluation, and longitudi-
 nal follow-up study, we consider electroneuromyo-
 graphic  measures   still  useful,  especially   MMCV.
 MSCV and  EMG  signs of muscular impairment are
 considered much less useful. Our present results, how-
 ever,  do not exclude  the possibility  that,  in  other
 forms of environmental toxicant central and peripher-
al neuropathy, particularly if the afferent pathways
are  severely damaged, studies  of  scalp-recorded
evoked potentials  may  be  a  helpful  diagnostic
 technique.

-------
ELECTROENCEPHALOGRAPHIC ANALYSIS
OF SUBACUTE  EFFECTS
OF METHYL-PARATHION IN THE MOUSE


 G. ZAPPONI, C. LINDSEY1, AND A. LOIZZO

 Laboratori di Farmacologia e Tecnologic Biomediche, Istituto Superiore di Sanita,
 Roma, Italy
     NeurotoxJc symptoms induced by large doses of
 anticholinesterase-type  pesticides  have  been  cor-
 related with biochemical alterations,  coniisting  of a
 reduction in cholinesterase activity and consequent
 increase in brain acetylcholine levels  (Bignami et al.
 1975). However,  there is no clear evidence of cor-
 relation between any long-lasting biochemical effect
 of single small doses of these drugs  and behavioral
 impairment.  It is also difficult  to  show  the  bio-
 chemical effects of small chronic doses until a state of
 severe intoxication is reached. In view of the impor-
 tance of the cholinergic systems in electrogenesis, re-
 cording of brain electrical activity has been used as a
 sensitive parameter to show discrete alterations of the
 functional state of the brain induced by organophos-
 phate  pesticides (Loizzo and  Longo,  1977). Electro-
 encephalographic (EEC) changes demonstrate central
 effects  more clearly than peripheral effects of drugs.
 Visual examination of EEC records yields qualitative,
 descriptive results; however, time series analysis, using
 power spectral techniques, allows quantitative evalua-
 tion of spontaneous nonparoxismal cerebral electrical
 activity.
     In previous experiments using these methods, a
 dose-related effect on hippocampal EEC of the rabbit
 was described  for  far-trans- tetrahydrocannabinol
 (THC) (Willinsky et  al. 1975). The minimal effective
 dose  of THC  (50 /ug/kg iv) was  equivalent to the
 threshold dose for psychic effects in man (Isbell et al.
 1967).  Using laboratory minicomputers, the method
 was improved  to demonstrate the depressant effects
 induced  by  an  anticholinesterase-type pesticide,
 methyl-parathion,  on the hippocampus of the rabbit
 at doses as low as O.OS to 0.1 mg/kg im (Loizzo et al.
 1976). During these  experiments no gross behavioral
 changes were observed in the animals.

 *Now  at  the Department of Psychobiology, Escola
 Paulista de Medicina, S&o Paulo, Brasil.
    A few studies have been undertaken to determine
the minimal toxic  levels of parathlon and related
compounds  using electrophysiological methods.
Santolucito  and Morrison (1971) recorded the elec-
trocorticogram  from anaesthetized monkeys which
had been receiving 0.3 mg/kg/day per os of parathlon
for 18 months. These authors found a reduction in
both high-voltage slow waves and low-voltage fast
waves, as well as Increased synchrony between hemi-
spheres. No significant alterations were found in mon-
keys fed  0.1 mg/kg/day of parathion. D&1 (1973)
reported  a  significant increase  in  electrical  brain
activity in rats fed for several weeks a diet containing
bromophos in doses as low as 100 ppm (equivalent to
10 mg/kg/day; LD50 for bromophos is about 1650
mg/kg po).  After 40 days treatment,  this dose also
induced a significant decrease in blood cholinesterase.
Finally,  Revzin (1973) found a dose-related decre-
ment in septo-hippocampal evoked responses in mon-
keys challenged with single doses of mevinphos, rang-
ing from 0.05 to 0.20 mg/kg im. No peripheral signs
of poisoning were seen at these doses.


    Investigations were initiated in our laboratories
to contribute to the establishment of threshold limit
values for substances  acting on the central nervous
system, and  also  to explore ways to  detect, at  the
earliest possible moment in biological time, changes
occurring in animal organisms from exposure to toxic
substances in the environment.
Methods
    Thirty DBA/2 male  mice  weighing 22  to 24
grams were chronically implanted with four cortical
electrodes  (anterior and  posterior  sensorimotor

-------
 482
                                                                      Zapponi et al.
                                               TABLE 1
                                     TREATMENT SCHEDULE
   Test Group
 No. of
Animals
Days 1,2
Days 3-5
Day 6
Control 1
Control tl
Exposure 1
Exposure II
Exposure III
Exposure IV
Exposure V
Exposure VI
4 Water 40 jul po Water Water
4 N
4
3
3 Sesame Oil
4
4(3)
4(3)
40 M! po Sesame Oil s
0.125
0.25
0.5
1.0
2.0
4.0

mg/kg methyl-parathion
administered po in
solution with sesame
oil (total volume)
<40Ml



Sesame Oil



 cortex). The mice were randomly assigned to control
 and experimental groups as indicated in Table 1. The
 testing cycle  consisted of six consecutive daily ses-
 sions  with  all mice receiving control treatments  of
 distilled water or sesame oil on days 1 and 2, experi-
 mental  groups  receiving treatments  of  methyl-
 parathion as shown in Table 1 on  days 3-5, and  all
 mice receiving control treatments on day 6.
     During each daily session, 120-sec epochs of EEC
 were recorded at 30 min intervals for two hours be-
 fore treatment and again at 30, 60, 120 and 240 mins
 after treatment.  Ninety dB intermittent noise was
 presented during recording  epochs to arouse animals.
 EEC signals,  bandpass  filtered between O.S and 40
 Hz, were recorded on analog tape for off-line com-
 puter (PDF-12) analysis. Sampling frequency was 128
 Hz over 8-sec epochs.  Using standard Fast  Fourier
 Transforms, power spectra, ranging from 0 to 63.5 Hz
 with  0.5-Hz  discrimination, were constructed and
 stored on digital  tape. On the  basis of a preliminary
 correlation  analysis  between  frequencies  (for
 methods, see Loizzo et al.  1978,  Zapponi et al.
 1978), six frequency bands were chosen: 0.5 to 3.5,4
 to  7, 7.5 to 12, 12.5 to 16, 16.5 to 20, and  20.5 to
 40 Hz. Mean power in the frequency bands was evalu-
 ated during 2-min time samples (corresponding to 12
 to  16 consecutive  artifact-free  power spectra).  Mean
 values,  representative  of pretreatment,  solvent-
 treatment,  or drug-treatment states, were compared
 using two-way analysis  of variance and the Duncan
 test (Senter 1965). Differences among dosage, time
 effects,  and animals  were evaluated. Absolute values
 and normalized values (the latter computed as a ratio
 of  treatment  to pretreatment  mean values in each
 animal) were used to reduce inter-animal variability.
Results


    Mice treated with water, solvent, and lower doses
of  methyl-parathion (0.125,  0.25,  0.5,  and  1.0
                                     mg/kg) resumed normal exploratory behavior in the
                                     cage  after a few  minutes  of immobility  following
                                     intubation. Mice given larger doses showed more evi-
                                     dent changes in gross behavior (long-lasting hypomo-
                                     bility, piloerection, sometimes lacrimation). An inver-
                                     sion of the weight increment  curve was also observed
                                     at the  end of treatment. Two animals out 'of eight
                                     died the third day of treatment, one showing a cere-
                                     bral abscess in the right posterior sensorimotor cor-
                                     tex.
                                        On visual inspection, EEC records appeared nor-
                                    mal in control animals and in animals given 0.125 to
                                    1.0 mg/kg (cf. Loizzo 1969). In animals treated with
                                    2.0 and 4.0 mg/kg, the  records sometimes appeared
                                    slower in frequency, although the amplitude and mor-
                                    phology  of waves appeared unchanged. No seizures
                                    were ever seen.
                                        Computer analysis showed that the smallest dose
                                     of methy-parathion (0.125 mg/kg) induced a short-
                                     duration enhancement of power in two bands of the
                                     spectrum (7.5 to  12  and 20.5 to 40 Hz). This en-
                                     hancement,  however,  was not significant.  Doses of
                                     0.25, 0.5, and 1.0 mg/kg showed a biphasic effect;
                                     i.e., these doses induced an enhancement of power in
                                     the 7.5 to 12 Hz range during the first 1 to 3 hr after
                                     administration followed  by a depression of power in
                                     the same frequency band lasting up to 24 hours. The
                                     administration of the second dose of the drug, after a
                                     brief inconsistent enhancement of power in the same
                                     part of the spectrum, accentuated the  trend towards
                                     depression. Fig. 1 shows the effect of 0.25 and 0.5
                                     mg/kg of methyl-parathion on the 7.5 to 12  Hz fre-
                                     quency band of the spectrum  in three animals. Ad-
                                     ministration of 0.25 mg/kg produced a significant de-
                                     pression in the mean frequency of the spectrum (cf.
                                     Hjorth  1975)  during the  first hour after treatment
                                     [control mean = 11.65 Hz, treatment mean =1053
                                     Hz; 2-way ANOVA:F(1,11) = 11.89, p <.01]. A
                                     linear regression analysis of the cumulative  effect of

-------
EEC Analysis of Methyl-Parathion in the Mouse



                                   A. 0.26 mg/kg
 10€-
        C   t  2hr   3hr
                              JU
 100-
24 hr T 26 hr
                                    B. 0.5 mg/kg
        C   tlhr
24hrt26hr
Fig. 1.  Temporal effect  of two doses of methyl-
parathion on power in the 7.5 to 12-Hz range of EEG
in mice.  Power is normalized with respect to predrug
control.  Asterisks indicate  significant effect  (p  <
0.05) relative to control.

0.5 mg/kg methyl-parathion 24 hrs after treatment on
three consecutive days indicated an increasing depres-
sion in the absolute power  of 4-7 Hz EEG activity
(Table 2). Note also the high within-subject variabi-
lity. Higher doses (2.0 and 4.0 mg/kg) induced a con-
stant depression  in  several  frequency bands. Both
total power and  the mean  frequency power values
were reduced. These effects,  which started 1  to 2
hours after the first drug dose, lasted as long as 24
                                                                483

                    hours, and in general became more marked after suc-
                    cessive treatment (Table 3).
                                                      Discussion
                        Pryor et al. (1966) and  Ebel et al. (1973) have
                    shown that  different inbred strains of mice may have
                    different  cholinesterase levels in various parts of the
                    brain. However, preliminary  work in our  laboratory
                    (unpublished) showed that the LD SO of methyl-para-
                    thion in DBA/2 male  mice was 39 mg/kg po.  These
                    results  were consistent with Wills (1970), and in-
                    dicated that DBA/2 mice  were not particularly sensi-
                    tive to  the drug, at least with respect to acute toxici-
                    ty. However, since  data obtained from  inbred strains
                    may be unreliable, further data are needed. EEG ef-
                    fects induced by the vehicle of drug administration
                    should  also  be considered.
                        Our results indicate  that  methyl-parathion in
                    mice induces EEG modifications which can be detect-
                    ed  and quantified at doses as low as 0.25 to 0.5
                    mg/kg po (approximately  1 to 2% of LD50). These
                    modifications consist primarily of an enhancement of
                    power at 7.5 to 12 Hz, and are followed  by long-
                    lasting depression in the same frequency band. For
                    low doses,  behavioral effects  are minimal.  Larger
                    doses (2 to 4 mg/kg, corresponding to 5 to 15% of
                    LD50)  generally produce  more  dramatic effects
                    (long-lasting  depression) on EEG rhythms and clear
                    changes in gross behavior.


                        These effects have been observed by Wsi(1973)
                    in  rats  administered  bromophos  chronically (50
                    mg/kg/day po) and by Vajda et al. (1975) in the EEG
                    of rats given a single dose of parathion (3.5 mg/kg po)
                    and recorded 7 days after treatment. De'si also report-
                    ed  a  decrease in brain and  blood cholinesterase and
                    conditioned responses of animals after several days of
                                              TABLE 2
              Regression analysis of effects on absolute power of 4-7 Hz EEG induced
                by 0.5 mg/kg methyl-parathion 24 hours after successive treatments.
                                     SS
                       df
                                          MS
Subjects
Linear Regression
Deviation from Regr.
Error
Total
8.214,004
898.837
170.287
2.286.164
11.569.272
2
1
1
31
35
4.107.002
898.837
170.267
73.747

55.69'
12.19*
2.31 (n.t.)


 Coefficient valuu: • - 1260.2; b - -190.6; Sb -1 54.6
 •p < .001

-------
  484
                                    Zapponi et al.
                                                TABLE 3
                Effects on EEG power, expressed as percent of control values, induced
                   by 2 mg/kg methyl-parathion 1 hour and 24 hours after dose II.
                                                      Frequency Bands (Hz)
                       0.5-3.5     4-7     7.5-12    12.5-16       16.5-20       20.5-40       Total Power
1 h rafter Dow II
24hrifterDo*ell
92
64"
87*"
74**
105
49*"
82"
52**
88**
51"*
07
52**
89*
56"*
* P < .06
"* P < .01
  chronic treatment.  Others  (Bignami and Gatti  1967
  for parathion, Kaloyanova-Simeonova 1961 for chlor-
  thion, Medved et al. 1964 and Russel et al. 1961 for
  Systox,1 Lewis  et al. 1973 for mevinphos) have re-
  ported changes Induced by  very low doset of anticho-
  linesterase-type pesticides on conditioned behavior in
  animals.
     Rider et al. (1969) failed to confirm either bio-
 chemical or clinical deficits in men treated orally for
 several  weeks with 19  mg/foro/day po of methyl-
 parathion. Note that this dose, administered to a man
 weighing 70 kg, is equivalent to 0.25 mg/kg, the ef-
 fective threshold dose used in our experiments.
     Several authors (see  reviews by G. Clark 1971;
 Bignami et al. 1975; Mertens et al 1975; Woolley, this
 volume) observed that cholinergic symptoms appear
 immediately following administration  of an acetyl-
 cholincsterase inhibitor. Despite continued cholines-
 terase  inhibition,  the  behavioral  and/or clinical
 symptoms soon disappear. The results of Warburton
 and Segal (1971) and Desi (1973), however, indicate
 that tolerance to anticholinesterase effects may not
 occur for certain conditioned behaviors. Low doses,
 moreover, can produce clear signs of poisoning under
 particular dietary conditions (Casterlirie et al. 1969).
 The question  remains  which behavioral deficits oc-
 cur at low doses that do not produce somatic signs of
 poisoning.

 Conclusion

    Organophosphate pesticides vary  so  widely  in
 systemic distribution, In affinity for different  types
 of cholines terase, and  in  metabolism that findings
 with one agent in one species cannot be generalized
 to other pesticides in the  group  or to other animal
 species. However, parallel studies  on the electrophys-
 iologtcal, biochemical, and  psychological (or clinical)
 effects of these drugs on laboratory animals or men
 are  rare. Metcalf and Holmes (1969) conclude that
 long-term exposure to organophosphorus compounds
 can induce irreversible or very slowly reversible brain
 disfunction.
    Our results show that it is possible, using inter-
disciplinary techniques for the simultaneous investiga-
tion  of  several  physiological,  biochemical, and
psychological  parameters,  to evaluate  threshold ef-
fects of these drugs  on brain function in laboratory
animals. Multivariate  statistical procedures,  such  as
discriminant analysis or factor analysis (Loizzo et  al.
1978, Zapponi et  al.  1978) may  also help in  deter-
mining  the most sensitive and cost-effective neuro-
behavioral  parameters  for  future research  on the
neurotoxicity  of pesticides and other chemical haz-
ards in the environment.
           ACKNOWLEDGMENTS
    This work was supported in part by a grant from
CEE (262-77-1-ENV I). The authors are  grateful  to
Drs.  E. K. and  K. F. Killam  and V. G. Longo for
advice  and criticism during preparation of the manu-
script.  Sampling programs for PDF 12 were supplied
by Dr. G. Rosadini. The valuable  technical contribu-
tions of E.  Deodati, S.  Fortuna, S,  Palazzesi, and
A. Pezzola are gratefully acknowledged.
lDiethyl-0-(ethylthioethyl)   phosphorothioate  2:1 withdlethyl-S
-------
ENVIRONMENTAL TOXICOLOGY REFERENCES
                                            485
Agadzhanyan, N.A. and Kalyuzhnyi, L. V. Materialy
      22 soveschchaniya po probl. VND., 1969:12.
      (Rus).

Albert, R.E., Shore, R.E., Sayers, AJ., Strehlow, C.,
     Kneip, IJ.,  Pasternack,  B.S.,  Friedhoff, A.J.,
     Covan, F. and Cimino, J.A. Followup of child-
     ren overexposed to lead. Envir. Hlth. Perspect.,
     1974,7:33-39.
Allison, T., Goff, W.R., Williamson, P.O. and Van-
      Gilder, J.C. On the neural origin of early com-
      ponents  of the human somato-sensory evoked
      response. Int. Symp, Cerebral Evoked  Poten-
      tials in Man, Brussels, 1974.
Amthauer, R. Intelligenz-Struktur-Test (I-S-T). Verlag
      fur Psychologic Hogrefe, Gottingen, 1953.
      and retinal ischemia. In H. Gastaut and J.Meyer
      (Eds.), Cerebral anoxia and the electroencepha-
      logram, Thomas, Springfield, 1961.
Beard, R.R. and Grandstaff, N. Carbon monoxide and
     cerebral function. Ann. N.Y. Acad. Sci., 1970,
     174:385-395.
Beard, R.R. and Grandstaff, N.W, Carbon monoxide
      and human functions. In B. Weiss and V.G.
      Laties (Eds.), Behavioral Toxicology. Plenum,
      New York, 1975:1-26.
Beard, R.R. and Wertheim, G.A. Behavioral impair-
      ment  associated  with small  doses of carbon
      monoxide.  Amer.  J.   Publ.  Hlth.,   1967,
      57:2012-2022.
Anderson, E.W., Andelman, R.J., Strauch, J.M., For-
      tuin, N.J., and Knelson, J.H. Effects of low
      level carbon monoxide exposure on the onset
      and duration of angina pectoris: A study in ten
      patients with  ischemic  heart disease. Ann.
      Intern. Med., 1973,79:46-50.


Anokhin,  P.K. (Current problems of electrophysio-
      logical research on the nervous system). Medi-
      cine Press, Moscow, 1964, 132 (Rus.).
 Beattie, A.D., Moore, M.R., Goldberg, A., Finlavson,
      M.J.W., Graham, J.F., Mackie, E.M., Main, J.C.,
      McLaren,  D.A., Murdoch, R.M. and  Stewart,
      G.T. Role of chronic low-level lead exposure in
      the aetiology of  mental retardation. Lancet,
      1975,1:589-592.
 Behse, F., Pach, J. and  Dorndorf, W. Bleipolyneurc-
      pathie. Z.Neurol., 1972,202:209-216.
Aronow, W.S. and  Isbell, M.W. Carbon monoxide
      effect on  exercise induced  angina  pectoris.
      Ann. Intern. Med., 1973,79:392-395.
Ayres, S. M., Mueller, H.S., Gregory, J.J., Gianelli S.,
     and  Penny,  L.L. Systemic  and  myocardial
     hemodynamic response to relatively small con-
     centrations  of carboxyhemoglobin  (COHb).
     Arch, environ. Hlth., 1969,18:699.
Bates, D. and Sundt, Jr., T.M. The relevance of pe-
      ripheral barorecepton and  chemorecepton to
      regulation  of  cerebral blood flow in the cat.
      Circ. Res.,  1976,38:488493.


Batuyev, A.S., Pirogov, A.A.,  and Tairov, O.P.,  An-
      alysis of the origin of the electrical responses of
      the associative fields of the cortex, Mekhaniz-
      my  VP mozga (Evoked potential mechanisms
      of the brain), 1971,28-40 (Rus.).

Baumgartner, G., Creutzfeldt,  O.  and Jung, R. Micro-
      physiology of cortical neuronei in acute anoxia
 Bender, W., Gothert, M., Malorny, G. and Sebbessee,
      P.   Wirkungsbild  niedrigei  kohlenoxid-
      konzentrationen beim menschen. Arch. Toxl-
      col., 1971,27:142-158.

 Benignus, V.A., Otto, D.A. and Knelson, J.H. Effect
      of low frequency  random noises on perform-
      ance  of  a numeric monitoring task. Percept.
      Mot. Skills, 1975,40:231-239.

 Benignus, V.A., Otto, D.A., Prah, J.D. and Benignus,
      G. Lack  of effects of  carbon monoxide on
      human vigilance.  Percept. Mot. Skills, 1977,
      45:1007-1014.

 Bergamlni, L., Bergamasco, B., Fra, L, Gandiglio, G.,
      Mombelli, A.M. and Mutani, R. Somato-senso-
      ry evoked cortical potentials in subjects with
      peripheral nervous lesions. Electromyography,
       1965,5:121-130.
 Beritov, I.S. (General Physiology of the Muscular and
       Nervous System,) Vol. 2. Moscow-Lord-
       1948. (Rus).

-------
486
         Environmental Toxicology References
 Betz,  E, Cerebral blood flow:  Its measurement and
       regulation. Physiol. Rev., 1972,57:595-630.
California Air Resources Board. Ambient Air Quality
      Standards, Sacramento, Ca., 1970.
 Bignami, G. and Gatti, G.L. Neurotoxicity of anticho-
      linesterase  agents.  Antagonistic  action  of
      various centrally acting drugs.  In  D. Davey
      (ed.)  Neurotoxicity of Drugs, 1967, 8:93-106.
 Bignami, G.,  Rosie, N., Michalek, H., Milosevic, M.
      and Gatti,  G. Behavioral  toxicity  of anticho-
      linesterase agents: Methodological, neurochemi-
      ical and neuropsychological  aspects.   In  B.
      Weiss  and V.G. Laties (Eds.), Behavioral Toxi-
      cology.  Plenum, New York, 1975:155-215.
Bokina, A.I., and Eksler, N.D.  (Electrophysiochem-
ical analysis of the  effect of some atmospheric pol-
lutants  on the central  nervous system).  Gigiyena
i  sanitariya (Hygiene  and  Sanitation),  1973,  12,
Broadbent, D.E. Noise,  paced  performance and vigi-
      lance   tasks.   Brit.  J.   Psychol.,   1953,
      44:295-303.
California Air Resources Board. Two-day hearing con-
      ducted  on health effects of lead air. CARB
      Bull., 1975,6(10):4-5.
California Air Resources Board. Standard for lead in
      air is  unchanged. CARB Bull., 1976, 7(2):3A
Casterline, Jr., J.L., Williams, C.H. and Key, I.E. Ef-
      fect of pesticide administration upon esterase
      activities in  serum and tissue of rats fed vari-
      able casein  diets.  Toxicol.  appl. Pharmacol.,
      1969, 14:266-275.
Catton, M.J.,  Harrison, M.J.G., Fullerton, P.M. and
      Kazantzis, G.  Subclinical neuropathy in lead
      workers. Brit. Med. J., 1970, 2:80-82.

Chapanis, A. The relevance  of laboratory studies to
      practical  situations.  Ergonomics,  1967,  10:
      557-578.
Broadbent, D.E. Effects of noise on behavior. In C.M.
      Harris  (Ed.),  Handbook  of  Noise  Control.
      McGraw-Hill, New York, 1957.
Christenson,  C.L., Gliner,  J.A., Horvath, S.M. and
      Wagner, J.A. Effects of three kinds of hypoxias
      on vigilance performance. Aviat.  Space Envi-
      ron. Med., 1977, 48:491-496.
Broadbent, D.E. Non-auditory effects of noise. Adv.
      Sci., 1961,17:406-409.
Broadbent, D.E. Decision and stress. Academic Press,
      New York, 1971.
Brown, I.D. The measurement of perceptual load and
      reserve capacity.  Trans.  Assoc. Indust. Med.
      Off., 1964,14:44-49.
Buchsbaum, M.D. and  Silverman, J. Stimulus inten-
      sity control and  the  cortical evoked response,
      Psychosom. Med., 1968, 30:12-22.
Burt, G.S. Use of behavioral techniques in the assess-
     ment of  environmental contaminants. In B.
     Weiss and V. Laties (eds.), Behavioral Toxicolo-
     gy. Plenum Press, New York, 1975, 241-263.
Clark,  D.L., Butler, R.A. and  Rosner, B.S. Dissocia-
      tion  of sensation and evoked responses by a
      general anesthetic in man. J.  comp. physiol.
      Psychol., 1969, 68:315-319.
Gark, D.L., Butler, R.A. and Rosner, B.S. Are evoked
      responses necessary?  A reply to Donchin and
      Sutton.  Commun.  Behav.   Biol.,   1970,
      5:105-110.
Clark, G. Organophosphate insecticide and behavior;
      A review. Aerospace Med., 1971, 42:735-740.
Coburn, R.F., Forster, R.E., and Kane, P.B. Consider-
      ation of the physiological variables that deter-
      mine the blood carboxyhemoglobin concentra-
      tion   in   man.  J.  Clin.  Invest,  1965,
      44:1899-1910.
Butollo, A. Unpublished thesis, Univ. of Vienna, Aus-
      tria, 1969.
Coburn.R.F. (Ed.). Biological effects of carbon mon-
      oxide. Ann. N.Y. Acad. Sci., 1970, 174.

-------
Environmental Toxicology References
                                            487
Colmant, H.J. Effect of low carbon monoxide con-
      centrations on the sleep and waking pattern of
      the  albino rat.  Staub-Reinhalt. Luft, 1972,
      32:84-87.
Cronbach,  L.J.  Essentials  of  psychological testing
      (2nd ed.), Harper and Row, New York, 1961.

Dargent, J. et Dongier, M. (eds.). Variations contin-
      gentes negatives. Congres et colloques de I'Uni-
      versite de Liege, 1969, 52:249.


David, O., Clark, J. and Voeller,  K. Lead and hyper-
      activity. Lancet, 1972, 2:900-903.
      Ledingham  (Eds.),  Carbon Monoxide Poison-
      ing. Elsevier, Amsterdam, 1967.
Dinman,  B.D.  Review  of electroencephalographic
      data. In:  Effects of Chronic Exposure to Low
      Levels of Carbon Monoxide on Human Health,
      Behavior  and Performance.  Nat.  Acad. Sci.,
      Nat.  Acad. Eng., SEN 309-01735-1, Washing-
      ton, D.C., 1969:29-31.
Donchin, E. and Cohen,  L. Averaged evoked poten-
      tials  and  intramodality  selective  attention.
      Electroenceph.  clin.   Neurophysiol.,  1965,
      19:325-335.
Davies, D.R. Physiological and psychological effects
      of  exposure  to  high  intensity  noise. Appl.
      Acoustics, 1968, 1:215-233.
Donchin,  E.  and  Lindsley,  D.B. (eds.)  Average
      Evoked Potentials: Methods, Results, and Eval-
      uations. NASA SP-191, 1969.
Davis, H. Principles of electric response audiometry.
      Ann.  Otol.  Rhinol.  Larygol.,  1976, Suppl.
      28:1-96.
Davison, A.N. Return of cholinesterase activity in the
      rat after inhibition by organophosphorus com-
      pounds.  1. Diethyl  p-nitrophenyl phosphate
      (E600,   paraoxon).  Biochem.   J.,  1953,
      54:583-590.
Donchin, E. and Sutton, S. The psychological signifi-
      cance of  evoked responses:  A  comment to
      Clark, Butler,  and  Rosner. Commun. Behav.
      Biol., 1970, 5:111-114.
Dressier, D.P., Shurnik, W.A. and  Bloom, S.B. Devel-
      opment of a smoke toxicity test. Toxicol. appl.
      Pharmacol., 1975,33:67.
De la Burde, B. and Choate, M.S. Does asymptomatic
      lead exposure in children have latent sequelae?
      J. Pediat., 1972, 81:1088-1091.
Dell, P. and Bonvallet, M. Mise enjue des effects de
      1'activite reticular par le millieu exterieur et le
      milieu interieur.  XX Intern. Physiol. Congr. I,
      Reviews, Brussels, 1956,
De"si,  I. Detection of early adverse effects of pesti-
      cides on  EEC  and CNS functions: The use of
      "reference substances." In M. Horvath and E.
      Frantik  (Eds.),  Adverse effects  of environ-
      mental chemicals and psychotropic drugs, Vol
      1. Elsevier, Amsterdam, 1973:125-140.
Desmedt, I.E. (Ed.). Progress in clinical neurophysio-
      logy, Vols.  1-9.  Karger, Basel, beginning 1977.
Detar, R.  and  Bohr,  D.F.  Oxygen  and vascular
      smooth muscle contraction. Am. J. Physiol.,
      1968,214:245-250.
DeValois, J.C. and Schade, J.P. An electrophysiologi-
      cal study of hlstoxic anoxia under normal and
      under  hyperbaric oxygenation. In  Blair and
DuBois, K.P., Doull, J., Salerno, P.R. and Coon, J.M.
      Studies  on  the  toxicity and  mechanisms  of
      action  of  p-nitrophenyl diethyl  thionophos-
      phate   (parathion).   J.   Pharmacol,,   1949,
      95:79-91.
 Dyer,  R.S.  and Annau,  Z.  Flash  evoked potentials
      from  rat superior  colliculus. Pharmacol. Bio-
      chem. Behav. 1977,6:453-459.
 Ebel, A., Hermetet, J.  and Mandel, P.  Comparative
      study of acetykholinesterase and choline ace-
      tyltransferase enzyme activity in brain of DBA
      and C 57 mice. Nature N.B., 1973, 242:56-58.
 Ellman, G.L., Courtney,  K.D., Andres,  Jr.,  V. and
      Featherstone, R.M.   A  new and rapid colori-
      metric  determination of  acetylcholinesterase
      activity.  Biochem. Pharmacol., 1961,7:88-95.
 Evans, H.I. and Carman, R. Scotopic vision as an indi-
      cator of neurotoxicity. Paper presented at 11th
      Rochester Int. Conf. on Environmental Toxici-
      ty, New York, June 1978.

-------
 488
         Environmental Toxicology References
 Fazlullah, S. and Ramamurthi, B. Lead intoxication
       and the nervous system. J.  Indian Med.  Ass.,
       1965,44:644-648.
 Feldman,  R.G., Haddow, J., Kopito, L. and Schwach-
       man,  H.  Altered  peripheral nerve  conduction
       velocity.  Am. J. Dis. Child., 1973, 125:39-41.
 Fisher, S.A. A "distraction effect" of noise bursts.
       Perception, 1972, 1:223-236.
 Fodor, G.G, and Winneke, G, Effect of low CO con-
       centrations on resistance to monotony and on
       psychomotor capacity. Staub-Reinhalt.  Luft,
       1972,32:46-54.
 Forbes, W.H., Dill,  D., Silva, H. and Deventer, F.H.
       The influence of moderate carbon monoxide
       poisoning  upon the ability to drive  automo-
       biles.   J.  Indust.  Hyg.   Toxicol.,   1937,
       19:598-603.
 Freeman, P.R. Table of d  and /)  Medical Research
       Council, Cambridge, 1964.
 Friedman, D., Simson, R., Ritter,  W. and Rapin, I.
      The late positive component  (P300) and infor-
      mation processing in sentences. Electroenceph.
      clin. Neurophysiol., 1975, 38:255-262.


Foster,  T.M., Creutzfeldt,  O.D. and Straschill, M.Z.
      Vergleich. Physiol., 1965,49:605.
Gastaut, H. and Tauinari, C.A. Clinical EEC, HI, Part
      A., Epilepsies. In A.  Remond (Ed.), Handbook
      of Electroencephalography and Clinical Neuro-
      physiology.  Elsevier, Amsterdam,  1975, 104.


Gelfan, S, and  Tarlov, I.M. Differential vulnerability
      of spinal cord structures to anoxia. J. Neuro-
      phyriol.,  1955,18:170-188.


Gerihuni, G.V. (Evoked potentials and mechanisms
      of discrimination of an external signal). Pavlov
      J.  Higher Nerv. Act., 1963, 13:882-90 (Rua).


GUchetti, A., Ciappi, 0. and Baccini, C. (Persistence
      of cholinesterase inhibition in rat brain after
      treatment with parathion: Pesticide residues in
      relation  to  public health.)  Ric.  Scl.  (Ital.),
      1966,36:1081.
 Giblin,  D.R. Somato-sensory  evoked potentials  in
       healthy subjects and in patients with lesions of
       the  nervous system.  Ann. N.Y.  Acad. Sci.,
       1964, 112:93-142.
 Giuliano, G,,  lannaccone, A. and Zappoli, R. Ricer-
      che  elettroencefalografiche  in  120  soggetti
      esposti al rischio di intossicazione cronica pro-
      fessionale  da  mercuric.  Lav.  Um.,  1965,
      17:293-361 (Ital.).

 Giuliano, G.,  lannaccone, A. and Zappoli, R. Studio
      elettroencefalografico  in  soggetti con  intossi-
      cazione  cronica professionale da benzolo. Lav,
      Um., 1967, 19;Suppl. 12:3-39 (Ital.).

 Giuliano, G., lannaccone, A. and Zappoli, R. Ricer-
      che elpttroencefalografiche in operai di calzatu-
      rifici esposti al  rischio di intossicazione da sol-
      vent i. Lav. Um., 1974,25:3-12 (Ital.).
Guiliano, G., Ronchi, O., Rossi, L. and Focardi, L.
      Polineuropatia tossica in operai dell'industria
      calzaturiera.  Studio elettromiografico  longi-
      tudinale. Lav. Um., 1975,27:35-41. (Ital.).


Grant, L.D. Research strategies for behavioral terato-
      logy   studies.  Envir.  Hlth.  Perspect.,  1976,
      18:85-94.
Graham, F.K. and Slaby, D.A. Differential heart rate
      changes to equally intense white noise and
      tone. Psychophysiol., 1973,10:347-362.
Grandstaff, N.  Carbon monoxide  and human func-
      tions. In  V.C. Xintaras, B. L. Johnson, and I.
      de Groot (Eds.). Behavioral  Toxicology: Early
      Detection of Occupational Hazards. HEW Pub.
      No.(N10SH)74-126,1974,292-305.

Green, D.M. and Swets, J.F. Signal Detection Theory
      and Psychophysics. Krieger,  Huntington,  New
      York, 1974.


Griffiths,  I.D.  Thermal  comfort:  A  behavioral ap-
      proach. In  D.V. Canter (Ed.), Architectural
      Psychology,  Heffer, Cambridge, 1970,  30-35.
Groll-Knapp, E., Wagner, H., Hauck, H. and Haider,
      M. Effects of low carbon monoxide concentra-
      tions on vigilance and computer-analyzed brain
      potentials.  Staub-Reinhalt.   Luft.,  1972,
      32:64-68.

-------
Environmental Toxicology References
                                            489
Grudzinska, B, Electroencephalographic patterns  in
      cases of chronic exposure to carbon monoxide
      in air. Folia Med. Carcov, 1963, 5:493-515.
Guest, A.D.L., Duncan, C. and Lawther, P.J, Carbon
      monoxide and phenobarbitone: A comparison
      of effects on auditory flutter fusion threshold
      and  critical  flicker  fusion  threshold. Ergo-
      nomics, 1970, 13:587-594.
Guy, A.W. and Chou, C.K.  System for quantitative
      chronic exposure of a  population of rodents to
      UAF  fields. Proceedings  of  the  USNC/URS1-
      IEEE  Meeting, Univ.  of Colorado,  Boulder,
      1975,37.
Guyton, A.C., Ross, J.M., Carrier, Jr., 0. and Walker,
      J.R, Evidence for tissue oxygen deiriand as the
      major factor causing autoregulation. Circ. Res.,
      1964, IS.Suppl. 1:60-69.
Haggendal, E. and Norback, B. Effect of viscosity on
      cerebral blood flow, Acta Chir. Scand, Suppl,,
      1966,364:13-22.
Haider, M.,  Groll-Knapp, E., Holler, H.,  Neuberger,
      M. and Stidl, H. Effects of moderate CO-dose
      on  the  central  nervous  system:  Electro-
      physiological and behavioral data, clinical rele-
      vance. In A.J. Finkel and  W.C. Duel (Eds.),
      Clinical Implications of Air Pollution Research.
      Publishing  Sciences   Group,  Acton,  Mass.,
      1976:217-232.
Halliday,  A.M.  Changes in  the  form  of cerebral
      evoked  responses  in  man associated  with
      various lesions of the nervous system. Electro-
      enceph.   din.  Neurophysiol.,   1967,  Suppl.
      25:178-192.
      Essays in  Toxicology, Vol. I. Academic Press,
      New York, 1969:  115-155.
 Hanks, T.G.  Human performance of a psychomotor
       test as a function of exposure to carbon mon-
       oxide.  In R.F. Coburn (Ed.), Biological Effects
       of Carbon Monoxide. Ann. N.Y.  Acad.  Sci.,
       1970:174-421.
 Hartley, L.R. Effect of prior noise or prior perform-
      ance on serial reaction. J. Exp. Psychol., 1973,
      101:255-261.
 Hatch, T.F. The role of permissible limits for hazard-
       ous airborne substances in the working environ-
       ment in the prevention of occupational disease.
       Bull. Wld. Hlth. Org., 1972,47:151-159.
Heistad,  D.D., Marcus, M.L., Ehrhardt, J.C. and Ab-
      boud, F.M.  Effects of stimulation of carotid
      chemoreceptors on total and regional cerebral
      flow. Circ. Res., 1976, 38:20-25.
 Hjorth, B. Time domain descriptors and their relation
      to a particular model for generation  of  EEC
      activity.  In G. Dolce  and  H. Kunkel (Eds.),
      CEAN:  Computerized EEC Analysis.  Gustav
      Fischer Verlag, Stuttgart, 1975.
Hockey, G.R.J. Effect of loud noise on attentional
      selectivity. Quart.  J.  exp.  Psychol.,  1970,
      22:28-36.
Hockey, R. Changes in information-selection patterns
      in multisource  monitoring as a  function  of
      induced arousal  shifts. J. exp. Psychol., 1973,
      101:35-42.
Halperin,  M.H., Niven, J.I.,  McFarland, R.A.  and
      Roughton, F.V.W. Variation in visual thresh-
      olds  during  carbon  monoxide and  hypoxic
      anoxia. Fed. Proc., 1947, 6:120-121.
 Horvath, S.M., Dahms, T.E. and O'Hanlon, J P. Car-
      bon monoxide and human vigilance: A deleter-
      ious  effect  of  present  urban concentrations.
      Arch, envir. Hlth., 1971, 23:343-347.
Halperin,  M.H.,  McFarland, R.A.,  Niven,  J.I.  and
      Roughton, F.J.W. The time course of the ef-
      fects of carbon monoxide on visual thresholds.
      J. Physiol., London, 1959, 146:583-593.
Hosko, M. The effect of carbon monoxide on the
      visual evoked potential  and the spontaneous
      electroencephalogram.   Arch,   envir.  Hlth.,
      1970,21:174-180.
Hammond, P.B. Lead  posioning:  An  old  problem
      with a new  dimension.  In  F.R.  Blood (Ed.),
Ikeda, T. Experimental carbon monoxide poisoning:
      Its electrophysiologieal  effects on  the  visual

-------
   490
       pathway  in cats. Folia Psychiat. Neurol. Jap.,
       1969,23:135-142.
 Innes, I.R. and Nickerson.M. Atropine.scopolamine,
       and related antimuscarinic drugs. In L.S. Good-
       man and  A. Oilman  (Eds.),  Pharmacological
       Basis of Therapeutics.  Macmillan, New York,
       1975:514-532.
 fsbefl,  H., Gorodetski, C.W., /asinski, D., Claussen,
       U.,  Spulak, F.  and Korte, F.  The effects of
       trans-tetrahydrocannabinol  in  man.  Psycho-
       pharmacologia (Bed)., 1967, 11:184.
 John, I.D.,  A statistical decision theory of simple
       reaction  time.  Aust,  J.  Psychol.,  1967,
       19:27-34.
 Johnson, B.L., Cohen, H.H., Struble, R., Setzer, J.V.,
       Anger, W.K., Gutnik.B.D., McDonough, T., and
       Hauser,  P.  Field  evaluation of carbon mon-
       oxide  exposed toll collectors. In C. Xintaras,
       B.L. Johnson and  I. deGroot (Eds.). Behavioral
       Toxicology,  HEW Pub. No. (NIOSH) 74-126,
       1974:306-328.
 Kaloyanova- Simeonova, F. Effet du chlorthion sur
      les reflexes  moteurs  conditionnels des  rats
      albinos. Activ. nerv. Sup. (Praha), 1961,3:284.
 Karczmar, A.G. Pharmacologic and therapeutic prop-
      erties  of anticholinesterase agents. In  W.S.
      Rood and F.G. Hofman (Eds.), Physiological
      Pharmacology,  Vol. 3C. Academic Press, New
      York, 1967:164-322.
Khachaturyan,  M. Kh. and Mitarevskaya. (Variation
      in the time parameters  of human  primary
      responses to light under the influence of low
      concentrations of capronic acid vapor.) Hygiene
      and Sanitation.  1969,6:76-80 (Rus).


Khananasvili, M.M., Kutuyev, A.B. and Silakov, B.L.
      (Evoked  potential mechanisms of the brain).
      Science Press, Leningrad, 1971:141 (Rus.).
Kimmel, C.A.,  Grant, L.D., West, G.L., Martinez-
      Vargas, C.M.,  McConnell, E.E., Fowler, B.A.,
      Woods,  J.S.   and  Howard,  J.L.   A  multi-
      disciplinary approach to the assessment of toxi-
      city associated with chronic lead. In: Proc. 17th
      Hanford Symposium, ERDA Symposium Series-
      1978, (in press).
         Environmental Toxicology References

King, B.D., Sokoloff, L. and  Wechsler,  R.L.  The
      effect of 1-epinephrine and 1-norepinephrine
      upon  cerebral circulation and metabolism  in
      man. J. Clin. Invest., 1952, 31:273-279.
Klinke,  R.,  Fruhstorfer,  H.  and  Finkenzeller,  P.
      Evoked responses as a function of external and
      stored information. Electroenceph. clin. Neuro-
      physiol.,  1968,25:119-122.
Knott, J.R. and Irwin, D.A. Anxiety, stress, and the
      contingent negative variation (CNV). Electro-
      enceph. clin. Neurophysiol., 1967,22:188P.

Kogan, A.B. (Electrophysiological indices ot central
      inhibition). In  I.S. Beritashvili (Ed.),  Garskiye
      Besedy (Awktalk), Vol. 2 (The Nature of Cen-
      tral  Inhibition),  Georgian  SSR  Acad.  Sci.,
      1956:390 (Rus.).

Kondratyeva, I.N. (Cyclical changes in the activity of
     cortical neurons after short-duration  stimuli).
     In: (Current Problems of the Electrophysiology
     of the Central  Nervous System), Moscow, Sci-
     ence Press, 1967, 148-159 (Rus.).

Kondratyeva, IN., Korolkova, T.A., Shulgina, G.I.,
     and Elkina, G.A.  (Changes  in single  unit dis-
     charge and evoked responses in the visual cor-
     tex of  a rabbit during execution  of  a  condi-
     tioned reflex).  Zhurn. vyssh. nervn. deyatel'n.
     (Journal of higher nerve  activity), 1970, 20,
     1000-9 (Rus.).

Kratin,  Yu. G.,  Isayeva, I.  Ye. Baltayeva,  A., Kur-
     banov, Sh. (The primary response  as an index
     of the functional state of the brain). In: Mek-
     hanizmy V? mozga (Evoked Potential Mecha-
     nisms of the Brain), Leningrad, Science Press,
     1971, 148-159 (Rus.).

Landrigan,  P.J.,   Baloh,  R.W.,  Barthel,   W.F.,
      Whitworth,  R.H.,  Staehlin,  N.W.  and
      Rosenblum,   B.F.  Neuropsychological  dys-
      function in children with chronic low-level lead
      absorption. Lancet, 1975,  1:708-712.

Lawther, P.J. Diskussionsbeitrag. Carbon monoxide,
      environmental  pollution  and  public health,
     European Colloquium, Luxemburg, Dec. 17-19,
      1973.

Lehotzky, K. and  Meszaros, I. Alteration of electro-
      encephalogram  and evoked  potential in rats
      induced by organic mercury. Acta pharmacol.
      Toxicol., 1974,35, No. 3, 180-184.

Lewis  J. Traffic  pollution and  mental efficiency.
     'Nature, 1970,225:95-97.

-------
Environmental Toxicology References
                                            491
Lewis, M.F., Mertens, H.W. and Steen, J.A.  Behav-
      ioral changes from chronic exposure to pesti-
      cides  used  in  aerial  application:  Effects  of
      Phosdrin in the performance of monkeys and
      pigeons  on  variable  interval  reinforcement
      schedules. Aerospace Med,, 1973, 44:290-293.
Loizzo, A. Correlati elettroencefalografici delTazione
      di alcuni  farmaci psicotropici sul topo (Mus
      musculus L.) libero non anestetizzato. 2°. BEG
      e comportamento del topo trattato con farmaci
      anticolinergici.   Studi  Sassaresi,  1969,  47
      (Suppl.), 72-91 (Ital.)
 Loizzo, A. and Longo, V.G. EEC effects of choliner-
      gic and anticholinergic drugs. In A. Remond
      (Ed.),  Handbook of EEC and Clinical Neuro-
      physiology,  Vol.  7C.  Elsevier,  Amsterdam,
      1977:72-91.
 Loizzo, A., Zapponi, G.A., Giuliani,  A.  and Longo,
      V.G.   C o m p u ter-assis ted  electroencephalc-
      graphic monitoring of pesticides effects in ani-
      mals. In M. HorvSth (Ed.), Adverse Effects of
      Environmental  Chemicals and Psychotropic
      Drugs. Elsevier, Amsterdam, 1976.
 Loizzo, A., Ceccarelli, P., Ortolani,  E., Palazzesi, S.
      and  Zapponi, G.A. (Methods of BEG classifica-
      tion in physiology and experimental pharmaco-
      logy).  Proc.  Spring  Meeting  of  Italian  EEC
      Soc., Firenze, June 8-10, 1978. (Ital.).
 Longo, V.G. Electroencephalographic atlas for phar-
      macological research. Rabbit Brain Research,
      Vol. II. Elsevier, Amsterdam, 1962.
      set in a reaction-time  task. J. Geront., 1974a,
      29:52-63.
Loveless, N.E. and Sanford, A.J. Slow potential cor-
      relates  of  preparatory set.  Biol.  Psychol.,
      1974b, 1:303-314.
Loveless, N.E. and Sanford, A.J. The impact of warn-
      ing signal intensity on reaction time and com-
      ponents of the contingent negative variation.
      Biol. Psychol.,  1975, 2:217-226.

Martinez, A.J., Taylor, J.R.,  Houff, S.A. and Isaacs,
      B.R, Kepone poisoning: Cliniconeuropathologi-
      cal study.  In  L. Roizin,  H.  Shitaki and  N.
      Grcevic (Eds.)  Neurotoxicology.  Raven Press,
      New York, 1977:443456.
McCallum, W.C.  Brain slow potential  changes  and
      motor response in a vigilance situation. In W.C.
      McCallum and J.R. Knott (eds.)  The Respon-
      sive Brain. Wright, Bristol, 1976:46-50.
McCallum, W.C. and Knott, J.R. (Eds.) Event-related
      Slow Potentials of the Brain:  Their  Relations
      to Behavior. Electroenceph. clin. Neurophysi-
      ol., 1973, Suppl. 33.
McCallum, W.C. and  Knott, J.R.  (Eds.). The Re-
      sponse Brain. Wright, Bristol, 1976.
McCallum, W.C. and Walter, W.G. The effects of at-
      tention and  distraction on the contingent nega-
      tive variation in normal and neurotic subjects.
      Electroenceph.   clin.  Neurophysiol.,   1968,
      25:319-329.
 Loveless, N.E. The effect of warning interval on signal
      detection and event-related  slow  potentials of
      the brain. Percept. Psychophys., 197S, 17:565-
      570.
McFarland, R.A. Low level exposure to carbon mon-
      oxide  and driving  performance. Arch, envii.
      Hlth., 1973,27:355-359.
 Loveless,  N.E.  Distribution of response to non-signal
      stimuli.  In  W.C.  McCallum and  J.R.  Knott
      (Eds.), The  Responsive Brain. Wright, Bristol,
      1976:26-29.
McFarland, R.A., Roughton, F.J.W., Halperin, M.H.
      and  Niven, J.I. The effects of carbon monoxide
      and  altitude on visual  thresholds.  J.  Aviat.
      Med., 1944,15:381-394.
 Loveless, N.E. Event-related potentials in selective re-
      sponse. Biol. Psychol., 1977, 5:135-149.
 Loveless, N.E. and Sanford, A.J. Effects of age on the
      contingent negative  variation  and preparatory
McFarland,   R.A.,  Forbes,  W.H.,  Stoudt,  H.J.,
      Dougherty,  J.D., Crowley, T.J., Moore,  R.C.,
      and Nalwalk, T.J. A study of the effects of low
      levels of carbon monoxide upon humans per-
      forming driving tasks. Final report by Guggen-
      heim  Center for Aerospace  Health and Safety

-------
492
      (Harvard University) to Coordinating Research
      Council and Environmental Protection Agency
      on Contract CRC-APRAC Project CAPM-9-69
      (2-70), 1972.
        Environmental Toxicology References

O'Donnell,  R.D., Chlkos, P. and Theodore, J. Effect
      of  carbon  monoxide  on  human sleep  and
      psychomotor  performance.  J. appl. Physlol.,
      19718,31:513-518.
Medved, L.I., Spynu, E.I. and Kagan, I.S. The method
      of determining the toxicity of small quantities
      of pesticides. Residue Rev., 1964,6:42.


Merlgan, W.  and Blick, D. Visual thresholds and elec-
      trophysiological responses in methylmercury-
      poisoned monkeys.  Paper presented  at  llth
      Rochester Int. Conf. on Environmental Toxlci-
      ty, New  York, June 1978.
Mertens,  H.W., Steen, J.A. and  Lewis, M.F. The ef-
      fects of nevinphos on appetitive behavior in the
      gerbil.   Psychopharmacologia  (Berl.),1975,
      41:47-52.
Metcalf, D.R. and  Holmes, J.H. EEC, psychological
      and neurological alterations  in humans  with
      organophosphorus exposure.  Ann. N.Y. Acad.
      Sci., 1969, 160:357-365.


Michael,  D.R.  and Holmes, J.H. Neurophysiological
      effects of hypoxia. In Cohen (Ed.) Biochemis-
      try, Ultrastructure and  Physiology of Cerebral
      Anoxia,  Hypoxia and Ischemia. Karger, Basel,
      1973.
Mikulka, P., O'Donnell, R., Heinig, P. and Theodore,
      J. The  effect of carbon  monoxide  on  human
      performance.  Ann.  N.Y.  Acad. Sci.,  1970,
      174:409-420.
National Academy of Science. Effects of chronic ex-
      posure  to  low levels of carbon  monoxide on
      human  health,  behavior,  and  performance.
      SBN 309-01735-1, Washington, D.C. 1969.
Nemeroff,  C.N.,  Grant, L.D., Biasette,  G.,  Ervin,
      G.N., Hairell, I.E. and  Prange, A.J, Growth,
      endocrlnologlcal,  and  behavioral deficits after
      neonatal  monosodlum L-glutamate  in the rat:
      Possible  involvement  of arcuate  dopamine
      neuron   damage.  Psychoneuroendocrinology,
      1977,2:179-196.
O'Brien, R.D, Insecticides: Action and Metabolism.
      Academic Presa, New York, 1967:55-82.
O'Donnell, R.D.,  Mikulka, P., Heinig, P. and Theo-
      dore, J.  Low level carbon monoxide  exposure
      and human  psychomotor performance.  Toxl-
      col. appl. Pharmacol., 1971b, 18:593-602.

Paul, D.D. and Sutton, S. Evoked potential correlates
      of psychophysical judgements: The threshold
      problem. A  new reply to Clark, Butler, and
      Rosner. Behav. Biol., 1973,9:421-433.
Paulson, O.B., Parving, H.H., Olesen, J. and Skinhoj,
      E. Influence of carbon monoxide and of hemo-
      dilution on cerebral blood flow and blood gases
      in man. J. appl. Physiol., 1973, 35:111-116.

Pentschew,  A. Morphology and  morphogenesis  of
      lead encephalopathy. Acta Neuropathol., 1965,
      5:133-160.

Permutt, S. and Farhi, L. Tissue hypoxia and carbon
      monoxide. In: Effects of Chronic Exposure to
      Low  Levels  of  Carbon Monoxide  on Human
      Health, Behavior and Performance.  Natl. Acad.
      Sci.,  SNB   309-01735-1,  Washington,   D.C.
      1969:18-24.

Peymer, I.A.(Time and spacial distribution of a field
      of evoked potentials and information process-
      ing  in  the  human brain).  Mekhanizmy  VP
      mozga  (Evoked  potential  mechanisms of  the
      brain), Leningrad,  Science Press,  1971, 21-28
      (Rus.).

Peymer, I.A. and  Modin, M.L. Electrophysiology of
      the Nervous System. Rusinov, U.S. (Ed.), trans.
      by Basil Haigh,  Plenum Press, New York, 1970.
 Picton, T.W.  and  Hillyard,
      evoked  potentials.  II.
      Electroenceph.   clin.
      36:191-201.
S.A. Human  auditory
Effects of attention.
Neurophysiol.,  1974,
 Picton, T.W. Hillyard, S.A., Krausz, H.l. and Galam-
      bos, R. Human auditory evoked potentials: I.
      Evaluation of components. Electroenceph. clin.
      Neurophysiol., 1974,36:179-190.
 Pitt, B.R., Radford, E.P., Gurtner, G.H., and Trayst-
      man, R.J. Interaction of carbon monoxide and
      cyanide on cerebral circulation in metabolism.
      Arch, environ, Hlth., 1979.

-------
Environmental Toxicology References

Polyanskiy, V.V.  The  relationship  between  after-
      discharges and evoked potentials in the  visual
      cortex of a rabbit. Zhurn. Vyssh. nervn. deyat.
      (Higher Nerv. Act., 1965, 15:903-1000 (Rus.).

Ponte, J.  and Purves,  M,J. The role of the carotid
      body chemoreceptors and carotid sinus barore-
      ceptors in the control of cerebral blood vessels.
      J. Physiol., 1974, 237:315-340.

Poulton, B.C. On increasing the sensitivity  of meas-
      ures   of   performance.   Ergonomics,  1965,
      8:69-76.
Provins,  K.A.,  Glencross,  D.J. and  Cooper,  C.J.
      Thermal stress and arousal. Ergonomics, 1973,
      16:623-632.

Pryor, G.T., Schlesinger, K. and Calhoun.W.H. Dif-
      ferences in brain  enzymes among five inbred
      strains of mice. Life Sci., 1966, 5:2105-2111.
Pueschel, S.M. Neurological and psychomotor func-
      tions in children  with increased lead burden.
      Envir. Hlth. Perspect., 1974, 7:13-16.
Ramsey, J.M. Carbon monoxide, tissue hypoxia and
      sensory psychomotor response in hypoxemic
      subjects. Clin. Sci., 1972,42:619-625.
Ramsey, J.M. Effects of single exposures of carbon
      monoxide on  sensory and  psychomotor  re-
      sponse. Am. Ind. Hyg. J., 1973, 34:212-216.
                                            493

      Brain Information  Service,  Brain  Research
      Institute, UCLA, 1973.
Regan, D. Evoked Potentials in Psychology, Sensory
      Physiology,  and Clinical Medicine.  Chapman
      and Hall, London, 1972.
Reiter, L., Talens, G.  and Woolley, D.  Parathion
     administration in  the monkey: Time course of
     inhibition and recovery of blood cholinesterase
     and  visual discrimination performance. Toxi-
     col. appl. Pharmacol., 1975,33:1-13.
Revzin, A.M. Subtle changes  in brain functions pro-
      duced by single doses of mevinphos (Phosdrin).
      FAA Off. Aviat. Med. Rep. AM 73-3, 1973.
Rider, J.A., Moeller, H.C., Puletti, E.J. and Swader,
      J.I. Toxicity  of parathion,  systox, octamethil
      pyrophosphoramide and  methyl-parathion in
      man.  Toxicol.   appl.  Pharmacol.,  1969,
      14:603-611.

Rolfe, J.M. The  secondary  task  as  a performance
      measure:  A survey of the literature. Inst. Aviat.
      Med., Farnborough, 474,1969.
Rosvold, H.E.,  Mirsky, A.F., Sarason, I., Bransome,
      E.D. and  Beck, L.H. A continuous performance
      test  of brain damage. J.  consult. Psychol.,
      1956, 20:343-350.
Rapela, C.E.  and Green, H.D. Autoregulation of ca-
      nine cerebral blood flow. Circ. Res., 1964, 15,
      Suppl.  1:205-211.
Roytbak, A.I. (Bioelectric phenomena in the cortex
      of large hemispheres), 1955 (Rus.).
Ray, A.M. and Rockwell, T.H. An exploratory study
      of automobile driving performance under the
      influence of low  levels  of carbon monoxide.
      Ann. N.Y. Acad. Sci., 1970, 174:396-408.
Roytbak, A.I. (Current Problems  of Electrophysio-
      logical Investigations into the Nervous System.)
      Medicine Press, Moscow, 1964,164 (Rus.).
Rebert, C.S. and Tecce, J.J. A summary of CNV and
      reaction time. In  W.C.  McCallum  and J.R.
      Knott (Eds.), Event-related Slow Potentials of
      the Brain:  Their Relations to Behavior. Elsev-
      ier, Amsterdam, 1973:  173-178.
Rechtschaffen, A. and Kales, A. A Manual of Stand-
      ardized Terminology, Techniques and Scoring
      System for Sleep Stages  of Human Subjects.
Russel, R.W., Watson, R.H. and Frankenhaeuser, M.
      The effect of chronic reduction in brain choli-
      nesterase activity on acquisition and extinction
      of a conditioned avoidance response. Scand. J.
      Psychol., 1961,2:21.
Santolucito, J.A. and Morrison, G. EEG  of rhesus
      monkeys following prolonged low-level feeding
      of pesticides. Toxicol. appl. Pharmacol., 1971,
      19:147-154.

-------
494
         Environmental Toxicology References
Schechter, G.  and Buchsbaum, M. The  effects of
      attention, stimulus  intensity, and  individual
      differences on  the average  evoked response.
      Psychophysiology, 1973,10:392400.
Schulte, J.H. Effects of mild carbon monoxide intoxi-
      cation. Arch, envir. Hlth., 1963, 7:524-530.
 Shevelev, I.A. (Functional  significance of  primary
      responses  of subcortical visual  centers.)  Zh.
      vyssh.  nervn. deyar. (Pavlov J.  Higher Nerv.
      Act.)1971,21:569-576(Rus.)
 Shilyagina, N.N. Proc.  6th,  All-union  Conf. on the
      Electrophysiology  of  the  Central  Nervous
      System, Science  Press, Leningrad,  1971,  294
      (Rus).
Schwent,  V.L. and Hillyard, S.A. Evoked  potential
      correlates of selective attention  with multi-
      channel auditory inputs. Electroenceph.  clin.
      Neurophysiol., 1975,38:131-138.
 Shugalev, N.P. Dynamics of the amplitude of primary
      and associative responses during conditioning.
      Pavlov J.  Higher Nerv. Act. 1970,  20:499-505
      (Rus.)
Segal, I., Saffer, D. and Segal, F. Diverse neurological
      manifestations of lead encephalopathy. S. Afr,
      Med.J., 1974,48:1721-1722.
 Shulga, T.M. New data for hygienic evaluation of CO
      in air. USSR Literature, 1964, 9:73-81.
Selye, H. The Physiology and Pathology of Exposure
      to  Stress.   Acta.  Inc.,  Medical  Publishers,
      Montreal, 1950.
 Shumilina, A.I. Proc. 6th, All-union Conf.  on the
      Electrophysiology  of  the  Central  Nervous
      System.  Science Press, Leningrad, 1971, 301
      (Rus).
Senter, R.J. Analysis of Data. Scott, Foresman, Glen-
      view, 111., 1969.
Seppalainen, A.M. Applications of neurophysiological
      methods in occupational medicine. A review.
      Scand. J. work, envir. Hlth., 1975,1:1-14.
Seppalainen,  A.M. and  Harkonen, H. Neurophysio-
      logical  findings among workers occupationally
      exposed to styrene.  Scand. J. work, environ.
      Hlth., 1976,2:140-146.
Seppalainen,  A.M. and  Hernberg,  S. Sensitive tech-
      nique   for  detecting  subclinical lead  neuro-
      pathy. Brit. J. ind. Med., 1972, 29:443-449.
Shute, C. and  Lewis,  P.  The ascending cholinergic
      reticular  system: Neocortical,  olfactory and
      subcortical  projections.  Brain,   1967,
      90:497-520.

Simpson, G.C., Cox, T. and Rothschild, D.R. The ef-
      fects of noise stress  on  blood glucose level and
      skilled  performance.  Ergonomics,  1974,
      17:481-488.
Sjostrand, T.L. Brain volume, diameter of the blood
      vessels  in the piamater, and intracranial pres-
      sure  in acute  carbon  monoxide  poisoning.
      Acta. Physiol. Scand., 1948,15:351-361.

Skinner,  J.E.  Neuroscience:  A laboratory manual
     W.B. Saunders, Philadelphia, 1971.
Seppalainen, A.M., Tola, S., Hernberg, S. and Kock,
      B.  Subclinical  neuropathy at "safe" levels of
      lead  exposure.  Arch,  envir.  Hlth.,  1975,
      30:180-183.


Seppalainen, A.M, and Tolonen, M. Neurotoxidty of
      long-term exposure to carbon disulflde hi the
      viscose  rayon  industry.  A neurophysiological
      study. Work, envir. Hlth., 1974,  11:145-153.
Shagut,  C.  Evoked  Brain  Potentials in  Psychiatry.
      Plenum Press, New York, 1972.
Skrebitskiy, V.G. and Voronin, L.L. Pavlov J. Higher
      Nerv. Act., 1966,16:864 (Rus).

Sluijter,  M.E.  The  treatment of carbon monoxide
poisoning by  administration of oxygen  at high  at-
mospheric   pressure.  Progr.   Brain   Res.,   1967,
24:123-182.

Smorto,  M.P. and Basmajian, J.V. Clinical Electro-
      neurography. Williams and Wflkins, Baltimore,
      1972,234.

Sokol, S. Visually evoked potentials: Theory, tech-
      niques,   and  clinical  applications.   Survey
      Ophthal., 1976,21:1844.

-------
Environmental Toxicology References
                                           495
Sokoloff, L. The action of drugs on the cerebral cir-
      culation. Pharmacol. Rev., 1959, 11:1-85.

Sokolov, E.N.    Perception  and the  Conditioned
     Reflex. Pergamon,Oxford, 1963.

Speth, R.C., Dettbarn, W.D. and Schmidt, D.E. Choli-
      nesterase (ChE) inhibition and synaptosomal
      acetylcholine (ACh) synthesis  and release fol-
      lowing in  vivo paraoxon (Px) administration.
      Pharmacologist, 1975,17:abstract 388.


Spyker, J.M. Assessing the impact of low level chemi-
      cals on development:  Behavioral and latent ef-
      fects. Fed. Proc., 1975, 34, N9,1835-1844.
Starr, A. and Achor, J.  Auditory brainstem responses
      in  neurological disease.  Arch.  Neurol., 1975,
      32:761-768.
Stewart, R.D. The effect of carbon  monoxide  on
      humans.   Ann.  Rev.   Pharmac.,   1975,
      15:409-423.
 Stewart, R.D., Peterson, J.E., Baretta, E.D., Bachand,
      R.T., Hosko, M.J. and Herrmann, A.A.  Experi-
      mental human exposure to carbon monoxide.
      Arch. envir.Hlth., 1970,21:154-164.

 Stewart, R.D.,  Fisher, T.N., Hosko,  M.J., Peterson,
      J.E., Baretta, E.D. and Dodd, H.C. Experimen-
      tal  human exposure  to  methylene chloride.
      Arch, envir. fflth., 1972, 25:342-348.

Stewart, R.D., Newton, P.E., Hosko,  M.J., Peterson,
      I.E. and Mellender, J.W. Effect of carbon mon-
      oxide  on time perception. Arch, envir. Hlth.,
      1973,27:155-160.
Stokinger, H.E. Ozone toxicity-A review of the litera-
      ture  through 1953. Arch. Ind. Hyg. Occup.
      Med., 1954,9:366-383.
Stokinger, H.E. Behavioral toxicology in the develop-
      ment of threshold limit values.  In C. Xintaras,
      B.L. Johnson and I. deGroot (Eds.), Behavioral
      Toxicology: Early  Detection of Occupational
      Hazards.  Hew  Pub. No.  (NIOSH)  74-126,
      1974: 17-20.
Sutton,  S., Tueting, P., Zubin, J. and John, E.R.
      Information delivery and the sensory evoked
      potential. Science, 1967,155:1436-1439.
Tecce, J.J. Contingent negative variation (CNV) and
      psychological processes in  man. Psychol. Bull.,
      1972,77:73-108.
Tecce, J.J. and Hamilton, B.T. CNV reduction by sus-
      tained  cognitive  activity  (distraction).  Elec-
      troenceph. clin.  Neurophysiol.,  1973, Suppl.
      33,229-237.
Tecce, J.J. and Scheff, N.M. Attention reduction and
      suppressed  direct-current  potentials in  the
      human brain. Science. 1969,164:331-333.
Tecce, J.J., Savignano-Bowman, J. and Cole, J.O.
      Drug effects on  contingent negative variation
      and eye blinks:  The distraction-arousal hypoth-
      esis.  In  M.A. Lipton, A. DeMascio  and K.F.
      Killam  (Eds.) Psychopharmacology:  A Gener-
      ation  of Progress.  Raven Press,  New  York,
      1978,745-758.
Teichner,  W.H., Arees, E. and Reilly, R. Noise and
      human  performance:  A psychophysiologlcal
      approach. Ergonomics, 1963, 6:83-97.
Tolonen, M., Hernberg, S., Nurminen.M. and Tiitola,
      K.A.   A follow-up study of coronary heart
      disease in viscose  rayon  workers exposed to
      carbon disulphide, Brit. J. ind.  Med.,  1975,
      32:1-10.

Traystman, R.J. and Carlson, S.M. Effects of hypoxic
      hypoxia  and carbon  monoxide  hypoxia on
      cerebral circulation. FASEB, 1974. (Abstract).
Traystman, R.J., Fitzgerald, R.S., and Loscutoff, S.C.
      Cerebral   circulatory  responses   to   arterial
      hypoxia in normal and chemodenervated dogs.
      Circ. Res. 1978, 42:649-657.
US DHEW. Air Quality Criteria for Carbon  Mon-
      oxide. Washington, D.C., NAPCA Publication
      No. AP-62, 1970.
Sutton, S., Braren, M. and Zubin, J. Evoked potential
      correlates  of  stimulus uncertainty. Science,
      1965,150:1187-1188.
US EPA. National Primary and Secondary Ambient
      Air Quality  Standards. Fed.  Reg., 30 Apr.
      1971,36(84):8186-8261.

-------
 496
         Environmental Toxicology References
 US PHS. Air Quality Criteria for Carbon Monoxide.
       Washington, D.C., Jan.  1969.
 Weiss,  B.  and Laties,  V. Behavioral Toxicology.
       Plenum,  New  York, 1975.
 Uttal, W.R. Do compound evoked potentials reflect
       psychological  codes?  Psychol.  Bull.,  1965,
       64:377-392.
 Whitfield, C.L.,  Ch'ien, L.T. and  Whitehead, J.D.
      Lead  encephalopathy  in adults.  Am. J. Med.,
      1972,52:289-298.
 Vajda, A., Groll-Knapp, E., Haider, M., Neuberger,
       M.,  Schmidt, H. and Steinringer, H.  Evoked
       potentials  and frequency analysis of EEC  as
       criteria for central nervous action of pesticides.
       Comm.   2°   Int.   Cong,   of  C.I.A.N.S.,
       Praha, 1975.


Vasilescu, C. Motor nerve  conduction velocity and
      electromyogram  in  chronic  lead poisoning.
      Rev. Roum. Neurol., 1973, 10:221-226.
 Wilkinson, R.   Some  factors influencing the effect
      of environmental stressors upon performance.
      Psychol. Bull., 1969, 72:260-272.
 Willinsky, M.D., Loizzo, A. and Longo, V.G.   EEC
      spectral analysis for the evaluation of the cen-
      tral effects  of  trans-tetrahydrocannabinol  in
      rabbits.   Psychoparmacologia  (Bed),   1975,
      41:123-126.
 Velasco, M., Velasco, P., Machado, J. and Olvera, A.
      Effects of novelty, habituation, attention and
      distraction  on  the amplitude of the various
      components of the somatic evoked responses.
      Int. J. Neurosci., 1973,5:101-111.
 Wills,  J.H.  Toxicity of anticholinesterases and its
      treatment.   In:  International  Encylopedia of
      Pharmacology  and  Therapeutics.  Section  13,
      Vol. 1. Pergamon, Oxford, 1970:357469.
Warburton,  D.M. and Segal, D.S. Stimulus control
      during  chronic reduction of  cholinesterase
      activity. Physiol. Behav., 1971, 7:539-543.
Weber, A., Jermini, C. and Grandjean, E. Wirk ungen
      niedriger  kohlenmonoxid-konzentrationen  auf
      die  flimmerverschmel-zungsfrequenz  und  das
      subjektive befinden des menschen. Int. Arch.
      Occup. envir.  Hlth.,  1975, 36:97-103 (Ger.).
 Winneke,  G. Kohlenoxid:    Humanexperimentelle
      untersuchungen  zur  fundierung  eines  MIK-
      wertes.  In:   Ges. Z. Ford. d.  Lufthygiene
      u.    silikoseforschung.    Girardet,    Essen,
      1974 (Ger.).
Winneke,  G. and Kastka, J.  Auswirkungen geringer
      kohlenmonoxid- konzentrationen  auf  senso-
      motorisches  Leistungsverhalten.  4.  Arbeit-
      stagung d.  Deutschen Ges. f.  Hyg. u. Mikro-
      biologie, Vortrag, Mainz, Oct. 2-4,1972 (Ger.).
Weerts, T.C.  and Lang, P.J. The effects of eye fix-
      ation and stimulus  and response location on
      the contingent negative variation (CNV). Biol.
      Psychol., 1973, 1:1-19.


Weinstein, S. and Annau, Z.  Technique for chronic
      cross circulation  of unrestrained  rats. J. appl.
      Physiol., 1967,23:601-604.


Weir,  F.W., and  Rochwell, T.H. An investigation of
      the effects of carbon monoxide on humans in
      the driving task.  Final report from Ohio State
      Research Foundation to Coordinating Research
      Council and Environmental Protection Agency
      on Contract 68-02-0379 and CRC-APRAC Pro-
      ject CAPM-9-69, Jan., 1973.
Winneke,  G.  and  Kastka, J.  Effects of  trichlo-
      roethylene  on signal detection and  auditory
      evoked potentials in man. Pro. 1st World Cong.
      environ. Med. Biol., Paris, July 1-5, 1974.
Winneke, G. and Kastka, J.  Testpsychologische und
      psychophysiologische   aspekte  der   trichlo-
      rathylenwirkung. Zbl. Bakt. Abt. I, Ref., 1975
      (Ger.).

Woolley, D.E. Effects of DDT and of drug-DDT inter-
      action on electroshock seizures in the rat. Toxi-
      col. Appl. Pharmacol, 1970,16:521-532.


Woolley, D.E.  Effects of DDT and of drug-DDT in-
      teraction on  electroshock seizures in  the rat.
      Toxicol. Appl. Pharmacol., 1970,  16:521-532.

-------
Environmental Toxicology References
                                             497
Wright, G.,  Randell, P.,  and Shepard, R.J. Carbon
      monoxide and driving skills.  Arch, environ.
      Mth., 1973,27:349-354.
Xlntaras, C, Johnson, B.L., Ulrich, C.E., TerrUl, R.E.
      and Sobeckl, M.F. Application of the evoked
      response technique in air pollution toxicology.
      Toxlcol. appl. Pharmacol., 1966a, 8:77-87.
Xintaras, C., Ulrich, C.E., Sobecki, M.F. and Terrill,
      R.E.  Brain  potentials studied by  computer
      analysis.  Arch,   envir.   Hlth.,  1966b,
      13:223-232.
Zappoli, R.  (Event related slow potentials (ERSPs)
      in neurologia.) Riv.Neurol., 1978,48:105-130
      (Ital.).
      da mercuric, Riv.  Neurol., 1966, 36:269-271
      (Ital.).
Zapponi, G.A., Ceccarelli, P. and Ortolan!, E. Alcunl
     metodi per la semplifkazione dell'  interpre-
     tazione dell' EEC. Proc. Ital. EEC Soc.,
     Firenze, June 8-10,1978 (Ital.).

Zenlck, H. and Reiter, L. (Eds.)  Behavioral Toxicolo-
     gy:    An  Emerging  Discipline.   EPA-600/
     9-77-042,  U.S.    Environmental  Protection
     Agency, Washington, D.C., 1979.
Zorn, H. A contribution to the diagnosis of chronic
      CO   intoxication.   Med.  Indust.,   1964,
      33:325-329.
Zappoli, R., Giuliano,  G. and lannaccone, A. Rilievi
      elettroencefalografici e clinic! in  120 soggetti
      esposti ad intossicazione  cronica professionale

-------
VIII.  SCALP DISTRIBUTION
       Section Editor:

       Emanuel Donchin
       Cognitive Psychophysiology Laboratory
       Departments of Psychology and Physiology
       University of Illinois
       Champaign, Illinois, U.S.A.

-------
USE OF  SCALP DISTRIBUTION AS  A DEPENDENT
VARIABL? IN  EVENT-RELATED  POTENTIAL STUDIES:
EXCERPTS OF PRECONFERENCE CORRESPONDENCE

E. DONCfflN
Department of Psychology, University of Illinois, Champaign,  IL,  U.S.A.
    This paper presents excerpts from preconference
correspondence related to the use  of scalp distri-
bution as a variable  in ERP studies. The following
people contributed to this correspondence:


    1. R.  Cooper,  Burden Neurological Institute,
      Bristol, England.
    2. L.  Deecke,  University  of Ulm,  Germany.
    3. W.J.   Freeman,  University  of  California,
      Berkeley.
    4. W.R. Goff, Veterans Administration Hospital,
      West Haven, Connecticut.
    5. T.W. Picton,  University of Ottawa, Ontario,
      Canada.
    6. W. Ritter, Herbert H. Lehman College, Bronx,
      New York.
    7. D. Ruchkin, University of Maryland School of
      Medicine, Baltimore.
    8. P. Tueting, Psychiatric Institute, New York,
      New York.
    The province of this discussion is the complex of
 technological, analytic, and interpretational problems
 encountered whenever two or more recording chan-
 nels are used in an ERP experiment. The  utilization
 of multiple channels affirms a faith in their nonredun-
 dancy. That  is,  the  investigator assumes the  data
 recorded from multiple channels will  provide more
 information  than he could gain from  recording any
 of these channels alone. The investigator expects that,
 at least on some occasions, systematic differences will
 appear in the recordings obtained at  different  elec-
 trode sites.  The  interest in  multiple  recordings, or
 scalp distribution (SD), is essentially  an  interest in
 the  differences  between  simultaneously recorded
 evoked responses.  The  differences must  be system-
 atic in order to be interesting; that is, they ought not
 to be attributable to the random fluctuations that can
be expected whenever any measurement process is
applied to any biological phenomena. In addition to
the criterion for systematicity, it is  also  necessary
that  the  differences, recognized as  intracranial, be
useful  in  resolving  some of  the  questions that
motivated  the recording of the ERP, Data become
information only if  the  data resolve some  uncer-
tainties. The implication of this last statement is that
there must be some relationship between the SD and
the independent variables manipulated or monitored
by the investigator.

    The above assertions  imply that for the success-
ful utilization of SD data, there will be at least the
following: (1) some reasonable rule for which the ex-
perimental hypothesis would, dictate the placement of
the recording electrodes; (2) procedures  for meas-
uring  the differences   between  the  waveforms
recorded   at  different  sites; (3)  techniques for
assessing  the  reliability of such differences (that is,
answers to such questions as: Is a given  difference
significantly  different from  zero? How  reliable  is
this  particular estimate of the  difference between
two  specific   electrodes?);  (4)   techniques  for
reporting SD data in a way that would communicate
to the reader both  the central  tendencies and the
variabilities of the data; (5)  finally, when an appro-
priate  picture of the scalp  distribution has been
obtained, procedures that wjll allow relating it  to
the experimental variables so as to obtain answers
to the experimental questions that motivated the SD
recording.

    There appear to be  two distinct, but closely
 related, rationales for the use of SD.  Both assume
 that the potentials  that  can be  recorded from the
 scalp  represent  the activity  of neural structures
 located inside the cranium.  Because different struc-
 tures have different patterns of activity at different
 times  and because  those  structures  are diversely

-------
 502
                                         Donchin
 located inside  the  cranium,  the potentials  that can
 be recorded at  any  point on the scalp at any time are
 quite likely to  be different from potentials recorded
 at other sites.  The differences  between scalp loca-
 tions  will   presumably  reflect   differences  in the
 activity  patterns of  the intracranial  sources. This
 is the central dogma shared by all ERP investigators,
 and  it is this central  dogma that underlies the re-
 cording of SDs.

     There are two discernible trends in the use of SD
 data. Crudely put,  one can  record SD data to make
 specific inferences about the intracranial location and
 activity pattern of neural generators. In this case, SD
 becomes  somewhat of a substitute for the stereo-
 tactically placed electrode, which one cannot readily
 use with humans. The other approach, while assuming
 a relationship  between  intracranial generators  and
 SDs, is focussed on establishing that different  SDs
 are  associated  with different values of some  inde-
 pendent  experimental  variable.  The differences are
 used as auxiliary variables in dissociating the effects
 that  experimental variables have on different evoked
 response  components.  Such  an analysis  depends
 strongly  on the assumption  that the differences in
 SDs  are  physiologically meaningful,  though  the
 specific form of the physiological relationship need
 not be known.

     As in any  classification, there is  much that is
 arbitrary, and no doubt the trends and attitudes are
 well  mixed  in most of our different investigations.
 The distinction, however, is valid, at least in the sense
 that  the investigator's primary interest strongly  dic-
 tates the  nature of  his data acquisition process  and
 the manner  in which he will assess the data. Those
 interested in making inferences  from SDs to intra-
 cranial  sources have tended to focus on "standard,"
 normative,   experimental  arrangements   in  which
 stimuli are presented under conditions which are as
 neutral as possible; the primary  interest has been
 mostly taxonomic, and the problems of measurement
 and  statistical inference, therefore,  have not been
 crucial. In most cases,  many subjects are run in a few
 fairly simple experimental conditions, the experiment
 being so designed as to minimize the between-  and
 within-subject variability. By averaging over subjects,
 an apparently standardized picture is obtained. The
 intellectual challenge in this approach is most often
 in developing rules of inference  from the SD to the'
intracranial generators.

    Investigators who try to use  scalp distribution as
an auxiliary  variable in  dissociating the effects of
experimental variables can ignore with relative impu-
nity the intracranial  generators. (Some even get away
with  to obvious a solecism  as reporting recordings
"from Wernicke's area"  when  referring to the scalp
location where  they happen to  have attached their
electrode.) On  the  other hand, the measurement
problem in this context is considerably more severe.
When  the  investigator's main  claim  is that the SD
observed  under condition A is different  than that
observed under condition B but not different than
that observed under condition C he  presumably has
a  technique for assessing the degree of  difference
between  SDs  and  has  assured  himself  that  the
difference  between A and B  is a "real"  difference
and,  what  is considerably more difficult, has con-
vinced himself that '.here is no difference between
conditions A and C.

    The previous paragraphs outline  the main issues
with  which a discussion of SDs is most likely to be
concerned,  and in fact the correspondence conducted
in preparation for the EPIC'IV conference touched
in a  variety of ways  on  all the  issues  enumerated
above. I shall use this framework to present excerpts
from the correspondence. It will become immediately
obvious that  the  correspondence is  inchoate. Many
issues  have  been very  briefly touched upon. Others,
even  though debated  in  some detail in the corres-
pondence  and  at  the  Congress are  still   far from
resolution.  Nevertheless, a remarkable consensus did
emerge on  a number of the issues. That this is the
case is best revealed through a perusal of the several
papers included in this section. They were written
after  the conference and represent a synthesis of the
writers' views prior to the meeting,  as  reflected in
their  contribution  to  the correspondence, coupled
with the effects of the discussions at  the conference.
The following pages represent: (1) excerpts of pre-
circulated   correspondence,  primarily  concerning
issues  not  covered in  postconference contributions;
(2) comments made by Cooper at Kanuga and the
response elicited from Allison;(3) papers contributed
by  Picton,  Ritter, Goff,  Freeman and  their asso-
ciates.

Preconference correspondence

Problem definition

    Donchin: Two major questions  arise  and form
the substance of this group's assignment. The first is
largely a methodological question and relates to the
manner in  which  the patterns of  spatiotemporal
variations  ought  to  be  measured.  The  problems
involved are the proper placements of the electrodes,
the manner in which the information from different
electrodes can be combined, the most economical and
effective way of expressing  the  patterns,  the most
efficient and useful ways of communicating results,
and the most appropriate statistical  procedures for
comparing patterns obtained on different occasions.

    The  second question, very substantive and dif-
ficult in nature, has to do with inferences that might
be drawn  about intraskull events from variations in
the scalp distribution of the potentials. What is the

-------
Scalp Distribution as Dependent Variable
                                             503
nature of the inferences that might be drawn legiti-
mately from variations in the amplitude distribution
of the potentials over the head? If lateral asymmetry
is  demonstrated,  can one infer differential activity
of the hemispheres? If some  component is  always
larger in the parietal leads, what can be inferred con-
cerning  the origin of this component? If the scalp
distribution of visual or auditory potentials is  dif-
ferent, can we indeed infer that either was generated
in the primary cortex? What should one make of the
fact that CNVs tend to show differential distributions
in the anterior-posterior axis?

    However we answer this set  of questions, even
if the inferences we can draw about internal brain
events from scalp distributions are minimal, the fact
that behavioral variables can consistently manipulate
scalp  distribution is of interest. I hope  that in our
discussions the utility of such variations can be easily
assessed.

Electrode placements

    Ruchkin: Correlation between  ERPs recorded
simultaneously  at   different  scalp  locations bears
upon the question  of the position  and  number of
recording sites to obtain a complete description of
ERP  distribution  upon  the scalp.  I believe that a
quantitative  answer  to  this   question  can  be
developed. The  ERP amplitudes  may be viewed as
a  function of three parameters:  time and the  two
spatial  dimensions   that  define location upon the
scalp.

    It is well known that in order to obtain a com-
plete  description of a waveform as a function of time,
it must be sampled  at a rate at least twice the highest
frequency of the waveform (in the time domain). The
same principle applies in the spatial domain. The loca-
tion of electrodes along a line in space should be such
that the average distance between electrodes is about
one-half  the length  corresponding  to  the  highest
spatial frequency  of the ERPs.  The  concepts  and
mathematics are  the same as used  for  signals as a
function of time. Instead of dealing with a graph of
a  waveform  plotted with time as  the  independent
variable (obtained  at a  single point  on  the scalp),
we deal  with a graph  of a waveform plotted with
distance along the scalp as  the independent variable
(obtained  at  a single  instant of  time).  From  an
abstract  point of  view, the  problems  of  charac-
terizing these two  waveforms are conceptually and
mathematically the same.

    I suggest that,  as part of the attack upon the
problem of how to optimize recording of scalp dis-
tributions, information be  obtained concerning the
spatial frequency spectra-in the same manner that
temporal  frequency spectra are obtained. I  do not
suggest that electrodes then be (blindly) placed at
"mathematically    optimum"  sites  on  the scalp.
Certainly as much use as possible should be made of
underlying  anatomy  and   physiology.  However,
analysis of  ERP spatial frequencies would  provide
us  with  a  statement  of the maximum  degree  of
resolution possible  and consequently with an  indi-
cation of  the number  and spacing of  electrodes
required  for a complete description of  the distri-
bution of ERPs over the scalp.

    Freeman: Assuming one can identify and isolate
an event, one can then ask: where is it located, what
is its  spatial  extent, and what is its spatial function
(variation  with distance in the intensity and/or time
of onset)? Further, what is its maximal rate of change
with distance, and, if it is periodic (in space), what is
its  maximal spatial  frequency?  If it is an event in a
neural surface such  as cortex, is the  spatial frequency
the same in  all directions, or is it a plane wave (the
difference between  a tray of pebbles and a sheet of
rods,   a  grating  or  an interference  pattern)?  As
Ruchkin states, the spatial interval between adjacent
electrodes in an array must in theory be less than half
the wavelength (one/spatial  frequency). In practice,
due to noise from various sources, it would be better
to  have  eight  or ten electrodes/wavelength rather
than  two  or three,  where  the wavelength is not
known or  is imprecisely  known, and the electrode
array   should extend over the  distance of at  least
four or five cycles if a periodic variation exists.

    Cooper:  Differential amplifiers appear  to  be
 necessary since high in-phase (common mode)  rejec-
 tion  is necessary. Thus, all  recordings are bipolar in
 the sense  that the measurement  is the difference
 between one electrode and a reference point, which
 usually has to be on the head, primarily because of
 the EKG.  Even if the EKG  is balanced  out,  the
 noncephalic electrode  can  still be influenced  by
 activity from the base of the brain just as an electrode
 on the foot still  picks up the EKG  (using the leg as a
 conductor).  If the  electrical field to be measured is
 small, say  confined to  the  occipital  region, then
 common  reference is good, but if  the  field is  large,
 the common reference  can  inject  signal. To decide
 which type of recording method  to  use, the field
 distribution must be  known; but  this  is what is to
 be determined.

    Just as EKG can be recorded from hand and foot,
 so  EEC activity, perhaps even alpha activity, can be
 picked up from, say,  the tip of the nose. Murphy's
 law  will  ensure that  the  worst  case will obtain,
 and it must be assumed that the reference is  active
 to  a  greater  or lesser extent. This  makes  inter-
 pretation very difficult.

    After many  years of trying I have come to  the
conclusion  that  there  is no way  of avoiding this
problem;  the problem can  be minimized, but  the
magnitude  of the minimum distortion  cannot be
determined.

-------
 504
                                        Donchin
     Deecke; A basic assumption  of all  potential
 mapping on the scalp is that a homogeneous dielectric
 layer  covers the  potential generators. However, the
 skull  varies in thickness. Since bone has the highest
 resistance of all the layers between cortical surface
 and  scalp  electrode,  differences in skull  thickness
 can result  in amplitude  differences. There are only
 few data in the literature, but it has been shown that
 a slightly thicker  skull can cause a marked attenu-
 ation, e.g.,  of the  alpha  rhythm. We measured the
 skull  thickness  of 51  skulls of an anatomical collec-
 tion  at  our recording locations (C3, C4, P3, Pz, P4)
 and found that the lateral parietal skull bone was 20%
 thicker  than the midparietal  one. At  the vertex, the
 skull  is  still thinner than midparietally  (posterior
 margin of the former fontanella). It is quite certain
 that  this factor contributes to  the frequent vertex
 maximum  of evoked  potential components.  Other
 factors  are  the massive  accumulation of cortical
 tissue of both hemispheres, including mesial cortex in
 the depth of the sagittal fissure, the  favorite locus for
 potential pickup in volume conduction.  Using data
 available in  the  literature, we  estimated a  possible
 attenuation  of 19.2% at  P3 and P4 as compared to
 Pz for potentials in the alpha range.

 Measurement of differences

     Tueting: I have been looking at  evoked potential
 waveforms  from different sites on the head (mono-
 polar with  linked earlobe reference) long enough to
 be rather perplexed about what is  really going on.
 For example, when  identifying a component by
 shape, size,  and approximate  latency, the latency of
 the component may change systematically from elec-
 trode  to  electrode  in a given plane.  Two interpreta-
 tions of  this observation, however, seem possible: (1)
 the component  actually does shift  latency  over the
 head; (2) the reciprocal masking of  overlapped com-
 ponents  having  different topographical  distribution
 may appear as a latency shift over the head.


    At issue is whether it is physiologically or physi-
 cally possible to conceive of a generator producing a
 component  that varies slightly in latency from elec-
 trode to  electrode in a given plane.  What kind of
 generator would  it be-a  widely distributed cortical
 generator triggered  at  different  latencies in separate
 areas?  A  discrete generator producing latency shifts
 because of  its orientation to  the surface? An alter-
 native  explanation  that components from any given
 electrode  site  are  overlapped  and  masking  one
 another (and that we  are simply seeing more or less
 of a given component  at an electrode site depending
 upon  the relative  amplitude  contribution  of these
masked  components)  can  be  tested.  Presumably,
experiments  could be  designed to  either  enhance
or eliminate  one  of the components of the compo-
site. Or,  with a good hypothesis of the theoretical
 latencies of the components, could principal compo-
 nents analysis unravel the components?

     The approach to  this  issue is important. I think
 it could affect the validity of measurements reported
 or used in principal component analysis. If N100 am-
 plitude  is considered and the computer is set  to
 measure the most negative peak between 90 and 110
 msec (at all electrode sites for all subjects  for  all
 conditions), the  results will not be valid if there are
 actually  two  components, a slightly more frontal
 N105 and a slightly more parietal N92.

     In  view of the  above, the  decision  of setting a
 latency search window, either for a human 1-mm-box
 counter or  for a computer, should probably not  be
 taken lightly.

     Deecke: After  the most  admirable efforts  of
 Freeman aiming  towards (and coming very close to)
 a  complete description of stimulus-dependent elec-
 trical events in a certain brain system, it is hard for
 the rest of us to continue evoked potential research
 in the old-fashioned way. However, let me enter here
 a  plea  foi simplicity. I think  that  the human eye-
 brain  system is  still a  very   good apparatus  for
 potential evaluation (with its superb ability to per-
 ceive invariances, Gestalwahrnehmung,  etc.), with
 only one  disadvantage: It is  always the  eye-brain
 system  of the  investigator that is biased.  Since we
 were aware of this, we tried to  force ourselves  to
 "double-blind analysis"  not only when comparing
 patients with normals, but also, for instance, in the
 evaluation of  laterality  differences.  This  should,
 in my opinion, be considered a general rule for scalp
 distribution research as well as preaverage editing
 and the  use  of  a  sufficient  number  of subjects
 (N  >  20)  because there  are  subject-specific dif-
 ferences.

 Problems of statistical assessment

     Donchin:  To get  the  discussion rolling, I shall
 make a  few provocative remarks. With  few excep-
 tions, we have all been treating the problem of scalp
 distribution much too simple-mindedly.  What  we
 normally do is measure the amplitude of components
 at different scalp loci, pften using only two "homo-
 logous"  scalp  locations such as C3 and C4. Broad
 generalizations about lateiality or the preponderance
 of  parietal  activity  are then based  on the  relative
 amplitudes at two such electrode locations. This has
 been dictated by instrumental  and  analytic restric-
 tions. However, the field  is rapidly  outgrowing its
 dependence on simple averagers as the use of sophis-
 ticated   computing  and  data  acquisition  devices
 becomes widespread. Therefore,  we ought to develop
an  approach to measurements of parameters of the
distribution of  the  potential  over  the head that
capitalizes on  this  increased data acquisition  and

-------
Scalp Distribution as Dependent Variable
                                             505
computational power. The problems involved in the
evaluation of distribution of a quantity in time, over
a two-dimensional space,  are  often  encountered in
geography, meteorology,  or  field-potential analysis
in physics. Should we not find out how these disci-
plines  deal with  the problems and  try  to emulate
what is useful?

    The statistical problems are formidable. To take
but one example, if you compare potentials recorded
simultaneously at different scalp  locations you can-
not, in all honesty, perform tests  that assume that all
measurements  represent  independent   samplings.
There  is considerable  correlation between different
electrode sites, and  these  correlations ought  to be
taken  Into account  in the description of the  distri-
bution. But  there are serious problems of presenta-
tion. What is  the  best way  to  convey your  ideas
about  a distribution? Do we need contour maps?
Do  we  need conventions for presenting  the data?
Are plots  of the type  Remond (1968)  published
necessary?

     Ritter: My inclination is, wherever  possible, to
use  the simplest,  most straightforward analysis and
presentation. If, for example, an investigator merely
wishes to  establish  asymmetry  for  a given class of
events (e.g.,  monaural stimulation, unilateral  motor
activities, language stimuli), then the use  of several
judiciously  placed  pairs  of  electrodes may suffice.
The demonstration  of asymmetry is  probably  best
accomplished  by  measuring  components  in  the
standard manner.

     To  conclude that two components  do not have
different distributions, however, it is not sufficient to
record from only a few sites. An interesting illustra-
tion is provided  by the negative wave elicited by  a
missing  stimulus.  We found  this component larger
centrally than parietally for both missing tones and
flashes (Simson et al.  1976), as was found by  Picton
et al.  (1974) for missing clicks.  But as  we reported
at  Brussels,  the negative  wave for missing flashes  is
different than that  for  missing tones  in that the
 former  has  two foci, one  near the vertex  and  a
 second, of even greater magnitude, on the lateral sur-
 face of the  scalp near the  border of the occipital
 lobe.  This component, then, appears to be modality
 specific, and the only way we could have established
 that was by placing  a fairly large number of elec-
 trodes across the scalp (13 in this instance).

     Goff: I certainly agree  that more sophisticated
 types of analyses should be tried  to express topo-
 graphy data, but the fact that they are more sophis-
 ticated  does not  automatically  mean that they are
 better for  the particular purpose. With regard to
 correlations  between different electrodes, a descrip-
 tion  of topography is essentially  a description of
 correlations   between  electrode  locations.   The
problem  is that,  so far, it has seemed practical to
describe the correlations only for amplitude.

    Responding now to Tueting,  both of her inter-
pretations  as to  the cause of shifts in component
latency are probably applicable to most components
most  of  the  time.  The reciprocal masking of over-
lapping components is probably the dominant cause
of what  appears  to be shifting component latency.
This certainly seems to be the case in our topography
study. To cite  but  two examples  where components
are known to be  masked by other  components:
(1) Under barbiturate anesthesia in man, the SER P4S
wave  drops out, revealing a large  primary negativity.
(2) The   negative  component  of the somatic  late
response   (SLR)  is  apparently  masked  in scalp
recordings by the vertex potential. An actual example
of  two  components  of approximately  the same
latency  is given  in our somatic topography paper
(Goff et  al.  1977) where the frontally distributed
P100 wave occurs about the same time as a parietally
focussed  P100 wave. We believe  that the  frontal
component is extracranial, while the  parietal com-
ponent is the scalp  reflection of the positivity of the
SLR.

    Donchin: The  problem of using scalp distribu-
tions as dependent  variables is not easily dismissable.
When the differences are large,  they may be easily
observable even with a few electrodes. We detected
laterality  differences in the motor potential using
two electrodes (Kutas  and Donchin 1974).  We can
now  evaluate P300 and slow wave distributions  with
three electrodes (Squires, K. et al.  1977). Fortunately,
however, in these cases the differences were obvious.
The issue, as Ritter points out, becomes more com-
plex  when one records from three electrodes and fails
to  find a difference. This is analogous to an earlier
period  of  CNV  research  (Donchin   1973)  when
investigators felt  that it  was sufficient to record
from the  vertex. Even if no differences are found
between  vertex  ERPs, it is  possible that there are
differences between frontal  and parietal  recordings,
one or the other dominating in different experimental
conditions.  The  vertex seems to  be  the pivot  on
which  the  frontal-parietal axis  rotates.  It  is  con-
 ceivable that the favored Fz-Cz-Pz axis may  show no
differences in distribution, while  the distributions are
in fact  quite  distinct lateral  to the  mldline. The
 problem  is  common to "negative" results; if you
 find no differences you never know whether (1) the
 differences do not exist or (2)  your measurements
 were  taken  in  an inappropriate  place,  time, or
 manner.


     Freeman:  Donchin's point is well made regarding
 use of spatial distributions of potential either as vari-
 ables dependent  on time-varying sensory, motor, and
 central event  states or as independent variables  for

-------
  506
                                         Donchin
  localization of "sources" with respect to anatomical
  landmarks. However, in both cases, there is an under-
  lying assumption that the neural event ("potential,"
  slow wave, fast wave, "bump,"  peak,  component,
  X-complex, ERP, SEP, etc.) is a unitary event (com-
  parable to a single action potential assignable to one
  neuron at an  unequivocable  location), or that a
  complex is the sum  of a set of such bumps. However,
  the  CNV or P300 wave  is probably  not a unitary
  event, but consists rather of  a set of similar  events
  overlapping in  time and  space, each  event having
  some  characteristic  waveform,  and  the weighted
  sum of their  potential fields giving  rise to one or
  more bumps at the scalp. Each  member of the  set
  of  events  might  involve  a  substantial number  of
  neurons, e.g.,  104  to 107  comprising part of a popu-
  lation; and on each  trial  in a series of events, the
  fraction and/or  location  of that part of the popula-
  tion may shift, or  it may  shift  with time during each
  trial of the event.  So it is reasonable to try  to ascer-
  tain  that the maximal  observable potential of a given
 sum of events is located in or over area 4, 7, or 17,
 although  the   underlying  neural  events   remain
 undefined and therefore  unlocalizable-volume con-
 ductor theory notwithstanding.

     The chief problem lies in the empirical descrip-
 tion  or definition of a "bump" as a certain function
 of potential in  time, as it is observed from a recording
 electrode near  a field maximum. The oscillograph
 screen  photograph or  ink tracing gives us a certain
 gestalt, and we tend to "see"  the same waveform  in
 records  from nearby points. This is  the "bias"  to
 which Deecke  referred. This is especially true  if the
 number of points is small  (four or less) and the spec-
 tral  distribution of the event is narrow, e.g.,  a "slow
 wave" or a "burst."  If larger numbers of electrodes
 (e.g., 32  to 64) or recording sites  for  sequential
 observations (e.g.,  100 to  500) are used, most such
 events  (in my experience) change shape  in a con-
 tinuous manner with distance  across the event, and
 the  gestalt  is  submerged  in  variance. I  think the
 success of Donchin et al. (1975)  in differentiating
 P300 and  CNV may be  due  in  part  to the  small
 number (two  or three) of electrodes  used. For the
 same  reason (that our mind's eye locks onto a simple
 waveform and induces us to see it again), we tend to
 decompose complex  waveforms into a sequence of
 "bumps" and then spin our wheels trying to "localize
 the source" of each of them.


    The analysis of principal components, in which
one  uses  location  in  an  observable complex as  a
stratifying variable to tease out the overlapping wave-
forms in a superimposed set of events, is  one alter-
native  way to decompose  the complex. One of the
two  main difficulties is that each component in an
overlapping set  may have a waveform that is causally
unrelated to its neighbors  but  is statistically related
  (in the sense that a somatosensory  and an auditory
  evoked potential may  be diphasic and  have a non-
  trivial  correlation  coefficient,  because  they  have
  similar mechanisms and waveforms and "look alike").
  In  this case, they  cannot be separated by a linear
  model. The second  difficulty is that neural  mass
  events  move in  the tissue, but the  electrodes  are
  fixed. Factor analysis and related techniques cannot
  distinguish  between variance  due to changes in  the
  location and intensity of activity. Each "factor" may
  be  a peculiar gemisch that fails to give clear informa-
  tion about  either the  time-varying locations or time
  courses   (time-dependent  amplitude)  of  neural
  events.

     Picton:  I have been hoping that someone would
  provide  a  simple  statistical  approach to  finding
  differences in scalp distribution. I do not feel compe-
  tent to propose such a statistical technique, and  I
  might be stepping  where  angels  fear to tread  since
  the more mathematically  inclined  members of  the
  correspondence  have been amazingly quiet in  this
  respect.

     Could one not apply  the principal components
  analysis described  by  Donchin (1966),  substituting
  a scalp location dimension for the time dimension?
  This could  be done by selecting  one, or a small
  number of  time points  and evaluating scalp distri-
  bution  for  these points. It might then  be possible
  to  determine significant differences in scalp distri-
  bution at these selected latencies using procedures
  described  in the first  part  of Donchin's paper and
  to describe  the  scalp  distribution differences using
  the   technique  of  principal  components  analysis
 described in the second part of that paper. However,
 it becomes   quite  difficult to arbitrarily  preselect
 one or several latencies, and it  would  be nice  to
 analyze the  data over both latency and scalp distri-
 bution.

     Having  raised  the   possibility  of  multivariate
 analysis, I  feel bound also to express my reservations
 concerning  these techniques. I must  admit that  if
 the access  to such data crunching devices were easier
 and  if my  understanding  were  deeper, perhaps I
 might  be less cautious. Generally, if the effect one
 is demonstrating is  large,  this  can be statistically
 proven  using fairly  simple  statistical  techniques
 rather  than the more powerful multivariate analyses.
 My  general approach has, therefore,  been to create
 the  largest possible effect and to  use  the  simplest
 possible statistical tests.

Distortion of signals in signal enhancement
 techniques

    Cooper:  Averaging is a poor method of recording
evoked  potentials. This is particularly  so in multi-
channel recording, since it is very easy to assume that

-------
Scalp Distribution as Dependent Variable
                                            507
if signals from two  brain regions, say Cz and  Fpz,
change  together in  the  average, then they change
together in  the  individual  trials. This need not be
so,  as  the spatial  distribution might  change during
the collection of the average-and  this is  what we
are trying to measure.

    More attempts  must be made to extract  data
from  individual  trials  by  correlation or  pattern-
recognition  techniques or, better still, by designing
experiments that give large amplitude responses that
can be seen  in the original record. From some recent
work   in  Bristol  (Weinberg  and  Cooper 1972,
Weinberg and  Papakostopoulos 1975),  I believe
this can be done. Averaging techniques, demanding
artificial repetition  of the  event,  as  they do, set
an experimental framework that is  at  variance  with
the  realities  of the  external  world where  "one
trial learning" is common.

Inaccuracies in the measurement of data

    Cooper: There has been discussion on the use of
statistics in  our work, and I agree with both Donchin
and Picton: Donchin for his desire to improve  relia-
bility in the face of variability and Picton for wishing
to  avoid using them  by looking for "sore thumb
effects." Both are right, of course, but we must  NOT
use statistics like a drunk and a lamp post; we  must
use them for illumination, not for support. We  must
use them, as it were, to  reduce the variability rather
than  improve the constant factor.  Let us  use  them
to design better experiments by showing where the
source  of variability lies so that we can reduce  it by
changes of paradigm or recording conditions.

    Freeman: The definition of the components in
an  empirical event  and  their localization to sets of
neurons requires a deep knowledge of the structure
and dynamics  of the  region  in  which the  event
occurs, to the extent that the locations, patterns of
spread, and time courses  can be predicted for par-
ticular  components. This view is set forth in  detail
in  Mass Action in  the Nervous System  (Freeman
 1975). It is, to be sure, a long way from the olfactory
bulb  to the association neocortex, but I will maintain
that  unless you develop and use a theoretical basis
for predicting waveforms from local neural anatomy,
topology, and dynamics, you will be wallowing in
principal components indefinitely. Of course, one of
the best, if  not the  best, sources of ideas in  the
development of predictions is the result of compo-
nent analysis, so my challenge is not directed toward
its use as a means, but  toward Us being regarded as
an end.

       Picton has  brought  into  better  focus  my
view of one of the roots of the difficulties we face
in the analysis of ERPs recorded at the scalp and pre-
 sented a possible way  to deal with the difficulty
directly.  The  sense of confusion  and uncertainty
stems from the lack of clear-cut templates (engrams,
gestalts, archetypal patterns) of what the components
of such events "ought" to look like.

    I perceive now an area of neglect in these studies,
and a positive approach to this difficulty. It begins
with the careful and painstaking parametric testing
of a specific event  in a small number of individuals.
That is, a stimulus-bound event should be established
in a set of well-defined and fixed  circumstances, and
the stimulus intensity (or perhaps some other of its
parameters) should be  varied  in  small steps  from
threshold on prolonged averaging to a supramaximal
level  at  which the form  of the  event is  fixed for
further increase. At each step, an adequate number of
events should be taken to determine (initially) the
intrinsic variability for fixed experimental conditions.
Thereafter, at  each step,  if sensory stimulation  is
being used, the event should be studied with pairs of
stimuli to determine the superposition characteristics
of the response. If at some step or over a small  range
of steps the event or a significant part of it displays
the superposition property, and if the input is a linear
function such as a pulse, step, or  sine wave, then the
event (or that part  of it) can be described as the sum
of a  set of linear  functions (exponentials, damped
sine waves, etc.). That is, there is a clear-cut expecta-
tion of what the components "ought"  to look like.

    At this point, the  task is to fit  a curve to the
event  waveform, i.e., the number of components,
their types, and the approximate values of their para-
meters  must be determined by  graphic techniques.
The  Fourier  transform of the event by  numerical
integration is often useful  here. When a curve has
been fitted to  one response from one individual at
one step, it becomes  a template for a set of responses
at that step. (An average of the set is often optimal
as the basis for the first template.) The uncontrolled
variation takes  two forms. If the same equation can
be  fitted to every response, the variation is expressed
 in  the continuous distributions  of the parameters.
 If some of the responses can  only be  fitted by the
 addition or  deletion of  other basis  functions (i.e.,
 using a different equation),  then  there  are one  or
 more discontinuous changes in the event.


     The procedure of fitting is extended to responses
over  a  range of steps and over a class of individuals.
The variation again occurs either by  variation  in the
 parameters or by differences in the set of basis func-
 tions. The result is the emergence of a set of response
 templates  for  a specific  set  of experimental  condi-
 tions. These  will  give the  basis  for  describing in
 detail the number and kinds of intrinsic variation, the
 types and properties of  the "noise,"  the degree of
 complexity of the responding system (in terms of
 the number and kinds of basis  functions), and the

-------
 508
                                         Donchin
 dimensions of the variance space; they will also pro-
 vide a basis  for  research and classification of indi-
 vidual  responses  that were  averaged to provide the
 data base for  analysis.

     The linear  response domain may be quite small,
 but my experience has shown that it can be extended
 quite broadly by  piece-wise linearization. Further, in
 each curve-fitting  procedure there are residuals, which
 often provide crucial insights into ways to restructure
 the templates. One should also expect to be sent back
 to the experiment  to redefine  the conditions so as
 to assist in the emergence of acceptable templates. In
 my experience, I have not really "seen"  an  evoked
 response clearly until I have tried to fit a curve to it.
 The  difference between  my intuitive expectation
 expressed  as  an analytic  function and  what actually
 has been recorded, when  plotted in the same  coordi-
 nates, has impressed me  with how faulty the intui-
 tions can be.

     One may suppose that this procedure may not be
 appropriate for  the complicated and seemingly highly
 nonlinear  and time-varying scalp-recorded responses
 to sensory  input.  However,  my students and I have
 begun with quite similarly appearing evoked  poten-
 tials  from  the  prepyriform cortex, olfactory bulb,
 hippocampus, superior colliculus, and cerebellum,
 and in each  case we have found  and extended  a
 linear domain. The neocortex and its sensory path-
 ways  are   unquestionably  more  complex,  but  I
 believe  that  the  stepwise procedures  for  pattern
 development  are  feasible  and  should  be explored.
 Certainly this is one systematic and well-defined way
 out of a morass of loosely structured empirical data.
 It  does not  require a  priori,  knowledge  of the
 dynamics of the responding systems, and in fact gives
 information about the dynamics that animal experi-
 menters may  not have, may crucially need, and may
 not be able to get from animals.

    Spatial analysis enters in two ways. At the out-
 set, the collection  of events  from multiple sites
 helps to determine the choice of a particular location
 for intensive analysis, because it is simpler in form,
 reproducible  across  individuals, or optimally con-
 forms  to  the superposition principle.  Later, when
 one has a  template, one can use it to determine the
 amplitude  and  latency distributions at the surface
 and begin to explore the overlapping of two or more
 events, perhaps  initially fused into  one template. As
 I have written earlier, one must have a template as a
 time  function before one  can  determine a  spatial
 distribution.

    In summary,  there is a systematic way  of de-
veloping clear-cut  expectations  of what  an  event
"ought" to look like. It begins with parametric and
paired-ttlmulus  testing,   and   proceeds   through
curve-fitting and the study of the variations  of the
 parameters. At each stage, it refers back to the experi-
 mental situation  for modification and  verification.
 Properly  used, it  can  provide  a rigorous basis for
 deciding what  is signal and what is  noise. That prob-
 lem  is  probably   the  major obstacle  in  applying
 multivariate  statistical analysis  directly  to  the digi-
 tized data. Finally, it does not depend on the  devel-
 opment of dynamic  models of  cortex, and indeed
 may  materially assist in  the  formulation  of  such
 models by animal experimentalists.

    Plcton: Freeman's excellent  comments  consider
 the problem of noise.  He suggests "supramaxlmal"
 averaging in  order  to determine any spatial distribu-
 tion function. This is most applicable to the analysis
 of the relatively constant responses such as the audi-
 tory brainstem potentials or the early somatosensory
 components. The later evoked potential  components
 will,  I  believe, be highly  refractive  to  such a sys-
 tematic  analysis since, as Freeman points out, they
 are "highly nonlinear and  time-varying." Even  such
 constant responses as the auditory brainstem poten-
 tials  have been reported  to vary over prolonged
 periods of time (e.g., Amadeo  and Shagass 1973).
 The problem of obtaining  the template  waveform is
 especially relevant to the very late components, such
 as the  P3, which  are  highly  affected by inevitably
 changing perceptual attitudes even under fairly rigidly
 defined circumstances. If one accepts relatively noisy
 data as inevitable, how big are the problems involved?
 If, instead of the  supramaximally averaged template
 waveform, one uses a relatively noisy waveform such
 as the vertex record, how  bad is any derived spatial
 function?

    This complete  derivation of the spatial function
 of the waveform is probably most needed when one
 is extrapolating to underlying  generators.  Perhaps
 a simpler and immediately more important problem
 is to  compare  the  spatial distributions of a compo-
 nent under two different conditions to  see whether
 they  are  indeed significantly different.  If so, one
 can then state with some  degree of certainty that
 they  must be  differently  generated. This in itself,
 even  without  the knowledge  of  the   underlying
 generators, is  an important conclusion. What  then
 is the most appropriate  analysis, and how does one
 compensate for the different amplitudes under the
 two conditions?

    Freeman:  The curve-fitting  approach,  which I
suggested  earlier  and  Picton  commented  on,  is
 particularly applicable to highly variable  late compo-
nents of ERPs.  The method requires two stages. In
the first stage,  a large  set of ERPs  is  averaged to
remove as much  of the  noise  as possible, while
removing  the  variability of the ERPs  as  well.  A
curve  is fitted to the  supramaximal average  that
constitutes a  matched   filter.  This same curve  is
then fitted to each ERP in the set or to subaverages

-------
Scalp Distribution as Dependent Variable
                                            509
of them by allowing the parameters in the equation
to vary;  i.e.,  the  filter  is now adaptive. The fits
are not as close as the superaverage, but they should
not be, because noise is  present in the subaverages.
The  end  result is that  the variance  in  the set of
ERPs is expressed in the matrix of coefficients of
the fitted curves.

    One way of looking at the procedure is to assume
that  the ERP is an information carrier that is modu-
lated  on  each trial and  is also obscured by back-
ground EEC. Supramaximal averaging permits us to
specify the  form  of the carrier, which is the same
across  trials (e.g.,  it is always a sinusoidal burst or
a  step), but the  details  change  on each trial (its
frequency, amplitude, delay, rise time, decay time,
etc.).  The superimposed  EEC obscures  the signal,
but the equation gives an expected generic form in
the search for  the signal.  The curve-fitting procedure
gives  a "best  estimate"  of the particular form of
each signal  while  separating it from  the "noise,"
i.e., the residuals (Freeman 1975, ch. 7).

    Goff: Using dipole analysis and attempting to
establish  the location of sources by application of
field potential theory is  irresistible. But  among the
problems  is one  aptly  put  by Freeman,  namely,
that  we  tend  to decompose  a complex waveform
into  a sequence of bumps and then try to  locate
each  one.  However, I do not know  how  else to
begin.  The problem  can be  approached in  two
ways. One way, the one that Vaughan  and others
use,  is to make assumptions  about the  location of
the source and then see if the empirical data seem
to fit. The inverse method is to attempt to determine
from   the  empirical  data  where  the  source  is,
assuming,  and this  is the crux of the matter,  that
there is a  single equivalent dipole generator, In some
cases  (e.g., postcentral gyrus primary activity), this
is probably a valid assumption, but in most cases we
simply do not know.  I  agree  with Picton that the
Henderson et  al. (197S) paper is excellent, particu-
larly because  of the elegant in vitro simulations of
dipole sources. We have  been using a very similar
technique in collaboration with Sidman.


    We now use a three-sphere model, which takes
into   account  the  different  resistivities of brain,
skull, and scalp, and find that it gives good agreement
with  data  where the location is known  in advance.
This method, like the Henderson et al. (1975) techni-
que, is based on the Wilson-Bayley equation (Wilson
and  Bayley  1950), although  we  use  a different
algorithm and think our method  is more stable  in
some situations.  But the  big question  is, do you
trust  the  answer  the computer gives you? There is
no objective way  of determining  this, but  you do
get  a number that  amounts to a goodness-of-fit
measure  and  at least gives some assurance that the
observed topography is consistent with the assump-
tion of  an  equivalent dipole located where the
method says  it is.  We are now using this technique
on  some of  the depth electrode data, but whether
it will  help us understand this complex set of data
remains to be seen. In any case,  these techniques
are  important and are worth pursuing.

Distortion of signals from brain to scalp
     Cooper   No  measurement represents  the true
value of a parameter since the measuring instrument
distorts the signal. However, this is not a major prob-
lem, as the accuracy we desire is only a few percent.
Our problem arises because we are forced to work
at  a distance from the source of the signals and do
not know the  transmission  characteristics of the
tissue  in between. Nor do we know the  shape and
amplitude of the sources of the signal.

     The one thing that we do know is that we are
dealing  with  a  system  that obeys fundamental
chemical and physical laws and that these laws are
inviolate in   the sense  that  they  must not be used
when the data fit into them and ignored when the
data do  not  fit.  They are there for all time and all
conditions of the experiment.

     If there  is a paradox (and there are many), then
the data collection and assumptions made  during the
interpretation of the data must be critically examined
since something must  be wrong.  For example, con-
sider acoustic responses  that apparently come from
deep structures by some kind  of volume conduction.
How can these data be reconciled  with the lack of
volume conduction that  makes  Freeman and our-
selves  use electrodes  with  spacing of a  few milli-
meters?  Or  how can the interpretation of the AEP
over  the central  fissure  as a dipole be  reconciled
with work showing that the so-called phase reversal
is not quite  180e but is a time shift and that one of
the components can be selectively blocked by certain
maneuvers?  Let us be clear that when we  talk  about
a dipole we  are stuck with a dipole and that it must
always  show the characteristics of a  dipole, such as
the rapid  attenuation of the field  with distance.
Why do  cortical  electrodes on the  surface  of the
frontal lobe  just above the eye show no sign  of the
blink artifact that is such  a  nuisance on the scalp
immediately   above these  cortical electrodes? Or
why can some large amplitude cortical activity not
be  seen  on the scalp immediately above the area of
cortex  involved? Scientific discovery is often made
by people realizing that  data and theory do not
fit-let us be  rigorous in our approach to this problem
and perhaps  we too can discover more about why and
how we pick  up EEC signals.

     In 1965, we in Bristol published our ideas about
scalp/cortex  relationships,  declaring that there  is
virtually no  volume conduction in  the brain  tissue

-------
 510
                                         Donchin
 and  that  scalp  activity is  only obtained  when  a
 relatively large area of cortex immediately below the
 scalp  electrode is involved  in synchronous activity
 (Cooper et al,  1965). We  still believe in these con-
 cepts,  since they  are the  only  ones that we can
 think of that explain our  data. Yet they are not
 satisfactory to explain the data presented by Vaughan
 and  Ritter (1970)  on the  sources  of the auditory
 responses or to explain the presence of the auditory
 brainstem  potentials  on the  scalp, both of which
 depend  on a great deal  of volume conduction. Or
 can the paradox be resolved in another way? Because
 we pick up the EKG in the foot, we  do not think that
 the heart is in the leg. Could it be that we are using
 the structures of the brain (ventricles?) as conductors
 to the source of the potential?

     Detailed measurement  of potentials might give
 us the answer but would it not be better to 'measure
 the direction of flow of current since  this is what we
 are  after-the  localization  of the changes in ionic
 currents? And what sort of attenuation do we get
 from artificial generators that are implanted in the
 brain? And how do the various  types of tissue and
 fluid between  scalp and cortex attenuate eye blink
 potentials?

    Only when these basic questions can be answered
 will we be  able to extrapolate the scalp data into cor-
 tical localization with more certainty.

    Donchin: I should  explain  in  more detail my
distinction   between  the  use  of scalp  distribution
as a dependent variable and its use for source locali-
zation.  Consider  for example  our  work  on the
lateralization  of  the motor  potential. We  were
interested in the degree to which lateral asymmetry
will  vary with the  hand used by  the  subject. As
it  happens,  it does (Kutas  and  Donchin  1974).  It
is   perfectly  reasonable  to  conclude  from  the
 nature  of  these  asymmetries  that  the  readiness
 potential is  large over the hemisphere contralateral
 to the  responding  hand,  and  from  our general
 knowledge  of the  motor control system, to  infer
 that  the readiness potential  may have  something
 to do  with  activity  in  the  motor  cortex.  Note,
 however, that we would be  hard pressed  to provide
 convincing  evidence that the potentials  we  record
 indeed  come  from  the motor cortex.  This  is  a
 speculative  inference  that is,  and this  is the  main
 point, not  truly necessary for the main  purpose of
 the  experiment,  which was  to determine  if there is
 a relationship between  the responding  hand  and
 scalp distribution. This is relevant for such questions
 as the  degree  to  which this "readiness"  activity
 represents nothing but generalized arousal (which
 ought not to be  differentially distributed among the
 hemispheres).  The  lateral asymmetry of  the  readi-
 ness potential differentiates  it quite nicely from the
 CNV.  Thus,  the scalp  distribution in  this case is
 used  merely  as  a  dependent variable,  allowing us
 to choose among competing hypotheses.

    Similar  use of the scalp distribution appears in
 such tasks as those described  by Simson et al. (1977),
 which allow differentiation between N100 and N190.
 Or the  demonstration in the paper by  Donchin et
 al.  (1975)  that   the  CNV  and  P300  have  quite
 different  scalp distributions.  In  the  latter paper,
 we again make no inferences about the  localization
 of the source of either the CNV or P300. All we are
 interested in  finding out is  whether the  two  have
 the same or  different scalp distributions. We feel
 of course that different scalp distributions  imply
different sources  and therefore that the two compo-
 nents are  different.  Whether  or not the mechanisms
for inferring the specific source are available, the fact
that the two are different is sufficient to make them
useful dependent variables.

-------
PROBLEMS  IN USING VOLUME CONDUCTION THEORY
TO LOCALIZE EVOKED-POTENTIAL  GENERATORS


R. COOPER
Burden Neurological Institute, Bristol, England
    One of the most important reasons for measuring
scalp distribution of evoked potentials is to determine
the brain regions that generate the potentials, for this
information is required to formulate theories of brain
function.

    The basic mechanism underlying the EEC is the
movement of an electrical charge (ions) across the
semipermeable membranes of neurons  in  the cere-
bral cortex. This ionic movement gives rise to current
flow in the external  medium  and to potential dif-
ferences that can be measured with voltage amplifiers.
The voltage measured, EEC or EP, depends upon the
distribution of current, which in turn depends upon
the relative  impedances of the  infinite number of
current paths available. Location of cerebral sources
is only possible if the distribution of current (or
voltage) is known  in  three dimensions. Inferences
of location based on scalp distribution are dependent
upon  assumptions  of  the  impedances of  various
tissues through  which  the current flows. Little rele-
vant quantitative work has been done, and most
theories are based  on potential  distributions deter-
mined  from  intracranial  electrodes in man and
animals (e.g., Cooper et al. 1965). The  main  finding
of  this  work was  rapid  attenuation with distance
from  the apparent generator, a common finding,
exemplified by  the high spatial frequencies referred
to by Freeman and Gerbrandt in this  volume. Attenu-
ation is probably due to the relatively high impedence
of tissue (mainly membranes) compared with the low
impedance  of  extracellular  fluid.  In   effect, the
generator  is  surrounded  with  an  insulator  that
restricts  the  field  to  the  immediate  extracellular
space.  Measurement of tissue and  fluid impedance
at  EEC  frequencies is difficult because electrodes
themselves have high impedances caused by electrical
double layers.


    Volume  conduction  in  brain   tissue  appears,
therefore, to be very small indeed. For this reason, I
consider  the  dipole models originally  suggested by
Brazier (1950) and  more recently by Henderson et al.
(1975), and others to be inappropriate. Nevertheless
an explanation is needed for the apparent far-field
effects described  by Jewett (1970 a, b) and others, in
which brain stem  potentials are recorded at the vertex
(referred  to  ear  lobes, which seems to me a rather
inappropriate reference  for auditory  potentials).
Another factor noted in intracranial EEC recording
is  the absence of blink or eye-rolling artifact, even
when subdural electrodes are located very close  to
the eyes. Recently I used averaging techniques trig-
gered by scalp blink potentials to measure intracranial
blink amplitude.  Fig. 1 shows  that  blink artifact is
attenuated  by a factor of about  30 between  an
electrode near Fpz and a subdural electrode immedi-
ately below. This suggests, however, that we should
never record EEC using scalp  electrodes since  the
electrocorticogram  is not 30 limes bigger. Cooper
et al. (1965) speculated why EEC can be seen in scalp
recordings: only  when areas of cortex are engaged in
synchronous activity does the attenuation apparently
become smaller. The high attenuation of blink poten-
tials must be due  to the three-dimensional current
flow caused  by closure of the eye (probably changing
the current paths) with preferential  paths leading to
apparent electrical screening of subdural electrodes.
These  data  show,  however,  that the brain  cannot
be assumed to be an isotropic medium with conduc-
tivity like  that  of  saline-an  assumption basic  to
dipole calculations.

     Brain stem  potentials  are  still unexplained,
although the small amplitude suggests the far  field
of i dipole (see  Allison, this volume). Another pos-
sibility which  would be difficult to prove  is that
the field is recorded via the relatively low-resistance
ventricular fluid  system in the  same way that EKG
can  be recorded from  the  foot and hand when,
loosely speaking, the leg and arm are used as con-
ductors tapping into the heart's potential field.

    Measurement of the magnetic  currents  using
magnetoencephalography, in which direction of cur-
rent flow can be determined,  may possibly throw
some much needed light on this subject.

-------
512
                                                                                               Cooper
    EOG-V
Fig. 1.  Average blink potentials (N-96) from  vertical
oculogram (EOG- V) and ttx subdural electrodes referred
to an average reference of 60 intracerebral electrodes
in a psychiatric patient. The averager was triggered by
the eye blink using a  Bereitschaftspotential-type pro-
gram that averaged activity before and after the blink.
Subdural electrode 1 was immediately beneath the ocu-
logram  electrode,  which was lust above the right eye-
brow, Electrodes 2 and 3 were 8 and 16 mm posterior
to electrode 1 (near the right frontal pole). Electrode 4
was approximately beneath the midpoint between F4
and C4. Electrodes 5 and 6 were located 8 and  16 mm
posterior to  reach sensory  motor cortex.  Only  the
anterior electrodes (I, 2 and  3) show a blink potential
with about a 30-fold attenuation.

-------
CALCULATED AND  EMPIRICAL EVOKED-POTENTIAL
DISTRIBUTIONS IN HUMAN RECORDINGS
T. ALLISON

Veterans Administration Hospital and Department of Psychology, Yale University, New Haven,
 CT, U.S.A.
    Cooper raised the issue of volume conduction
in evoked potential  (EP) recording. He noted that
under some conditions there is little apparent spread
of field potentials in  brain tissue. This conclusion
derives mainly from  recordings of spontaneous EEC
activity in  man  and animals  (reviewed by Cooper
et al   1965).  Clearly,  there are situations  in which
volume conduction  appears  minimal.  Recording
EPs from the  cortex or deep structures, one observes
very sharp attenuation of EP amplitude as the elec-
trode is moved away  from a focus. On  the other
hand, as Cooper noted, the fact that one can record
brainstem  evoked potentials  from the scalp argues
for an appreciable  degree of volume  conduction.
I believe this apparent paradox  can be understood
if one  takes into account the circumstances in which
scalp recordings are made.

    Consider  the distribution of potentials generated
by a dipole source located under a plane and oriented
perpendicular to it. The  potential recorded at  the
surface is given by
                       PD
                 (X2+D2)3/2

 where P is the potential of the dipole point source, D
 is the distance from the dipole to the surface, and X
 is the distance  along  the surface (Brazier  1949).
 Cortical  dipole   sources  are  usually  superficial.
 Polarity  inversion is  typically  seen  at  depths  of
 0.3-1.0  mm  (e.g.,  Schlag  1973) and, in  theory,
 potential  distributions should  be calculated for  a
 D within this range.  An  EP generator  is never,
 however,  a  point source, but is an area of tissue
 which, in the simplest case, lies in a plane. Such an
 extended  source  can be simulated by increasing D.
 For  D equals 1 cm,  the equation yields the potential
 distribution shown as a solid curve in Fig. 1. There is
 a rapid  attenuation of  amplitude with increasing
 distance.  For example,  2 cm  from  the source the
 potential  is  less  than 10% of  its maximal value. If
the source is moved nearer the surface, the  peak is
even steeper, while for deeper sources it is flatter.
The region of rapid attenuation of amplitude near the
source corresponds to the "near field" of Jewett and
Williston  (1971).  As  the distance from the source
increases, amplitude  approaches  an asymptotically
low level and is approximately a linear function of
distance.  This  region corresponds to  Jewett and
Williston's "far field."

    Under appropriate circumstances, the amplitude
of human EP components  can agree well with cal-
culated values. The dashed curve in Fig. 1 is the
amplitude of SEP component N55 (G.D. Goff et al.
1977) as recorded  from the pial surface of a patient
during  localization of the  central  sulcus.  An  8-
electrode array spanned the median nerve representa-
tion area of the postcentral gyrus in a medial (elec-
trode 1, to left of curve) to lateral (electrode 8, in right
part of  curve) direction.  The  SEP at electrode
5 is shown in the inset. N5S is the negativity at about
100 msec (it is increased in latency in most epileptic
patients). N55 was chosen for this purpose because it
is large, easily quantified, and (unlike shorter latency
components) appears to  be undistorted by  other
activity   in  the   same latency  range.  Agreement
between  calculated and empirical amplitude  is  fairly
good (and could be improved further by simulating a
more  extended source by increasing  D slightly).
Electrodes  1-7 are within the near field while elec-
trode 8 is in the far field.

    Now consider the comparable potential recorded
from the scalp; it is illustrated in the inset as recorded
from  C4, where its scalp amplitude was largest. Its
latency is earlier than that seen at the cortical surface,
as  is  often the case in scalp-pial EP comparisons
(Broughton 1969; Allison  et al. 1977).  The  scalp-
recorded potential is miniscule  relative to its pial
counterpart. The  distance between C4 and  the cor-
tical surface is about 1.5 cm; the calculated amplitude
for this distance is indicated by the shaded triangle on
the curve. In addition to the effect of distance per se,

-------
 514
                                              Allison
 there  is a further order-of-magnitude  decrease as a
 result  of attenuation  by dura, skull and scalp. The
 empirical amplitude  is  indicated by the unshaded
 triangle  and  corresponds to  an apparent 0 of about
 4.5 cm. Scalp topographies  (G.D. Goff et al. 1977),
 together  with other  pial recordings (unpublished),
 indicate  that  N5S is generated in the  crown  of
 the postcentral  gyrus.  Thus  its amplitude, as  re-
 corded from the  scalp, is  greater   than if it were
 generated in sulcal cortex (Vaughan  1974). In other
 words, this component represents a "best case" situ-
 ation  for volume  conduction  to the scalp. Yet,  if
 plotted  on  the  same amplitude scale as  the pial
 recordings, it would hardly be visible, although  by
 scalp-recording  standards it  is a  relatively  large,
 highly   reproducible   component.   Because   the
 averaging technique is now  commonplace, one may
 tend to forget how easy it is to  record  potentials
 quite distant in the far field. An extreme case is the
    brainstem auditory  evoked potentials, which can be
    several hundred microvolts in amplitude in brainstem
    recordings  in  cats  (Jewett  1970) but are  usually
    less than a microvolt  as  recorded from the human
    scalp.

        Thus, the  question is  not whether there is, or is
    not, volume conduction—we can say definitely that
    there is. Rather, the question is where the recording
    electrode is in  relation to the source. If it is close
    to the source, i.e., in the near field, small differences
    in location have a large effect on amplitude, and one
    is inclined to believe  there is little volume  conduc-
    tion. If one is  recording  at a distance, i.e., in the
    far field, and if the  apparent distance is even greater
    as a result of interpolation of tissue of low  conduc-
    tivity, one is then likely to conclude  that the head
    is a  good conductor  and that it does not much matter
    where the electrode is placed.
   x
   ut
   &
       100
       80
       60
       40
       20
        0 L-
                0   50   100  150
                      tntw
                                                                             —— THEORETICAL
                                                                             	EMPIRICAL
                I	L
_L
                  SURFACE
. DIPOLE SOURCE
 I     '	I	1	1	
                                          J	1
                                                    1012

                                                   DISTANCE, cm
Fig.  1.  Theoretical distribution of potential along a plane surface produced by a dipole source compared to
empirical distribution of pial surface recorded EP. For the pial recordings, interelectrode distances were deter-
mined from photographs made during surgery, with the electrode array in place.  Inset:  SEPs to left median
nerve stimulation recorded from pial surface of right postcentral gyrus f electrode 5) and scalp (L4). Scalp re-
cording was preoperative. Details in text.

-------
METHODOLOGY  AND MEANING OF  HUMAN EVOKED-
POTENTIAL  SCALP DISTRIBUTION STUDIES

T. PICTON
Department of Medicine and  Experimental Psychology, University of Ottawa, Ottawa,
Ontario, Canada
D. WOODS
Department of Neurosciences, University of California at La Jolla, CA, U.S.A.
D. STUSS
Aphasia Research Center, Boston University School of Medicine, Boston, MA, U.S.A.
K. CAMPBELL
MRC Applied Psychology Unit, Psychophysiology Section, Cambridge,  England
    Scalp distribution studies of human event-related
potentials  involve  the recording  and evaluation of
such waveforms at multiple scalp  locations. Such
studies provide information helpful in determining
the possible generator sources of the scalp-recorded
events and in comparing electrical  events recorded
under different physical or psychological conditions.
This paper will attempt to describe and illustrate the
techniques of recording and displaying scalp distri-
bution data and the procedures and problems of their
interpretation.

Recording

Electrode location

    Most scalp distribution  studies  utilize the Inter-
national  Federation  10-20 system  of  electrode
location (Jasper  1958).  In this system, standard
electrode sites are determined by dividing distances
along  the  head into prespecified  ratios.  The use of
ratio measurements makes the system more adaptable
to different sizes and shapes of heads than systems
based  on absolute meaurements.  Electrodes located
outside of the standard 10-20 sites are often utilized
in evoked  potential research for two reasons. First,
a greater spatial concentration of electrodes might be
necessary  to  delineate  small  differences or asym-
metries  in scalp distribution. Vaughan (1974) sug-
gested  the  minimum practical  spacing  of scalp
electrodes  should be approximately  2.5 cm. Jeffreys
and Axford  (1972)  utilized 2.5-cm  interelectrode
distances in  their evaluation  of the visual  evoked
potential scalp distribution. Peronnet et al. (1974),
in their studies of  the auditory vertex potential,
used a coronal chain of 13 electrodes  rather  than
the 7  provided by the basic 10-20 system. Theoreti-
cally, at least two electrodes are needed in the spatial
period  of an evoked  potential component  (see
Ruchkin's  correspondence in Donchin, this section);
unfortunately, the spatial frequency of most scalp-
recorded components is unknown and cannot be cal-
culated prior to mapping. A second reason for using
nonstandard  electrode locations is that the optimal
recording sites for certain evoked potentials may lie
outside the basic  10-20 system. Several researchers
(e.g., Matsumiya et  al. 1972; Brown et al.  1973;
Megela  and Teyler, this volume) have attempted to
record  meaningfulness-sensitive  evoked  potential
components  from electrodes overlying Wernicke's
area between the temporal and parietal electrodes of
the 10-20 system. The optimal recording location for
the somatosensory evoked response elicited by pero-
neal nerve stimulation is a few centimeters posterior
to the vertex (Tsumoto et al. 1972) and by median
nerve stimulation  a few centimeters behind the C3
or C4 electrodes (Calmes and Cracco 1971).

   Fortunately the  electrode labels  used in the
10-20 system were developed so as to allow the incor-
poration  of additional electrode locations. Fig.  1
represents  a possible  modification  of the  10-20
system  to provide for a greater concentration of
electrodes on the  scalp and for more accurate elec-
trode location for specific purposes. The modification
      •F
                                       Cb
Fig. 1. Possible extension of International 10-20 System
providing greater spatial concentration  of scalp -'"
trades.

-------
516
                                    Picton et al.
involves  the  addition of  anterior frontal, posterior
frontal, postcentral, preoccipital, and inferior tem-
poral lines of electrodes. The only necessary numeri-
cal nonsequiturs are the T3a and T4a electrodes. The
system  is  fairly simple .in measurement, provides
interelectrode distances of between 2 and 4 cm, and
can be easily described. It should be noted that using
more closely spaced electrodes need not provide any
further meaningful  information. Interindividual vari-
ability in brain topography may reduce the advantage
gained by the closer spacing of the electrodes.

Recording montage
    Because of the  low amplitude of the  evoked
potential signal and the pervasiveness  of environ-
mental electrical noise,  all  ERP recordings are per-
formed  using  differential amplifiers. As has  often
been pointed out (W.R.Goffet al. 1969,Cracco  1972,
Vaughan  1974), it is  usually preferable to  record
scalp  activity relative to a distant  indifferent elec-
trode  rather than  between adjacent closely spaced
electrodes.  Closely spaced bipolar  montages  may
occasionally be helpful. Polarity reversal might reveal
a  point of maximum amplitude or locate a  fairly
focal   evoked  potential component.  Yet,  even  in
such cases, referential montages can provide the same
information.  Bipolar  recordings   are  particularly
difficult to  understand if multiple  ERP components
of similar latency are recorded at different electrodes.

Reference electrodes

    The choice of an  indifferent  reference  can be
extremely difficult. There is probably no place on
the human  body   on  which an electrode will not
pick up, through volume conduction, some electrical
activity  of  the brain.  Indeed,  the  indifference  of
many  reference electrode locations is a matter  of
some  difference of opinion. For example, Vaughan
and  Ritter  (1970)  considered  the  nose  to  be
indifferent  for   the  auditory  vertex  potential
recording, and  using a nasal reference demonstrated
a  phase  reversal of the Nl  and P2 components  of
the auditory ERP  across the Sylvian fissure.  They
therefore suggested that the auditory vertex  poten-
tial derived from bilateral vertically  oriented  dipole
generators in the  primary auditory  cortices of the
transverse temporal planei. However, W. R. Goff et al.
(1969)  using a linked-esr reference and Kooi  et al.
(1971)  using  a chest reference  found no evidence
of polarity  reversal across  the  Sylvian  fissure, and
Kooi  et al.  moreover  found that  nasal electrodes
picked   up  vertex   potential  activity,  possibly
deriving  from  the  frontal pole.  The phase reversal
found  with a  nasal reference might therefore have
been  caused by the activity at the reference site.
Vaughan (1974),  however, has  suggested that the
chest   reference  might  actually  pick  up activity
from the base of the  brain,  generated by the vertical
auditory cortex dipole, and might Itself, therefore,
be considered an active reference. In our own work
(Fig. 2), we have found that, using an ankle reference,
there  is a definite  moderate amount  of activity at
the nose and minimal activity at the chest reference
during the period of the auditory vertex potential.
In the five subjects examined, the N100 component
in nose-ankle  recordings was between  IS  and 50%
(mean  28%)   of the  vertex-ankle measurement,
whereas  there  was  no  significant  activity  at  this
latency  in chest-ankle recordings. It is  difficult to
presume that all the  activity recorded in such a nose-
ankle montage represents distant volume-conducted
activity  at the  ankle. The relatively  large amount of
activity  picked up at the nose might be due  to the
increased  longitudinal  as  opposed to  transverse
current  flow in the nose and  nasal sinuses, allowing
the recording of frontal activity at a distance. There
are similar controversies about  reference electrodes
in visual ERP  recording (e.g., Michael and Halliday
1971).  For  the most  part,  however, the  chest-
reference electrode  as described by  Stephenson and
Gibbs (1951) seems to provide a relatively indifferent
reference  for  the late  components of  the  human
evoked potential. At earlier latencies, other references
might be better because of the possible muscle and
spinal cord activity recorded from the chest.

Evaluation and display

 Identification of components

    Scalp distributions are plotted for each of the
components of the recorded evoked potential. The
identification of distinct EP  components, however,
is a  very  major difficulty. There  are  two general
approaches  to  this  problem,   neither of which is
completely  satisfactory. The one  most commonly
used  is  to  pick out  a "peak" event,  identify its
latency, and evaluate the scalp voltage distribution
at that latency. There is, however, no way of knowing
that this waveform peak represents a discrete physio-
logical event and not a recording artifact  of several
physiological  events of  similar latency and over-
lapping field distributions. Fig. 3 illustrates some pro-
blems of overlapping components. In the vertex and
central regions of the scalp, a distinct negative-positive
wave can be observed with mean peak latencies of 104
and 203  msec for eight  subjects examined.  At the
mastoid and ear, a small negative component can be
identified with a mean latency of 152 msec. In certain
of the subjects, a smaller positive wave can also be dis-
tinguished at some 48 msec prior to this negative peak.
In most of the subjects, these  components can also
be recognized in the temporal region. They probably
represent a concomitant physiological event occurring
in lateral and inferior temporal regions at the same
time  as the  vertex potential. A similar temporal
component of the auditory evoked response has been
described by  Wolpaw  and Penry  (1975), who by

-------
Methods and Meaning in Topographical Studies

sophisticated computer subtraction procedures could
identify  a  positive-negative complex  in  temporal
electrodes with peak latencies of approximately 110
and 160 msec.

    A second approach  to the identification of EP
components  involves  the  formal  mathematical
analysis  of  the recorded  waveforms. Using  fairly
simple mathematics, one  can fit the evoked potential
waveform with  a series of exponentials or sine waves
or  other mathematical  functons. Such a possible
component structure is, however, quite arbitrary and
not necessarily  related to the underlying generators.
Using more  sophisticated techniques, the orthogonal
"principal  components"  of a  waveform can be
obtained (John et  al. 1964, Donchin  1966).  These
components  represent those  parts of the  evoked
potential that  contribute the major  portion of its
variance. As such, they are determined in part by the
experimental manipulations  during  the  recording,
Thus, the principal  components of the visual evoked
potential as derived by  Donchin  in 1966 delineated
only those parts of the waveform that varied with the
differences in light  intensity used in his experiment,
and  had little  relation to  the largest EP waveform
peaks, which remained fairly constant across different
                                                                                                   517
              C«
              T,
              M,
              NECK
              NOSE
              EDO
                                  SOdBSL
                                  200 msec
                                  1kHz ton*

                                  360 mMC
                                  AFTER R-WAVE
                                  OFEKG

                                  R. ANKLE
                                  REFERENCE


                                  266X

                                  RESIDUAL EKG
                                  SUBTRACTED
                                  300mi»
                                  SWEEP
Fig. 2.  Human auditory evoked potentials recorded
using  an ankle reference for two different subjects.
The electrooculogram was recorded between supra-
orbital and  infraorbital ridges.   All other channels
were  recorded using a reference  electrode on  the
right ankle.
                                                          100
                                                                                  106
                                                          1kH2
                                                                                  ZSOHz
                                                     Fig. 3.  Lateral  scalp distribution for the auditory
                                                     evoked potential obtained  using equally  loud (80
                                                     phon) tones of  250 and 1000 Hz.  Each  tracing
                                                     represents the average over 100 trials.
intensities. With sufficient experimental manipulation
in both  physical and  psychological  domains,  it is
probable  that  such principal  component  analysis
might more fully and more meaningfully resolve the
structure  of the evoked potential waveform.  The
relationship  of the principal components to physio-
logical processes, however, is difficult to evaluate and,
as has been previously pointed out (John et al. 1964,
John et al. 1973, K. Squires et al.  1977), a principal
factor  should not  be  construed  as  representing  a
physiological system. Indeed, the  orthogonality of
the components, while  greatly helpful  to statistical
analysis, is quite  different from the interdependence
of  physiological  systems  (see  also  Donchin   and
Heffley, this volume).

Data display

    Once the EP components have been  identified,
their scalp distribution can be evaluated. The techni-
que most often used, at present, Involves the selec-
tion of a latency or peak event,  the calculation of
individual waveform magnitudes  at  each  electrode
relative  to some  selected reference (e.g.,  the vertex
electrode magnitude), the averaging of these relative
magnitudes  over a  number of  subjects, and the
interpolation  of  contour  lines.  This is  Illustrated

-------
  518
                                                                                          Pic ton et al.
    SUBJECT C.S.
                                                       N1
                     MEAN OF SIX SUBJECTS
                     1 secSOdB 1kHz TONE
                     EVERY 3 sec. CHEST REF.
                                                               o  a
                                                                      6.1uV
                                                                      4.3UV
                                        3.4uV
 Fig. 4. Scalp distribution of human auditory evoked potential to 1-sec, toneburst. Waveforms from one subject are at
 the left and the mean data from six subjects are shown on the right.
 in Fig. 4. The average evoked potentials for a sample
 subject should also be given, since this allows for the
 examination of  the  potentials at all  latencies,  at
 least in that one subject.
 Time dimensions

    The major disadvantage of a display system such
 as that  described in the  preceding paragraph is that
 the scalp topography is  evaluated only  at  selected
 discrete  latencies. The  continuous time dimension
 is not portrayed, and important EP information may
 thus  be lost (Weinberg,  this volume). Without the
 time  dimension,  such things as  travelling waves  or
 spread of excitation cannot be evaluated. The  tem-
 poral  factor  can be added to such  contour maps
 using  a moving film technique as demonstrated by
 Walter  Freeman. If one spatial  dimension  is given
 up, a  spatio-temporal map may be derived from the
 EP  data  of  a  one-dimensional electrode  array
 (Re'mond  1968). Interpolation  of spatio-temporal
contours produces  a map of the evoked potential
showing the sequencing of its waveforms at different
locations. This type of  map  should be  based on
referential  records as in  Renault and Lesevre  (this
volume). Earlier maps based on differential recordings
between adjacent electrodes and using microvolts per
centimeter  rather than  simple microvolt contours
 were somewhat more difficult to understand.
 Multivariate statistical analysis

     The scalp distribution of an EP component is a
 set  of  measurements  of  that component recorded
 simultaneously at a number of scalp locations.  The
 techniques of multivariate analysis  are necessary for
 the proper statistical evaluation of such information.
 Multivariate statistical techniques, in particular that
 of principal component analysis, have been applied
 to evoked potentials  in a number of ways (Donchin
 1966).  Evoked potential data in the time dimension
 have been analyzed  to determine  which temporal
 components of the waveform are changed by experi-
 mental  manipulation  (e.g.,  Donchin  1966,  Suter
 1970, Chapman 1973). Evoked potential data across
 both temporal  and  spatial  dimensions have been
 analyzed  into principal  components in order  to
 distinguish different brain states (John et al.  1973,
 Thatcher  and John  1975),  to  define  independent
 psychophysiological processes (Squires,  N. et  al.
 1975, Donchin et al. 1975), and to evaluate evoked
 potential  asymmetries in  newborn  infants (Molfese
 et al. 1976). The application of multivariate analysis
 to scalp distribution data  will be considered briefly
 using two examples.

Differences in scalp distribution

    Evoked potential components with significantly
different  scalp distributions  must  derive from  dif-
ferent sources. Either different cells are  involved in

-------
Methods and Meaning in Topographical Studies
                                            519
 the generation of the scalp-recorded potential or the
 active cells are differentially responsive. The statis-
 tical evaluation  of scalp  distribution differences is
 therefore  an  important  procedure in the differen-
 tiation of possible source mechanisms for the evoked
 potential.  To  illustrate this, we shall consider whether
 the scalp distributions of the onset response and sus-
 tained  potential  evoked by  auditory stimuli (Fig. 4)
 are significantly  different,  and  if so,  along  what
 dimension or  dimensions this difference lies. The pro-
 cedure used is based on the principal component anal-
 ysis technique as described by Donchin (1966) except
 that electrode location is used  instead  of the  time
 dimension, and replications are over different subjects
 rather  than for the same subject. The scalp  distribu-
 tion of the Nl, P2, and SP components for each sub-
 ject was  expressed  as  a  percentage  of the vertex
 magnitude. This procedure decreased the interindivi-
 dual variability.  It was  also necessary since only six
 channels of dc data  could  be  recorded  at  any one
 time.  The principal  component  analysis  was per-
 formed on the  17 measurements in  order to obtain
 the scalp locations contributing most to  the variance
 among the three evoked potential components. The
 first four principal components, explaining 75% of
 the total  variance, were selected  as being  possibly
 meaningful (having eigenvalues >  1, and each contri-
 buting >  10% of the total variance). An analysis of
 variance was  then performed for the combined  scalp
 distribution   measurements  weighted by  the  four
 orthogonal dimensions  defined by  these  principal
 components.  Only the first  factor showed any  signi-
 ficant  differentiation of EP  components (p < .01).
 According to the  Tukey  a  posteriori testing proce-
 dure,  Nl  (0.23 on the  factor-weighted  measure-
 ments)  and   P2  (0.02)  were both  significantly dif-
 ferent (p < .05) from the SP component (-0.66) but
 not significantly different from each other. The first
 principal component  was  positively weighted for T6
 (0.80), T3 (0.77), T4 (0.72), Oz (0.69), T5 (0.50),
 F3  (0.41), P4 (0.34), F7  (0.23), and Pz (0.12) and
 negatively weighted for Fz [(1/N)-0.30], F8 [(1/N>
 0.23], and F4 [(1/N)-0.18], with the other electrodes
 having  minimal  effect  (< 0.1).  Thus the  analysis
 suggests that the sustained potential is somewhat  more
 frontal and much less posterior and temporal than the
 onset vertex potential.


     This  conclusion  is  supported by independent
 data from another  experiment wherein  Nl and SP
 were measured only at the Fz and Pz electrode  loca-
 tions. For the twelve subjects tested, the Fz/(Fz  + Pz)
 ratio was significantly (p < ,01) less for the Nl  com-
 ponent (0.60 SD 0.05) than for  the SP  component
 (0.68  SD 0.08). Thus the auditory sustained potential
 has a  different  scalp distribution from  the  onset
 response,  and must, therefore, derive from a some-
 what different generator.
Identification of distinct psychophysiological
events

    Scalp distribution data may also be used as an
added dimension to distinguish one brain event from
another.  To  illustrate  this use of scalp distribution
information,  the evoked potential to auditory feed-
back  stimuli in a conceptual learning  task will be
examined (Stuss  1976). Eleven subjects were asked
to  determine  repeatedly  along  which  of five in-
formational dimensions  (e.g., color or shape) they
should sort a block of 9 to 12 complex visual stimuli.
The subjects made a hypothesis as to the possible
sorting criterion, responded on that trial according to
this hypothetical criterion, and were then given audi-
tory feedback as to whether the response was correct
(1 -kHz toneburst) or incorrect (4-kHz toneburst). With
repeated trials, the subjects soon learned the correct
criterion and then  responded appropriately until an
incorrect feedback stimulus signaled a new block of
trials with a different  criterion. Prior  to  averaging,
the  evoked  potential measurements were grouped
across multiple blocks  of trials  according  to  five
possible   psychological   conditions:   "preinsight,"
"insight,"   "confirm,"   "overlearn,"   and   "false
feedback"  (the criterion change). The evoked poten-
tial to the  feedback stimuli prior to or at the time of
criterion  discovery  (the  "false  feedback,"  "prein-
sight," and "insight"  conditions) contained a definite
late positive wave  occurring with a mean peak latency
of approximately 650 msec that was termed the "P4"
in  order to distinguish it  from an earlier "P3" wave
(350 msec).  This P4  component is  not definitely
recognizable in the response to feedback in simple
discrimination  tasks  not requiring learning (e.g.,
Squires, K. et  al. 1973), and in our paradigm it was
essentially nonexistent in trials after insight had  been
obtained ("confirm"  and "overlearn"  conditions).
Since the amplitude of the P3 wave changed similarly
during the different conditions, it was possible  that
the P4  represented  merely  a continuation  of the
psychophysiological process  underlying the P3 com-
ponent.  The scalp distribution of the two  compo-
nents  were, however,  distinctly different,  the P4
being  of greater  amplitude  in the parieto-occipital
regions  (Fig. 5).  These  impressions of the raw  data
were supported when the five basic evoked potential
measurements  (prefeedback  CNV, Nl, P3, N3, P4)
were submitted to an analysis of the principal compo-
nents of the variance over the five conditions of the
experiment  and  the  six  electrode  locations  (cf.
Squires, N.  et al.  1975; Donchin  et  al. 1975). It
would, of course, have been  preferable to submit the
evoked potential data for such an analysis in  a  more
arbitrary and extensive fashion, i.e., at multiple set
latencies as in the Donchin (1966) article. Such a pro-
 cedure was, however, beyond the current capacity of
 our laboratory. A  repeated-measures two-way analysis
 of variance  was  then performed in order to assess

-------
  520
                                     Picton et al.
                    INSIGHT
                      CONFIRM
 Fig, 5. Evoked potentials to feedback stimuli in one subject for two conditions of the conceptual learning task. Each
 tracing is the average over 12 trials.
 the  significance  and the pattern of differentiation
 between the  first three components (Table  1). The
 "P4" and  "P3" components were similarly affected
 by experimental  condition, both differing markedly
 from "CNV." Each of the three components had dis-
 tinct scalp  distributions. The "P4"  component was
 fairly widespread in  parietal, occipital,  and central
 areas. The "CNV" was parieto-occipital, and the "P3"
 showed a  marked vertex predominance. There are
 several  reports in the  literature describing positive
 waves  that occur after the P3 component of the
 evoked  potential, and their possible relationship to
 the  P4 component  that we observed is unknown.
 Jenness (1972) described a P4 component associated
 with the learning of a more accurate auditory dis-
 crimination, but this was somewhat earlier in latency
 (400  to 500  msec) and occurred in relation to the
 task stimulus  rather than to the feedback stimulus.
 K. Squires etal. (1977) reported a positive "slow wave"
 component to the evoked  potential to  a detected
signal,  probably  analogous  to the  transient return
to baseline  of a  preceding  negativity as postulated
by  Naatanen   (1975)  and   Wilkinson  and  Ashby
(1974).  This slow wave, however, had a more gen-
eralized scalp  distribution  and  a longer duration
(cf. Hillyard et al. 1976) than the P4 recognized in
our paradigm.
Interpretation of scalp distribution data

Inferences as to possible sources

     Scalp distribution studies, in effect, involve the
recording of an electrical field at a distance from one
or  several generators located within  a volume con-
ductor. The generation of the electrical field depends
upon the separation of charge within this conducting
medium. In the nervous system, such charge separa-
tion derives largely  from polarization and depolari-
zation of neuronal  and glial  membranes. The nature
of the electrical field depends upon the extent and
relationship  of the activated membranes, and upon
the impedance and geometry of the volume conduc-
tor. Lorente de No (1947), in his brilliant treatment
of  neurophysiology, identified "open," "closed,"
and "mixed"  electrical  fields (Uinas and Nicholson
1974). Recently, Rail (1970) has extended the origi-
nal  concept  of  open  and  closed  fields and  has
described what might be called a "semi-closed" field.
This is generated by a nuclear mass with a low resis-
tance pathway entering at  one point ("punctured
symmetry"). Such a neuronal aggregate will generate
a field at a distance characterized  by a widespread
single polarity as opposed to the  dipolar nature of
the  open field.  Finally,  there is what might  be

-------
Methods and Meaning in Topographical Studies
                                              521
         Table 1. Principal Component Analysis of Feedback Response in Conceptual Learning Task
Principal component
Percentage of total variance
Weighting




Effects of





Effects of






of raw variable CNV
N1
P3
N3
P4
condition HSDa
Preinsight
Insight
Confirm
Overlearn
False FB
scalp location HSDa
Cz
Oz
F3
F4
P3
P4
I"P4"
54.7
0.06
0.12
0.32
0.56
0.90
1.02
1.33
0.84
0.04
0.29
2.37
0.82
1.56
1.01
0.60
0.17
1.07
1.41
II"CNV"
24.7
0.95
0.29
0.16
-0.05
0.10
0.48
-0.74
•0.95
-1.13
•0.99
•0.18
0.75
•0.73
•1.32
•0.18
•0.21
•1.36
•0.99
III"P3"
11.4
0.14
0.08
0.91
0.32
0.31
0.85
1.97
1.68
0.73
0.91
2.97
0.81
2.66
1.14
1.55
1.09
1.54
1.94
 8HSD: th< "honestly significant difference"  between  the meant at p<05 at determined by Tukey procedure after an Initial
 analytit of variance.

 Meaturementt uted for evaluating the effectt of condition and icalp location were obtained by multiplying raw variable measure-
 ment (In 0V re beiellne) by "weighting" factor at determined by principal component analytit.
termed a "diffuse field" generated through the over-
all  separation  of charge  between  intracellular  and
extracellular spaces. In an  area  of neurons  that  is
diffusely excited, there is a net intraneuronal accumu-
lation of positive charge with a resultant extracellular
negativity. Depolarization  of gllal membranes through
the release  of  potassium will  also contribute to the
extracellular negativity. Such  an electrical  field  is
similar to that of a  simple current source or sink and
can  be recorded at a distance, albeit rapidly falling
off (Somjen 1973).

    An electrical  field recorded at the scalp in rela-
tion  to some  physical  or psychological  event  may
derive from numerous, spatially separate, generators,
each  of which may create its own type of.electrical,
field.  It is, therefore, probably impossible to derive
the nature  of  the source even  with extensive knowl-
edge  of the resultant field. Other sources of Informa-
tion are often  necessary for the understanding of the
evoked  potential and its scalp distribution: extensive
animal  correlative recordings  (e.g.,  Arezzo  et al.
197S),  human  intracranial  recordings (e.g., Allison
et al., in press) and human clinicopathological corre-
lation (e.g.,  Michel and  Peronnet  1974).  At times,
even  such  sources are  Insufficient  for  the  under-
standing  of a recorded  potential waveform; e.g.
human  Intracranial  recordings  define  no  distinct
generator for the auditory-vertex potential although
the multiple latency shifts  suggest  the  presence  of
several sources  (W. Goff et al., In press).

    The Interpretation  of the possible origin of the
auditory vertex potential  is Indeed a prime example
of the  difficulties in  the inference of source from
scalp  distribution data (Picton et al. 1974b; Hillyard
and Picton, in  press). It has been variously  postulated
that   this potential derives  from  a single thalamic
generator (Smith  et  al.  1973),  from  deep-seated
dlpoles  In the  primary auditory cortices  (Vaughan

-------
  522
                                     Picton et al.
 and  Ritter 1970), or from widespread  cortical acti-
 vation in frontal and central areas (Kooi et al. 1971).
 Computational  procedures to derive  the  equivalent
 dipole for the vertex potential  (e.g.,  Smith et al.
 1973) are of little assistance since the interpretation
 of this  equivalent dipole remains  quite difficult.
 As Henderson et al. (1975) point  out, it is impossible
 to distinguish using surface recordings  a  real dipole
 lying  below a convex surface (such as  the thalamic
 or primary cortical dipoles) from synchronous activ-
 ity in the surface  itself (such as  widespread cortical
 activation).  In  visual  and somatosensory systems,
 "parallel  late  waves"  occur  in  both  the  primary
 sensory areas and  the  frontocentral region (Donald,
 in press),  and this possibility  also  occurs  in the
 auditory system (cf. Wolpaw and Penry 1975, Arezzo
 et al. 1975).  In all probability, the primary  cortex,
 secondary areas in  the temporal lobe, and widespread
 areas of frontal and parietal association cortex are all
 active during the time human auditory vertex poten-
 tial appears on  the scalp, and all may contribute to
 some extent to the scalp-recorded activity.

    The interpretation of the possible source of the
 auditory sustained potential is equally  difficult. As
 we have shown, its scalp distribution differs from that
 of the vertex potential elicited by the onset of the
 stimulus.   The  etiology  of  such  differentiation,
 however,  is somewhat obscure. One might postulate
 that the sustained potential is generated, like the on-
 set response, in both primary  and association areas,
 but with  different contributions of each source.
 Mapping down  the side of the scalp (Fig. 3) showed
 no sustained  activity below the  temporal region in
 five out of eight subjects. In three subjects,  and in
 them only at the  lower frequency, a sustained  posi-
 tivity was noted in the inferior electrodes. This might
 be due to a sustained dipole source  in  the primary
 cortex, the difference between frequencies being due
 to a different orientation of the primary cortex in the
 more lateral region of the temporal plane  concerned
 with the lower frequencies. One might also postulate,
 however, that the sustained potential, like the CNV,
 is  generated mainly in the frontal association cortex
(e.g.t  Jarvilehto  and Fruhstorfer 1973) and that the
positivity  represents  the  recording from  below an
activated convex plane, the differences in  frequency
being because of the relatively larger potential elicited
by lower frequency stimuli (Picton and Woods 1975).
Most  probably, the auditory sustained potential, like
the onset response, is generated in both  the primary
auditory cortex and association  areas,  the  relative
contributions of each source differing somewhat for
each component.

Distinction of psychophysiological events

    Although  the  derivation  of the  intracranial
sources from scalp distribution data can be extremely
 difficult, it is  somewhat simpler just to differentiate
 components of the evoked potential on the basis of
 their scalp distribution. Scalp-recorded events with
 different voltage distributions must derive  from dif-
 ferent sources, and therefore the evaluation of scalp
 distribution  becomes a very  powerful tool in  the
 dissection out of distinct psychophysiological events.
     Probably the most  important use  of such scalp
 distribution studies has been in the evaluation of the
 late  positive  component  of  the evoked  potential
 (Hillyard  et al. 1976).  As was shown by Vaugban
 (1969),  Vaughan  and  Ritter (1970), Picton  and
 Hillyard (1974), Ritter et al. (in press), and Hillyard
 et al. (1976),  the  late positive wave has a different
 scalp distribution from the earlier components of the
 sensory evoked potential. Its  scalp topography is also
 distinct  from  that of the preceding CNV (Donchin
 et  al.  1975). Scalp distribution studies have  further
 shown  that the  late  positive potential  does  not
 represent a unitary process. Courchesne et al. (1975)
 have demonstrated  that novel, unpredictable visual
 stimuli elicit more frontal late positive components
 than those elicited  by expected signal stimuli. In the
 conceptual learning  task  reported in  this paper,  a
 distinction between the "P3"  and "P4" wave of the
 feedback EP was made on the basis of latency  and
 scalp distribution. This distinction makes it possible
 to  hypothesize that  the two  waves  reflect separate
 psychophysiological processes, possibly the apprecia-
 tion of feedback information  (P3) and  its utilization
 in  conceptual  learning  (P4).  Such a  hypothesis is
 tentative and will require further experimental testing
 to  differentiate  the  psychological determinants  of
 these physiologically distinct late positive waves.


 Conclusion
    The study of the  scalp distribution of human
event-related potentials is a difficult yet extremely
helpful approach to our  understanding of  these
waveforms.  It  provides evidence that  can be used
together with other  experimental data to delineate
the possible intracranial  sources of scalp recorded
events.  Perhaps  most  importantly  at the  present
time,  it aids in the differentiation of distinct psycho-
physiological events occurring within the brain.
Acknowledgments

    The research for this paper was performed in the
laboratory of Drs. Galambos and Hillyard at the Uni-
versity of California and in the Faculty of Medicine
at the University  of Ottawa, The research was sup-
ported by  the Medical  Research Council of Canada,
the Ontario Mental  Health Foundation, NASA, and
the National Science Foundation.

-------
INTRACRANIAL SOURCES  OF EVENT-RELATED
POTENTIALS

W. RITTER
Lehman College, City University of New York, New York, NY, U.S.A.
    Important  objectives  of current research  in
human event-related potentials (ERPs) are to ascer-
tain  the intracranial location and functional signifi-
cance of the various components that can be observed
in scalp recordings. Achievement of these objectives
is  important in providing data pertinent to psycho-
logical  theories of cognitive processes (Hillyard and
Picton, in press; N. Squires et al. 1977) and  in cli-
nical assessment.  In addition, this information will
hopefully provide the basis for future research on the
mechanisms  through  which  these   functions are
accomplished. At present, we  have only the faintest
idea  what  these mechanisms  might be, and th«
task  of discovering them seems awesome, especially
where complex psychological processes are involved.

    Many ERP  investigators  maintain that  neuro-
physiologists  will eventually  conduct unit  studies
that explain the mechanisms underlying the functions
ERPs reflect. This belief  is comforting because it
relieves  the  investigator  of  any responsibility for
thinking about mechanisms, but even a cursory con-
sideration  of  complex  psychological  processes-in
particular, how  brain  processes produce conscious
experience-exposes  the   naivete  of that   belief.
Although  neurophysiological  achievements   rank
among the most  satisfying and elegant accomplish-
ments of the biological  sciences, they do not even
begin to  suggest how  biophysical events constitute
conscious  events. In  one  of  the  few attempts to
deal  with  this question, John (1976) suggests both
the complexity of what  must be explained and the
concepts  of  brain mechanisms that incorporate but
go beyond neural activity. The end of the search,
however,  remains a mystery. Given the complexities
and difficulties brain scientists will encounter as more
information  is accumulated,  the attempt  to  locate
sources of ERPs is a small  step. The modesty  of the
goal, however, is not correlated with its difficulty.

    The actual sources of ERPs consist of complex
configurations  of biophysical activity that produce a
net separation of charge across a particular area of
brain tissue.  At present, investigations are essentially
limited to circumscribing brain sites within which the
separation of charge occurs. There are two ways in
which this has  been attempted. The  first is  to use
ERP scalp distributions  to test  specific hypotheses
based on known anatomical and other facts (Vaughan
1974). Though quite useful at times, the limitations
of this  approach are considerable, as  detailed  by
Picton et  al.  (this volume). The other  method for
localizing sources of ERPs is to record directly from
the brain either in man (e.g., W. Goff et al., in press) or
animals  (e.g., Arezzo  et  al.  1975).  The  use  of
animals  permits  extensive  explorations (including
successive  approximations  to  whatever extent  is
required)  that  are  precluded  in human  subjects,
whereas the use of human subjects is essential  for
establishing that  the ERPs being studied are identical
to those recorded from the scalp and for examining
ERPs unique to man.

    A particularly helpful clue in localizing sources is
contained  in  the  method of transcortical recording.
Dipoles that occur in the cortex produce waveforms
that   invert  in  polarity when  recorded separately
above and below the cortex with reference to some
distant  point. Although  the  relationship  between
geometrical configuration of the actual  source and
inversion is not known, the fact of the inversion is
important.  In all likelihood,  most cortical  ERPs
recorded at the scalp derive from such dipoles.

    Kelly  et al.  (1965)  demonstrated  the use  of
transcortical recordings in localizing sources. Median
nerve stimulation was studied in monkeys by placing
electrodes at many sites, above and below the cortex,
and  recording in monopolar fashion using  a single
(bone)  reference. Upon stimulation,  ERPs  were
obtained at all  recording sites. The only recordings
that   showed inversions  above  versus  below the
cortex,  however, were in the somatosensory hand
area.  Outside this area, ERP components  were of
the same polarity above and below the cortex. The
interpretation of the similarity  in polarity is that
both  recordings reflect volume conduction from  a
distant source. Since a given pair of electrodes above
and below the cortex are in about the same geometri-
cal position with respect to a distant dipole, com-
ponents recorded from them are of similar polarity.
To test this,  some cortical areas where no inversion
occurred were excised and filled with cotton soaked
in saline. Subsequent stimulation produced the same
ERPs recorded above and below the cotton, as were
obtained prior   to  the excision, making clear  the

-------
  524
                                           Ritter
 volume conducted nature of these potentials. When
 cortical  excisions  were  filled with saline-soaked
 cotton  in  those  areas  where inversions  did  occur,
 subsequent stimulation failed to elicit inverted poten-
 tials. Inversions of ERP components above and below
 given portions of cortex, therefore, may be taken as
 an  indication  that  those portions  of cortex  are
 sources for those components.

     The transcortical recording technique has been
 used  in monkeys  to  study  sources  of auditory
 (Arezzo et al. 197S) and motor (Arezzo and Vaughan
 1975)  ERPs. In studying auditory  ERPs, they  found
 that potentials  could  be elicited from all monopolar
 epidural electrodes  (up to 60) spread out across  the
 cortical  surface.   However,   inversions   between
 recordings above and below the cortex were confined
 to three areas: superior temporal plane, motor cortex,
 and lateral surface of the superior temporal cortex.
 There were two especially interesting features of this
 study.  Pint,  the volume conduction of several com-
 ponents that emanated from  the  superior temporal
 plane (each with somewhat different locations  and
 orientations)  could be traced from their  source to
 surface  cortex.  Second,  complications  of effects
 of three main sources on surface  recordings  could
 not  be  disentangled  without depth  recordings,
 demonstrating  the   limitations    of  using  scalp
 recordings by themselves for inferring  intracranlal
 sources. In the study  of  motor potentials, specific
 source locations, anterior and posterior to the central
 sulcus,  were identified for various  components asso-
 ciated  with  wrist  extensions. These workers  are
 currently using polarity inversions to  locate ERP
 sources within specific cortical laminae. These studies
 are promising, therefore, for circumscribing discrete
 locations for sources of cortical ERPs.

     To unravel  the  mechanisms ERPs reflect, it will
 be necessary  to establish  the  functions with which
 given potentials are associated.  The goal is not only
 to identify  biophysical configurations that produce
 ERP components, but to determine how those con-
 figurations  achieve  functional  significance.   Scalp
 distributions can be useful in providing information
 about the functions related to ERPs. If two compo-
 nents have different distributions, it can be inferred
 that  they have different sources and, by  implication,
 different functions.

    A good  example  of  this  pertains  to  the  N2
 component  of auditory and  visual ERPs.   Simson
 et al. (1976)  had subjects detect randomly omitted
 stimuli  in a train  of tones in one  condition and  in
 a train  of  flashes  in  another condition.  The N2
elicited  by  omitted stimuli had a  central  focus  in
 the auditory  condition and a  preoccipital  focus  in
 the visual condition (Fig. 1). By contract, P3 elicited
by  omitted  stimuli had a parietal  focus  in  both
 conditions.  The  N2 component had a distribution
 different from  P3 in each modality, suggesting that
 the functions of N2  and P3  are distinct.  Since N2
 preceded P3 in both modalities, it seemed reasonable
 to infer that N2 reflects  the detection of missing
 stimuli  (or  target selection)  and P3 reflects some
 other process.
          AUDITORY
                                  VISUAL
 Fig. I.  Scalp distributions of N2 and P3 associated
 with  omitted auditory and visual  stimuli  (grand
 mean  of eight subjects).  Dots represent electrode
 placements,  all  referred to  the nose.   Successive
 isopotential  curves  depict  areas in which  response
 amplitude was 90, 70, and 50% of maximum.
     If the  above interpretation is valid, then  N2
should  also  be associated  with target  selection in
vigilance tasks where targets are physically present.
Simson et al.  (1977) conducted a vigilance experi-
ment where subjects were required  to detect random
changes in pitch embedded in a train of  tones in one
condition and  random changes  in the orientation of
flashes  in  another condition.  Since,  due  to  over-
lapping latencies of the two components, N2 elicited
by physically present targets is often obscured by P2,
which is elicited  by any physically present stimulus,
ERPs  associated  with  nontargets  were  subtracted
from ERPs associated with targets in each condition.
The resultant waveform looked similar to  that elicited
by omitted stimuli, The N2 component associated
with targets again had a central  focus in the auditory
condition and a  preoccipital   focus in  the visual
condition. As with the omitted stimulus  experiment,
P3 had a parietal focus In both conditions (Fig. 2).

-------
Intracranial Sources of ERPs
                                            525
       AUDITORY
                                VISUAL
 Fig.  2.  Scalp distributions of N2 and P3 in audi-
 tory and visual vigilance tasks based on waveforms
 derived by subtracting ERPs of nontargets from
 ERPs of targets (grand mean of eight  subjects).
 Electrodes and isopotential curves as in Fig. 1,
    Results of these experiments, therefore, are con-
sistent with  the  hypothesis  that  N2  is related  to
discrimination  of  targets  within  a  given  sensory
modality, whereas P3 is associated with some other
process that is not modality specific. The implications
of these findings for the functional significance of P3
are discussed elsewhere (Ritter, this volume).

    In  conclusion,  the  purpose  of investigating
location  of specific ERP components, along with
related  matters  such  as latency  and conditions of
occurrence, is  to  provide  information  useful  for
discovering brain mechanisms. Subsequent  indenti-
fication of actual sources of specific  ERP  compo-
nents will  no doubt constitute a substantial under-
taking.  But that  is not enough. The elaboration of
brain mechanisms requires  knowledge of the func-
tional  significance,  as well  as the neuroanatomical
source, of the components  under study. Determina-
tion of the functional significance of ERPs, moreover,
may  be a more difficult  problem than localizing
sources.

     Consideration of function raises other important
questions. Are the biophysical substrates of ERP com-
ponents and their  associated functions identical, or
are the sources of ERPs but part of larger and more
complex configurations that underlie these functions?
Do the fields generated by ERP sources perhaps have
functional significance, or are ERPs essentially by-
products of the brain mechanisms we seek to under-
stand? Whether these  brain events, in turn, are related
to conscious experience, and if so how, is so beyond
our grasp  as to preclude even the most elementary
speculation. The road ahead seems to get steeper the
further one looks.  The difficulties of the task, how-
ever, are more than  matched by the importance of
 the goal.

-------
 SCALP  TOPOGRAPHY  IN THE LOCALIZATION  OF
 INTRACRANIAL EVOKED POTENTIAL SOURCES1

 W. GOFF, T. ALLISON, P. WILLIAMSON, AND J. VAN GILDER
 Neuropsychology Laboratory, Veterans Administration Hospital, West Haven,  and
 Departments of Neurology and Neurosurgery, Yale University School of Medicine, New
 Haven, CT, U.S.A.
     Topographical analysis is being used increasingly
 in  evoked  potential (EP) work. Some investigators
 have used it to try to separate intracranially generated
 from extracranially generated  potentials e.g., Mast
 1965, Cracco and Bickford 1968, W. Goff et al. 1969,
 Picton  et  al.  1974);  others have used  it to infer
 neural origins of presumed intracranial components
 (e.g., Vaughan  1969,  Vaughan and  Ritter 1970);
 still others use it primarily  to infer whether or not
 EP  components have common neural origins (e.g.,
 Donchin et al.  1975, Simson  et al. 1976) without
 emphasizing  their location. We have  recently com-
 pleted  an  extensive topographic study including  all
 three types of analyses mentioned above (G.D. Goff
 et al. 1977, Allison et al. 1977).

     Somatic, auditory, and visual EPs were compared
 in  the same  subjects  at all  10-20 system electrode
 locations  plus supplementary  locations  to  assess
 ocular and  myogenic  activity.  We have  also been
 recording  intracranially from  multi-contact depth
 electrodes  and directly from the cortical surface  in
 man. Bringing together the scalp and intracranial data
 permits a comparative  analysis of potentials recorded
 from the surface and depth of the brain with their
 reflection on the  scalp. While the topic of this sym-
 posium  panel is EPs as a dependent variable  without
 necessarily  making inferences about cerebral origins
 per se, we take the position that  in practice inferences
 about origins  are  implicit in any study in which
 topography is  analyzed. Thus, our analysis provides
 information relevant to critical questions for this
 panel and for EP research in general: How useful are
 scalp topographies  in the localization of EP genera-
 tors? Under what circumstances are they accurate or
 inaccurate?

    Since all of our cortical surface recording has
 been  to  somatosensory stimulation, we will report
 only somatosensory evoked potential (SEP) results.
 We constructed  SEP scalp topographies for  ampli-
 tude,  using  base-to-peak  measurements; latency
 changes between most  electrode locations were so
 small  that  isolatency  plots  were unin forma live.
 Somatic stimulation was median nerve shock  to the
 right  wrist.  Generally,  we  obtained  six to  eight
 replications,  using averages of 64  trials per subject.
 The consistency of the data justified making measure-
 ments on "grand" averages for each subject.

    Component identification  is a  problem even at
 one electrode location,  and identification of corre-
 sponding components across locations is a still bigger
 problem. Component identification was aided by
 obtaining a  grand average  across subjects for each
 location and using  this  average as a guide for indi-
 vidual subjects. Examination of coronal and anterior-
 posterior arrays and construction  of latency  histo-
 grams were  also useful  in  identifying  components.
 In  addition  to problems of identification, arbitrary
 decisions have  to   be  made  about  measurement
 methods  in  cases where the component appears to
 be  distorted in various  ways by  other  activity.
 Examples of methods used in  this  study to make
 base-to-peak  measurements are illustrated in Fig. 1.
 These problems do  not appreciably affect the area
 of maximal amplitude of a particular component, but
 they do affect the total  region enclosed by isopoten-
 tial lines. In the interpretation of any scalp topo-
 graphy, the particular measurement methods used in
 its determination must be kept in mind.

    First, we will consider scalp-brain relationships
 for early SEP components. Fig. 2 shows schematically
 the first 40  msec of the SEP recorded at P3. Also
 shown schematically are P20 and N35 components
 recorded  anteriorly  as illustrated at Fpl. Below (in
 Fig. 2) are shown the topographies of these compo-
 nents. It is generally accepted that  P15 reflects the
subcortical afferent  volley and that N20, P2S, and
Supported by Medical Research Service, Veterans Administration, and USPHS MH-05286, RR-05358, and
NSF GB-5782.

-------
Comparison of Scalp and Epicortical Recordings
                                                         527
                     AMPL.,
                      Mv   TRANSFORM
                           +2  4
                                       33
+7
+8
t10
+6
+0
9
10
12
8
2
75
83
100
67
17
                          +6    8

                          +5    7


                          +4    6
67

58


50
              tions an equal amount so that the most negative value
              at any location became zero. Adding or subtracting a
              constant from all values does not change the relative
              amplitude across electrode location.  After the base-
              line shift transformation, the largest amplitude value
              for  the particular component  was taken as 100% as
              shown for Pz.   Other locations were then converted
              to appropriate percentages.  The areas within which
              the  component was 75% or greater, and between 50
              and 75% of its maximum amplitude were plotted as
              shown  in the  schematic head.   Isopotential  lines
              were determined by  linear  interpolation between
              electrode locations.   (From G.D. Goff  et al. 1977.)
                                                                          P15
                                                                             P20
                                                                    Fp1-
      N20
6  10 20 30 40
Fig.  1.  Schematic illustration of measurement meth-
ods used to derive the scalp  topography of an SEP
"components."    The  topography  based on these
measurements is indicated below.  In this and all sub-
sequent figures, EP recording is referential to the ears,
positivity  is  up, and stimulation is  to  left  or right
median nerve as indicated.  Examples of base-to-peak
amplitude  measurement  are  shown at several loca-
tions.  At  Cz the  component is not visible; at T6 it
is visible but has  a base-to-peak amplitude of zero;
at C4  it has a negative amplitude; at O2 it appears
only as an inflection on  another component.  Note
that at C4,  what  was clearly a positive component
at other locations has a negative value because the
small  positivity was apparently riding on  a larger
negativity.    A similar problem is illustrated at T6
where the  positivity has  a zero value even though it
is visually apparent.
To compensate for these cases, we transformed ampli-
tudes by shifting the baseline for all electrode loca-
              Fig. 2. Schematic of SEP early components and their
              topographies. X's indicate locations where in any sub-
              ject the component was 90% or greater of the maxi-
              mum amplitude; this indicates across-subject variability.
              Numbers in parentheses  are the number of subjects
              (out of total  of 12)  in whom the component was
              observed. (Adapted from G.  D. Goff el al 1977.)

              P30 reflect  the primary cortical response. But after
              we first  recorded these  components (in 1960), we
              wondered why N20 and P30 were  recorded  maxi-
              mally  so  far posterior to the presumed location of
              the central sulcus. In the present topographic study,
              we noted a positive component, P20, anterior to
              the central  sulcus;  P20 corresponded  in latency to
              N20. We also noted a less precise but possible corre-
              spondence  between  the  posterior P30  and the
              anterior N35. These topographies suggested a polarity
              reversal of N20 and P30 across the central sulcus,
              a suggestion that is consistent with the observations

-------
 528
                                       Goff et al.
 of  Broughton  (1969)  recording from the cortex.
 The anterior-posterior relationships can be seen more
 clearly in  selected  subjects  for  whom the closely
 spaced electrode  arrays shown  in Fig. 3 were used.
 The results indicate  a  polarity  reversal across  the
 central sulcus, an indication  that in turn  suggests a
 dipole, oriented  parallel to the  surface of the cortex
 and normal to  the  bank of the central  sulcus, as
 Broughton suggested.
   P20
   N20
  N3S
  P30
                    0 20 40 60
                       msec
 Fig. 3. SEP topography from a selected subject (Column
 C) compared to the group average (Column A). The po-
 tentials in Column  B were recorded from the array
 shown on  the upper head of Column C.  (Adapted from
 Allison et al, in press.)
     We compared  these  scalp results to direct pial
 surface recordings as shown in Fig. 4. Scalp records
 from Fz and  P4 were compared to transcortical and
 pial surface recordings anterior and posterior to the
 central sulcus. Cortical surface recordings were from
 the   right  hemisphere.   Both  surface  and  scalp
 responses were evoked by left median nerve stimula-
 tion. Electrode  2 was a  transcortical electrode with
 the  depth  probe in white matter immediately under-
 lying  the  surface  electrode.  The polarity  reversal
 across the central sulcus between electrodes 1 and
 surface 2  is clear. The lack of polarity reversal from
 surface to depth in  the  postcentral  gyrus indicates
 that N20-P30 are not generated  immediately under
 this electrode, which was located on the posterior
 aspect of  the postcentral  gyrus. Notice  also that
 no P25 is seen. That this is due to the posterior place-
 ment of the electrode is suggested by the results from
 another patient (Fig. 5) in  whom a more favorable
 electrode   placement was  achieved.  The  polarity
 reversal of N20-P30 between postcentral electrode
 6  and precentral electrode  4 is  clear.  At electrode
 5, which  was very  close to  the central sulcus,  a
large  P25  was apparent;  this component  does not
 reverse across  the  central  sulcus  but  diminishes
 posteriorly  and  anteriorly.  This  finding  suggests
 that P25 is generated at, or very near, this electrode
 and  that the dipole source  is oriented orthogonally
 to the surface of the gyrus.
   Fz

    1  -S



SURFACE
       _r

   2

 DEPTH   \
                           ANT.
                                                                                               LAT.
                                                             0 2040 60 80 100
                                                                   msec
                        CORTICAL STIMULATION
                          B) HAND FLEXION
                          IB FOREARM DORSIFLEXION
                          Q ARM ABDUCTION
                          USEYELID MOVEMENT-
                            NECK TWIST
 Fig.  4. Comparison of scalp and cortical records from
 SEP early components. Boxed letters indicate location
 of electrodes from  which cortical stimulation elicited
 the motor responses  listed  below.  C.S.  indicates  the
 location  of the central sulcus of the right hemisphere.
 This and all subsequent cortical surface SEPS shown are
 grand averages  of three  replications of 48  individual
 responses. Left median nerve stimulation. (From Allison
 et al., in press).
                        ANT.
                                          LAT.
  0  20 40 60 80 100
        msec
CORTICAL STIMULATION
 A WRIST-FINGER FLEXION
 B CHIN-MOUTH MOVEMENT
 C MOUTH MOVEMENT TO LEFT
 D MOUTH CLOSING
 E MOUTH MOVEMENT TO LEFT
Fig. 5. Cortical surface recording of SEP early compo-
nents. Other details as in Fig. 4. (From Allison et al., in
press).
    Fig. 6 summarizes our theories about the origins
of the  primary  complex, N20, P20,  P25, P30, and
N30,  and explains  the  observed topographies. N20
and P30 actually represent the "depth" side of the
equivalent dipole source located orthogonally to the
bank of the central sulcus and parallel to the surface

-------
Comparison of Scalp and Epicortical Recordings
                                             529
of the postcentral gyrus as shown. Thus, at the sulcus,
there is a "null" point, which explains why the com-
ponents  are  not well  recorded  at central Scalp
electrodes,  but  are better  recorded in  the parietal
area.  The  "surface"  side  of  the  dipole for these
components is  seen  in scalp recordings  anterior to
the  central sulcus,  as  P20 and N30. P20 is  usually
clearly  seen, while  N30 is  seen  only  in  selected
subjects  and   is  frequently  obscured  in  scalp
recordings by the larger  N35 (Fig. 2), whose origins
are   unclear.   P25,   the   equivalent   dipole   of
which  is orthogonal  to  the surface of  postcentral
gyrus,  is best  recorded  from  a  restricted  area of
central scalp near C3 or C4 (Fig. 2).
                              P25
                                         P30
                                     N20
Fig. 6.  Probable  cortical origins of SEP early compo-
nents. (Adapted from Allison et al., in press.)

     Parenthetically, this interpretation of the origin
of  N20,  P20, P30,  and N30 differs from  that of
Papakostopoulos  et al. (1975). They suggested  that
the   precentral   and  postcentral  components   are
generated  independently  in  motor   and   somatic
cortex  because  these components  were  differen-
tially affected by  flexion of the  left  index finger.
Amplitude measurements from their Fig. 2  indicate,
however, that N20 and  P20 were unaffected by
finger movement while P30 and N30 were essentially
abolished. In neither case were the postcentral com-
ponents (N20 and P30) differentially affected com-
pared to their precentral counterparts (P20 and N30).
Thus, the results of Papakostopoulos et al. (1975) do
not contradict our  interpretation  that N20-P20 and
P30-N30 are both generated in the posterior bank of
the central sulcus.

     How accurate  are these scalp topographies for
 source  location? They are reasonable  for a compo-
nent like P2S, where there is a restricted source, the
equivalent  dipole  of which is  orthogonal  to  the
cortical surface. They can be accurate for compo-
nents like N20 and P30, where the equivalent dipole
is  parallel  to the cortical surface  if, and only if,
potentials can be recorded  from both sides of the
dipole unobscured by other components. Then, the
null  point of the polarity reversal gives a reasonable
indication of the source. If either side of the dipole
is  obscured,  the  topography can be misleading. We
were  puzzled by the  parietal amplitude maximum
of N20 and P30 (W. Goff et al. 1962) because we did
not recognize their polarity reversal correspondence
to the anterior components until Broughton's cortical
recordings caused us to look for it.

    An example  of a scalp  topography that is vir-
tually  useless in locating cerebral origins  because of
temporal  coincidence  of multiple  intracranial  and
extracranial  EPs  comes from  scalp-brain compari-
sons  of SEP late  components. Fig.  7  shows the
topography of P100 followed by the somatic vertex
potential N140-P190. The vertex potential is focused
centrally  around Cz and  is rather diffusely distri-
buted. P100 is shown here  as having  two foci.  This
topography is based on subjects who do not show a
large  frontal PI00, such as that shown in Fig. 8. A
topography based on a subject like that of Fig. 8
would not show  a dual focus since the frontal PI00
swamps the smaller posterior P100. Note  that in this
subject,  P100 is  still large  at F7,  F3, and Fz.  One
obtains a dual focus from subjects  such as shown in
Fig. 9 who have a moderate frontal P100. Note also,
that in contrast  to  the previous figure,  the frontal
P100  is considerably diminished at F7 and F3 and
essentially  missing  at  Fz.  At  C3,  its  amplitude
increases again and continues to inciease posteriorly.
We could not separate the frontal and posterior P100
on the  basis of latency. The  distribution of the
frontal P100 suggested an extracranial response. This
suggestion was supported by the fact that the frontal
PI00 shows great adaptation over repeated recordings;
the  posterior P100 does not.  The frontal P100 is
larger in naive  subjects  compared to  experienced
subjects and also is larger for unpredictable stimuli
compared  to stimuli delivered  at a fixed repetition
rate.

     The extracranial nature of the frontal P100 and
the  source  of the posterior PI00 can be verified in
direct brain recordings. We have  examined records
for those patients who  had  frontal  depth probes and
find  no  indication  of an  intracranially generated
frontal PI00.  The  posterior  PI00  appears to be
Intracranially generated, as shown  in  Fig. 10, which
shows  early and  late  SEP components  recorded
directly from the  pial surface of the  postcentral
gyrus. A scalp response from Cz is shown for  com-
parison; note the difference in amplitude calibrations.
At electrodes 2 and 3, where we record the primary

-------
 530
                                       Goff etal.
 complex maximally (100-msec time base), we observe
 on the 500-msec time base a large amplitude posi-
 tivity  at 100 msec followed by a negative potential
 peaking at approximately  200 msec. Note also, that
 this P100-N200 complex  is seen only  where  the
 primary complex is  seen and  only  in  response  to
 contralateral  stimulation.  At  medial  and  lateral
 locations,  or in response  to ipsilateral  stimulation,
 P100-N200  is not apparent, and  the waveform is
 more  similar to that of  the  scalp-recorded vertex
 potential.  We have termed  this cortical P100-N200
 complex the somatic late  response or SLR. Fig.  11
 presents similar data  from another subject in whom
 we were able to place an  array of electrodes across
 the  central  sulcus and  roughly  orthogonal to  it.
 At electrodes 11, 13, 18,  and 15, which progressed
 from the  parietal area anteriorly towards the central
 sulcus, there are increasingly well-developed primary
 components  on the   100-msec time  base  and  an
 increasingly  well-developed P100-N200 sequence  on
 the  500-msec  time  base.  Most important,  there is
 a  polarity   reversal  of  the primary components
 between postcentral  electrodes  18 and 15 and pre-
 central electrode  7  and  a similar polarity  reversal
 of  the SLR between  the  same  electrodes. These
 results have been replicated in other subjects.

    We can  now interpret what was a very puzzling
 part of our  topographical analysis before  we were
 able to record  directly from the cortex.  There  are
 two PI 00 components, as the  topography based  on
 those subjects with moderate or small PI 00 indicates
 (Fig. 7); the  frontal P100 is apparently extracranially
 generated,  probably in periocular  and possibly fore-
 head musculature, but usually  does  not give rise  to
 an  overt eye blink.  The  posterior P100  reflects  an
 intracranial response, apparently generated in neurons
 either  identical to, or coextensive with, those giving
 rise to the  primary  N20-P30 complex.  Like N20
and P30, the  polarity reversal across the central sulcus
indicates that the equivalent dipole  of these genera-
tors  is  oriented  orthogonally  to  the bank of the
central  sulcus and  parallel  to the lateral surface of
the cortex. This finding explains the posterior parietal
maximum of the scalp topography of PI00, similar to
that of N20 and  P30. On  the basis of scalp topo-
graphies alone, it is not apparent that one side  of a
dipole is observed since the  polarity reversal of P100
across the central sulcus is, it  seems, obscured by the
frontal, extracranial PI00.  Finally N200, the nega-
tivity of the SLR, is obscured in scalp recordings by
the positivity of the vertex potential (PI90). Thus,
recognition  of the  existence of the P100-N200  SLR
and of the location of its source would not be possi-
ble from scalp recordings alone.


    In  summary, a comparison of topographic  data
with  the  results  of direct brain recording indicates
that topographic analysis can be quite  accurate in
localizing the intracranial generator  in  the  case of
components generated by focal sources in which the
equivalent   dipole   is oriented orthogonally  to the
cortical surface, as  for example is the case for  P25.
Further, it can be accurate for a source whose equiva-
lent dipole  is oriented parallel to the surface of the
cortex  if,  and only  if,  activity  generated by  both
sides of the dipole is sufficiently  uncontaminated by
other potentials to be observed adequately. To know
what  kind of a source one is dealing with in order to
interpret the scalp topographies, however, one needs
to  record directly  from the brain. If, on  the other
hand, one  can record directly from  the brain, one
does not need scalp topographies for the localization
of sources. Thus, our data indicate that, in the general
case,  scalp  topographies  alone are not  particularly
useful  in determining  the  locus of generation of
cerebrally  evoked  potentials.  This  conclusion is
consonant  with a  similar  conclusion  reached for
movement  potentials by Cerbrandt  et  al.  (1973).
Topographies can be useful as a dependent variable to
indicate  similarity  or difference in  the source of
components without drawing  any conclusions about
precise cerebral origins.
      P100
      (7)
         Fig. 7. Scalp topography of some SEP late components. (Adapted from G.D. Goff etal. 1977.)

-------
 Comparison of Scalp and Epicortical Recordings
                                                                                             531
                                        0    200    400
                                               msec
                                                                                    0    200   400
 Fig. 8. Scalp SEPs recorded from a subject showing a large frontal PI 00 wave, largest at FPl and FP2 and apparently

 generated in periocular and forehead muscles. (From G.D. Goffet al. 1977.)
F,
T,
0    200   400      0    200   400
200   400
   msec
                                                                0    200    400
                                                                                     0    200    400
Fig. 9. Scalp SEPs recorded for a subject showing a relatively small PI 00 wave. (From G.D. Goffet al.  1977.)

-------
532
                                                     Goff et al.
      SEP • SCALP
                                                                               PT NO. 1
                                                                    RIGHT HEMISPHERE

                                                                      MEDIAL
                                                                                         ANT.
                                                               Q] WRIST FLEXION
                                                               J] FOREARM FLEXION
                                                               (3D ARM ABDUCTION
                                                               IE EYELID MOVEMENT • NECK TWIST
            0  20 40 60  80 100
0  100  200  300 400
       HUM
      Fig. 10.  SEPs recorded from postcentral gyrus. Motor responses elicited by electrical stimulation
      (boxed numbers) are indicated. C.S. indicates location of central sulcus. Solid traces are respon-
      ses to left, and dotted traces are responses to right, median nerve stimulation. (From W.R. Goff
      et al, in press).
                                                  SEP • PIAL SURFACE
         ffl MOUTH MOVEMENT
         If HEAD MOVEMENT
         JjFOREARM- INTERNAL ROTATION
             AND FLEXION
         M FINOER AND WRIST EXTENSION
         U MIDDLE FINDER EXTENSION
                  AND TWITCH
     POST.
          ODHAND-INTIRNAL
              ROTATION, THUMB TWITCH
          Zl SHOULDER TWITCH
          SlJAW CLOSE, SWALLOW
                                               PTN0.2
                                            0  20  40   M   •»  100      0   100   200   MO  400
      Fig. 11.  SEPs recorded from pre- and postcentral gyrus. Details as in Fig. 10. (From W.R. Goff
      et al., in press).

-------
SPATIAL  FREQUENCY ANALYSIS OF AN EEG  EVENT
IN THE  OLFACTORY BULB1

W. FREEMAN
Department  of Physiology-Anatomy,  University  of California  at Berkeley,
Berkeley,  CA.,  U.S.A.
     The  purpose  of this paper is to describe the
derivation of the spatial spectrum of the EEG re-
corded transdurally over the surface of the olfactory
bulb in cats and rabbits. The information is needed to
determine the optimal spacing of electrodes in surface
arrays.

    There is a close parallel here to the computation
of the  temporal spectrum of the EEG  (Matousek
1973) recorded monopolarly with a single electrode
at a point on the  surface. From a single  channel, a
continuous  record made with  photographic  film
suffices to show the maximal and minimal rates of
change of EEG potential with respect to time (tem-
poral frequencies). When such a signal is digitized to
compute its temporal spectrum, the digitizing time
(the  interval between measurements) is  made less
than  half the wavelength of the maximal  frequency,
and the duration of the digitized record is made to
exceed  the  half-cycle length  of the minimal fre-
quency.

    It is not similarly possible to record EEG poten-
tials in a  continuous manner with respect to space.
The potential must be recorded concomitantly at a
set of discrete points from an array of electrodes, and
the interelectrode  distance constitutes the digitizing
interval with respect  to the spatial variable. If the
maximal and minimal rates of change of EEG poten-
tial with respect to distance (spatial frequencies) are
not known beforehand, the  interelectrode spacing
and array length  must be determined empirically.
From a  set of  digitized measurements of potential
made simultaneously with a linear array of electrodes,
the spatial spectrum is obtained by the same compu-
tations used for the temporal spectrum.

     In this study the focus of interest was not in the
EEG as a potential function but  in the spatial analysis
of a  particular type  of  event or wave complex that
manifested a distinctive and recurring neural activity
pattern in the olfactory bulb, i.e., the sinusoidal burst
that  commonly accompanies each inspiration  and
that Adrian (1950) termed the "induced activity" as
distinct from the temporally intervening "intrinsic
activity."
    The olfactory bulb is a part of the paleocortex,
with roughly the size and shape of a pea in the
rabbit  and a lima bean in the cat. The outer  layer
facing the dura contains the afferent axons from the
olfactory receptors comprising the primary olfactory
nerve (PON). The innermost  layer surrounding the
ventricle  consists of efferent and centrifugal axons,
of which  the most  prominent set forms the lateral
olfactory  tract (LOT). The PON axons synapse  on
the  apical dendrites  of  large excitatory neurons,
called mitral cells, which send their axons in the  LOT.
The  basal  dendrites of the mitral cells form reciprocal
synapses with inhibitory interneurons, called granule
cells, in  the middle  layers of the bulb (Shepherd
1972). The  negative feedback loop formed by the
mitral  (excitatory) and granule (inhibitory) cells is
the principal mechanism for the sinusoidal oscillation
of the induced wave  in response to receptor  input
(Freeman  1975). The potential field  manifested at
the surface of the bulb is a dipole field generated by
the granule cells that alone have the requisite location
and  cytoarchitecture to generate such a field  (Rail
and Shepherd 1968).

     Spatial analysis  of the bulbar-induced wave was
accomplished in four stages. First, records were  made
from single electrodes  implanted at various depths in
the  bulb  to determine the temporal  characteristics
of the  induced wave in both anesthetized and waking
animals. Second, arrays of 60 electrodes (6x10) in a
rectangular  matrix  were implanted on  the lateral
aspect  of the bulb  to determine the spatial  distri-
bution of the induced wave over the bulbar surface.
Third,  linear arrays of 64 electrodes (1x64) were
implanted over identified foci of induced activity to
determine  the spatial frequency spectrum. Fourth,
the spatial spectrum of the granule cells was predicted
from volume conductor theory.

     Chronic implanation was used  primarily be-
cause the  induced  wave  could be found most  re-
liably  when the animals were waking, healthy, and
motiviated,  e.g., by hunger, curiosity, or fear.  In
quiescent  and anesthetized animals, the amplitude of
the EEG was low, and  the induced waves were usually
not clearly distinguishable from intrinsic activity.
 Supported by Grant MH06686 from the National Institute of Mental Health.

-------
 534
                                        Freeman
 Methods

     Observations were  made  on adult rabbits  and
 cats  of both sexes.  Single electrodes were inserted
 stereotaxically through  burr holes in the skull over
 the bulb. Electrode arrays were placed on the lateral
 surface  of the bulb  following orbital exenteration
 and removal of the bone overlying the bulb. In acute
 experiments  under   anesthesia  (pentobarbital  40
 mg/kg I.V.) the dura was also removed, but in chronic
 implants it was left intact.

     The arrays were placed on the  bulb  through an
 opening in the skull cut with a drill just large enough
 to receive them. The orbit was filled with agar  and
 closed with dental cement. Array  recordings were
 rhonopolar with  respect to reference electrodes over
 the cerebellum and  in the floor of the  orbit. The
 orbital reference was useful in abating the receptor
 potential  from  the   mucosa  without significantly
 diminishing or altering the spatial distributions of the
 induced waves.

    The recording  electrodes were stainless-steel wire
 insulated with  Formvar.  Single-channel  recordings
 were   from pairs  of wires 250  micrometers (ftm) in
 diameter, which were implanted on opposite sides of
 the bulbar  dipole  field. The  6x10  electrode arrays
 (4x7  mm) were of 250-jtm wires at spatial intervals
 of 800 /urn  (center to  center).  The  techniques
 for making  the  arrays for implantation  have been
 described  (Eastman   1975).  The  1x64  array was
 made  of 25-p.m wire. The interelectrode distance
 (center to center) was 38 to 42 im and the width of
 the array was 2.4 to 2.6 mm.

    Signals were amplified (10k gain), filtered (3  dB
 falloff at 10 and 300  Hz), multiplexed, and digitized
 at  1-msec intervals for each channel in 100-to 900-
 msec  epochs. Records were displayed on  an oscillo-
 scope  to check for artifacts (e.g., movement poten-
 tials,  open  channels,  polarized electrodes) and to
 determine  the precise time locations of induced
 waves. Data were then written on magnetic tape  for
 off-line processing.

    Prior to each  recording session, animals  were
 deprived of food (but not water) for 24 to 48 hours.
 They  were placed  in a closed  box with continuous
 inflow of charcoal-filtered air and given time to settle
 down. Odors were  introduced  without interrupting
the air stream. Six to eight records were made of
induced  activity  before  and  during the  odors  of
food or of a chemical substance (e.g., butyric  acid,
amyl   acetate,  clove  oil,  methyl salicylate).  Most
records were taken  from  1 to 4  weeks after implanta-
tion, after which the amplitudes of activity tended to
deteriorate.
     Postmortem dissection was done to verify  the
 placement  of electrodes and  the  extent of tissue
 reaction. Regrowth of bone under the  array, dural
 fibrosis, and pressure marks in the bulb were observed
 in some animals. In those animals  giving the results
 described here, the dura was normal  in appearance
 and uniformally translucent over the  full  extent of
 the  array, and the  bulbs were slightly  flattened over
 their lateral aspen in conformance  to  the  planar
 arrays.

     Off-line data  processing of records from 6x10
 arrays was  as follows. The data comprising a single
 burst of induced  activity  lasting  80 to 300 msec
 were read off magnetic tape into a  CDC  6400, and a
 graphic display  of  the 60 waveforms was made with
 the Cal  Comp plotter.

     For the great majority of block records showing
 relatively constant  waveforms  over the  array,  the
 spatial  distributions of the amplitude and time  of
 occurrence of the sinusoidal burst were determined.
 First, the mean  amplitude for each  channel was sub-
 tracted  from  the signal  on that channel. The root-
 mean-square  (rms)  amplitude was  found  for each
 channel, and the 60 values were used to make a con-
 tour plot and perspective  drawing  by second-order
 extrapolation  of the  amplitude distribution  of  the
 induced wave.

     Second, an  ensemble average was made over the
 60  waveforms. A product-moment  correlation coef-
 ficient was computed between  the  ensemble average
 and  each of the 60  waveforms. This was repeated
 after the waveform from each  electrode was lagged
 +1, +2, -1, and  -2  msec. The five correlation coeffi-
 cients from  each site were fitted with a parabola, and
 the most likely  time lag of each waveform from the
 ensemble waveform was taken as the time difference
 between the apex of the parabola and zero time.

     The dominant  frequency of the sinusoidal burst
 was found by the  Fast Fourier transform, and each
 time lag was  expressed  as  a phase with respect to
 zero phase for the  ensemble. The 60 values of phase
 were displayed  in contour plots  and  perspective
 drawings as in the case of the amplitude distributions.
 In combination, these displays  described the surface
 location,  distribution,  and  time  of  occurrence
 (relative to  the  mean) of the EEC  burst manifesting
 the induced  activity.

    The same computations were applied to the data
 from 1x64 arrays, except that the display was in the
 form of amplitude and phase as functions of distance
along the array. The data sets were each normalized
to zero  mean  and unit  variance, and  the auto-
covariance and  spatial spectra  were computed and

-------
Spatial Frequency Analysis

plotted for amplitude and phase at intervals of 0.2
cycle/mm  from  0  to  12.5  cycles/mm.

    The correctness of the connections of the arrays
of  electrodes and  the  amplifiers was  verified  by
immersing  each  array in a shallow dish of Tyrode's
solution and applying an  alternating current at 40
Hz across  the conducting volume. The two  current
electrodes  were  placed one at each end of the array,
and the reference electrode was placed several centi-
meters  to  one side of the array. When records were
made and  processed in the standard way, any errors
in  the  sequence  of connections were  revealed by
irregularities in the plots of potential versus distance.
The data were rearranged by the computer.

     The precision of the measuring system was deter-
mined by recording a sine wave in Tyrode's solution.
In  this  case, one current electrode was placed  close to
the reference electrode and the other  several centi-
meters  distant  from the array on its  opposite side.
With monopolar recording, the expected normalized
amplitude  distribution  was unity  and  the expected
phase distribution was zero. For an rms amplitude of
70/M.V, simulating a typical burst of induced activity,
 the standard deviation of amplitude from unity was
 ±1.1%  (±0.8juV) and the standard deviation of ampli-
 tude from unity was ±0.012 radian (0.7 degree, which
 is 0.002 cycle, equivalent at 40 Hz to 0.05 msec). On
 this basis,  the system was judged to be capable of dis-
 criminating lag  time differences in EEC patterns on
 the order  of 1/20 of the digitizing time interval (I
 msec). For low-amplitude or low-coherence signals,
 the precision fell off rapidly.

 Results
 Spatio temporal patterns of induced waves in
 the bulbar EEC
     The typical form of the  olfactory bulb  EEC and
 its relation to  respiration are shown  in  Fig. 1. With
 each  inspiration there  was a surface-negative,  deep-
 positive (Fig.  la, lower trace, downward  positive)
 shift in potential that reversed with expiration (down-
 ward in the upper trace). When it occurred, the sinus-
 oidal  burst was seen to start and end slightly  after
 the start  and end  of inspiration.  When  the low fre-
 quency wave  was filtered out  and  the respiratory
 rate was slowed (Fig. Ib), the distinction between the
  intrinsic (between bursts) and induced activities was
  most clearly seen.

      Fig. 2  shows an example of the  60 waveforms
 recorded  from  the surface of the bulb and anterior
 olfactory  nucleus of a rabbit during a  single burst.
 Fig. 3 shows the amplitude and phase distributions
 derived from the olfactory bulb of a cat. The bottom
 of each frame  represents the anterior aspect of the
 left bulbar surface, and  the right edge corresponds
 to the ventral edge of the bulb.
                                             535

    In each of 6 cats  and 12 rabbits, the following
features  were  found.  There  were  one to  several
localized domains of high amplitude activity either
within the borders of the array (Fig. 3) or (by extra-
polation)  adjacent  to the  array. These  foci  were
irregular in size, shape, and location, varying unpre-
dictably  from animal to animal. However, for each
animal they were relatively constant in location and
form  over  several weeks of observation. Variations
were  analogous to  those  of a signature that has a
certain pattern, but is never twice the same. On visual
inspection, the variation in pattern appeared to be no
more  or less  between  two successive bursts during
presentation  of a certain odor than  it was  for two
bursts recorded at different times or on different days
in response to different odors.  There was no apparent
relation between  the size,  shape, or location of foci
and the type of odor presented, the overall amplitude
or duration of the burst, or the kind of motivational
state, except that bursts were reproducibly observable
only  if some form of motivation was present  (as
implied by food deprivation, nociceptive stimulation,
exploratory behavior, aversive behavior, etc.).

    The phase plots over the bulb were somewhat
less regular than the amplitude  plots. That is,  the
pattern for each animal was clearly recognizable from
trial to trial,  but  the irregular  variations were greater
than  those for amplitude. The maximal and minimal
values for phase most often occurred near to, but not
at, the amplitude peaks.  A typical pattern  revealed
a  phase maximum  and phase  minimum on  opposite
sides  of an  amplitude peak  in a  focus of  activity
(Fig.  3). In about half the animals, there was a per-
sistent weak  phase  gradient across the array in  the
direction  of  PON  axonal  conduction (anterior to
posterior), with phase lead (positive values) anteriorly
and phase lag (negative values with  respect to the
mean) posteriorly. In records  from the other animals,
there was  a  weak dorsoventral phase gradient or no
gradient.

 Spatial spectral analysis of induced waves

    The fine structure of active foci recorded at the
 surface was  explored  by placing 1x64 arrays  across
 them. Initially, five of the animals with implanted
 arrays  were  re-anesthetized with pentobarbital.  The
 relatively weak induced activity present under anes-
 thesia  was recorded and  processed.  The  amplitude
 and  phase distributions were less precise due to the
 lower amplitudes,  but the spatial distributions were
 similar to those found in the waking state.
      With this assurance,  a  6x10 array was  placed
 on the bulb  of  an  anesthetized  animal and  the loca-
 tion  of a focus was determined. Then the 6x10 array
 was  replaced with a 1x64  array running from anterior
 to posterior  across the  maximum of the focus. Five
 animals (three rabbits and two cats) were implanted
 in this manner.

-------
 536
                                                                                     Freeman
                                                  (a)
                                                                                   500/iV
                                                                                   500 msec

F/£.  7. /a,/ ££0 bipolarly recorded flower trace,  negative upward) from the olfactory bulb of a waking rabbit (inner
with respect to outer layers) and respiration recorded (upper trace, inspiration upward) with a thermocouple in the
ipsilateral nares. EEC filter settings:  0.1 Hz/1 kHz. (b) EEC from anesthetized rabbit after tracheotomy with pump-
controlled airflow through nares. Filters settings:  10 Hz/1 kHz. From Freeman 1976.
F16  E31   3/7/75
                             T2908    FOOD 3
                                                           BIN  760 TO 840
Fig. 2. Concomitant EEC records of an induced -wave in a waking rabbit from a 6x10 electrode array (4x7 mm)
placed over the left bulb and anterior olfactory nucleus. Left side is anterior, top is dorsal.  Duration of record, 80
msec. Calibration in tiV.

-------
 Spatial Frequency Analysis
                                             537
    F21 AVERAGE OF 12 SETS
    F2I AVERAGE OF 12 SETS
                                                           AMPLITUDE
                                                                              WIN-18.3   MAX-60.3
    PHASE
                                                           PHASE
                         MIN--.119  MAX -.152
 Fig. 3. Upper frames: Contour (left) and perspective (right) plots of amplitude distribution of induced waves over the
 left olfactory bulb of a cat recorded with a 6x10 array (4x7 mm). Left side is dorsal; bottom is anterior. Range of
 rms amplitudes, 18 to 60 \iV. Lower frames: distribution of phase values. Range, -0.119 to +0.152 radian (6.8 to 8.7
 degrees).
    Induced activity was recorded and processed as
before in the  presence  of filtered air, the odor of
food, and a chemical odor. Amplitude and phase dis-
plays (Fig. 4) were  consistent with  the appearance
of cross sections through contour plots as previously
displayed (Fig. 3).

    The Fourier transform was taken of single ampli-
tude and phase plots and of the autocovariance func-
tion of sets of three to six plots placed in series. In
the  latter  procedure,  individual  records were de-
trended  to remove  amplitude and  phase gradients
extending the  full length of the array. The lower cut-
off spatial frequency was l/(2.5  mm), or 0.4 cycle/
mm,  and the  upper frequency was  l/(2x40jum), or
12.5  cycles/mm. The procedure was also  applied to
records from anesthetized animals and from arrays in
Tyrode's solution as controls.

     A set  of ten pairs of spectra for amplitude and
phase from a cat (F42) are shown in Fig. 5. Records
were taken in pairs on  different  days over a period
of 2 weeks. Each pair was during exposure to filtered
air and  to an  odor. Maximal  amplitude  and  phase
values were observed between 0.4 and 1.0 cycle/mm.
The locations  of  peaks varied unpredictably  from
trial to trial. A comparison is made in Fig. 6 between
the  spectra for amplitude  from  ten animals  under
anesthesia  (upper frame), five waking animals (upper
five  curves in  the lower  frame)  and five  control
records in Tyrode's solution (lower five curves in the
lower   frame).   Some  records  from  anesthetized
animals showed  spectral peaks near 0.6/mm,  but
most  were  relatively flat, consistent with the  low
amplitude  of the  induced activity in  that  state.
Control spectra were uneven over most of the spatial
frequency range. This was the result of the normali-
zation  to  unit   variance, which  accentuated  the
variations  due to noise. Activity present in waking
animals in  the band from 0.6 to 1.0 cycle/mm  was
not present in control and anesthetized records.

    The data are summarized in Fig. 7 as the averages
of ten spectra from anesthetized animals, ten control
spectra, ten spectra from F42, and ten spectra from
five waking animals (two from each). The significant
feature is  a spectral peak in amplitude spectra from
waking animals  near 0.8  cycle/mm that Is  absent
from  the  other  spectra. There is  no corresponding
peak in the phase spectra.

Theoretical evaluation of the spatial spectrum

    The bulbar field of potential in response to elec-
trical stimulation of the PON or LOT  was found to
have the same dipole structure as the field of the EEC
(Freeman  1972, 1975).  When  a recording micro-
electrode  was inserted  perpendicular to the bulbar

-------
  538
                                                                                               Freeman
       •F42 E1003 10/28/75 T6004 AMYL  1  ELEC C
    .04
20    30   40    50
     CHANNEL
                                         60
                                               70

   -.02

   -.04
      BIN1  TO  100 FREQ-332


 Fig. 4. Distribution of rm$ amplitudes (upper frame)
 and phase (lower frame) of an induced wave recorded
 from the surface of the olfactory bulb of a waking cat
 (F42) with a horizontal 1x64 array 2.6 mm  in width.
 Right is anterior.
 surface through the epicenter of an evoked response,
 the  potential with  respect to a distant point at  the
 crest of the initial  surface-negative peak of the gran-
 ule  cell  response (and the peak-to-peak amplitude of
 the  subsequent damped sine wave oscillation) went
 to a maximum at a depth of 600 to 700/tim, to zero
 at 800  to 1000 /zm, to  a minimum at 1100  to
 1300 jxm,  and toward zero at  greater depths (data
 points in Fig. 8). The depth profile was fitted with a
 curve  of potential  (solid curve, Fig. 8),  which was
 generated by a core conductor simulating the current
 source-sink distribution  of  the  granule cell  popula-
 tion (dashed curve, Fig. 8).

     As  previously  described (Freeman 1972), the
granule  cell was treated as  a cylindrical  core con-
ductor oriented perpendicular to the bulbur surface
with its midpoint lying at a  depth of 1000 ^.m. The
synaptic  input was to the upper  half of the cylinder,
and the lower half was passive. During an EPSP, there
was inflow into the upper half and outflow from the
lower half. The sink-source  relation was reversed
during an IPSP. In either case, the source sink density
was  approximated by a double exponential curve of
fixed charge density, q
                               zn
            = .ezn/X,
                                              0)
where z was the axis of the cylinder oriented perpen-
dicular to the bulbar surface and X= IVOjLtm was the
experimentally  determined  length constant of  the
granule cell (dashed curve in Fig. 8).
                                                                     AMPLITUDE
                                                                        234

                                                                           cycles/mm
                                                                      PHASE
                                        01      2      3      4      S      6

                                                        cyclM/mm


                                     Fig. 5. Ten spectra of amplitude and phase from a wak-
                                     ing cat (F42), taken in pain on separate days. The rms
                                     amplitude and phase  values  were  normalized  to zero
                                     mean  and unit variance prior to spectral analysis. The
                                     variation in spectra was not related to odor.

                                         The  surface coordinates  of the bulb were x and
                                     y.  The  potential, vm(x),  at  a  set of points repre-
                                     senting m recording sites at xm on the surface along x
                                     (where ym = 0) for a single granule cell at x0 = 0 and
                                     yo = 0 was given by
                                      m
,
-------
Spatial Frequency Analysis
                                                                                          539
                2341
                   CYCLES/mm
           ANESTHETIZED, AMPLITUDE
         1
       2345
         CYCLES/mm
WAKING/CONTROL, AMPLITUDE
Fig.  6. Upper frame: amplitude spectra from four cats
and six rabbits under pentobarbital anesthesia (40 mg/
kg i.v.). Lower frame: upper five traces, amplitude spec-
tra from  two cats and three rabbits in the waking state
with implanted 1x64 arrays; lower five traces, ampli-
tude spectra from records of sine wave signals recorded
in Tyrode's solution at 40 Hz and 70 pY rms.
surface  peak of potential. If the width of the active
focus was reduced, the width of the surface peak of
potential was also  reduced  (Fig. 9). Over most of
the range, the relation between the width of the focus
and the half-amplitude width of the field of potential
was linear.  For small widths approaching  zero, the
half-amplitude   diameter  of  the   potential   field
approached  a limiting value. By extrapolation, if a
single granule cell were activated, its field of potential
recorded at the surface would be bell-shaped with a
half-amplitude  width corresponding to  the limiting
value. The minimal-width curve of surface potential
with distance from  the epicenter is shown in Fig. 10,
lower frame, solid curve. The points were computed
from equations (1)  and (2). They were fitted by non-
linear regression with an empirical curve
                ,-x/X,
                   ,-*/X2
                                              (3)
 The values of the coefficients are given in the legend
 for Fig. 10.
                                                     ANESTHETIZED
                                                               1.0   2.0   3.0  4.0  5.0   6.0
                                                                      CYCLES/mm
                                                                      AMPLITUDE
                                                            WAKING  / x-.
                                                                      A
                                                          CONTROL  /AV
                                                                F42

                                                     ANESTHETIZED
                                                                        \
                    1.0   2.0   3.0  4.0   5.0   6.0
                          CYCLES/mm
                            PHASE
Fig. 7 Averaged spectra for amplitude and phase from
ten anesthetized animals, ten records from a waking cat
(F42),  ten control records, and ten records from wak-
ing animals.
          "I	'—7T
                                                 -200
                                                  200 -
                     DEPTH, jam

 Fig. 8.  Points show the means and standard errors of
 the peak amplitude (Nl) of the averaged evoked poten-
 tial  with  depth in micrometers from the surface on
 PON stimulation. Dashed curve is the postulated source
 -sink distribution for the granule cell as a function of
 depth as in equation (1). Solid curve is the potential as
 a function of depth calculated from equation (2). From
 Freeman  1972. \ length constant  of granule cell; zo,
 site  of  membrane current reversal; ZIS, zero isopoten-
 tial surface.

-------
 540
                                                                                            Freeman
         POTENTIAL /    X  V<*.V»
        *>Y/  ^
                             1000
           STANDARD DEVIATION (0 ), (y)
                                        1600
Fig.  9. Relation between the half-amplitude width of a
surface  distribution  of potential (ordtnate) and  the
width of an active focus of granule cells expressed by
the standard deviation ,  \  of a normal distribution.
For near-zero width, the width of the surface field of po-
tential approaches a limiting value. From Freeman 19 72,

     The activity of a single granule cell was treated  as
a spatial delta function, and its surface field of poten-
tial  as  an  impulse  response. If a  distribution  of
granule  cells were  activated by  an input  u(x), the
potential fields of all the cells would be superimposed
in the volume conductor to give the observed field
of potential v(x). This operation  of summation was
represented by convolution
    v(x)-_/u(x)g(x-X)dX
                                             (4)
Alternatively the operation was expressed by taking
the two-sided Fourier transform of equation  (4).
    V(ju) "
                     G(juj)
(5)
The spatial frequency,  «x (radians/mm), was 2 * fx
where fx was the spatial frequency in cycles/mm.

When the input was a  spatial delta function, then
U((jx) •  1 and

                                             (6)
                                             (7)
    VCo)x) '

Then, from the Fourier transform of equation (3)

                2\!               2X3
               	" k2    ,   ,	
                       1        *220)  2 + 1
                                                               0       0.5      1.0      1.5      2.0

                                                                      SPECTRUM, eycht/mm
                                                                                    SURFACE


                                                                                   -DEPTH
                          0.0      1.0      1.8      Z.O
                         POTENTIAL FIELD, mm
 Fig. 10, Lower frame: symbols show the potential at a
 function of distance in mm from the epicenter of an
 active focus of near-zero width, calculated with equa-
 tions (1) and (2). Solid curve and triangles: predicted
 potential for an array of electrodes at  the surface
 11000 micrometers from ZIS). Dashed curve and rec-
 tangles: predicted potential from an array at a depth
 of 500 urn from ZIS.  Curves are from equation (3),
 Solid curve: A / » 670 urn.  A? • 190 pm. *; - 1.50,
 ki " 0,33, Dashed curve: \ / "- 480 pm, X •> - 9.3 'fim,
 * ;- 1.03, k2- 0.08.
     Upper frames: Fourier transforms of the  curves in
 lower frame from equation 17) and normalised to
 unity at fx - 0, These are the transfer functions des-
 cribing the attenuation by  the volume conductor of
 the EEC amplitude from surface and depth records
generated by differing  spatial frequenetei of granule
 cell activity.

-------
Spatial Frequency Analysis
                                             541
The  amplitude function  from equation (7)  was the
transfer function or weighting curve (Fig. 10, upper
frame, solid curve), which showed the attenuation
effected  by the  volume conductor on the  surface
field of  potential as a function  of the  spatial  fre-
quency of granule cell activity. This curve predicted
that  the spectrum  of the potential recorded at the
surface should  fall sharply  between  0.5  and  1.5
cycles/mm and should be essentially zero at higher
frequencies.

    In comparing the spectral curve in Fig.  10 with
those in Fig. 7, it should be noted that the experi-
mental data were set to zero mean and unit  variance
(and in some cases were also detrended). This  had
the effect of passing the data through a low-cutoff
spatial  frequency filter  with  a  characteristic  fre-
quency of 0.4 cycle/mm. The purpose  of this proce-
dure was to reduce  the zero frequency peak of the
experimental curves, which otherwise resembled the
solid  curve in Fig. 10 (upper frame), in  order to
accentuate (for display  purposes)  the spatial vari-
ations at  rates above 0.5 cycle/mm, up to the limits
of resolution at about 1% of maximal rms amplitude.
Hence the  decline  in amplitude  spectra shown in
Fig.  5 to 7 between 0.6 and 1.0 cycle/mm  is com-
parable to the decline in the solid  curve in  Fig. 10
(upper frame) over the same frequency range.

    The  question  whether a  better  resolution of
spatial frequency of the  EEC might be attained by
recording in a plane parallel to the surface, but at
a depth near the maximum of the granule cell dipole
field, was considered. The expected distribution of
potential with distance from the epicenters at a simu-
lated  depth of 500 um  is shown in Fig. 10, upper
frame, by the dashed curve. The Fourier transform is
shown in  Fig. 10, lower frame, also by the dashed
curve. The cutoff spatial  frequency  was raised to 1.5
to 2.0 cycles/mm.
Discussion

     The experimental results indicated that induced
EEG waves recorded  at the surface of the olfactory
bulb  were  spatially  nonuniform  and  appeared as
localized active foci.  The foci were irregular in out-
line, averaged 1  to 3 mm in half-amplitude width,
varied in shape  and  location between  animals, but
were relatively invariant for each animal. These activi-
ties  were  recorded  in  waking  animals with 6x10
arrays. The interelectrode distances were 0.8 mm and
the  size  of the arrays was 4x7 mm. The area of the
array (28 mm2) was about 16% of the estimated total
surface area (176 mm2) of the  bulb in the cat. The
rabbit bulb was smaller, but only part  of the array
could be fit over the  bulb; the rest of the array over-
lay  the anterior olfactory nucleus, so that the array
covered about 17% of the bulb of the rabbit. The area
in which induced  waves were  observed above half-
amplitude was less than one-fourth of the electrode
area in all animals. By extrapolation, less than 25% of
the  bulb  was  responsible  for  the  appearance  of
induced waves. In some preparations  there was  no
induced activity within the array. The trauma of sur-
gical implantation may have been responsible for this,
and perhaps for the  spotty distribution  of induced
activity as well.

     The spatial frequency  spectrum across such foci
was determined with  concomitant records of poten-
tial from 1x64  electrodes averaging 2.5 mm in length
and 40jum in interelectrode distance. The maximal
observed spatial frequencies in records from waking
animals ranged  from 1.0 to 1.5 cycles/mm (Fig. 5 to
7). This upper  limit was consistent with the  known
volume conductor  properties  of  the  neurons gen-
erating the EEG  (Fig. 10). This implies  that the
desired interelectrode distance  for spatial analysis is
from 0.5 to as  low as 0.33 mm, or perhaps even less,
for  recording surface distributions  of the induced
waves.

     The recording system used  was limited  to  64
channels. An 8x8 array at  0.5-mm spacing is 3.5x3.5
mm, and the area (12 mm2) is 7% of the cat bulb or
17% of the rabbit bulb.  An  8x8 array at 0.3-mm
spacing is  2.1x2.1 mm(4.4 mm2), which covers  2.5
and 6% of the bulbs,  respectively,  of the cat  and
rabbit. These dimensions would appear to suffice for
the study  of a single  active  focus,  but they pose
difficulty in the localization of a focus, particularly
under anesthesia when the induced waves tend to be
poorly defined.

     Theoretical analysis shows that  little is  to be
gained in spatial resolution of the EEG by attempting
to place the electrodes within the bulb rather than on
its surface. The attendant  surgical damage would be
prohibitive.

     The upper limit  of the spatial frequency of the
EEG does  not address the question of whether neural
activity in  the bulb occurs at relatively higher spatial
frequencies. Indeed the input synaptic region of the
bulb in the glomerular layer has a mosaic appearance
(Freeman  1975), suggesting that  the  PON input is
spatially "coarse grained" by parcellation into synap-
tic subdomains, in a fashion analogous to the parcel-
lation of input by the compound facetted eye of an
insect. The average diameter of the glomeruli in  the
cat is 135 jim.  If this is treated as a half wavelength,
the implied spatial frequency of bulbar activity is 3.7
cycles/mm. If such spatial variations in activity level
occur, they are inaccessible in EEG  recordings due
to  the smoothing effect of the volume conductor in
which the generating neurons lie. The  presence  of
activity between 0.6 and 1.0 cycle/mm In the spectra
from  waking but not  from  anesthetized cats and

-------
 542

 rabbits indicates that neural activity does occur with
 spatial rates of change in  excess of 1.0 cycle/mm,
 although its  manifestation in  the  EEC is sharply
 curtailed.
     The detailed application of these techniques to
other structures  and spatial activity  patterns in  the
brain (Petsche  1972)  appears  to be  conceptually
straight  forward  in terms  of  defining  particular
reproducible events  in  the EEC, measuring their
time courses and spatial distributions, and computing
their temporal and spatial spectra. The rate of fall-off
of spatial spectra with increasing spatial frequency
depends on  the  geometry  and particularly on  the
depth of the generating neurons, so that the forms of
the spatial frequency spectra for other events cannot
be predicted without detailed anatomical and electro-
physiological specifications. However,  for purposes of
modelling volume conductor properties, particularly
in gauging the attenuation of spatial frequencies in
going from cerebral to scalp recordings, the value for
spatial frequency of  1 cycle/mm may  be particularly
useful.
                                         Freeman
 Summary
     An approach for experimentally measuring and
 theoretically predicting  the  spatial  frequency spec-
 trum of a component  of the EEC is described. A
 particular  recurring waveform in the  EEC of  the
 olfactory  bulb (the "induced wave" of Adrian) of
 cats and  rabbits  is described in  terms of its time
 course, spatial distribution at  the surface of the bulb,
 and depth distribution  within the bulb. The spatial
 spectrum  is computed  from  the Fourier  transform
 of amplitude distributions determined concomitantly
 from 64 electrodes spaced along a line on the bulbar
 surface at  intervals of 40 jim. The observed spectra
 fall  off rapidly  in amplitude between  0.6 and  1.0
 cycle/mm.  This  result is consistent with the known
 volume conductor properties  of  the  neurons gen-
 erating the  bulbar EEC.  This does not exclude the
 occurrence  of neural activity  at higher spatial fre-
 quencies, but  it  implies that such spatial  variations
 are not reflected in the EEC, and that electrodes need
 not be spaced  closer together  than 0.3  to 0.5 mm to
 extract the  available  information  about the spatial
structure of the bulbar EEC waves.

-------
SCALP DISTRIBUTION REFERENCES
Adrian, E.D. The electrical activity of the mammalian
   olfactory bulb. Electroenceph. clin. Neurophysiol.,
   1950,2:377-388.

Allison, T., Goff, W.R.,  Williamson, P.O.  and Van
   Gilder, J.C. On the neural origin of early compo-
   nents of the human somatosensory evoked poten-
   tial.  In  J.E. Desmedt (Ed.), Prog.  clin.  Neuro-
   physiol. Vol. 7, Somatosensory evoked potentials
   and their clinical uses. Karger, Basel, (in press).

Allison, T., Matsumiya,  Y.,  Goff, G.D. and Goff,
   W.R.    The scalp  topography of human  visual
   evoked  potentials.  Electroenceph,  clin.  Neuro-
   physiol., 1977,42:185-197.

Amadeo,  M.,  and Shagass,  C.  Brief latency click-
   evoked potentials during waking and  sleep in man.
   Psychophysiol., 1973, 10:244-250.

Arezzo, J., Pickoff, A. and Vaughan, H.G., Jr. The
   sources  and intracerebral distribution of auditory
   evoked  potentials  in  the alert rhesus  monkey.
   Brain Res., 1975,90:57-73,

Arezzo, J. and Vaughan, H.G., Jr.  Cortical potentials
   associated   with voluntary  movements in  the
   monkey. Brain Res., 1975,88:99-104.

Brazier. M.A.B. The electrical fields at the surface
   of the  head   during sleep.  Electroenceph. clin.
   Neurophysiol., 1949,1:195-204.

Brazier. M.A.B. A study of  the electrical fields at
   the  surface of  the  head.  Electroenceph. clin.
   Neurophysiol., 1950, Suppl. 2:38-52.

Broughton, R.J.  Discussion after H.G.  Vaughan,  Jr.
   The  relationship of brain activity to  scalp record-
   ings of event-related potentials. In E.  Donchin and
   D.B. Lindsley  (Eds.), Average evoked potentials:
   Methods, results and evaluations. NASA SP-191,
   Washington, D.C., 1969:79-84.

Brown, W.S., Marsh, J.T. and Smith,  J.C.  Contex-
   tual  meaning effects on speech evoked potentials.
   Behav.  Biol., 1973,9:755-761.

Calmes, R.L. and  Cracco, R.Q. Comparison of soma-
   tosensory and somatomotor evoked responses to
   median  nerve  and digital nerve stimulation. Elec-
   troenceph. clin. Neurophysiol., 1971, 31:547-562.

Chapman,   R.M.  Evoked  potentials  of  the brain re-
   lated to thinking.  In F, J.  McGuigan and  R.
   Schoonover (Eds.), The psychophysiology of think-
   ing: Studies of covert  processes. Academic Press,
   New York, 1973.

Cooper, R., Winter, A.L., Crow, H.J,  and Walter,
   W.G. Comparison  of  subcortical,  cortical and
   scalp activity  using chronically indwelling elec-
   trodes in man. Electroenceph. clin. Neurophysiol.,
   1965,18:217-228.

Courchesne, E.,  Hillyard, S.A. and  Galambos, R.
   Stimulus novelty, task relevance and  the  visual
   evoked   potential in man.  Electroenceph. clin.
   Neurophysiol., 1975,39:131-143.

Cracco, R.Q. Traveling  waves  of  the human  scalp-
   recorded somatosensory evoked response: Effects
   of differences in recording technique and sleep
   on  somatosensory  and somatomotor  responses.
   Electroenceph.  clin. Neurophysiol.,  1972,  33:
   557-566.

 Cracco, R.Q. and Bickford, R.G.  Somatomotor  and
   somatosensory evoked responses: Median nerve
   stimulation  in man. Archs. Neurol.,  1968,  18:
   52-68.

 Donald, M. Parallel late waves in visual and somato-
   sensory evoked  response. In J.E. Desmedt (Ed.),
   Progr.  clin. Neurophysiol., Vol. 7, Somatosensory
   evoked potentials  and  their clinical  uses. Karger,
   Basil, (in press).

 Donchin, E. A  multivariate approach to the analysis
   of average evoked potentials. IEEE Trans, biomed.
   Eng.,  1966,  BME-13:131-139.

 Donchin,  E.  Data analysis techniques  in  evoked
   potential  research.   In E.  Donchin  and D.B.
   Lindsley  (Eds.),   Average  evoked  potentials:
   Methods,  results, and evaluations. NASA SP-191,
   Washington, D.C., 1969:199-236.

 Donchin, E.  Methodological issues in CNV research,
   a  review.  In W.C.  McCallum and  J.R.  Knott
   (Eds.),  Event-related slow potentials of the brain:
   Their relations to  behavior, Elsevier, Amsterdam,
    1973:3-17.
 Donchin,  E., Tueting, P.,  Ritter,  W., Kutas, M.  and
   Heffley, E. On the independence of the CNV and
   the  P300  components of the human  averaged
   evoked  potential.  Electroenceph.  clin.  Neuro-
   physiol., 1975,38:449-461.

 Eastman,  C.  Construction  of miniature electrode
   arrays  for  recording  cortical  surface potentials.
   J. Electrophysiol. Tech., 1975, 5:28-30.

-------
 544
                   Scalp Distribution References
 Freeman, W.J.  Depth recording of averaged evoked
    potential  of olfactory  bulb.  J. Neurophysiol.,
    1972,35:780-796.

 Freeman, W.J.  Mass action in  the nervous system.
    Academic Press, New  York, 1975.

 Freeman, W.J.  Quantitative  patterns of  integrated
    neural  activity. In J.C.  Fentress (Ed.), Simpler
    Networks and Behavior, Sinauer Assoc., Sunder-
    land,  Mass.,  1976: 280-296.

 Gerbrandt, L.K., Goff, W.R. and Smith, D.B. Distri-
    bution of the human  average movement potential.
    Electroenceph.  clin.   Neurophysiol,,   1973,  34:
    461-474.

 Goff, G.D., Matsumiya, Y., Allison, T. and  Goff,
    W.R.  The scalp topography  of human somato-
    sensory  and  auditory evoked potentials. Electro-
    enceph. clin.  Neurophysiol., 1977,42:57-76.

 Goff, W.R., Matsumiya, Y., Allison, T. and  Goff,
    G.D.  Cross-modality  comparisons   of  average
    evoked potentials. In  E.  Donchin and D.B. Linds-
    ley (Eds.), Average evoked potentials: Methods,
    results, and evaluations.  NASA  SP-191, Washing-
    ton, D.C., 1969:95-141.

 Goff, W.R., Rosner, B.S. and  Allison, T. Distribution
    of cerebral somatosensory evoked responses  in
    normal man.  Electroenceph.  clin. Neurophysiol.,
    1962,14:697-713.

 Goff, W.R.,  Williamson,  P.O.,  Van Gilder,  J.C.,
    Allison,  T.  and  Fisher,  T.C. Neural  origins  of
    long latency  evoked potentials recorded from the
    depth  and cortical surface of the brain in  man.
    In J.E. Desmedt (Ed.), Progr. clin. Neurophysiol.,
    Vol. 7, Somatosensory evoked potentials and their
    clinical uses. Karger, Basil, (in press).

 Henderson, C.J.,  Butler, S.R.  and  Glass, A.  The
   localization of equivalent dipoles of EEG sources
   by  the  application   of  electrical  field theory.
   Electroenceph.  clin.  Neurophysiol.,  1975,  39:
    117-130.

 Hillyard,  S.A.,  Courchesne,  E., Krausz,  H.I.  and
   Picton, T.W, Scalp topography of the "P3" waves
   in  different  auditory decision   tasks.  In  W.C.
   McCallum and J.R. Knott  (Eds.), The  responsive
   brain, Wright .Bristol, 1976:81-87.

 Hillyard, S.A. and Picton, T.W. Event-related poten-
   tials  and selective information processing in  man.
   In  J.E. Desmedt (Ed.), Progr. clin. Neurophysiol.,
   Vol.  6,  Cognitive components  in event-related
   cerebral potentials, Karger, Basil, (in press).

Jarvilehto,  T. and  Fruhstorfer,  H. Is  the sound-
   evoked DC potential  a contingent negative  vari-
   ation?  Electroenceph.  clin. Neurophysiol., 1973,
   Suppl. 33:105-108.
 Jasper, H.H. The ten twenty electrode system of the
    International   Federation.  Electroenceph.   clin
    Neurophysiol., 1958, 10:371-375.

 Jeffreys, D.A. and Axford, J.G. Source locations of
    pattern-specific  components  of  human   visual
    evoked potentials. Exp. Brain Res., 1972,16:1-40.

 Jenness, D. Auditory evoked-response differentiation
    with discrimination  learning in humans. J. comp.
    physiol. Psychol., 1972, 80:75-90.

 Jewett,  D.L.  Volume-conducted potentials  in re-
    sponse to auditory stimuli as detected by averaging
    in  the  cat. Electroenceph. clin.  Neurophysiol.
    1970a,  28:609-618.

 Jewett, D.L.  Human  auditory evoked  potentials:
    Possible brain  stem components  detected on the
    scalp. Science,  1970b, 167:1517-1518.

 Jewett,  D.L. and Williston, J.A. Auditory-evoked far
    fields averaged from the scalp of humans. Brain
    1971,94:681-696.

 John, E.R. A model of consciousness. In G. Schwartz
    and D.  Shapiro (Eds.), Consciousness  and self
    regulation: Advances in research, Vol.  1. Plenum,
    New  York, 1976:  1-50.

 John, E.R.,  Ruchkin,  D.S. and  Villegas, J. Experi-
    mental background:  Signal  analysis and behavioral
    correlates of evoked potential configurations  in
    cats. Ann. N.Y. Acad. Sci., 1964, 112:362-420.

 John, E.R.,  Walker, P., Cawood, D., Rush, M. and
    Gehrmann,  J.  Factor analysis of  evoked poten-
    tials.  Electroenceph. clin.   Neurophysiol.,  1973
    34:33^3.

 Kelly, D.L., Jr., Goldring, S. and O'Leary, J.L. Aver-
   aged evoked somatosensory responses  from ex-
   posed  cortex  of  man. Archs.  Neurol.,  1965,
    13:1-9.

 Kooi, K.A., Tipton, A.C. and Marshall, R.E. Polarities
   and field configurations of the vertex components
   of the human auditory evoked response: a reinter-
   pretation.  Electroenceph.   clin.  Neurophysiol.,
   1971,31:166-169.

 Kutas, M.  and Donchin, E.  Studies of squeezing:
   Handedness, responding hand, response force,
   and asymmetry  of  readiness  potential.  Science,
   1974, 186:545-548.

Llinas, R. and Nicholson, C. Analysis of field poten-
   tials in the central nervous system. In A. Remond
   (Ed.), Handbook of electroencephalography  and
   clinical neurophysiology. Vol. 2 Part B. Amster-
   dam, Elsevier, 1974:61-83.

Lorente de  No, R. A  study  of  nerve physiology.
   Studies from the Rockefeller Institute for Medical
   Research, Vol. 132, 1947.

-------
Scalp Distribution References
                                            545
Mast, T.E. Short-latency human evoked responses to
   dicks. J. appl. Physiol., 1965,20:725-730.

Matousek, M. Frequency  and  correlation  analysis.
   In A. Remond (Ed.), Handbook of electroencepha-
   lography and clinical neurophysiology, Vol.5,Part
   A, Amsterdam, Elsevier, 1973:3-109.

Matsumiya, Y., Tagliasco, V., Lombroso, C.T. and
   Goodglass, H. Auditory evoked response: Meaning-
   fulness of stimuli  and interhemispheric asymme-
   try. Science,  1972, 175:790-792.

Michael,  W.F. and  Halliday, A.M. Differences be-
   tween  the  occipital distribution of  upper and
   lower  field  pattern-evoked   responses in  man.
   Brain Res., 1971,32:311-324.

Michel, F. and Peronnet.F. Extinction gauche au test
   dichotique;  Lesion  hemispherique  ou  lesion
   commissurale? (apport  de 1'etude des potentiels
   evoque's acoustiques). In F. Michel and B. Schott
   (Eds.), Les syndromes de disconnexion calleuse
   chez  1'homme.   Hopital  Neurologique,  Lyon,
   France, 1974.

Molfese,  D.L., Nunez,  V., Seibert, S.M. and Raman-
   aiah, N.V. Cerebral asymmetry: Changes in factors
   affecting its development. Ann.  N.Y.  Acad. Sci.,
   1976,280:821-833.

Naatanen, R. Selective attention and evoked  poten-
   tials in  humans-a critical review. Biol. Psychol.,
   1975,2:237-307.
Papakostopoulos,  D.,  Cooper,  R.  and  Crow, H.J.
   Inhibition of  cortical evoked potentials and sen-
   sation by self-initiated movement in man. Nature,
   1975,258:321-324.

Peronnet, F., Michel, F., Echallier, J.F. and Girod,J.
   Coronal topography of human auditory evoked
   responses.  Electroenceph.  clin.  Neurophysiol.,
   1974,37:225-230.

Petsche,  H. BEG topography. In A. Remond  (Ed.),
   Handbook of  electroenceph.  and  clin.  neuro-
   physiol.,  Vol.  5 Part  B. Amsterdam, Elsevier,
   1972:3-75.

Picton,  T.W.  and  Hillyard, S.A.  Human auditory
   evoked  potentials II: Effects of attention. Elec-
   troenceph. clin. Neurophyjlol., 1974,36:191-199.


Picton, T.W., Hillyard, S.A. and Galambos, R. Cor-
   tical  evoked   response} to  omitted  stimuli.  In
   M.N.  Livanov  (Ed.),  Major  problems  of  brain
   electro physiology, USSR  Academy of Sciences,
   1974a:  302-311.
 Picton, T.W., Hillyard, S.A, Krausz, H.I. and Galam-
   bos,  R.  Human auditory  evoked  potentials:  I.
   Evaluation of  components. Electroenceph, din.
   Neurophysiol.,   1974b,  36:179-190.
Picton,  T.W.  and  Woods,  D.L.  Human  auditory
   evoked sustained  potentials. Electroenceph. clin.
   Neurophysiol., 1975,38:543-544 (abstract).

Rail, W. Dendritic neuron theory and dendrodendritic
   synapses  in  a simple  cortical  system.  In  F.O.
   Schmitt (Ed.), The Neurosciences, Second Study
   Program, Rockefeller  Univ.  Press, New  York,
   1970:552-565.

Rail, W. and Shepherd, G.M. Theoretical reconstruc-
   tion  of  field  potentials   and  dendrodendritic
   synaptic interactions  in olfactory bulb. J. Neuro-
   physiol., 1968,31:884-915.

Remond,  A. The importance  of topographic data in
   EEC phenomena and an electrical model to repro-
   duce them.  Electroenceph. clin. Neurophysiol.,
   1968, Suppl. 27:29-49.

Remond,  A. and Lesevre, N. Variations in average
   visual evoked potential as a function of the  alpha
   rhythm phase ("Autostimulation"). In W.  Cobb
   and C.  Morocutti (Eds.),  The evoked potentials,
   Elsevier, Amsterdam, 1967:42-52.

Renault, B. and Lesevre, N, Variation des potentials
   e'voques  par des  patterns  (P.E.P.)  chez 1'homme
   dans une situation de temps de re'action: Influence
   de la vigilance et de la motricite. Electroenceph.
   clin. Neurophysiol., 1975,5:360-366.

Ritter,  W.,  Simson,  R. and  Vaughan, H.G. Topo-
   graphic analysis of task-related cerebral potentials.
   In. I.E. Desmedt  (Ed.), Progr. clin. Neurophysiol.,
   Vol. 6,  Cognitive  components in event-related
   cerebral potentials, Karger, Basil, (in  press).

Ruchkin,  D.S, and Sutton, S. Latency characteristics
   and trial by trial variation of emitted potentials.
   In J.E. Desmedt (Ed.), Progr. clin. Neurophysiol.,
   Vol. 6,  Karger,  Basil, (in  press).

Schlag, J. Generation  of brain  evoked  potentials.
   In  R.F.  Thompson  and M.M. Patterson (Eds.),
   Bioelectric Recording Techniques. Part A. Cellular
   processes and brain potentials, Academic  Press,
   New York, 1973:273-316.

Shepherd, G.M,  Synaptic organization of the mam-
   malian  olfactory  bulb. Physiol. Rev.,  1972,  52:
   864-917.

Simson, R., Vaughan,  H.G.,  Jr.  and  Ritter W.  The
   scalp topography of  potentials associated with
   missing visual or  auditory stimuli.  Electroenceph,
   clin. Neurophysiol., 1976,40:3342.

Simson, R, Vaughan, H.G.,  Jr. and Ritter, W.  The
   scalp topography of potentials in auditory  and
   visual  discrimination tasks.  Electroenceph.  clin.
   Neurophysiol., 1977,42:528-535.

Smith, D.B.D,, Lell, M.E., Sid man, R.D. and Mavor,
   H.  Nasopharyngeal  phase  reversal  of cerebral

-------
 546
                 Scalp Distribution References
   evoked  potentials and  theoretical  dipole impli-
   cations. Electroenceph. clin. Neurophysiol., 1973,
   34:654-658.

Somjen, G.G.  Electrogenesis of sustained potentials.
   Prog. Neurobiol., 1973,1:201-237.

Squires,  K.C.,  Donchin, E., Horning, R.  and  Mc-
   Carthy, G. On the influence of task relevance and
   stimulus probability on ERP components.   Elec-
   troenceph. clin. Neurophysiol., 1977,42:1-14.

Squires, K.C.,  Hillyard,  S.A. and  Lindsay, P.H. Cor-
   tical potentials evoked by confirming and discon-
   firming  feedback  following an auditory discrimi-
   nation. Percept. Psychophys., 1973,13:25-31.

Squires, N.K., Donchin, E. and Squires,  K.C. Bisen-
   sory stimulation:  Inferring decision-related  pro-
   cesses from the P300  component. J. exp. Psychol.:
   Human Perception and Performance, 1977, 3:299-
   315.

Squires, N.K.,  Squires, K.C. and Hillyard, S.A. Two
   varieties of long-latency positive waves evoked by
   unpredictable auditory stimuli  in  man. Electro-
   enceph. clin. Neurophysiol.,  1975,38:387-401.

Stephenson, W.A. and Gibbs,  F.A.  A balanced non-
   cephalic reference electrode. Electroenceph.  clin.
   Neurophysiol., 1951,  3:237-240.

Stuss,  D.T. Electrophysiological correlates in human
   concept learning. Unpublished  doctoral disserta-
   tion, Univ. of Ottawa, 1976.

Suter,  C.M. Principal component  analysis of average
   evoked  potentials.  Expl.  Neurol.,  1970,  29:
   317-327.

Thatcher,  R.W. and John, E.R. Information and
   mathematical quantification of brain state. In N.
   Burch  and  H.L.  Altshuler (Eds.),  Behavior and
   brain electrical activity. Plenum, New York, 1975:
   303-324.

Tsumoto, T., Hirose, N., Nonaka, S. and Takahashi,
   M.  Analysis of somatosensory evoked potentials
   to  lateral  popliteal nerve  stimulation  in  man.
   Electroenceph.  clin.  Neurophysiol.,  1972,  33:
   379-388.

Vaughan, H.G.  The relationship of brain activity to
   scalp  recordings of  event-related  potentials. In
   E.  Donchin  and  D.B.  Lindsley  (Eds.), Average
   evoked potentials: Methods, results, and evalu-
   ations.  NASA SP-191, Washington, D.C., 1969:
   45-94.

Vaughan, H.G.  The analysis of scalp-recorded brain
   potentials. In R.F.  Thompson and M.M. Patterson
   (Eds.),  Bioelectric  recording techniques, Part B:
   Electroencephalography and human brain poten-
   tials. Academic Press, New York, 1974: 157-207.

Vaughan, H.G. and Ritter, W. The sources of auditory
   evoked responses recorded from the human scalp.
   Electroenceph.  clin. Neurophysiol.,  1970, 28:
   360-367.

Weinberg, H.  and Cooper, R. The recognition index:
   A pattern  recognition technique for noisy signals.
   Electroenceph.  clin.  Neurophysiol., 1972,  33:
   608-613.

Weinberg, H, and  Papakostopoulos,  D. The frontal
   CNV:  Its  dissimilarity  to  CNVs  recorded from
   other  sites.  Electroenceph. clin.  Neurophysiol.,
   1975,39:21-28.

Weinberg, H., Walter, W.G. and Crow, H.J. Intracere-
   bral events in humans related to real and imaginary
   stimuli. Electroenceph. clin. Neurophysiol., 1970,
   29:1-9.

Wilkinson, R.T. and Ashby, S.M. Selective attention,
   contingent  negative  variation and  the evoked
   potential Biol. Psychol., 1974, 1:167-179.

Wilson, F.N.  and Bayley, R.H. The electric field of
   an  eccentric  dipole in  a homogeneous  spherical
   conducting medium. Circulation, 1950,1:84-92.

Wolpaw, J.R. and Penry, J.K. A temporal component
   of  the auditory evoked  response. Electroenceph.
   clin. Neurophysiol., 1975, 39:609-620.

-------
IX. ALTERNATIVES TO SIGNAL AVERAGING
     Section Editor:
     Hal Weinberg
     Brain Behavior Laboratory
     Department of Psychology
     Simon Fraser University
     Burnaby, British Columbia, Canada

-------
ALTERNATIVES TO  SIGNAL AVERAGING:
EXCERPTS  OF  PRECONFERENCE  CORRESPONDENCE

Prepared by H. Weinberg, Panel Chairman and Section Editor
    The following  correspondence was  conducted
prior to EPIC IV by the panel on Alternatives to Sig-
nal Averaging. Members of the panel included:

    1.   H.  Weinberg (chairman), Brain Behaviour
        Laboratory, Simon Fraser University, Van-
        couver, British Columbia, Canada.

    2.  L. K. Gerbrandt, Department of Psychology,
        California State University, Northridge Cal-
        ifornia.

    3.  R. Herning, Langley Porter Neuropsychiatric
        Institute, San Francisco, California.

    4.  E.  R. John, Brain  Research  Laboratories,
        New York University Medical Center, New
        York, New York.

    5.  P. Naitoh, Naval  Health  Research  Center,
        San Diego, California.

    6.  J. Peters, Brain Behavior Laboratory, Simon
        Fraser  University,  Vancouver,  British  Col-
        umbia, Canada.

    7.  D.  S. Ruchkin, Departments  of Physiology
        and Computer Science,  School of Medicine,
        University  of Maryland,  Baltimore, Mary-
        land.

    Some  of the material covered in the correspon-
dence was later amplified and presented at the confer-
ence; such material has been omitted. Other portions
of  the  correspondence have been abridged  and at
times paraphrased for brevity and  proper transition.

    Weinberg:  The development of circulated corres-
pondence concerning alternatives to signal  averaging
implies that  there are limitationi to this technique.
Vigorous use of  signal averaging over the past ten
years, I believe, has made one thing evident: the ob-
fuscation of relationships between  independent  and
dependent variables.

    Let me offer these questions as keys to subjects
that I hope we can  develop further through  corres-
pondence  and conversation:   What  is meant by
"signal"?  Is there  really  "noise"  in  the nervous
system? What constitutes the stimulus? Is it defined
by characteristics of the input or by its measurable
influence on the nervous system? Is the nature of the
signal in signal-to-noise ratios dependent on methods
for detecting that signal? Is the electrode an uncon-
trolled averaging system? Precisely what information
is  lost through signal summation?  Can the logic of
signal averaging be extended so as to extract the sig-
nal related to the common  element in variations of
stimulus  input, e.g., if we average  the  signals to
different stimuli, all of which fall into the same ab-
stract category, do  we  have a signal reflecting that
abstraction? Can changes in the variability of a signal,
given a stable mean, also be considered a signal, i.e.,
a proper response of the nervous system? Can spatial-
ly disparated signals be  averaged to extract the com-
mon  elements?  Are the methods  underlying topo-
graphic representations  of voltage gradients methods
of spatial averaging? These  are questions that  arise
in consideration of signal averaging techniques. There
are other obvious questions that arise in consideration
of other  techniques. For example, do Fourier trans-
forms reveal anything about when something is hap-
pening as distinct from  what happens? In pattern re-
cognition techniques, how is the pattern to be detect-
ed determined in the first place?


    Naitoh (as amplified below) suggested that, for
the purpose of discussion, a distinction be made be-
tween signal extraction  techniques   and  statistical
techniques designed to  establish whether signals are
"real" or are different  from other signals. This dis-
tinction raises questions in my own mind, especially
when  dealing with procedures  like discriminant
analysis.  The methods  of extraction are  themselves
statistical treatments of the data, the results of which
are statements of the probability of the occurrence of
the  signal  in relation to  concommitant  variability.

    I have separated the techniques into two lists, the
distinction between which may itself be a topic for
discussion:

     1.  Extraction procedures:  averaged evoked po-
       tentials,  median  evoked  potentials,  cross

-------
 550
                 Preconference Correspondence
        correlation techniques. Wiener filtering, aver-
        aging  by  cluster,  spectral  analysis  using
        Fourier  transforms, toposcopic. and  topo-
         graphic  representations,  recursive  filters.

     2.  Statistical procedures: multiple regression an-
        alysis,  discriminant  analysis,  factor analysis,
        multiple and partial coherence and phase an-
        alysis,  ANOVA  and MANOVA approaches.

These are only partial lists; I offer them only as some-
thing with which we can begin.

    Naitoh:  1 feel that two aspects of any evoked po-
tential (EP) research must be discussed: (1) how are
the EPs to be extracted and what measures  are to be
chosen for their analysis, and (2) what sorts of statis-
tics are to be used on the EPs once  they have been
distinguished from the noise? Sometimes a distinction
between mathematics required to extract  EPs and
mathematics required to evaluate them, is  indeed
fuzzy (e.g., stepwise multiple, discriminant analysis).
But such a distinction is critical.

    For amplification  and discussion, I would like to
present two lists, neither of which I cpnsider compre-
hensive. First, methods of extracting EPs:

    1.  Median evoked potentials.

    2. Pattern recognition  with  cross correlation.

    3. Recurrence frequency function.

    4.  A posteriori Wiener filtering - This method
       may  be well   suited for  developing  an  EP
       template as it gives, insofar  as possible, a
       background-EEC free EP.

    5.  Sayer's phase-forcing method  - This method
       is  an extension of time-series  analysis. It is
       well  suited .for single-trial analysis and also
       for detecting very small EPs.

    6. Averaging  by  cluster  (Saltzberg method).

Secondly,  methods of  statistically  analyzing the
features of EPs:

    1. Stepwise multiple  regression analysis.

    2.  Stepwise multiple  discriminant analysis.

    3.   Hotelling's T2 or Mahalanobis' D2 - These
       methods have a limited  scope of applications
       but are very instructive to  those concerned
       with multivariate statistics.

    4.  Factor analysis, including  discriminant prin-
      cipal component analysis.

    5.  Walsh analysis.
     6.  Basis function analysis — This is an analysis of
       EP on the basis of a few  biologically "mean-
       ingful" wavelets.
     7.  Equipotehtial mapping (analogous  to  topo-
       scopical analysis).

     8. Multiple, partial, and pairwise coherence and
       phase analysis.

     9.   Information matrix - Multidimensional or
       simple Markovian.

   10.  ANOVA and MANOVA approaches.

       Weinberg:  I  would  like to consider further,
specifically,  the  question  of   whether  spatially
disparate signals can be meaningfully averaged as an
index  of the commonality of the signals occurring at
different sites. There is no obvious reason why  the
logic of the averaging  could  not be applied to  the
averaging of spatially disparate signals: after all, there
is no more reason  to  believe  that signals recorded
from the same  site on  successive occurences  of an
input  have commonalities.  In  a sense,  the  same
question arises with  regard to the averaging of signals
ostensibly recorded  from the same site  in different
subjects. Or to push  the logic of signal averaging to its
extreme, consider this:  If it is true that an event-re-
lated potential indexes the attributes of the input,
and if different signals from different inputs (stimuli)
are averaged, then the resulting signal should be that
associated with the  commonality of the inputs,  a
commonality  that could  presumably be  an  abstrac-
tion, i.e., not associated with the absolute properties
of  the stimuli but with the  relative properties of the
stimuli.

    But why average? Why obscure differences in the
signal when those differences could be critical for an
understanding of  how different sites  respond  to
input? To answer that signal averaging extracts  the
common  signal from background noise is, I believe,
unsatisfactory because current techniques allow us to
identify the signal in single occurrences.
    Gerbrandt: My first reaction to Weinberg's ques-
tion as to whether signals that are spatially disparate
can be  averaged  for the purpose of extracting the
common element was another question: How can one
distinguish  between  the multiple events  that are
superimposed onto the same space and time? Theta
rhythms have  usually  been thought  of as homo-
geneous events and as a functional entity.  Studies
now indicate that theta rhythms arise  from multiple
origins within  the  hippocampus  (Gerbrandt  et al.
1974, Winson 1974, Bland et al. 1975)  and that some
of these sources  are independently manipulated by
specific  experimental  variables  (Gerbrandt  et al.

-------
 Preconference Correspondence
                                            551
1975). Thus, the many interpretations of the func-
tion of theta rhythms may have arisen because theta
rhythms  are of heterogeneous functional origin. The
only thing that theta rhythms may have in common is
their periodicity; the close phase coupling may mere-
ly be  a device needed to preserve functional indepen-
dence  of   circuitry  (analogous  to  multiplexing)
through a process of phase encoding.

    Similarly, many disagreements have arisen over
the proper  distribution and  function of motor po-
tentials  (MPs)  in  humans  (Vaughan  et al,  1968,
Deecke et al. 1969, Gerbrandt et al.  1973) and many
of these differences in  result  and interpretation seem
to have resulted from the superimposition or multiple
MP components (Gerbrandt 1974). Thus, any activity
that occurs with an abrupt negativity around the time
of a movement can be called N2, although it could be
a change in  the lateralization of Nl,  a new command
potential, or reafference mistakenly thought to be oc-
curing before EMG onset. CNV-like events and the
P300 wave,  once  thought to be  functional entities
called the "expectancy wave" and the  "uncertainty
wave" are now breaking down into a myriad of distri-
butions,  each affected  by different experimental vari-
ables (Gerbrandt 1977, Ritter et al. 1974). Studies of
single-unit  activities during  the  readiness potential
period indicate that different types of unit activity
occur simultaneously under  the  same readiness po-
tential,  and  changes in unit  activity associated with
changes in directions of movement are not  reflected
in the readiness potential (Tanji  and Evarts 1976).

    It may  be then that the  extraction of a common
element from spatially disparate slow potential signals
(even when  they are recorded by a single microelect-
rode) represents nothing more than the fortuitous art-
ifact  of  hundreds  of parallel elements happening  to
statistically  have only  one phenotypic property  in
common - voltage rather than  function.  The fol-
lowing analogy may help  to explain the issue: Imag-
ine we search for the event-related sounds of a typical
picnic.  Would the sound of the  picnic basket open-
ing,  the  table cloth unfolding, and the  ants chewing
be termed a "common element," or is a picnic a com-
plex enough happening to result in hundreds of inde-
pendently generated sounds  that nonrandomly sum-
mate into the typical picnic spectrogram?

    Naitoh: Another  fundamental  question will  be
related to the EPs tie with basic neurophysiology and
with  behavioural psychology. For such applications,
it may be necessary to consider changing correlational
values as the time frame of observation is extended to
days, or even months.

     The fundamental question for time-series analysis
of brain waves in general is whether the researcher be-
lieves the statement, "God plays dice." If God plays
dice, all records will be noisy in one sense or another.
For a given noise, a set of rules that the noise follows
is developed. Later, the rules define the signal. If God
docs not play dice, there is somewhere in this world a
pure signal,  a  deterministic  event, a nonfuzzy thing.
Only sadly our vision is blurred so that we cannot see
this signal. Thus, in terms of signal analysis, there  is
always  noise  in  the nervous system.  Some  may
define  it  simply  as  "unwanted event";  others may
consider that noise is all we get. In psychophysiology,
"stimulus" is  defined from both standpoints by re-
ferring  to stimulus  specificity and response speci-
ficity, i. e.,  a  given stimulus produces the responses
that are unique to the stimulus.


    Herning:  The signal-noise model of signal aver-
aging has been a useful procedure for the extraction
of signal buried in noise, but may not have been that
useful in understanding the  dynamics of cortical pro-
cessing  in humans. A reason for this is the concept  of
noise in the signal averaging  framework.
     I propose that a distinction be made between ex-
 traction noise (i.e., noise  that  a given statistical or
 extraction technique cannot explain) and meaningful
 biological noise (i.e., noise that is inherent in the phy-
 siology of the organism from which one is recording).
 Extraction noise  in signal averaging is the part of the
 brain response that is averaged out. This noise may or
 may  not be meaningful biological noise or perhaps
 may be part of a dynamically changing signal. Extrac-
 tion noise in the use of autoregressive model-fitting
 in the detection of single-trial event-related potentials
 (ERPs)  is the variance not  accounted  for when a
 model  is being fit to the background EEC. In  this
 case, the extraction  noise is only  that part  of the
 background  EEC not  fit  by the stochastic model.
 Since  autoregressive model fitting in this example is
 modeling background EEC activity and signal aver-
 aging is largely ignoring background  EEC activity, ex-
 traction noise  has different meaning for each tech-
 nique.  Thus the  quantification of  extraction noise
 may be specific  to the extraction or statistical tech-
 nique used. Meaningful biological noise, on the other
 hand, is actual noise in the nervous system and  the
 physiology of  the organism,  such as EMG, EOG, or
 other brain activity that is not  related to the process
 under study. This noise may  obscure the brain activ-
 ity one is attempting to record.


     Signal averaging, by its  nature, has been a mul-
 tiple-occurrence  technique. That is, signal averaging
 requires a number of single trials or epochs to extract
 the  signal  or  experimenter-meaningful part  of the
 brain's response. Since a number of single-trial ERP
 extraction  techniques have already been developed,

-------
 552

 an important question is whether the brain response
 to the same set of external events, say in a given CNV
 paradigm,  is similar from trial to trial.  My experi-
 ence  with  the use of autoregressive model building
 techniques and stepwise discriminant analysis in the
 detection of auditory EPs  is  that the  event-related
 cortical response  may not be completely similar from
 trial to trial and its detectability is closely related to
 the nature of the on-going EEC activity at  the time of
 the stimulus.

     CNVs, motor potentials, P300s, and  other slow
 electrical potentials would seem to be ideal for study
 with  single-trial  techniques  since they occur over
 longer periods of time than pure sensory  evoked re-
 sponses and since they are sometimes observable in
 the raw EEC record.  However, since  these slow po-
 tentials have been deemed to represent more complex
 cortical processing,  they  may be more variable from
 trial to trial.

     As with signal averaging, some alternate extrac-
 tion or statistical  techniques are also multiple-occur-
 rence  techniques and may be subject  to some of the
 same criticisms as signal  averaging. Such techniques
 include  multiple  regression, ANOVA, MANOVA,
 factor analysis, and discriminant  analysis. Although
 these techniques can actually be used to  classify or
 extract single-trial ERPs, more information is requir-
 ed  than a single epoch of EEC data. For example,
 stepwise discriminant analysis (SWDA) and cross-cor-
 relation  detection methods  must build their class-
 ification rule  or template from a number of single
 trials.  In addition, the SWDA is not sensitive to la-
tency changes in single ERPs.  Other  methods,  like
Wiener  filtering and autoregressive model  fitting, re-
 quire EEC data from  only a  single epoch  to extract
 or detect ERPs. These  techniques are not as yet as
sensitive as a multiple-occurrence technique. Certain-
ly more research in this area is needed.

    Ruchkin: Evoked potentials can usually be char-
acterized  as consisting of  the  sum  of a  response,
sychronized with the stimulating event that elicits it,
plus ongoing, spontaneous neuroelectric activity that
is not related  to the stimulus event. For data proces-
 sing purposes, the event-related activity may be refer-
 red to  as "signal" and  the  spontaneous  activity as
 "noise."

    There are  many sources  of variability of ERPs.
 At  the tingle-cell level, there may be random fluc-
 tuations in  membrane potentials,  sometimes causing
 stray action potentials.  Presumably,  these  neuro-
 electric events are due to random variations in met-
 abolic and other such activity and may not be direct-
 ly related to conscious or subconscious (e.g., autono-
 mic, reflexive) information processing.  It appears  rea-
 sonable to call such activity noise since: (I) It may in-
                 Preconference Correspondence

 terfere with recording of signal activity. (2) The oc-
 currence of random events such as stray action po-
 tentials could  interfere with  information processing
 by the central  nervous system, and probably the CNS
 functions  in  a  manner  that  will  attenuate  such
 random effects.

    External events elicit  a sequence of neuronal ac-
tivity that may be reflected by ERPs. It appears rea-
sonable to call  this activity signal since: (1) It is not
entirely random and, to some  extent, can be brought
under  the control of the experimenter. (2) It may be
related to other observations concerning the behavior
and state  of the  subject.  Of course, there are many
types of CNS signals that  we try to attenuate or tend
to ignore  because they are not readily related to ex-
ternal events.

    There  are  difficulties with the definitions pre-
sented above. For example, from the point  of view
of an  experimenter observing ERPs, ongoing autono-
mic neuroelectric activity  is interference, but from
the point of view of the CNS, it is  signal. The prob-
lem is that the  brain  is  a multiprocessor and only
limited means are available for distinguishing the ac-
tivities of the various subsystems.

    As another example, trial-by-trial alterations  in
the state of the  subject that  affect ERP generation
may be viewed as the source of either signal or noise.
If the experimenter wishes to attenuate the effects
of  such  alterations  and  can  neither  sufficiently
control the  state  of the subject nor correlate the ERP
fluctuations with behavioral  observations, then the
ERP fluctuatuons effectively  amount to noise.  How-
ever, if the experimenter is interested in analysing the
ERP  fluctuations and/or  it  is  possible  to relate
the fluctuations to alterations  in the behavioral state
of the subject,  then they can be  viewed as signal. In
effect, one man's signal may be another man's noise.


    When  the  state of the subject cannot be deter-
mined by  direct report, the problem of changes  in
state is difficult  to deal with, and interpretations of
results can  be  murky. Various  ad hoc approaches
have been devised; however, they are not guaranteed
to always yield correct results. They work best when
there  is preliminay information available concerning
the character of the data. For examples: (1) There
may be reason  to believe that  the shape of the wave-
form  is constant, but its latency varies. In this case,
latency-compensated averaging or recognition index
procedures  might be  used. (2) There are a limited
number of  states of  interest  and data are available
that allow  initial averages to be  computed for each
state.  In  this case, discriminant procedures may be
used to classify  the data prior to averaging.

-------
Preconference Correspondence
                                            553
    Gerbrandt: Weinberg raised  the  question: "In
pattern recognition techniques, how Is the pattern to
be detected determined" and what can It be used for?
In my first dealings with such techniques, I developed
a  theta detector to quantify the records  of theta
rhythms in the rat.  I then used the detector to syn-
chronize computer averages of theta rhythms (ATRs)
at a given electrode site.  The resulting waveforms are
analogous to averaged evoked potentials. By synchro-
nizing to  theta rhythms at a given electrode site and
concurrently averaging ATRs at other electrode sites,
measures  of the  average phase, frequency, and am-
plitude can be derived  (Gerbrandt et al. 1974. Green
eta!,  1960).

    In  subsequent Investigation  of phase shifts be-
tween ATRs, I  learned more about the advantages of
the ATR  and the distortions  It Is subject to (Ger-
brandt  et  al. 1974, Lesevre et al.  1967,  Ruchkin
1971). Some  of the problems encountered could  be
large enough to result In unreliable averages, but they
could  be  corrected by a program that sorts  single
trials  on the basis of frequency modes and modes of
amplitude  heterogeneity. Analogous  programs  for
sorting on the basts of amplitude heterogeneity have
been  developed (Ruchkin 1971, John et al.  1973).
When  frequency modulation Is a  problem, It can  be
corrected  by scaling all events to the same frequency.
Analogous  programs  have   been   developed   to
minimize  amplitude  and waveform  distortions due
to  latency  jitter  (Woody  1967, Ruchkin  1974).

    I  then began wondering  what else could be done
with endogenous  event-related potentials (EERPs),
The possibilities are numerous and seem to break into
three categories.
    First, EERPs can be studied as primary depen-
dent variables. Once the pattern recognition scheme Is
developed for the endogenous event, it can be averag-
ed into EERPs by "auto-triggering" (where the trig-
gering event is Itself formed into an EERP) and relat-
ed to concurrent EERPs at other electrode sites. For
example, I have studied the topographic distributions
of theta rhythms (Gerbrandt et al. 1974) and how
they are affected by relevant experimental variables
(Gerbrandt et al.  1975). Similar work has been re-
ported  for the human  alpha  rhythm (Magnus and
Ponsen  1965, Remond  et al.  1969).  Less well  de-
fined   electrical patterns,  such as seizure spikes,  the
CNV, or P300, could also serve as templates that auto
•trigger  EERP averages of themselves and concurrent
events  at other electrode  sites whenever cross cor-
relations rise  about a criterion. This technology is al-
ready in use (Woody 1967, Ruchkin 1974). Even un^
known waveshapes can be  extracted by searching for
a highly correlated  waveshape that  is  common to  a
set  of  data samples,  using  the waveshape  as  a
first-level  template  for auto-triggering EERPs, then
using the  auto-triggered EERPs as second-level tem-
plates, etc. (Woody and Nahvi 1973). When the de-
tected events are sudden enough  in onset, duration,
or off-set, such as Is the case with seizure spikes, sin-
gle-unit activity,  or spindle bursts,  the  events  can
serve merely to synchronize EERPs in order to topo-
graphically search  for  clues about their  anatomical
and  physiological  origins.  I term these potentials
"trigger-referenced EERPs.*'  Examples in the litera-
ture are the cerebral motor potentials, trigger-refer-
enced to EMG onset (Vaughan et al. 1968, Deecke et
al. 1969 .Gerbrandt 1974). By extending the average
backwards as well as forwards in time  from the trigger
event, possible causal relations can be found between
the  activities represented by  EERPs  and the trigger
reference.

     In the second category of EERP studies, the de-
tected event is also synchronized to exogenous trig-
ger,  and the detection  process is used to  correct dis-
tortions In amplitude and waveshape that occur  due
to latency jitter in relationship to the exogenous trig-
ger.  This technique has been used for early  (Woody
and  Brozek 1969) and late (Ruchkin 1974) compo-
nents of  the average evoked  response. It should be
used more frequently  in  EP studies to  assure that
changes In amplitude  as a function  of Independent
variables are not just artifacts of latency jitter.

     In the third category of EERP studies, EERPs are
secondary data that can be used as  probes  or state
variables that  affect primary  events  such as average
EPs, single-unit  activity, or other types  of EERPs.
The   timing relationship  of  the  background state
(EERP) may or may not be considered, A number of
studies  have shown, for example,  that the phase of
occipital alpha rhythms is Important in determining
the  waveshape  and  amplitude  of  the visual  EP
(Magnus  and  Ponsen  1965,  Remond and  Lesevre
1967, Peacock 1970). Walter (1968) has suggested
that background delta activity may be an important
determinant of CNV amplitude. The possibility  also
exists that some components  of EPs  are just  artifacts
of background states (EERPs) that are superimposed
upon  the EP under  certain  conditions (Peacock
1970). Clearly, EERPs are going to be In widespread
use  in the field of EP  research, although they are in
limited use at present.

     Peters: To separate an evoked response (signal)
from the  ongoing background EEC (noise), the  two
are generally subjected  to the signal averaging process,
the  theory being that  the random noise that  is not
time-locked to  the stimulus will average  to zero,
leaving only the  signal. Unfortunately, the averaging
process does  not  provide any  information  with re-
spect  to  trial-to-trial  variability  of  the  waveforms.
This information can be obtained by simultaneously

-------
 554
                 Preconference Correspondence
calculating the variance and standard deviation along
with  the  average (Walter 1972). However, once ob-
tained, the  variance  gives no indication  of whether
the variability is in the noise or the signal, and if in
the signal whether it is due to  amplitude or latency
changes.

    The first possibility, that the variability might be
due to noise, has been examined by filtering each trial
with  a brick-wall digital low-pass filter prior to calcu-
lating the average and SD. The filtering did not sig-
nificantly affect either the average  or SD during the
CNV interval. The evoked  responses (particularly to
SI) were altered by  the last filter,  which cut off all
activity above 4 Hz. This would seem to  support the
contention that the increased variability seen follow-
ing SI is not due specifically to noise. On this basis,
one  could speculate  that  if  the pre-S variability  is
taken as a baseline, then increments in variability be-
yond that level might be  due to signal alone.  This
cannot, however, be taken as fact based on this limit-
ed data.

    Given that the above hypothesis does hold up,
the data  still  say  nothing about whether the vari-
ability is  one of amplitude or latency. This question
might possibly be  addressed  by examining the vari-
ability of the amplitude over time as opposed to at a
specific  point in time.   Specifically, given N data
points and the average amplitude and SD associated
with  each, an average of both amplitude and SD over,
say,  the first  10  points results  in what  is often  re-
ferred to  as mean amplitude and the mean SD associ-
ated  with it.   If,  in addition, the SD of the average
and SD are calculated, one is conceivably dealing with
amplitude changes over time.   The applicability of
this calculation to  latency variability is presently be-
ing investigated.

    A third  point  of interest is the minimal change
induced  in  the CNV by  selectively  filtering  single
trials  prior to averaging.   Although this  did not sig-
nificantly affect SD, the reduction in noise associated
with  the  average  may significantly reduce measure-
ment errors. The most frequently used  measures of
CNV  magnitude have  been peak  amplitude to a visual-
ly smoothed curve, area, and mean amplitude in some
interval (say  100 to 200 msec) prior to S2 (Peters et
al.  1976), all  of which  have been  chosen primarily
for their  reduction in error due  to the ongoing EEC
associated with the average. If prefiltering can reduce
this noise still further without altering CNV morphol-
ogy  and/or  amplitude, then  measurement accuracy
should improve.

    This discussion is only a preliminary review  of
the analysis  that is currently being pursued. The pos-
sibilities discussed here must  be compared to single-
tilal analysis before any final statement may be made.
     Weinberg;  Let me  raise one other more specific
issue - the question of whether topography can give
us information  about responsiveness  to  input, i.e.,
can  it extract  signals from the  EEG. The  greatest
weakness of topography has  been its  inability  to
easily  represent  topographic  relationships  as  they
occur over time. At least two dimensions must be re-
served  for designation of site, and a third for the de-
pendent  variable, usually voltage, leaving the fourth
dimension, time,  dangling. It seems  to  me  that if
there were some  simple method of reducing four-
dimensional  information  accumulated  with  topo-
graphy into  an  easily digestible  two-or even  three-
dimensional  display,  this  technique would  tremen-
dously enhance  our understanding of the total pat-
tern  of activity within the brain that is responsive to
input.
    One way in  which this might be done is to use
correlations for the purpose of collapsing two dimen-
sions. What topography attempts to do is describe the
spatial change in  voltages over a two- or three-dimen-
sional surface by extrapolating voltage gradients from
observed voltages distributed in space, and the re-
lationship of the  gradient changes to time. Specifical-
ly, what is meant by "relationships"? In this context,
the term refers to the ways in which voltages in dif-
ferent  parts  of the  brain covary.  And this can be
determined through correlations. But correlations be-
tween what? It is very difficult, if possible, to  corre-
late everything with everything else and still retain a
simple picture of relationships.
     A  partial  solution  is  to  identify  an arbitrary
 standard within the brain, such as vertex change in re-
 sponse  to an input, which is then correlated with ac-
 tivity from numerous other sites. Once a standard is
 selected,  the  question can  be  asked:  What  is the
 degree  to which activities  in  different parts of the
 brain differ with respect  to the standard? These cor-
 relations would be an index of the  relationship be-
 tween the activity recorded from different sites in
 the  brain, all  of which  are compared  to the same
 standard. This pattern can then be displayed as chang-
 ing  correlations over time and plotted  on  a two-
 dimensional  surface  representative  of the array of
 electrodes, using the Z  axis  for time. This should
 work nicely for the changes that occur after a signal,
 for some given small period of time, but it is not a
satisfactory   arrangement for  the  observation  of
long-term changes in  a spontaneous record. Neverthe-
less, when used for  changes occurring after known
inputs,  on  successive  occasions, the  correlations
plotted  with  techniques  of  back-suppression, at
multiple sites,  give  a kind of topological representa-
 tion of voltage relationships  occurring over wide-
spread areas of the brain.

-------
MULTIVARIATE ANALYSIS OF
EVENT-RELATED  POTENTIAL DATA:
A TUTORIAL  REVIEW

E. DONCHIN AND E.F. HEFFLEY, III
Cognitive Psychophysiology Laboratory, Department of Psychology,
University oflllinois, Champaign, IL, U.S.A.
Introduction

Purpose

    This review is concerned with the choice among
techniques that can be used to analyze data acquired
in  studies  of event-related potentials (ERPs). Many
analysis procedures are available, and their appropriate-
ness has been a matter of some dispute (Donchin
1969, 1976; Vaughan  1974; John et al. 1978).  Yet
no technique is inherently good or bad. Techniques
exist to extract information from data, and the choice
depends on the degree to which any given technique
makes the data informative.  Data yield information
when they can be made to reduce uncertainties that
investigators have with respect to specific questions.
The choice of an analysis technique depends, there-
fore, on the experimental questions and on the data.

    This review describes several of the available  ERP
data analysis procedures within a  consistent frame-
work. Focus is on the application of multivariate sta-
tistical techniques. Many investigators are apprehen-
sive of multivariate procedures even when such tech-
niques are particularly appropriate.  Through an in-
tuitive rather than formal development, we hope  to
demonstrate  the  motivation for  multivariate  ap-
proaches to ERP data analysis (for a more rigorous
presentation, see Glaser and  Ruchkin  1976).  Our
purpose is to show that multivariate techniques are a
natural and logical extension of the more commonly
employed univariate techniques.
    Most,  if not all, data analysis techniques that
have been applied to ERPs can be described as linear
1 Preparation of this paper was supported by ARPA,
ONR  contract  N-000-14-76-C-0002; Earle Heffley
holds  an NSF Graduate  Fellowship, NSFSM1  77-
21998.  The helpful comments of Connie Duncan-
Johnson, Richard Horst, Pat Michie, Greg McCarthy,
and  Dan  Ruchkin  are  gratefully acknowledged.
combinations; the diversity of approaches masks an
underlying common structure. Making this structure
explicit  should  facilitate  meaningful  comparisons
between techniques and provide a heuristic for select-
ing  the techniques best suited for different experi-
mental problems.

The ERP as a vector
    Fig. 1 presents four hypothetical ERP waveforms.
Each waveform is simply the sequence of voltages
recorded at each of the time points. In the example,
there are 15  such measurements for each waveform.
Thus, a waveform is represented by an ordered set of
numbers [x(l j), x(2j), ..., x(i j),.... x(Pj)], where
j identifies the ERP, and i=l,. . .,Pindexes the time
points.  These primary data vectors  (recall that an
ordered  set of numbers  is called  a  vector) are the
average  ERPs computed from the  EEC segments
recorded during each trial of an experiment. Average
ERPs can be arranged in a data matrix, as shown in
Fig. 1, where rows represent the ERPs. There will be
one row-vector for each combination of values of the
independent variables in the experiment. The columns
represent the time points. The measurements at each
time point are the primary dependent variables in the
study. Thus,  there are as  many dependent variables as
there  are  time  points.  The  number of  variables
depends, therefore, on the sampling rate used during
digitizing. It is this many-dependent-variable situation
that compels consideration of multivariate statistical
techniques (Tatsuoka 1971, Donchin  1966).

    The data matrices generated in ERP experiments
are fairly large. When the data from all subjects for
several  electrode positions  are  accumulated over
several  experimental conditions and replications,  a
matrix  with  many hundreds of rows  is created. If
single-trial vectors   are   considered,  the  matrices
are gargantuan.

-------
556
                                                                  Donchin and Heffley
 lo o 1011111110 ie it it it 111110 a ]
      ERPVECTOR 1
             ERP3
              I 0 0 4 I I I 4 14II It It 11 It 10 0 1
                  ERPVECTOR 2
                         ERP 4
 I « 010111111
     ERPVECTOR 3
      ERP  !
                  ERP VECTOR 4


    ERP DATA MATRIX
0 0 10 11 II 11 10 It 11 11 11 1t II 10 0
00 4 I I I 4 14 II 10 It 1t II 10 0
0010K1I1I4I7II7I40
004SII4ITII7I40
                    > I 7 I * 10 II 12 11 14 II
                    TIME POINT
Fig. 1. Four simulated average ERPs are shown along
with their corresponding vector representations.  Be-
low the ERPs, the vectors have been grouped into the
average ERP data matrix.
ponent. The combination rules are quite diverse, but
most can be described as linear combinations.

Linear combinations

    A linear combination of the elements of a vector
is formed quite simply. For a vector with P elements,
a set of P coefficients is required, each one correspond-
ing to one point  in  the vector. Forming  a  linear
combination involves two steps  (Fig. 2). First, each
data value  x(i), i=l,  .  . ., P, is multiplied by the
corresponding  coefficient a(i).  Then these products
are summed to yield the linear combination y(l). A
different linear combination of the same data points
x(l) through x(P) can be obtained with a different set
of coefficients b(l) through b(P), producing a new
value y(2),  as shown in Fig. 2.
     While number of rows in the matrix is determined
 by experimental  design and cannot be reduced, the
 number  of columns may well be  reducible since the
 time points in themselves are not of primary interest.
 Rather, ERP studies tend to focus on the behavior of
 derived dependent variables called components. The
 term component, which will appear frequently in the
 following discussion, is used to describe the elementary
 processes of which the ERP is presumably constructed
 (see Donchin et  al., 1978, for a discussion of this
 concept). The number of components is generally
 much smaller than the number of time points. Each
 component  therefore affects  the values measured at
 many time points, and at each point one may measure
 the  contribution  of many  components. The measure-
 ments  over  different  time points are, then,  not
 independent.  Furthermore, if it were possible  to
 measure  the contribution of each component to each
 time point,  then  it would be possible to express the
 measures of  each component as a single value  that
 combines  the  component's  contribution  at  the
 appropriate  points. The number  of columns in the
 data matrix could thereby be reduced to the number
 of components.
    This has, in fact, been standard operating proce-
dure in the ERP laboratory. Investigators combine the
elements of the data vectors to yield one value per com-
                                                    (ERP ROW VECTOR]   X
                                                             COEFFICIENT
                                                             [  VECTOR  J
                                      COMPOSITE
                                      VARIABLE
(X2a2)

(X2b2)
                    (X3a3)
                    
-------
Multivariate Analysis of ERP
                                                                                               557
K, i X12
K21 X22
i



r

•ll • •• «1N
'21 • • • »2N
'31 • •• «3N

•M • •• »PN
-
1

• EHP
M
Y,, ... Y1N
Yj, ... Y2N

YM1 •• • YMN
1 N~
COMBINATION
1 N
COMBINATION
                                                      The peak measurements may then be subjected to an
                                                      analysis of variance to test for significant differences
                                                      across experimental treatments. We do not, of course,
                                                      imply that this paragraph describes the actual proce-
                                                      dure  used by an investigator executing a peak analy-
                                                      sis.  We do assert that whatever procedure is used in
                                                      practice,  it  can  be  described  formally as a linear
                                                      combination.
    IERP DATA MATRIX)  X
                    [COEFFICIENT
                    L  MATRIX  J
I COMPOSITE 1
L VARIABLES]
 Fig. 3.  The process of forming a linear combination
 of a single ERP is generalized to the case where N
 linear  combinations are computed for each of the M
 ERPs from an experiment.

 used for inferential purposes. There will be an inter-
 action between the inferential intent of the investigator
 and the measurement he chooses to use. One purpose
 of this paper is to elucidate these interactions.

    In  the remainder  of the paper, four analysis
 techniques  are described, two traditional (peak and
 area measures) and two multivariate (principal com-
 ponents and discriminant analysis). In each case, the
 rules  that  determine the coefficients that define the
 analysis and then its advantages and disadvantages are
 discussed.  An intuitive understanding of these rules
 should  enable an  investigator  to determine which
 techniques  are appropriate given the nature  of the
 data and the questions addressed by the research.

 Analysis techniques

 Peak analysis

    Peak measurement is probably the  simplest ap-
 proach to quantification of differences in amplitude
 of ERP components. A peak is defined  either as the
 largest  or  the smallest voltage value  in an interval
 representing the latency range of the component being
 measured. The measured values  are expressed either
 as the  difference   between  successive  peaks (e.g.,
 N100-P300) or  as the difference between the peak
 and the average of several points (e.g., NIOO-baseline).

    Peak measurement can be easily expressed as a
 linear combination. We begin with a time function ex-
 tending over an epoch. First, we select a time segment
 in the epoch so  that only values x(K) through x(L)
 are considered.  Next,  the maximum  (or minimum)
 value x(M) is found between points  K and  L. The
 coefficient a(M)  for the point at the peak  is set  to one.
 All other coefficients  are set equal to zero. Forming
 a linear combination with these coefficients is clearly
 equivalent  to peak measurement. Fig.  4 illustrates
 measurement of  the two peaks for the ERP from Fig. 1.
    Peak  analysis, like  all  techniques,  should  be
evaluated  in terms of the nature of ERP data. First,
peak  analysis  assumes  that  it is  not necessary to
measure  more than  a single  point  to  adequately
quantify  the   magnitude  of  a   component.  This
assumption  restricts  its  applicability,  as discussed
below.  Furthermore,  because  peak  measurement
includes contributions from all components that affect
the voltage at a time point,  the investigator may in
fact be measuring effects due  to  components other
than the one he believes accounts for the peak.

    The  principal  advantage of peak measurement is
its  face  validity.  It  is intuitively  appealing.  The
method is a direct analog of the visual inspection pro-
cedure for gaining impressions of  data trends. More-
over,  it is almost the only measure  that is easy to
obtain  from  x-y  plots.  Many   investigations are
restricted to this method, as it requires nothing but
an x-y plotter, a ruler, and enough time for this rather
tedious job.  An  additional  advantage  is  that the
magnitude measurements are fairly independent of
differences in latencies of components across ERPs.

    The technique has several disadvantages. It is not
easy to avoid experimenter bias in  identifying a peak,
                                                     o o w 11 11 11 10 11 11 u 11 11 i« 10 o

                                                     00 4 I I I 4 14 II It » II 1* 10 0

                                                     00 10 II II II 4 I 7 I I 7 « 40

                                                     004IIIIITII7I40

                                                               TIM! POINT
                                                                                     0

                                                                                     PMK
                                                          II I

                                                          I I

                                                          PIAK
                                                              (ERP DATA MATRIX)
                                              fcOCFFICIINf]  f PEAK 1
                                              L  MATRIX J  [MATRIX]
                                                   Fig. 4. Peak measurements are made for the ERPs
                                                   illustrated in Fig. L

-------
 558
                                                                               Donchin and Heffley
 because waveforms are often not as sharply delineated
 as in idealized examples. Also, it is often difficult to
 define a  baseline  or reference peak  in the data,
 particularly when there  are few trials in an average or
 when  rapid  oscillations are  superimposed on slow
 potentials that begin  during the baseline period and
 continue throughout the epoch (Squires et al. 1977).
 (Baselines will  be  discussed later.) Another problem
 appears when several peaks are measured in the same
 waveform. These peaks  are often not independent,
 and the  application  of  inference techniques  that
 assume independence may  be inappropriate. Com-
 ponent overlap presents a serious problem  in peak
 measurement.  A typical example  is the effect that
 the "slow wave" (Squires et al. 1977) has  on the meas-
 urement of P300 amplitude. A similar vexing problem
 is the dissociation of ERP components from contingent
 negative variation (CNV) (Donchin et al. 1975). Fig.
 5 illustrates  a  case  in  which a slower  component
 accounts for differences in another component meas-
 ured  by  the base-to-peak method.  Unfortunately,
 such relationships, while probably prevalent, are not
 always so clear in real data.
   UNDERLYING COMPONENTS
OBSERVED ERPS
                                           12
      19
                        -S
      19
                                           14
      19
      19
Fig. 5.  Baseline-to-peak measurements are made for
ERPs  composed of two underlying components: a
constant biphasic waveform and a varying slow proc-
ess.  The numbers represent peak values relative to
the initial baseline (arbitrary units).  The measures on
the observed ERPs suggest  that  the sharp positive
(downward) component varies across the four ERPs
when in fact it is constant.
    The peak measurement technique suffers from
other limitations. Because the measurements are based
on a single point, moderate amounts of nonsystematic
variance can obscure real differences. Further, peak
measurement is not very applicable when measuring
slow potentials such as the CNV or readiness potential
(RP). Finally, peak amplitudes reveal nothing about
the waveshape of the component being measured.

     It  should  be  noted that, despite  its apparent
simplicity, peak  measurement presents  challenging
problems in statistical estimation. The waveforms are
rarely  smooth and  the precise latency at  which the
component achieves a  maximum cannot  often  be
defined unequivocally. The relative order of magnitude
of different peaks may  well be preserved even when
the errors  of measurement  are large. However, when
these measurements are  used to define the latency of
components, the uncertainty in peak definition may
lead to  serious difficulties. Other  techniques  have
been used  to identify peaks (e.g., locating the inter-
section of the tangents of positive and negative slopes).
Whatever the solution, the investigator should avoid
seduction by the deceptive simplicity of the measures.

    In  summary, peak  measurement techniques are
most applicable when the data analyst is dealing with
distinct components exhibiting  clear differences  in
ERP records. Peak analysis was probably an adequate
approach  for the earlier, exploratory phase of  ERP
research when large, dramatic effects were sought. In
general, the procedure will be valid when the processes
underlying the  ERP are well understood or possibly
when the experiment is unidimensional in scope (e.g.,
when stimulus  intensity  is the only variable). Peak
analysis is  limited in scope when the investigator is
attempting to dissect complex ERPs, especially when
components overlap.

Area measures
    The measurement of  the  area  under selected
portions of the ERP curve is similiar to peak analysis.
Areas are  computed  by  integrating the  voltages,
relative  to  a reference baseline. Limits can be set so
that the interval encompasses either an entire compo-
nent or some portion of it.  The area is also a linear
combination of the  ERP. The first step is to select a
range of points K  through  L that correspond to the
duration of  the  component  or subprocess being
studied. The coefficients for points K through L are
then set to one; all other coefficients are set to zero.
A baseline  interval is chosen for reference. Computa-
tion  of areas  by linear combination  for  the  four
sample ERPs of Fig. 1 is illustrated in Fig. 6.

    Area analysis is based upon specific assumptions
about  the  data.  Unlike  peak  measurement,  area
analysis is based on the premise that the amplitude  of
a component is accurately gauged by combining values

-------
Multivariate Analysis of ERP
                                                                                                 559
   0 0 10 1C II 16 10 16 16 19 1* II 16 10 0

   00 4 8 • 8 4 14 IB 19 19 18 IB 10 0

   00101i1l1«4«7t>7«40

          I84e71>7(40

             TIME POINT
0 0
            [ERP DATA MATRIX)
   00       70 IW

     0       33 118

   10       MM

   10       33  M

   1  0       AREA

   1  0

   1

   0

   0

   0

   0

   0
   0
   AMA

[COEFFICIENT]   ["AREA"!
[ MATRIX J " [MATRIXJ
 Fig.  6.  Area measurements are made for the four
 ERPs from Fig. 1. Each area is a linear combination
 of the corresponding ERP vector.
 measured at several time points. If the analyst chooses
 an appropriate interval, then it is likely that most of
 the contribution of the component to the total ERP
 will  be included.  Unfortunately, this  measurement
 procedure does not deal adequately with component
 overlap. As the area limits are extended, the  probabil-
 ity  increases  that  contributions from other compo-
 nents will be included.

     Area measurement offers many of the advantages
 of the  peak methods. The technique is easy  to under-
 stand  and involves relatively  little  computation,
 although a ruler and x-y plotter no longer suffice. In
 addition, some degree of latency jitter can be tolerated,
 provided the area limits are appropriately  adjusted.
 Areas  are  more  accurate  representations of  the
 magnitude  of slow components than are peak meas-
 ures. Also, they are less susceptible to random varia-
 tions if enough time points are included in the interval.

     Area measures share many of the disadvantages
 of peak methods. Determination of integration limits
 is often difficult, or even arbitrary, because compo-
 nents cannot  be accurately  resolved by visual  inspec-
 tion.  Component  overlap and intercorrelation can be
 serious  problems,  and such  complications may  be
 very difficult to identify. As with peak detection, the
 technique is sensitive to the  definition of the baseline.
 Finally, the procedure yields no information about
 component waveshape.

     In summary, whenever the investigator  is analyz-
 ing ERPs with independent, well-defined components
 and  clearly defined baselines, area measurement is
 valid. However, when component overlap is suspected,
 or when a more detailed analysis of the waveform is
                                                  needed,  then one of several  multivariate techniques
                                                  might yield better results.
                                                  Principal component analysis (PCA)

                                                      Component  overlap: At this  point, it  will be
                                                  instructive to review the nature of linear combinations
                                                  for peak and area component measures. Peak analysis
                                                  implies that the  magnitude of a component  can be
                                                  adequately  represented by  the peak value  of the
                                                  corresponding wave. Thus it assigns a coefficient equal
                                                  to 1.0 to the time point at  which  the wave shows a
                                                  maximal  deflection (coefficient vector A in  Fig. 7).
                                                  For area measures, the assumption  is  that the sum of
                                                  values  over an  interval  characterizes  component
                                                  magnitude. These points are given equal weight by
                                                  assigning  1.0 to their coefficients (vector B in Fig. 7).
                                                  Weighting these  points equally over  the interval K
                                                  through L  results in  incorporating an undetermined
                                                  contribution of  the magnitude of adjacent or over-
                                                  lapping components into the  measure. The greater the
                                                  component overlap,  the more serious the problem.
                                                  Clearly, it would be preferable to weight the values in
                                                  the ERP in proportion to the contribution the compo-
                                                  nent of interest  makes to  the variance at each  time
                                                  point. The contribution  at each point is determined
                                                  by the component's waveshape. Such  a set of weights
                                                  is represented by vector C  in Fig. 7. Coefficients are
                                                  highest in the region of the ERP where the amplitude
                                                  is determined to  the greatest  extent by Component II.
                                                  In the region where  Components I and II have the
                                                  greatest degree of overlap,the  weights are small, reflect-
                                                  ing the assumption that Component  II accounts for
                                                  only a part of the amplitude variation  in this interval.
                                                                                    CHI
                                                  ERP
                                                  (A) PEAK WEIGHTS
                                                                         M
                                                  (B) AREA WEIGHTS
                                                  (C)'IDEAL* WEIGHTS
                                                  Fig 7. Three coefficient vectors for measuring Com-
                                                  ponent II are represented graphically.

-------
 560
                            Donchin and Heffley
     Principal component analysis (PCA), followed by
 an analytic rotation, is one procedure for obtaining
 such a  set  of weights. The pattern  of covariation
 among data points within an ERP is used in determin-
 ing the coefficients. PCA is a technique for extracting
 a small number of components,  each representing
 systematic influences on many time points, from the
 total variance in the ERP data matrix.

     Example:  two-dimensional  ERPs:  If  measure-
 ments at two points are correlated, it is likely that
 they are  influenced by  a common  process.  For
 example, the values of individual points constituting
 an ERP component  tend  to increase and  decrease
 together, or  covary, as the  magnitude  of the compo-
 nent is modulated. This suggests that the matrix of
 covariances between time points might serve as a basis
 for  determining ERP components. If a  number of
 adjacent time points exhibit high covariance,an ERP
 component can be assumed to be influencing them
 jointly. Patterns in the covariance matrix are com-
 plicated by component overlap, but if two overlapping
 components are not highly correlated across all experi-
 mental  variables and electrode sites,  then principal
 components  and  associated techniques  should be
 successful in separating them.

     A  few  examples may  help in  visualizing  the
 relationship   between   variance,  covariance,  and
 principal components. Each example will involve a set
 of average ERPs, each  made up of only two time
 points, x(l) and x(2), representing the peak amplitudes
 of the components. These  ERPs can be represented
 by two-element vectors consisting  of  values  for  the
 variables x(l) and  x(2). We can  plot the ERPs in a
 two-dimensional space with  x(l) as the abcissa. In  the
 first example  (Fig.  8a), x(l) is the value at a time
 point strongly influenced  by  an  underlying ERP
 component, while x(2) is a  time point from a part of
 the sampling epoch not influenced by that component.
 The variance  within these waveforms,  as represented
 in our two-element vectors, occurs almost  entirely
 along axis XI. Points x(l) and x(2) exhibit little
covariance. If we   are interested in  a  measure  of
Component I in any given waveform,  then it suffices
 to measure the value along XI. Thus, the variances
and  covariances show  that the component  varies
along XI, and further, that XI  is  the only axis  (or
 time point) the  component influences. We  can use
 this information to derive a measure of the magnitude
 of the component  in any  given  waveform, which in
 this case is simply the value along the XI axis.

    In the second example,  x(l) and x(2) are affected
 by independently varying components (Fig. 8b). In
 this case, there is considerable variance  along both XI
 and  X2, and  so neither variable alone characterizes
 the total variance in the data matrix.  Further, these
 two variables do not covary, i.e., the value of x(l)ls
     X2
            (A)
                    f
                      Xl
     X2
            (B)
                     XI
     X2
                     Xl
Fig.  8. Scatter plots represent measurements at two
points along an ERP. (a) ERP varies a great deal at
point x(l) but is relatively constant at x(2).  (b) Var-
iance is present at both points, but x(l) and x(2) are
unrelated, (c) Both x(l) and x(2) are unrelated,  (c)
Both x(l) and x(2) vary. In addition,  the two vari-
ables exhibit strong covariance.  Axis AI is the di-
mension  of maximum variability. Axis A2 is ortho-
gonal to Al.

not systematically  related to the value  at x(2). Two
separate components can be assumed to Influence the
ERPs. If, for any given  waveform,  we wish to assess

-------
Multivariate Analysis of ERP
                                            561
the magnitude of Component I, then XI provides the
appropriate measure, and we can ignore the variance
along X2. Similarly, Component II is measured along
X2.   Thus,  we have two independent components,
one identified with XI, the other with X2.

    The   third example   illustrates  the  case  that
motivates the use of PCA and rotation techniques in
ERP  research.  In  this example, x(l) and x(2) are
affected  by  the same underlying component.  Both
variables  exhibit a  considerable  amount of variance.
However, because they are influenced by a common
process, or ERP component, they covary quite strong-
ly, as revealed by the elongated  diagonal scatter plot
in Fig. 8c. Although the two variables  are not perfect-
ly correlated,  there is substantial common variance.
The  imperfect correlation may occur  because one or
both of the points  are affected by other components
or because  noise  and error cannot  be completely
eliminated. As we  sample two points at shorter and
shorter temporal intervals, the  correlation between
them will increase.

    If there is only one  process, then it  would be
convenient to find the  axis along which  it can be
measured, as XI was used in the example in Fig. 8a.
This is a crucial point: one component is to be con-
sidered as a single  dimension, although its existence
is manifest at many time points. At each time point,
that portion of the variance that is influenced by the
single  ERP component  is extracted.  The maximum
variability occurs along Al as in Fig. 8c rather than
along XI or X2 as in Fig. 8a, although Al is deter-
mined by both axes.  By examining the relationship
of Al to the  two  original axes, the component's in-
fluence at each of  the  time  points can  be deter-
mined.   Furthermore,  a measure  of the compo-
nent can be  obtained for each ERP by projecting the
data  points  from  the  two-dimensional  time-point
space onto the Al axis in  the new component space.

    There are two different ways to characterize the
data in terms of Al. The new axis can be more or less
"like" either of the original axes. A useful measure of
the  similarity, or  association, between  the  derived
axis and  each  of the original axes is the cosine of the
angle between the original and the derived axes. There
are,  of course, as many such angles for each  derived
axis as there are original axes. The cosines of these
angles, which are equivalent to correlation coefficients,
are called component loadings in PCA jargon. A load-
ing is therefore a measure of the association between
the  principal component  and the original axes, and
for each principal component there are as many load-
ings as there  were original  axes. There is, also, for
each of  the  ERPs  in the data-base  a measure that
expresses its value on the axis  Al. This measure is
called a component score. For each principal compo-
nent, there are as many component loadings as there
were original axes (time points) and as many compo-
nent scores as there were  observed ERPs. In this con-
text, we are no longer interested in values at time
points but  rather in components of variance. There-
fore, a new space spanned by the derived components
rather than by  the time  points  is developed. Each
component will  be represented  by an  axis in the
component space. Amplitudes can then be measured
along Al  just as  they were with  XI  in Fig. 8a.
These   measurements  are   represented  by  the
component scores.

    While  Al accounts for most of the variance in
the data, there may remain a significant percentage of
variance still to be explained. This can be viewed by
observing the residuals, or the representations of each
ERP after  the influence of Al is removed from the
original  data. It is possible to obtain a second  axis A2
that meets the  criteria that (1) it accounts  for the
largest possible  percentage of the residual variance,
and (2)  it is orthogonal to Al. There will  then be two
new axes for the ERP data, each accounting for an
orthogonal source  of variance in  the data. Note that
in the  example  the  variance along A2  does not
represent   a second ERP  component.  Rather this
variance reflects  the variability in the two time points
that is not due to Component I.

    Extensions  to multidimensional data: For two-
dimensional data, the process is thereby exhausted. If
the original data are multidimensional, however, the
process  can continue, successively adding orthogonal
axes. The number of axes can theoretically be equal
to the number  of dimensions,  i.e., the number of
digitized voltages in the sampling epoch. However, in
practice, due to the large  measure of correlation
between data points,  the percentage of  variance ac-
counted for by successive principal components dimin-
ishes rapidly. For  ERP data, most of the variance is
accounted  for by  six  to eight components (Donchin
1966, Donchin et al. 1975, Squires et al. 1977).

    The result of this procedure is a representation of
each  ERP  in  a space that might have five or six
dimensions as opposed to the raw ERP space that had
P dimensions, where P may be quite large (as 50 to
250  time points are typically measured). The value
along each new axis for each ERP (i.e., the component
score) is a linear combination of the original P points.
The  principal components of variability among points
are determined  by analyzing patterns of covariation,
and  the relative  contribution of each component to
each of the original ERPs  can then be assessed.

    A complete  PCA involves at least four major
steps. First, the  P x P variance-covariance matrix is
computed. Second, the principal  components are ex-
tracted from this matrix.  This produces a component

-------
562
                          Donchin and Heffley
 matrix showing the loading, or influence, of the prin-
 cipal components at  each  time point. The loadings
 matrix will have  one  row for each principal compo-
 nent and  one column for each time point. Another
 way to think of  the weights  is as hypothetical basic
 ERF components. Because these basic ERP compo-
 nents tend to be somewhat correlated across observed
 ERPs,  and  also  because time points within each
 electrode  often covary, it is usually the case that prin-
 cipal components exhibit weights across the entire
 epoch. In order to remedy this undesirable situation,
 an  additional  step is performed. A Varimax rotation
 (Kaiser 1958) is performed as a third step to better
 derive  the basic ERP  components from the loadings
 matrix. The  Varimax procedure  has the effect of
 maximizing high  loadings and minimizing low ones
 while maintaining orthogonality. The result in practice
 is to concentrate the  high  loadings  for each compo-
 nent to  a restricted  region  of the epoch, thereby
 producing distinct basic ERP components.

     The  fourth stage  locates  each observed ERP in
 the new, reduced, rotated space. The transformation
 is accomplished by multiplying each  observed ERP
 vector by a coefficient vector derived from the rotated
 loading vector for each component  output from the
 Varimax step. The component loadings, which reflect
 the  association between  each time  point  and each
 particular  component, are used to develop the coeffi-
 cients needed to form linear combinations. Each raw
 ERP vector yields a composite score  representing a
 measure of the magnitude of a specific component in
 a specific  ERP. If M principal components were re-
 tained and rotated by the  Varimax procedure, then to
 each ERP  there would correspond M such linear com-
 binations.  These measurements can then4be subjected
 to an analysis of variance  to test for significant differ-
 ences across experimental  treatments.

    Principal  components  as linear  combinations:
 Before examining  the  assumptions  and restrictions
 of PC A, as described,  it is appropriate to relate PC A
 to the discussion  of coefficients, illustrated in Fig. 7.
 The objective of  PCA is  to produce a set  of weights
 for each  component that will permit assessment of
 the  contribution  of each component to  amplitude
 variance of each point in the  ERPs. The linear com-
 bination of raw ERP time points to form  composite
 scores for  each component for  the ERPs is represented
 in Fig. 9.  The raw ERPs  (Fig. 1), now listed as devia-
 tions from the grand mean,  are given. The weights
 were determined  by a PCA of the covariance matrix
 for these waveforms followed  by a Varimax rotation?
2  Each of  the  four  waveforms was  replicated  20
times,  with  independent  uniform random  noise
(-0.01, +0.01) added to each point in order to avoid
computational difficulties. The  arithmetic precision
of the computations  is  greater than  indicated  by
the figures.
  |04 04 tO 4* 4J la 44 44 M IJ i.« 04 04 14 04J
  PIoo 0.0 -i.t-44-4444^4 u u 14 '* u 44 u 04)
  10.0 0.0 14 44 4J 4J-1.l-4J4.l-U4J-U4.t-MO.tl
  [o-O OH -14 -44 -44 -44 11-4J 44 4J 44 44 44 44 O.oj

              TIMi POINT
   40 40
   40 40
   •40 41
   •40 40
   •40 41
   •40 40
   41 4t
   41 41
   40 -40
   40 -40
   41 -40
   40 -40
   40 -4>
   4t -40
   .40 40

  OOMPOMNT
            [ERP DATA MATRIX)
        ,
COEFFICIfNT
L  MATRIX
1-hssr]
 Fig. 9. Component measurements for the ERPs from
 Fig. 1 are made using the PCA-Varimax method. The
 grand mean ERP has been subtracted from each raw
 ERP before the covariance matrix was factored. Note
 that coefficients are maximal in  the  region where a
 particular component dominates.
The result  of the linear combination is a matrix of
component scores, as shown in the right of Fig. 9.

     As noted briefly above, the nature of the numbers
yielded  by PCA and their role in data analysis must
be understood. The analysis yields two sets of values:
the coefficients (i.e., the component loadings) required
for forming the linear combinations and the actual
linear combinations (i.e., the component scores) com-
puted  for  specific  ERPs. In practice, use is often
made of the set of component loadings used  in com-
puting the coefficients. These loadings, a set of which
is available for each component, represent a measure
of  the  association  of each time  point  with each
component. For  ERP data,  the  loadings obtained
after a Varimax rotation tend  to be  large over
restricted regions of the epoch and small elsewhere.
This  suggests  that  the principal  components can
be  identified with underlying ERP components. A
separate set of principal  component loadings  will
be  extracted for each independently varying ERP
component.

    To  illustrate these concepts,  the ERPs displayed
in Fig.  1, and the ERPs labeled "observed ERPs" in
Fig. 5 were subjected to PCA. In Fig. lOa, the  compo-
nent loadings from  PCA of Fig.  1 ERPs are  plotted.
Because the two ERP components varied independent-
ly,  two principal components were extracted. Only
one  component was extracted from the Fig.  5 ERPs
(Fig. lOb). Note that this  set of loadings represents
the influence of the second "underlying component,"

-------
Multivariate Analysis of ERP
                                            563
       (A)
 Fig. 10.  Component loadings, (a) Component loading
 vectors from PC A of the  waveforms presented in
 Fig. 1 are plotted, fb) the single principal component
 vector from PCA of  the 'observed ERPs'  in Fig. J
 is illustrated.
The PCA did not extract a component corresponding
to  the  first  "underlying component" because this
component  did  not vary across ERPs. It is notable
that even though visual inspection suggests that the
waveforms  in  Fig.  5 differ in  their peak-to-peak
amplitudes,  the  PCA correctly  extracted the under-
lying "slow component."
     Fig. 11 presents plots of loadings obtained  in
 several  different  experiments.  In  each  case, the
 "reference" ERP is the grand mean ERP obtained by
 averaging across  all  subjects, all electrodes, and all
 experimental conditions. It is clear that, despite the
 diversity of the  experiments, the PCAs  reveal  a re-
 markably uniform structure for the ERP components.
     While  PCA  identifies  systematic  sources  of
 variance in the data,  it is not necessary that these
 sources correspond to the experimental manipulations
 of interest to the  investigator. This correspondence
 must be established by analyzing the linear combina-
 tions, or component scores.  It is these linear combi-
 nations that are most  clearly analogous to the base-
 to-peak measure discussed previously.  It is often true
 that some components are, and some  are not, affect-
 ed by experimental variables.  Thus, for example,
 Squires et al. (1977) found  that components corre-
 sponding to P300 (component 2 in Fig.  lla) are af-
 fected by  the probability that the eliciting stimulus
 will appear. Similar plausible relationships were es-
 tablished for other poststimulus components.  On the
 other hand, component 4 in  Fig. lie, while account-
 ing for 12% of the total variance, was not affected by
 any experimental  manipulation, and represents such
 factors as  differences in baseline  levels between sub-
 jects and between electrodes.
    It is very important to note that components re-
vealed by  PCA  depend on the variance in the data.
Some part of this variance is induced by the experi-
mental manipulations. For example,  some  of  the
structures  of Fig. 11 are restricted to the region im-
mediately following a stimulus. Others show a compo-
nent that corresponds either to the CNV or to P300
appearing  between stimuli.  In Fig. lla and b, only
one component appears in this region. Fig. 1 Ic shows
two  components in the same range because experi-
mental  manipulations  generated  two  sources  of
variation in the same time region.
    Several assumptions underlie application of PCA
to ERP data. First, it is important  to note that the
principal components model is a linear model. In other
words, it is assumed that individual ERP components
do not interact; i.e., they simply sum together to pro-
duce the complex ERP waveform.
    Second,  the analysis  assumes that  the major
sources of variance  are orthogonal. In most cases,
independent  manipulation  of ERP components of
interest will be characteristic of a good experimental
design. In addition,  the different scalp topographies
of components  will enhance  their  independence.
Nevertheless,  there will be cases in which two under-
lying  ERP components are highly correlated across
experimental variables and electrodes. In such cases,
PCA  will produce a loading vector and component
scores based on the compound waveform. Therefore,
it is not always possible to equate a principal compo-
nent with a specific deflection in the ERP. Therefore,
PCA  is not a magic and foolproof way of analyzing
experimental data. It is a  tool  that must be used in
conjunction with good design  based  on a plausible
theoretical approach. However,  examination of the
loadings matrix will  reveal  the nature of the principal
components.  Independence  does  not  necessarily
imply orthogonality, but it has been our experience
(as shown in Fig. 11) that an orthogonal model can
satisfactorily  represent many  data  bases. Further,
techniques are available for examining the assumption
of orthogonality (Harman 1960, Mulaik  1972). To
produce a nonorthogonal solution, the analyst follows
the principal components extraction with an oblique
rotation instead of the orthogonal Varimax rotation.

     Third, component variability is assumed to be in
 the amplitude  rather than  in the time domain. There
 might be considerable difference in the time at which
 a component peaks so that the amplitude at a given
 time point varies because of latency rather than "real"
 amplitude  differences.  The  principal components
 technique will extract a component associated with
 this latency-based amplitude variability, as illustrated
 in Fig.  12. In this example, latency  variability of a

-------
 564
                                                            Donchin and Heffley
                                                 REFERENCE ERP
  REFERENCE ERP
                                                                COMPONENT
                                                 REFERENCE ERP
  [A]          3
               /\  4
  COMPONENT/ \  /s  6
   LOADINGS /  \ /
                                                 COMPONENT
                                                  LOADINGS
 REFERENCE ERP
                                                 REFERENCE ERP
 COMPONENTA
/\         x-x
/ \   4   /    \
                                                 COMPONENT
                                                  LOADINGS
                                                            4
Fig. 11. Component loadings and reference ERPs from P300.CNV, and RP studies, (a) Squires et al. 1977, (b)
Duncan-Johnson and Donchin 1977, (c) Donchin et al. 1975, (dj McCarthy and Donchin 1976, (e) McCarthy
and Donchin, in press.
constant-amplitude ERP component produces a load-
ing  vector that is  nonzero in the entire range over
which the component varies. Its biphasic form results
from the fact that the waveform in the right segment
of each ERP is above the grand mean but in the left
segment it is below the mean. As the example demon-
strates, neither the  component nor the variance of its
amplitude will be adequately represented. Unfortun-
                                    ately, PCA will not discriminate between variance due
                                    to latency changes and true magnitude  differences.
                                    Careful examination of average ERPs, however, should
                                    make such effects apparent. If latency variability is
                                    evident, then an adaptive filtering algorithm proposed
                                    by'Woody (1967) might  be applied to adjust the
                                    latencies of components  before the PCA is done
                                    (Ruchkin and Glaser, this section;Kutaset al. 1977).

-------
Multivariate Analysis of ERP
                                           565
        RAW WAVEFORMS
         COMPONENT LOADINGS
Fig.  12.  Raw waveforms for component with fixed
amplitude that varies in  latency across observations
and  the component loadings resulting from PCA of
the data.3
ure the degree to which knowledge of one of the two
variables  provides information  about the  other. The
mean crossproducts (Fig.  13a) are formed  by sum-
ming the results of cross multiplication of the values
of x(l) and x(2) for all observations. No transforma-
tion is performed  on the  raw data. The  covariance
(Fig. 13b)  is computed in a similar manner except
that the  mean  value of each  variable is  subtracted
from each observation before forming the crossprod-
ucts.   In ERP terms, this  implies  that  the  grand-
mean ERP is subtracted from each average ERP vec-
tor prior to the computation of covariances between
corresponding ERP points, The effect is to move the
data swarm within the two-dimensional space so that
it  is centered on the origin.  For the correlation ma-
trix (Fig. 13c), the correlation  coefficient is obtained
by standardizing  each variable so that  all variables
have an equal variance prior to the formation of the
crossproducts.  The mean of each variable is sub-
tracted from the  values of the variable, and the re-
sulting differences are divided by the standard devia-
tion of the variable.  After these  two transforma-
tions of the data in Fig.  13a, the data are centered on
the origin and all dimensions have an equal variance.
    Association  matrix  selection:  In  the foregoing
discussion, the principal components were assumed to
have been extracted from a covariance matrix, Covari-
ance, however, is  one of many possible indices of
association that might  be computed between two
variables.  Correspondingly, a PCA might be applied
to different association matrices. Thus the data analyst
must choose  among association matrices before per-
forming the PCA.   Most PCA programs allow a con-
venient choice among  the mean crossproduct, co-
variance, and correlation matrices. The consequences
of this choice are  substantial, as it affects  both the
component loadings and the  component scores that
will emerge from the analysis.

     Each matrix is obtained by applying a different
transformation to  the data prior to the computation
of the association  index.  These transformations are
illustrated in the three  scatter plots in Fig. 13. The
two-point "ERPs" plotted are similar to those shown
in Fig. 8. The  raw ERPs  are plotted  in  Fig.  13a as
points in two dimensional space. The points define a
swarm that has a specific locus in the  space, depend-
ing on the means of the two variables. The extent of
the swarm along each axis depends on  the variance of
the data in each dimension. Association indices meas-
        consisted of a fixed-amplitude half-sine wave
 replicated 20 times at each of 10 different latencies.
 Independent  uniform random  noise (-0.01, +0.01)
 was added  to each  point in order to avoid compu-
 tational difficulties.   The arithmetic precision  of the
 computations is greater than indicated by the  figure.
    The effects of these transformations on the result
of a PCA are quite complex. A proper understanding
of these effects must guide the choice of the associa-
tion matrix to which the PCA is applied. Of principal
importance is the fact  that the substrate in the PCA
analysis is the variance of the analyzed data. The total
variance of the raw data is analyzed  when the mean
crossproducts are  used. Part of this total variance de-
rives from the differences between the means of the
different variables.  This  portion  of  the variance  is
removed in computing the covariance matrix. Another
portion of the  variance is contributed by the differ-
ences between the variances of the individual variables.
This source is removed upon computation of the cor-
relation coefficient. Quite  different variances are
factored, then,  when a PCA is applied to each of the
three matrices.
     That the mean is not removed from the data in a
 crossproducts analysis has at least two effects on the
 results of a PCA.  First, portions of the ERP that
 have a large base-to-peak amplitude will emerge as
 components even  when they are not affected by the
 experimental variables. Second,  in  most cases the
 loadings of the first component extracted in such an
 analysis roughly duplicate the waveform of the grand-
 mean ERP.

     Analysis of the covariance matrix will yield prin-
 cipal components  that correspond  to the  variance
 around the grand-mean ERP. The  extent to which the
 Individual ERPs differ from the  grand mean,  rather

-------
   S66
                           Donchin and Heffley
                    X2
                                       X1
          (B)
                    X2
                                      Xl
         (0
                   X2
                                      X1
Fig.  13.  Scatter plots representing two points along
an ERP.  (a) Computation of mean crossproducts re-
quires no transformation of these raw data,  (b) ERPs
after the  transformation necessary for covariance
computation has been applied.  Note the grand mean
for the ERPs is now (0,0).  (cj Transformations in-
volved in correlation computation have been applied
to the ERPs.   The grand mean has been subtracted,
and  each point has been divided by its standard de-
viation across ERPs.  The scale along the axis is now
in standard deviation units, unlike the raw measure-
ment scales for (a) and (b).
 Thus, time points at which differences are small, and
 possibly unreliable, can receive large loadings due to
 the standardization of values at each time point.
    These considerations must be kept in mind when
interpreting the  component scores  obtained  from
PCAs applied to the different matrices.  Of particular
importance are  the h plications to the interpretation
of the "polarity" of the scores. Only for a cross-pro-
ducts analysis will a  negative factor score  be a sure
indication  of   the  polarity  of the corresponding
component  in  the original  ERP. In  a  covariance
analysis, the score for a given ERP component repre-
sents its amplitude relative to the grand mean.  For
example, a  positive peak lower in amplitude than the
corresponding peak in  the grand mean will have a
negative component score.
    In general, PCA of covariance matrices is most
useful in ERP experiments. The data analyst is inter-
ested only in those ERP components that are affected
by  the experimental manipulations. The  differences
between ERPs relative to the grand mean are of prime
interest.  Furthermore, since the  values at all time
points represent voltage,they all have identical scales,
and therefore there is no need to scale the data by  the
standard deviations. The covariance matrix is therefore
preferable to the correlation matrix.
    Advantages and disadvantages:  PCA provides
 information  about  the  amplitude  variability and
 waveshape of components. The technique also pro-
 vides valuable information when components overlap,
 provided  they are not highly correlated. Given a set
 of measures on the orthogonal principal components,
 the analyst can then employ an analysis of variance to
 assess  the effects  of experimental variables on  the
 independent  components. An important  feature  of
 this method is its adequacy regardless of ERP compo-
 nent waveform. It handles fast and slow components
 equally well.  Another advantage is the existence of a
 well-developed  statistical theory  to  guide the data
 analyst. Finally,  though this  procedure  requires a
 substantial amount of computation,  it is possible  to
 quickly summarize results from complex experiments
 that yield  hundreds of individual ERPs.
 than  the  absolute amplitude of the  different ERPs,
 will determine which components will be extracted.
 Analysis of the correlation matrix will usually yield
 components of a similar nature, except that the load-
 ings will be more uniform across the extent of a com-
 ponent rather than reflect the component waveshape.
     As usual,  there are also disadvantages  to  the
 method. The power and nature of the technique may
 not be intuitively obvious, nor is it immediately clear
 how the conclusions of visual inspection are translated
 into numerical  values by PCA. A more serious dis-
 advantage   is  that   PCA  confounds   latency  and

-------
Multivariate Analysis of ERP
                                                                                                567
 amplitude variability. The results should therefore be
 interpreted with great care whenever latency variability
   substantial  across ERPs.  Finally, PCA  requires
is
significant computing power. While such procedures
can be easily performed at most computer facilities,
the requirements  exceed the power of laboratory
computers.
 Discriminant analysis

    Discriminant analysis is a classification procedure.
 Detailed treatment of its application to ERP data has
 been given  by Donchin and  Herning (1975)  and
 Squires and  Donchin (1976).  It is similar to PCA in
 the sense that it provides yet another rule for develop-
 ing  a linear combination  of  the  observations. The
 difference lies in the fact that in PCA the coefficients
 are obtained through an analysis of the total variance
 in the ERP data matrix, disregarding the independent
 variables that group the ERPs to produce component
 scores.  In discriminant analysis it is assumed that the
 ERPs can be objectively classified into two or more
 groups that correspond  to values of the experimental
 variables. Between-group variance of the ERPs is then
 analyzed directly. Again, a linear combination of each
 ERP vector  in  the data matrix is produced, but this
 time the linear combinations are selected to account
 for  an  increasing proportion of the  between-group
 rather than the  total variance.
     The  discriminant   analysis   strategy   can  be
 described with a simple example. Suppose a number
 of two-point, x(l) and x(2), ERPs are collected under
 two different levels of sunspot activity, high and low.
 The data points are  plotted with different  symbols
 for  high  and low sunspot ERPs (Fig. 14).  Ignoring
 differentiation  between  data points  and analyzing
 total variation, as in PCA, results in line Al accounting
 for maximum total variance. However, with the dis-
 criminant  analysis   procedure,  which  maximizes
 between-groups   variance,  the  variance along A2
 would be of interest.  Each ERP can then be projected
 on this line.  The next step would be to  determine a
 criterion point along the line that runs  through the
 group means.
     Two-group discriminant analysis provides a set of
 coefficients  for  linear  combination  with  the  ERP
 vectors to produce  a single composite variable that
 can  then be evaluated  against  the  discriminant
 criterion. The matrices involved are illustrated in Fig.
 15. The example is based on the ERPs from Fig. 1.
 ERPs  1 and 3 (group A) are discriminated from ERPs
 2 and 4  (group  B).  Stepwise  discriminant analysis
 (Dixon 1975, Jennrich 1977) will use only a subset of
                                                      X2
                                                                  A2
                                                                               XI
                                                    Fig. 14.  A scatter plot of two-point ERPs illustrates
                                                    one of the differences between discriminant analysis
                                                    and principal components analysis. Symbols * and 0
                                                    represent ERPs  from  two  different  groups.  PCA
                                                    would select Al as the principal axis because maximal
                                                    variance in the  total data base occurs along  this
                                                    dimension. Discriminant analysis would select A2 as
                                                    the primary dimension because it maximizes between
                                                    group variance.
                                                     the P time points in the ERP data matrix to construct
                                                     a discriminant function.  It will  select variables (or
                                                     time points) for inclusion such that the first variable
                                                     selected accounts for the  maximal proportion of the
                                                     between-group variance, the second variable accounts
                                                     for the maximal portion of that between-group vari-
                                                     ance  that  remained unaccounted for aftej  the first
                                                     step, and so on. This selection process continues until
                                                     some specified condition  is met. This condition can
                                                     be the  number of variables included, the amount of
                                                     variance accounted  for,  or the  point at which  no
                                                     improvement is observed in some criterion.
                                                         The sources of between-group variance are inde-
                                                     pendent, but in general will not be orthogonal.  The
                                                     discriminant function  is simply a set of coefficients
                                                     for selected time points. The discriminant score is the
                                                     result of the application of the function to any ERP.
                                                     Thus,  the  discriminant score is a sum of several
                                                     weighted amplitudes along the ERP epoch plus a con-
                                                     stant. Again,  the essential similarity of the multi- and
                                                     univariate techniques is evident. Discriminant analysis
                                                     is  a procedure for selecting time  points at which the
                                                     amplitudes  are to  be measured  and for obtaining a
                                                     relative weighting of these amplitudes.  It does  so by
                                                     attempting  to  provide  optimal  separation between
                                                     two  groups of ERPs.  So, it differs from the conven-
                                                     tional  base-to-peak measures  merely in the way in
                                                     which  the amplitudes are  selected for measurement.

-------
568
                          Donchin and Heffley
  |"o o 10 11 11 ie 10 ie 11 i» i» 11

   DO 4 S > • « '« '• « 1» »•
   o o 10 11 it i<
                                    29.A

                                    217

                                    0.0

                                    0.0

                                    0.0

                                    00

                                    it

                                    na

                                    04

                                    0X1

                                    00

                                    0.0
                AUGMENTED  |     [DISCRIMINANT] [DISCRIMINANT]
              [ERP DATA MATRIX]    * [ FUNCTION J"[  SCORE   J

Fig. 15. Discriminant analysis of the data from Fig. 1
is illustrated.4  ERPs 1 and 3 are discriminated from
ERPs 2 and 4.  A  column of ones is added to the
ERP data matrix so that the constant for the dis-
criminant  function  is added  to  the combination of
values at the time points.  The  constant appears at
the bottom  of the discriminant function  vector.
The discriminant analysis was limited to three points.


     Once the discriminant function has been comput-
 ed, it can be employed in several ways. It  can be ap-
 plied to other sets  of ERPs collected under the same
 general experimental paradigm. Besides group classifi-
 cation, the   composite variable  produced  by  the
 function can be used as a measure of ERP magnitude.
 This  is particularly true  when one  of the data sets
 used  to develop the function is characterized  by a
 very small ERP (see Donchin and Herning 1975). The
 discriminant function thus defines a continuum along
 which the resemblance of an  ERP to the average for
 one group or the other can be assessed. Discriminant
 analysis also identifies  the regions of the ERPs that
 best differentiate between  groups.  The time points
 selected identify  these regions and can  serve as a
 guide to visual inspection.

    Example of application: Squires and Donchin
 (1976) investigated the performance of stepwise dis-
 criminant  analysis (SWDA)  using single-trial  ERPs
 elicited by auditory stimuli in an experiment designed
 to  measure the effect  of the a priori probability of
 stimuli on  the magnitude of the P300 component.
 Two groups  were  used to develop  the  discriminant
 function. One contained  ERPs associated with  a rare
 event;  the  other  contained  ERPs  associated with a
 4 The four waveforms were each replicated 25 times
 with independent uniform random noise [-2.0, +2.0]
 added to each point in order to avoid overflow/un-
 derflow computational errors.
 frequent  event.  The  SWDA  program  was  able  to
 classify correctly as "rare" or "frequent" 89% of the
 single  trials  (on the  average) if information  from
 multiple electrode  sites was used. The  data from alt
 subjects were  then pooled to generate a "subject-
 independent"  discriminant  function  and  used  to
 classify data  obtained from seven new subjects. These
 data thus served as a training set for development of
 the  discriminant function. The subject-independent
 function  proved  sufficiently  general  to correctly
 classify 81% of the single trials from new subjects.

     Examination of misclassified trials can be instruc-
 tive. In Fig. 16 averages for incorrectly classified trials
 are shown with the averages  for correctly classified
 trials for  the two  groups upon which the  function
 was based. Trials on which the rare event (loud signal)
 was presented, but classified  as  frequent,  actually
 evoked, on the average, an ERP characteristic of the
 response to frequent stimuli (soft signal). This result
 suggested that the ERP elicited by the stimulus does not
 depend entirely on the a priori probability of the stim-
 ulus. Subsequent work has shown that expectancies
 that depend  on the precise structure  of the stimulus
 series must be considered (Squires et al.  1976).

    Advantages and disadvantages: SWDA objectively
 assesses  ERP  differences along  any  independent
 dimension, provided data can be  obtained at points
                                                                LOUD SIGNAL
                                                                       SWDA CLASSIFICATION
                                                                           LOUD	
                                                                           SOFT	
                                                                 SOFT SIGNAL
                                                      Fig.  16.  A verage ERP waveforms sorted according to
                                                      stimulus presentation and discriminant analysis classi-
                                                      fication. Note the average ERP for misclassified loud
                                                      (rare) stimuli exhibited no P300 and, therefore, re-
                                                      sembled responses evoked by soft (frequent) stimuli.
representing  the  extremes  of the dimension.  The
power of the technique to extract differences makes
it  applicable  to  analysis  of single-trial  ERP data
(Donchin 1969, Donchin  and Herning 1975, Squires
and  Donchin  1976).  Finally, there  is  a statistical
theory that guides the interpretation of the statistics
derived by  the technique.

-------
Multivariate Analysis of ERP
                                             569
    There  are  several  disadvantages to SWDA. The
most obvious is the need to have an independent basis
for grouping ERPs so that a discriminant function can
be developed. If the coefficients are computed on a
training set for later use on another ERP data base,
then the analyst  must be sure the ERPs for the two
groups within the training set  differ significantly. If
the Investigator is using  discriminant analysis to test
for differences  between  the groups upon which the
function was built, then caution must be exercised,
because the power of the technique may produce mis-
leading results.  It is possible that the analysis may dif-
ferentiate between groups on  the  basis of random
variation in the sample rather  than systematic differ-
ences that might be present in the population (Lachin
and Schachter  1974). We can be confident that the
function yields a meaningful differentiation between
groups  when either the original sample is large or
satisfactory  classification  is   achieved when   the
function is applied to data not used in the training set.
     The need for cross validation of the discriminant
 function  must  be  emphasized.  The  discriminant
 analysis programs always yield a function. Even if the
 two groups  in the training set are not different, the
 program will develop a rule for distinguishing between
 the  groups, capitalizing  on  random  distinctions
 between the groups. In general, the weakness of such
 a function will  be apparent,  since it will provide a
 very weak classification of the training set. However,
 it should never be assumed that a strong classification
 of the training set is sufficient  to establish the quality
 of the discriminant function.  The user must demon-
 strate the validity of the discriminant function. The
 most  direct  way for doing so is to apply the  function
 to an altogether new set of data. This test  set must
 consist of data that belong to  one or the other of the
 classified groups. The  function  should classify the test
 set at least as well as it classifed the training set. If the
 test  set  is  classified weakly  (in  the sense that a
 random classification would have  done as well), the
 function  is not providing a valid discrimination.  This
 test requires, of course, that the analyst have sufficient
 data  to generate a test set. On occasion this is not
 possible. In some  cases  there are barely sufficient
 data in the  training set  to develop the discriminant
 function. At least two procedures are available to test
 such limited data.
      The jackknife (or leave-one-out) test uses the
  training set itself to generate a test set. It proceeds to
  compute  a set of discriminant functions, each with
  one case removed from consideration. It then applies
  the discriminant  function  to the held-out case. A
  table is obtained in which  each case is classified on
  the basis of a discriminant function computed without
  use of that particular case. This procedure is a fine
check but should not be  overrated. The effect that
any  given case has on  the discriminant function is
relatively  weak, so  the repeated  computations  of
discriminant  functions using the same  data cannot
but produce results that are very similar to those ob-
tained by the original computation. Only if the within-
group variance is very large (relative to the between-
group variance) would the jackknife yield results that
are very different  from  the  original computation.
Since the procedure is embedded  in the  packaged
SWDA programs, there is no great cost in its use, and
it may well give useful information, but it should be
interpreted with great caution.
     As an alternative to the jackknife test, a random-
ization test may  be applied  to the training set  to
assess the  original  discriminant  function.  In this
procedure (for an example see Donchin and  Herning
 1975), the  training  set is used  to generate two new
groups by randomly assigning each case to one of the
two groups, disregarding the original group member-
ship. A discriminant function is  then obtained for the
randomly shuffled data. This procedure  is repeated
several times, and  a  distribution  of discriminant
functions is obtained.  This distribution provides an
estimate  of the discriminant function that  can  be
expected from  the data  set under purely  random
classification. If the original function falls within this
range, it is probably  invalid.
     As  with  principal  components  analysis,  the
 relationship of discriminant analysis to visual inspec-
 tion  may not  be clear,  although  the investigator
 should  be able  to see the differences between wave-
 forms in  tracings. An additional disadvantage is the
 need to transfer the  raw data to a  reasonably large
 system for analysis. Once a discriminant function has
 been computed for an adequate training set, however,
 the computation of discriminant scores involves only
 a few  multiplications and a  summation which can
 easily be done on a small laboratory computer system
 or even a portable calculator.
 Baseline in ERP analysis

     The treatment of baseline in ERP research  has
 always been a vexing problem. All amplitudes must
 be measured relative to some value. Peaks and areas
 are always determined with respect  to a reference
 point. In PC A, subtraction of a  baseline value from
 each  ERP is necessary to reduce error variance. In
 fact,  if  the ERPs  are not adjusted  by preliminary
 baseline  subtraction, the  first extracted  component
 will usually reflect the differences between the level

-------
 570
                           Donchin and Heffley
 of the individual waveforms. In discriminant analysis,
 baseline  subtraction  helps  minimize  within-group
 variance.  Traditionally, the baseline has been defined
 as  the average value over a portion of the recording
 epoch just  preceding  the stimulus that evoked the
 ERP.  However,  the  assumption  that  this average
 represents a true "zero  point" is sometimes untenable,
 Often there is too  much variability  in prestimulus
 baselines. Another problem  is the presence of event-
 related activity in these intervals. The CNV and RP
 are common confounding factors. The  problem  of the
 fluctuating  baseline  is sometimes circumvented by
 measuring  peak-to-peak  rather  than  base-to-peak
 amplitudes.  However, accurate interpretation of peak-
 to-peak measurements  is dependent upon knowledge
 of  the correlation  between components  and  the
 degree to  which components overlap.


    The first problem to be faced, then, is determina-
 tion of a reliable measure of the baseline. One possible
 solution is use of trimmed averages. The midmean has
 been used by Donchin  et al. (1973). The midmean is
 obtained by eliminating all values in  the upper and
 lower quartiles. The arithmetic mean of the remaining
 values  is then computed. This measure is unaffected
 by  extreme  deviations  in the data.  Similar statistics
 are described by Tukey (1977).


    The  problem of one component  overlapping or
 correlating with another must  also  be considered.
 PC A seems to offer the best approach to this problem.
 A  plot of  component  loadings should  make  the
 degree of ERP component overlap clear. The loadings
 will also help reveal highly  correlated ERP compo-
 nents.  ERP  components  that covary strongly will
 appear within a single  PCA loading vector. The ERP
 data analyst might also examine discriminant analysis
 results to  determine whether a baseline is free of ERP
activity.   If  points  within  the  baseline  interval
contribute substantially to the discriminant function,
 then one  must carefully check  for the presence of
 ERP components in that portion of the epoch.

    The  problems associated with baseline determi-
 nation in ERP data analysis should not  be ignored. The
 consequences  include  missing significant differences
 in components because of variability  in the baseline
 and incorrectly identifying  the  varying component
 because components  overlap  or are highly correlated.


 Practical aspects of multivariate analysis

    ERP  investigators  have been slow  to adopt  multi-
 variate techniques, despite their advantages. To some
 extent, this reluctance may be due  to a tradition
 determined  by  the  nature  of the instrumentation
 prevalent in this field.  Special-purpose averagers with
 an x-y plotter as the major output device naturally
 lead to an emphasis on visual inspection. Even though
 more and more general-purpose computers are being
 used  in  ERP laboratories, there  is a  reluctance to
 venture beyond the confines of the laboratory. This
 section is meant to serve as a guide to transfer of data
 from the laboratory computer to a larger facility for
 analysis.

    The  large central computer facility offers several
 advantages over minicomputer systems for multivariate
 data analysis. One  significant  advantage is the avail-
 ability of standard statistical packages such as BIOMED
 (Dixon 1975), SPSS (Nie  et al. 1975), and SOUPAC
 (Dickman 1974). These packages tend to be adequately
 documented and relatively error free.  Although such
 procedures as PCA and discriminant analysis could be
 performed on a minicomputer, their implementation
 is certainly not easy. Advanced  programming  skills
 are required  to adapt programs  that  require  large
 amounts  of  computer memory  to  the   restricted
 program  space  of  a minicomputer.  Further, some
 knowledge of numerical analysis is required to ensure
 reasonably  accurate computations. The result of all
 this effort  may be  a procedure  that  is impractical
 because  it takes many hours  to execute,  depending
 upon  the  particular  minicomputer.  In  general,
 resources are  probably better  spent in  developing an
 efficient  procedure  for transferring data to a larger
 computer.

    We have been  able to process ERP data at a wide
 variety of computer facilities.  In each case, all that is
 required  is  that the  data be  written on a standard
 "industry-compatible" digital  magnetic tape. A typi-
 cal specification is  for a nine-track tape drive with a
 recording density of 800 or 1600 bits per  inch. Such
 drives are readily available for all minicomputers. The
 cost is relatively high, but as we have noted elsewhere
 (Donchin and Heffley 1975), the  lack of such a tape
 drive  cripples the ability of the investigator to fully
 realize the advantages of computerizing a laboratory.
 The use  of direct computer-to-computer  commu-
 nication  is  an  alternative. It is possible to attach a
 laboratory  computer to a  central  computing facility,
 an arrangement most often accomplished via tele-
 phone. At present, most of these lines permit relative-
 ly slow communication (300 baud) and the transfer
 of large data bases is, therefore, some what impractical.

    Once a standard tape drive  is  available,  ERP
single trials  and averages  along with  other psycho-
physiological or behavioral data  can  be written on
digital  tapes. If the tape drive is well  aligned,  then
any data written on the tapes  are, in principle, read-
able by any corresponding tape  drive  regardless of
manufacturer. While the physical bit pattern recorded
on a tape can be read with ease, different computers
maintain  different  protocols  for organizing data on

-------
Multivariate Analysis of ERP
                                                                                                 571
 tapes. The differences lie in such 'factors as the size
 and structure of data  blocks and the codes used to
 represent numbers and characters. These factors are
 termed the logical format,of the tape and are fairly
 independent of the physical recording method;

     A comprehensive discussion of the logical formats
 used for digital magnetic tape is beyond the scope of
 this paper. However, some general information will be
 offered  to help guide 'investigators  who  are  not
 familiar with the details of data transfer between
 computer systems. First, data may be stored in either
 symbolic or binary formats. Symbolic format repre-
 sents each data point as a series of characters as if the
 data were output on a line printer. Binary represen-
 tation  is direct  output of data  in the internal code
 of the computer. There is a tradeoff between  the
 two representations. Data in symbolic format are more
 easily  transferred from one  computer  to another
 because  they  are  usually' written  using standard
 coding schemes (ASCII or EBCDIC) that almost.all
 systems accept. Binary format usually requires special
 processing to handle, differences in such characteristics
 as  computer word size and number  representation.
 The binary format is, however, more efficient. Binary
 representation  requires less space on  a  tape and is
 faster because the conversion from internal code to
 characters is not needed. In general, it is practical to
 store the average ERPs from an experiment in sym-
 bolic  format, but because  of the volume of data,
 single trials are best stored in binary format.
    Another factor is the format in which numbers
are grouped into records, and sometimes blocks; on
the tape. A problem arises  because  the  command
most   readily  available  to  the  programmer, the
FORTRAN  "WRITE" statement, causes data  to be
output in a logical record- format that is specific to
the particular minicomputer and operating system. In
addition, data output in this manner typically con-
sume a great deal  of tape, because a data array is
usually broken apart and output in  chunks of about
128 bytes with relatively large gaps 'between  these
records. Unfortunately, most  minicomputer software
systems  do not  offer the FORTRAN  programmer
commands to output an entire array (EEC channels
by time  points) as one  record. Such commands are
usually available within  the operating system and so
can be interfaced  to the FORTRAN program via an
assembly  language subroutine call (Donchin and
Heffley 1975), Using this strategy, data will be written
on tape faster and more efficiently than with standard
FORTRAN commands.

     The user must then face the problems associated
 with  reading data into the larger  computer  at  the
central computer facility. Resolving these problems
will usually require three steps.  First, data are read, at
best one full array at a time,  into the computer's
memory. Second, for  binary representation, conver-
sion from the minicomputer to large computer nu-
meric format is done. If the data are stored on tape as
integers, rather  than as real numbers, then this step
involves'simply extending the sign bits, assuming two's-
complement arithmetic. Third, data can then be out-
put to a storage device  on the large computer from
which they can be read by any of the analysis pro-
grams available' on the system.

    Several major statistical packages are available for
multivariate, analysis. It is not worthwhile in this con-
text to recommend one package over trie others. Each
has its advantages and disadvantages. In most comput-
ing centers, the statistical consulting staff tends to
favor one or another  package for reasons that have
more to do with the history and training of the staff
than with rational considerations. Users are wise to
use whichever package receives maximal support at
their  installations. (It  is possible, of course, to write
a PCA,  or any other multivariate program, by utiliz-
ing any  of the readily  available mathematical, or
matrix manipulation,  subroutine  packages  such as
SSP or IMSL.) An unfortunate characteristic of most
packages,  with  the exception  of SOUPAC,  is that
none includes a convenient routine for performing a
repeated-measures analysis of variance (ANOVA). It
so happens that most ERP experiments require such a
design, since the comparisons are usually within sub-
jects  and  across experimental  conditions; I.e., each
subject  is  tested repeatedly  in  all  experimental
conditions. The error  terms in such an ANOVA are
quite different  from those used in standard  designs
(see Keppel 1973). One ought to avoid using between-
subject  variance as the  error term. It is important,
therefore,  to  have  a  repeated-measures ANOVA
program that can analyze the factor scores yielded by
the  PCA.  Of  the  three  large  statistical  packages,
BIOMED, SPSS and SOUPAC, only SOUPAC offers a
repeated-measures  ANOVA  program.  In our own
laboratory, the  package  called  ALICE (Walker et al.
1976) has proven useful for this purpose?

Summary

    We have reviewed four different techniques for
the analysis of ERP data (peak and area measurement
and  principal component and discriminant analysis)
and indicated that all four involve the formation of
linear combinations of  ERP amplitudes at selected
time points. The techniques differ with  regard to the
manner  in which the time points are selected and to
the coefficients that are  used to scale the voltages as
 5 ALICE can be obtained from its developers (ALICE
 Associates, 29  Wellesley Avenue, Natjck, Massachu-
 setts 01760). Versions for several computer systems
 are available.

-------
 572
                          Donchln and Heffley
they enter into the amplitude measure. The strategy
adopted  in  each technique  makes It  more  or  less
appropriate for any given ERP problem.

    Conventional procedures  determine the selection
of time points largely through visual inspection of
records. Components are Identified, more often than
not, by inspection, and then peak amplitudes or areas
are  measured.  Multivarlate  procedures  derive  the
measures  through an analysis of either  total  or
between-group variance  In  the data acquired In  any
given  experiment. Peak and  area measurements are
relatively easy to compute and have direct Intuitive
appeal, but they prove difficult to  use when the data
bases  are  large,  the ERPs complex, and  the experi-
mental design multifactored.  Multivarlate techniques
provide a useful solution in such cases.

    The  ERP data analyst's task typically consists of
three  steps. First, ERP components are Identified. In
this context,  we have  discussed the value  of principal
component analysis combined with visual Inspection.
Second, the magnitude  of  components Is measured.
Computations of ERP peak and area measures, PC A
component  scores,  and  discriminant  scores are all
valuable  approaches. Finally,  differences  between
ERPs  are  assessed.  This  step  usually  Involves  a
repeated-measures analysis of variance for peak, area,
or principal component scores. Differences are direct-
ly assessed in discriminant analysis. As was noted,
analysis depends upon the determination of a reason-
able baseline.
    We wish to emphasize that our Intent was not to
provide a cookbook for data analysis; any such cook-
book would be worthless, There  It an unfortunate
tendency,  derived  from the  strong  Influence  of
hypothesis testing,  to seek procedures that, through
some manipulation  of the data, arrive ultimately at a
magic number known as the level of significance. We
prefer to view data analysis as a heuristic that provides
the user with a set  of tools that allow  rich and inter-
active manipulation of data. The  same data  can  be
viewed  from many vantage points. The waveforms
need  to be examined and measured In many different
ways until  a coherent and Intellectually satisfying
picture   emerges.   In   this  endeavor,  multlvarlate
techniques,  at least In our experience, have proven a
valuable aid,

-------
BEFORE AVERAGING:  PREPROCESSING  SLOW
POTENTIAL DATA WITH A WIENER FILTER1


P. NAITOH AND S. SUNDERMAN
Naval Research Center, San Diego, CA, U.S.A.
      Two models for the generation  of event-re-
lated potentials (ERP) may be distinguished: (1) an
event-locked model, and (2) a latency-variable model.
The event-locked model assumes that each stimulus
event initiates slow potentials after a certain latency,
that this latency remains constant from one trial to
another, and  that, within  each  ERP, rank-orders
of amplitudes and latencies for negative  and positive
peaks  remain unchanged. The latency-variable model
assumes  that the  initial latency for an ERP varies
from one trial to another, but that amplitude and re-
lative latencies for these waves, within each ERP, re-
main unaltered from one trial to another.

      If ERPs  were generated  according  to the
latency-variable model, averaging over trials would be
improper for estimating a  true waveform because
averaging techniques require a fixed latency for gen-
eration of ERPs.  A variety of methods have  been
used to capture  ERPs  under the  latency-variable
model (e.g., Childers and Pao 1972, Pfustscheller and
Cooper  1975, Weinberg  and Cooper 1972,  Woody
1967, Sayers et al. 1974). The results seem to confirm
the validity of the latency-variable model by showing
considerable variability in  the initiation time of ERPs
from one trial to another.

      Most researchers,however, prefer the averaging
method  (thus adopting, willingly  or unwillingly, the
event-locked model) because it offers conceptual as
well  as  computational  simplicity.  A  persistent
problem in averaging, however, is that a large number
of trials must be averaged to obtain true ERP  wave-
forms that are relatively free from the noise of spon-
taneous electroencephalographic  (EEC) activity. A
linear, non-time-varying filter derived from the the-
orem  of Wiener and Khintchin (Wiener 1964, Walter

'This study was supported, in part, by the Department
of the Navy, Bureau of Medicine and Surgery, under
Work Unit MROOO.01.01-6006. The views presented
in this paper are those of the authors. No endorse-
ment by the Department of the Navy has been given
or should be Inferred.
1969) has been said to provide a significant improve-
ment  in the averaging method when  this  "Wiener
filter"  is used to preprocess individual records before
averaging.


       The  purpose of this paper is to describe and
evaluate the Wiener  filter as applied  to single-trial
data. The filter was first used on an averaged evoked
potential (EP) based  on artificially generated "toy"
data (Walter 1969).  Later,   Nogawa et  al. (1973a,
1973b, 1973c) published a FORTRAN  algorithm for
the filter and applied it to visual (V) EPs from human
subjects. They noted that Wiener  filters caused a
significant   loss  of fast  frequency  components in
VEPs. This  appreciable loss of faster frequency com-
ponents in  Wiener-filtered VEPs created some con-
cern among researchers. Some felt that the filter was
too "heavy" because  it reduced, in additon  to noise,
some peaks in VEPs that were often of interest. To
overcome these problems, a modification of the fil-
ter was suggested by Doyle  (1975) and by Walter
(1975), as  well  as  Nogawa  (personal communica-
tions). A recent  paper by Ungan and Basar (1976)
showed, however, that loss of faster frequency com-
ponents in Wiener-filtered  EPs remains a problem.
       Despite some  persistent problems  with  the
 Wiener  filter, it could still be very useful in signal
 processing (see, for example, Rosen et al. 1975), be-
 cause it has some desirable attributes. First, the filter
 could be applied to remove noise from single records
 without  introducing phase shifts. Second, effective-
 m.w  of the  filter  to  enhance signal/noise ratios
 could  be elevated  by comparing the filtered single
 record with a corresponding raw record.
 Description of the Wiener filter

       The Wiener filter is a digital filter in which the
 gain or filter weight for each frequency band varies
 from  1 to 0, depending on signal/noise ratio. If a

-------
 574
                         Naitoh and Sunderman
 frequency band contained only "response" or "sig-
 nal," the filter weight would be 1, passing all of the
 activities in that frequency band; the  filter weight
 of 0 would be assigned to frequency bands that con-
 tained nothing  but "noise."   In  the Wiener filter
 analysis, the response or signal would be those  fre-
 quency components phase-locked to  the time of the
 stimulus event over all of the available trials for aver-
 aging.


        Although  the Wiener filter for  the latency-
 variable model could be  written  (Walter, personal
 communication), the filter as discussed in  this paper
 is able to  handle only those cases where response is
 assumed  to  have non-time-varying waveform from
 one trial to another with a response that starts with a
 fixed latency from the time of stimulus event (event-
 locked model). First, the ERP is written as:
                                               (1)
 where Xj(t) is the  observed voltage at time t of the
 ith trial, r^(t) is the "response" voltage at time t of
 the ith trial, and n^t) is the "noise" at time t in the
 ith trial. This model states that an observed voltage in
 the EEC record after a stimulus event is composed of
 response and noise.
       For each trial, the measurement of voltage of
the EEC record is repeated many times at regular in-
tervals to produce a time series of Xj(t). For instance,
examples in  this paper used EEC  records that had
256 data  points  measured every  15.625 msec (a
sampling rate of 64 samples/sec).


       The second step in Wiener filtering is to do a
discrete Fourier analysis (or a harmonic analysis) of
Xj(t).  Mathematical  details of the analysis can be
found  in discussions of  time  series  analysis  (e.g.,
Anderson 1971, Bloomfield  1976, Jenkins and Watts
1969,  Walter 1969). In this study,  the analysis was
performed up  to  the  highest frequency admissible
with the sampling rate  (Nyquist frequency) using the
Fast Fourier  Transform (FFT) algorithm of Nogawa
et al. (1973c).  The analysis yielded sine  and cosine
Fourier coefficients for each frequency, totaling 129
cosine  coefficients and  127  sine coefficients for the
256-data-point  record.  These coefficients were kept
in computer core. Storage of these coefficients as well
as the  average overall EEC records meant that  phase
information  related to stimulus onset was retained
for each frequency  component. Eight EEC records
were evaluated; thus, after  FFT, there were 256 x
8  = 2048 Fourier coefficients  in  computer  core.
Computation  of these  coefficients was followed by
calculating spectral estimates centering at 0.25,0.50,
 0.75,  and so on, to 32  c/sec (Nyquist frequency).
 The data yielded 128  "line" spectral estimates per
 EEC record.

 _  The third step is to compute an average spectra,
 S(X) [SP(N) in Nogawa's notation]. There were eight
 spectra, one spectrum  for each EEC record (trial),
 each spectrum  ranging from  0.25 to 32 c/sec. To
 obtain  S(X), the  eight  spectra  were averaged at
 each frequency over the 128 line spectrum estimates.

    The fourth  step  is to compute  a spectrum of
 averaged  EEC records  or averaged  ERPs, S(7J), or
 PAVA in Nogawa's notation.

    In  terms of the model represented by equation
 (1), the  mathematical  manipulations for  averaging
 spectra can be expressed as:
                                                             S(X)-S(r) + S(n)
                                              (2)
 where_S(r) is the spectral intensity of the response,
 and S(n) is  the averaged  spectral intensity of the
 noise. It should be observed that S(X) is not exactly
 equal to the sum  of S(r) and "S(n) because  S(X)
 under this model  has a third term that approaches
 zero only when a sufficient number of trials are aver-
 aged.

       The spectrum  of the averaged ERPs is quite
different from  S(X)  because the process of averaging
N number of EEC .records reduces the variance of
noise  in  these  records by  1/N (see Regan 1972,
Walter 1975). Thus,  for  the model the  spectrum of
the averaged ERPs is:
                   - S(r) +
                              N
      Equations 2 and 3 provide an estimate of two
unknowns. S(r)  and  S(n),  the  spectra of response
and  noise. Further algebraic  manipulations  are  re-
quired to solve  this simultaneous equation for S(r):


           s(r)g (NxS(X)) -SQQ        (4)

                       (N-l)
and for S (n):
               S(n) =  S (X) - S (r)           (5)
      Then, a   ratio  of  the estimated  signal-to-
noise spectra may be computed for the time series
X as follows:
               S(r)          S(r)             (6)

            S(r) + S(n) ~   S"(X)

-------
 Preprocessing with Wiener Filter
                                                    575
where H is the gain factor for the Wiener filter of the
individual EEC record or trial. In the examples used,
we computed a  ratio for each line spectrum. This H
would  be 1 if the record contained  nothing but re-
sponse, or 0 if S (r) were 0.

     The H may now be used to "correct" Fourier
coefficients,  C.  Thus, a new set   of Fourier  co-
efficients, Y, would be defined as:
                   Y =  HxC
(7)
For instance, cosine and sine coefficients at 2.5 c/sec
were 2.625 and 0.833 before  correction, and H for
this frequency band was estimated to be 0.718. Then
Y would be 1.885 (= 2.625 x 0.718) for the corrected
cosine  value  and 0.598 for the  corrected  sine co-
efficient.

   The fifth  step is to regenerate the Wiener-filtered
single  KEG record (or trial)  using these corrcctctl
Fourier  coefficients.  This  is  done  by  obtaining
the  Fourier  inverse  (again, details arc available in
the previously cited references). The Fourier inverse
can be performed so  that either the response or the
noise can be regenerated in the time domain.

     Doyle (1975) has described the difference bet-
ween Wiener filtering of individual EEC records, (sec
equation 6) and the averaged ERP.  For averaged (A)
ERPs:
    H(Ave) =
                     S(r)
               S(r) +     S(n)
                       N
(8)
Application of Wiener filter to ERP data

       In the present study, the Wiener filter  was ap-
plied to vertex  EEC record (referenced  to linked
mastoids) supplied by H. Weinberg. The EEC was re-
corded  with a  time constant of 5 sec and with high-
frequency amplifier cutoff set at 30 H/. Experimental
conditions are described by Curry et al. (this volume).
We utilised data from one subject under conditions of
"standard" and "word-speak."

       One set of H  was computed for the standard
condition, and  another set  for the word-speak condi-
tion. In cases  where N x S(X)  (where N_was  the
number of EEC  records) was smaller than S(X) and
hence (S(r)  became  negative  (see  numerator of
Equation  4),   we arbitrarily set the  corresponsing
H for that frequency to bo zero.

       Fig.  1   shows the result of Wiener filtering
individual  EEC records obtained under  the word-
speak condition. Fig. 2 shows the results of analyzing
the calibration  and word-speak conditions.
                                     4      /
                                          V
                 J	I
                  SI    S2
                                        SI   S2
         Fig. 1. Wiener-filtered individual records in compari-
         son with  raw EEC records for the word-speak condi-
         tion The left column shows eight individual EEC re-
         cords  before  filtering. The right column  shows the
         same EEC records after Wiener filtering.  Wiener filter-
         ing seems  to have  removed  most  high-frequency
         noise,  still  retaining, however,   some features  of
         evoked poteneials to SI. Wiener filtering  appears to
         aid in judging strength of dc response  in each EEG
         record; records 3 and 6 exhibited much  smaller dc
         shift response than other single EEG response records.
         Note  that  EEG amplitudes for the Wiener-filtered
         records are amplified twice more  than  the raw EEG
         records (see calibration markers.)

-------
 576
                                                     Naitoh and Sunderman
                                             JTO/zV
 B
                                                              B
             S1
S2
S1
S2
h'lg.  2.  Analysis  of the  calibration condition (left column) and of the word-speak condition (right column).
Graph A of the left column was obtained by superimposing all eight KEG records, and the average was plotted im-
mediately below.  Graph B was a super-imposition of eight Wiener-filtered l\KG records, and below it is the average
of these eight Wiener-filtered /;'/;'(/ records. Comparisons of the two averaged responses show that Wiener filtering
tended  to attenuate the  evoked potential to SI and also resolution associated with S2. Graph A of the right
column was a result of superimposing all eight /:'/;'(/ records under the condition of word-speak, and its average
is shown just below. Graph li of the right column is a supcrimposition of individually Wiener-filtered MG records
(see  1-lg. 1),  and  their average is shown  lust below. In this case, Wiener filtering did not result in loss of the
evoked  potential  component to  Si, indicating  that this evoked potential component occurred consistently in
almost all l''h'G records. Note that raw KKGs were plot ted with one-half the amplification used for  all other plots
(see calibration marks).

-------
Preprocessing with Wiener Filter
                                           577
           Table 1. Effect of Smoothing over Frequencies on Wiener Filter Gain H

c/sec
0.25
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
9.00
10.00
11.00
Sum
%
% Loss
Calibration condition
S(r) ''b
3142
-
124
203
27
19.
53
26
5
-
15
9
3623
28.8
23.6
S(n)b
1764
-
986
821
518
670
440
263
247
-
200
83
5992
47.6

S(Total)b
4906
2738
1110
1024
545
689
493
289
252
225
215
92
12578
100,0

H
0.840
0.000*
0.112
0.198
0.050
0.027
0.107
0.089
0.017
0.000*
0.068
0.098



Word-speak condition
S(r)"
1996
528
763
145
178
17
-
17
28
17
9
12
3710
34.7
4.4
S(n)b
937
2269
623
869
533
626
-
241
123
106
106
75
6508
60.9

S(Total)b
2933
2797
1386
1014
711
643
474
258
151
123
115
87
10692
100.0

H
0.681
0.189
0.551
0.143
0.250
0.026
0.000*
0.064
0.183
0.138
0.080
0.139



   1 Dash Indicates negative value of S(r); for such cases, H was set to be zero and marked by '.
   b Spectrum Intensities S(r), S(n), and S(total) were expressed In terms of squared microvolts.
     A check of the calibration (in microvolts) along
 the  Wiener I'iltcrcd  REG records in Fig.  1  revealed
 that the filtering resulted in an overall reduction of
 EEC amplitude, especially  of high-frequency EEC.
 An  examination of computer  printouts of  two sets
 of H  revealed  many cases  of  negative S  (r) due to
 the  fuel that N x S(X)  was smaller than  S(X) espe-
 cially  In spectral activities faster than 10 H/,, Walter
 (1975) and Nogawa (personal communication) sug-
 gested that awkward cases of negative  S (r)  might be
 avoided  by (1) using a larger number of EEC records
 (hence stabili/.ing the value  of N x  S(X) in Equation
 4),  and (2) smoothing  across frequencies  to  make
 the variability  of the spectral estimate smaller and
 more stable. The results  are listed in Table 1.
        Except  at 0.25 c/sec, we smoothed  five line
 spectra  to estimate the spectrum of response and
 noise  factors. Smoothing over frequencies produced
 more  satisfactory  H values than those used in pre-
 paring Fig. 1 and 2. Table 1 suggested also that 30%
 of spectral activities in  the average of spectra could
 represent  the  response  under the  calibration condi-
 tion, and  35 percent under  the word-speak condition.

 Discussion

        Careful  examination of H indicated that we
 would require  at  least  ten EEC  records for satis-
factory  Wiener  filtering.  This means that Wiener
filtering would  not substantially Improve  the chance
of obtaining valid averaged ERP under experimental
circumstances where a large number of EEC records
(trials) could not be  obtained. This  would suggest
that averaging methods are about as efficient as the
Wiener  filter,  and  the  slight improvement due  to
Wiener  filtering may  not justify  the  computational
cost and complexity (Ungan and Basar 1976).

       However, there appear to be occasions when
use of the Wiener filter  might nevertheless be desir-
able.  One such occasion would be when the ERP is
superimposed  on quite strong spontaneous EEC  ac-
tivity, such  as  alpha. Nogawa et al. (1973b, 1973c)
showed that  spontaneous background EEC signals
could be effectively removed from  averaged ERPs
with  a  Wiener niter.  Wiener  filters could,  therefore,
be applied  to  remove some  distortions  of averaged
ERP  that  are due to differences in background EEGs
seen  in  different sleep stages.

        Another important application of the Wiener
 filter could be  to separate response from noise.  For
 instance,  a comparison of the noise spectrum during a
 prestimulus  baseline with   that during stimulation
 would indicate  if the stimulus event altered the noise,
 i.e., the spontaneous EEC.

-------
 578
                          Naitoh and Sunderman
        Furthermore, the Wiener filter  might also be
 useful when spontaneous EEGs of one cerebral hem-
 isphere differ considerably from those of the other
 hemisphere. The Wiener filter could substantially re-
 duce the risk of observing erroneously an altered ERP
 of one  cerebral hemisphere because of contamination
 by unusual background EEC.

        It  should be mentioned here that the Wiener
 filter should not be regarded as an esoteric mathe-
 matical method  of analysis. Use of spectrum analysis
 for removing noise in  ERP records (e.g.,60-Hz noise)
 is a very familiar technique. A DEC PDF 12 computer
 program set of ANECDOTE, a DECUS program pre-
 pared by Cooper et al. (1973), has a well-known data
 manipulation program that includes Fourier analysis.
 In addition, it allows the deletion of certain harmonic
 components, and resynthesis of remaining harmonic
 components, allowing a return to the time domain.
 This process of Fourier transform, followed by man-
 ipulations of Fourier  coefficients,  followed by the
 Fourier inverse is similar to  the Wiener  filter tech-
 nique.

   A  major difference between  the  Wiener  filter
 algorithm  in this paper and  those found elsewhere is
 the ready  accessibility of all Fourier coefficients for
 further  manipulation. Coefficients are stored in com-
 puter core. Thus, we  can combine other spectrum
 analysis approaches with Wiener  filtering. For ex-
 ample, Sayers et al. (1974) computed the harmonic
 of an auditory (A) EP to a strong stimulus, retaining
 the phases of these harmonics. They then  obtained
 another AEP, but  in response to a  much weaker
 stimulus, resulting in a poorly defined AEP. They did
 the harmonic analysis on this poorly defined  AEP,
but  replaced phases  of the  harmonics  by those
 obtained from  the AEP  to  the  strong stimulus,
 the welt-defined AEP, and then computed the Fourier
 inverse.  They noted  that  this "phase-forcing"  re-
sulted in an appreciable improvement in determining
 the AEP to weak stimulation. An obvious extension
 of this  technique would be  to employ  this  phase-
 forcing  on the Wiener filtering of individual EEC re-
 cords to compensate for latency jitter  from one re-
cord to  another.
       As  clearly indicated  by  Ungan  and Basar
(1976), one unresolved problem of the Wiener filter is
related to the length of the time series included in the
analysis. Is 0.2 sec duration after the stimulus event
needed for Wiener filtering, or 0.5 sec, or even 4 sec?
This question appeared to be resolved by the trial and
error method. In fact, it appears that critical infor-
mation  about  the duration of a signal or response  in
EEC records can be obtained by carefully examining
the size of H and S(r) (Nogawa, personal communica-
tion).  Suppose a signal persisted for 0.5 sec. H and
S(r)  would  reach maximal  value using  an analysis
epoch of about 0.5  sec, and then  diminish as the
length of the epoch increased. Thus, it appears pos-
sible  to estimate signal  duration by systematically
incrementing the analysis epoch until maximal values
of H and S (r) are obtained.
Summary

       Wiener  filtering was  described as it applies  to
nonaveraged EEC signals. A FORTRAN algorithm
developed by Nogawa and his colleagues  was modi-
fied and applied to a sample set of data. Wiener filter-
ing of individual  EEC records resulted  in  clear re-
sponse waveforms, but such improvements  were ac-
companied by  significant loss  in  faster  frequency
components when compared to  the averaged ERP.
The loss in faster frequency components resulted  in
the loss  of some significant  peaks  in the  averaged
ERP, which were easily seen with the usual unfiltered
averaging technique.

       It was suggested that the Wiener filter should
not  be  used as  a method  of replacing the usual
averaging techniques,  but rather  as a preprocessing
method  that  has some supplementary advantage  to
averaging. Application of the Wiener filter combined
with  other spectral analysis  methods was  also dis-
cussed as an additional method for increasing useful-
ness of the Wiener filter.

Acknowledgment

       The authors thank  Laverne  C.  Johnson for
editorial  comments   and  Donald  Walter  of  the
University of California, Los Angeles, and D. Nogawa
for his helpful suggestions.

-------
SIMPLE DIGITAL FILTERS FOR  EXAMINING
CNV AND P300  ON  A  SINGLE-TRIAL BASIS
D. S. RUCHKIN AND E. M. GLASER
Departments of Physiology and Computer Science, School of Medicine, University of Maryland,
Baltimore, MD, U.S.A.
    Evoked potentials can usually be characterized as
consisting of the sum of a response, synchronized with
the stimulating event that elicits it, plus  on-going,
spontaneous neuroelectric activity that is not related
to the stimulus event. For data processing purposes,
event-related  activity is  referred to as "signal,"  and
spontaneous activity as "noise." The noise is often of
sufficient intensity to obscure the signal.  Computa-
tion of an average response over several trials is usual-
ly used to attenuate noise interference.

    Average  response  computation  has  long been
known to be  a valid procedure when the signal is in-
variant from trial to trial. If the signal is not invariant,
then averaging across trials can result in an imprecise
or  misleading average  waveform. Procedures have
been devised  to deal with such a situation (Burns and
Melzack  1966; Woody  1967;  Ruchkin 1968, 1971;
 Ruchkin and Sutton, in press; Glaser and Ruchkin
 1976). A useful first step in dealing with trial-to-trial
variation of  the  signal  is to  inspect the  data on a
single-trial basis.  Such inspection can provide insights
 that lead to  a fuller understanding of the nature of
 the data and help in devising further, more refined
 data processing procedures.

     In order  to  profitably examine  single-trial data,
 some attenuation of the noise is desirable. This can be
 achieved by  operating  upon  each evoked potential
 with  a suitable linear filter,  provided (1)  the power
 spectra of the signal and noise differ sufficiently, and
 (2)  the  signal amplitude is  sufficiently  large  with
 respect to the noise.

     These two conditions can occur for slow event-
 related potentials  such  as the contingent negative
 variation (CNV) and P300 components recorded from
 the scalp of man. Much of the power in CNV and
 P300  waves  is concentrated at frequencies below 10
 Hz, while the spectrum of the background electro-
 encephalogram extends to frequencies well above 10
 Hz. If the potential's  amplitude averaged over the
 SOO msec preceding the start of a trial is used as a
 baseline, the  root mean  square (rms) level of the noise
usually ranges from 9 to 13 juV, while the CNV peak
amplitude ranges from 5 to 15 ^V and the P300 peak
amplitude ranges from  15 to 25  juV. Such data can
be effectively low-pass filtered so that examination
of the CNV and P300 components in single-trial rec-
ords is feasible.

Methods

    Off-line filtering of the digitized data was imple-
mented on a general purpose digital computer. The
data were  previously stored  on  magnetic tape. The
sampling rate was 62.5 samples/sec.

    The basic filtering algorithm  consists of averaging
together the amplitudes from 2L + 1 adjacent time
points. The smoothed or filtered  output at time point
mT is defined by
        x(mT) =
                     1
L
I
r[(m+k)TJ
(D
 The interval between time samples is denoted by T.
 r(mT) denotes the amplitude of the recorded evoked
 potential at the m^ time point.  x(mT) denotes the
 amplitude of the filtered version of the evoked poten-
 tial. L  specifies the number of adjacent time points
 used. Thus, x(mT) is computed by averaging together
 the  evoked potential amplitudes at time  point mT
 plus  the amplitudes at the L time points preceding
 point mT and (he  L time points following mT. The
 computation  is  temporally  symmetric:  x(mT)  is
 obtained from an equal number of equally weighted
 time points preceding and  following mT.  An impor-
 tant  consequence of this temporal symmetry is that
 the  resulting filter causes no phase shift. The filtered
 data are therefore free of phase distortion.

     The transfer function, W(f), which expresses the
 filter's  gain as a  function of frequency,  is given in
 Equation 2 and is plotted with the solid curve in Fig.
 1 (Oppenheim and Schafer 1975, Glaser and Ruchkin

                  «in(2L+|),rrF
                  (2Ul)simrTF           W

-------
 580
                                                                       Ruchkin and Glaser
 _    1.0
 N
 5    0.8
 I 0.707
 g    o.a
 c
      "
        o
      -0.2
                     Wff)
                     [Wlfll2
                          1
                     0  I2L»1IT
0.31 0.44
                NORMALIZED FREQUENCY (f/f0)

 Fig. 1.  Plot of filter gain as a function of normalized
 frequency for  the one-step filter (solid line} and the
 two-step filter (dashed line).
 The  frequency  f0=l/[(2L+l)T]  is the first  (lowest)
 frequency at which W(f) passes through zero. f0 is a
 useful parameter for characterizing the filter. It relates
 filter bandwidth to both the sampling interval and the
 number of time points  used for smoothing or averag-
 ing. Filter gain, W(f), is plotted versus normalized fre-
 quency, f/f0,  in Fig. 1. The figure indicates  that the
 averaging algorithm produces low-pass filtering. We
 use the half-power frequency, at which point the filter
 gain is 0.707 (or, equivalently,-3 dB),to indicate the
 upper cutoff frequency of the filter.  Thus, the filter's
 pass-band is from 0 to 0.44fo.

     We refer to Equation 1 as a "one-step" filter. It is
 readily implemented on a digital computer.  Its per-
 formance may be adequate in many situations. How-
 ever, inspection  of Fig. I  indicates that it has a trou-
 blesome property.   The transfer function has a rela-
 tively high secondary peak in the frequency range be-
 tween I.2t0 and I.6f0. This means  that strong noise
 components  in  that range may  not  be effectively
 attenuated.

     A way of dealing with  this problem is to reapply
the same averaging procedure specified by Equation 1
to the x(mT) waveform, the resultant of the original
filtering operation.  It can be shown that the  transfer
function corresponding  to this "two-step" filter algo-
 rithm  is  [W(f)]2  (Oppenheim and Schafer 1975,
Glaser and Ruchkin  1976). The dashed curve in Fig. I
 is a plot of [W(0] 2 versus f/fiv The pass-band for this
 filter extends from 0 to 0.3If0 and the secondary
peak magnitude  in the 1.2 to I.6f0 frequency range is
much smaller than for the one-step filter (by a factor
of approximately 0.212). It should be noted that the
two-step filter is also temporally  symmetric and so it
too has zero phase shift. A total of 4L+1 time points
from r(mT) are utilized  In order to compute the am-
plitude at each time point of the two-step filter out-
 put,  2L  points preceding and  2L points following
 time mT.

 Results

    An example of the use of one-step filters is pro-
 vided  in  Fig. 2. The  data arc  from  a  paradigm in
 which the second of a pair of flushes elicited u P300
 potential  (Ruchkin and Sutton 1973 and in  press).
 The interflash  interval  was 880 msec. The effective
 pass-band of the recording system was  from  O.I  to
 30 H/.. The data were digiti/cd over an cpocli starting
 500 msec prior lo  the  first flash  and extending for
 2500 msec after the first flash.

     Evoked potential records for three selected single
 trials and  the  average  obtained from these and 81
 other  trials are  plotted in each of the four columns,
 The top row is a plot  of the data when no filtering
 was used. The  second, third, and fourth rows are
 plots of the data after one-step low-pass filtering with
 half-power  frequencies  of 5.5,  3.1, and  1.6  H/,
 respectively. The filters were implemented as follows.
 Since  the sampling interval was 16 msec, f0 = 62.5/
 (2L+1) H/.  and the half-power (-3 dB)  frequency is
 0.44fQ «  27.5/(2L+1) H/.. Thus Ihe  5.5-H/. low-pass
 filter ulili/.es 5  time points (L=2), the  3.1-H/  filter
 utili7.es 9  time  points  (L=4), and (he  1.6-H/.  filter
 utili/.cs 17 time points (L=8).

     Inspection of the average response waveforms in
 the fourth column suggests that much slow potential
 detail  is preserved when the 0- to  5.5-Hz low-pass fil-
 ter is used.  Use of narrower band  low-pass filters,
 such  as a  1.6-H/. filter, results in  considerable wave-
 shape  distortion. In our experience,  a  5.5-H/. low-
 pass filter is generally  suitable for simple  inspection
 of single-trial P300 andCNV data.

    While  one-step filtering  is often  satisfactory, a
 two-step filter  with the same half-power frequency
 will attenuate high frequency noise more effectively.
 This can be inferred from Fig. 1. Note, however, that
 a two-step filter requires about 50% more  computer
execution time  than a  one-step  filter with the same
half-power frequency.

    We have found two-step filtering is sometimes
necessary  when estimating P300  peak  latencies  in
 single-trial records. We found that when a peak detec-
 tion procedure  is utilized, preliminary filtering of the
 data  with a 3.1-Hz low-pass filter  generally yields
most  satisfactory  results (Ruchkin  and Sutton,  in
 press). For subjects exhibiting low level alpha activity,
a one-step filter with L=4 is adequate. However, due
to the secondary peak  in the  transfer function, high
intensity alpha in  the  9- to 11-Hz band is not suffi-
ciently  attenuated.  For subjects with  such alpha ac-
tivity,   use  of  a  two-step  filter with  L=3 and a

-------
Digital Filters for CNV and P300
                                                                                               581
                            SELECTED SINGLE-TRIAL RECORDS
                                                                   AVERAGE OF
                                                                    84 TRIALS
     NO
   FILTER
  LOW PASS
   5.5 Hz
  LOW PASS
    3.1 Hz
/"V  \  /
   LOW PASS
     1-6Hz     r t      V
                 SB
                                                                 4  500 m$K
                                                                                  t    t
                                                                   t    t
                                           71
                                                                          FLASH
                                                                       PRESENTED
 Fig. 2.Plots of single-trial evoked potential records for three selected trials and the average evoked potential for
 84 trials. No filtering was used for the records in the top row. The results of applying single-step low-pass filter-
 ing with half-power frequencies of 5.5, 3.1, and 1.6 Hz are illustrated in the second, third, and fourth rows. The
 time epoch is 3 seconds.
 half-power frequency of 2.8 Hz has yielded satisfac-
 tory results.

 Discussion

    The rationale for using the low-pass filters describ-
 ed above is largely pragmatic and empirical. The filters
 are readily implemented. Their capability for attenu-
 ating noise while passing P300 and CNV activity ap-
 pears to be satisfactory. They are not "optimal" in
 the  sense  of least mean square error  filters (Wiener
 1949, Walter 1969, Nogawa et al. 1973, Doyle  1975).
 However, they do not require the detailed knowledge
 of signal and noise power spectra nor the degree of
 computing effort  that are necessary for implementa-
 tion of such "optimal" filters.
                                     Summary

                                         Linear filtering procedures that can make exami-
                                     nation of CNV and P300 components on a single-trial
                                     basis feasible are described. The filtering procedures
                                     are readily implemented  by general purpose digital
                                     computers. Examples of the use of the filters are pro-
                                     vided.

                                     Acknowledgment

                                         This work was supported in part by a U.S. Public
                                     Health Service NINCDS Grant  NS11199 to D. S.
                                     Ruchkin, We thank Dr. Samuel Sutton of the Biomet-
                                     rics Research Unit, New  York State Department of
                                     Mental Hygiene, for his criticism and advice.

-------
 IMPLICIT SPATIAL AVERAGING OF SURFACE
 MACROPOTENTIALS

 L. K. GERBRANDT
 Department of Psychology, California State University, Northridge, California, U.S.A.
     Macropotentials are, by definition, population
 estimates of spatiotemporally localized events. The
 representativeness  of these population estimates  is
 crucial in the accurate determination of the functions
 corresponding to these macropotentials. A major con-
 cern of many investigators has been the neuroelectric
 information lost or confounded due to the summation
 of time-locked  signals.  This paper concerns  some
 recent work that suggests that single-trial macropoten-
 tials recorded from cortical surfaces may constitute a
 spatial average and may therefore be responsible for
 an even greater loss of theoretically relevant informa-
 tion than that associated with signal averaging in the
 time domain.
    As Klemm  (1976) has noted, the hippocampal
theta rhythm (rhythmic slow activity, or RSA) is one
of the readiness stales that occurs in preparation for
adaptive movements. Preliminary to the further study
of the function of hippocampal steady potentials and
RSA in voluntary  movement (Vanderwolf et  al.
1975), we studied the surface and depth topographies
of RSA in the rat (Gerbrandt et al.  1974; Gerbrandt
et al. 1975). Half-cycles of RSA were detected elec-
tronically and the resultant detection pulses served to
synchronize  the sweep of an averaging computer to a
constant phase of the RSA monitored at a phase refer-
ence electrode located over the neocortex; thus, mul-
tiple channels of averaged RSA were obtained as
shown in Fig. 1. In  Fig. 2, the scalp topography of
the average RSA amplitudes and phase are shown. At
               Lam
Fig 1. Ctx - neocortical phase reference site indicated by dashed traces in A and R Lam = laminar electrode,
125 H>"  above hippocampal pyramidal cell layer, indicated by solid traces in A and B.  A t right, traces A and B are
computerized averages of the raw KK(i (50 samples taken during epochs indicated by solid marks under KKG). A
was obtained with RSA at neocortical phase reference  used as a trigger for zero-crossing  "theta" detector,
while B used laminar electrode as the trigger.  Calibration bars = 100 n V (solid) and 300 ^.V (dotted). Phase shift
in 8=34°, in B = 32°. Trace duration for A, B = 500 msec.

-------
Implicit Spatial Averaging
                                             583
                                   00 10  M  "0  20  D

                                   RELATIVE AMPLITUDE. H Of «
Fig. 2.  A indicates RSA  averages (50 EEC samples)
obtained from  active recording sites identified in  B
(earthed reference). B depicts a rostral view ofneoconex
with epidural recording  sites; underlying hippocampus
depicted as horseshoelike structure between hregrna and
lambda points.  The cross-shaped nine-electrode array
with 2-mm interelectrode spacings is centered (electrode
"x") at  3  mm behind bregma and 2 mm lateral to the
midline; candidate indifferent electrodes (triangle  and
square) are  located  over  the  olfactory bulbs  and
cerebellum. C is a plot of the percent amplitudes of the
waveforms in A relative to the waveform recorded at site
x. Calibration line (far left) = 200 u V.

the scalp surface, phase  shifts of 4° are detected with.
electrode  spacings of 1 mm, and in the same distance,
amplitudes do not change more than 25% of the peak
amplitude. This  poor  spatial  resolution of latency
(phase)  and amplitude is not too surprising, consider-
ing that neocortically monitored RSA is generated in
the dorsal hippocampus rather  than  in the neocortex
(fig. 3); the dorsal hippocampus lies about 2 to 3 mm
below the surface of the neocortex. Indeed, a 4° shift
is  detectable with about a  100-Aim  movement of an
electrode  horizontally  along  the  surface  of  the
hippocampus  (Winson  1976),  although amplitude
gradients  are no more  than two times steeper than
those registered along the neocortical surface (Bland
et al. 1975).
    This tenfold increase in latency  on the surface,
impressive  as it is relative to the adjacent neocortex,
is  suboptimal  compared to  the  resolution and accu-
racy of information at rij>ht angles to  these cortical
surfaces. A gradual phase advance  of  the type shown
in Fig. 4 occurs  at an average of  4°  with as little as
10/jm of laminar electrode movement. In this vertical
plane, it is also common to observe a tenfold change
in amplitude  with an electrode movement  of only
400 Mm. Finally, a change in the  mode of induction
of RSA (spontaneously or tactile-induced RSA) does
not experimentally alter the  amplitude or phase gradi-
ents of RSA observed along the surfaces of the neo-
cortex or the hippocampus. Yet, the dentate gyrus of
the hippocampus  selectively changes in phase  as a
function of these variables,  without concommitant
changes in  the CA I region. If a factor analysis of the
phase and  amplitude gradients is utilized, even sub-
regions of the single-layered pyramidal cells of the CA
1 region show variances that are  independent of the
other CA 1 sub-regions. Furthermore, different inde-
pendent  variables selectively control different hippo-
campal sub-regions (Gerbrandt et  al. 1975). The ob-
servation that certain independent variables selective-
ly affect only some sub-regions of dendritic branches
is understandable, considering the laminated and heter-
ogeneous separation of afferents along the  axis of
each hippocampal  neuron (Fig. 5).  However, what is
surprising is that we in fact did expect that independ-
ent variables that influence such a complex structure
could be estimated from the recording of surface
macropotentials,  or  indeed  of  macropotentials at
any single electrode location.

    In summary, these findings may have two impor-
tant  suggestions  concerning the  information lost
because  of implicit spatial averaging of scalp macro-
potentials.  First,  even if it were  possible that all of
the information  "tapped"  by an  electrode passing
vertically along  the cortical (hippocampal) cell axis
were  broadcast upward along the surface  plane, the
considerable loss of spatial resolution (10  to 100 X)
Fig. 3.  To the left is a scaled schematic of a parasag-
ittal section  at 2.0 mm lateral to  the midline. Each
point is a scaled  100-nm interval.  At right, RSA av-
erages of 20 samples, with the dotted  traces repre-
senting recordings from the contralateral homotopic
phase  reference electrode and solid traces taken from
the depths indicated at left. Note the in-phase activi-
ty recorded  to a depth 200-nm above the pyramidal
cell layer, with an approximate phase reversal evident
at  the  hippocampal fissure.   All  calibration bars =
JOOfiV.

-------
 584
                                                                                              Gerbrandt
                                                       of information, demonstrated here, would  result in
                                                       multiple superimpositions that  are  likely to cancel
                                                       some of the real signals.  This  type of information
                                                       loss is a spatial analog of the loss due to the temporal
                                                       smearing, which  has been discussed. Unfortunately,
                                                       few of  the variations experimentally induced among
                                                       the multiple  generators of RSA are  seen in activities
                                                       monitored at the cortical surface. Thus, spatial averag-
                                                       ing imposes the additional difficulty that only a small
                                                       subset of the variables thai delcrmine each function
                                                       are likely to be sampled at the cortical surface.
Fig.  4. At the right is a scaled schematic parasagittal
section through the hippocampus at 2.0 mm lateral to the
midline. To the left, RSA averages are shown (SO trials
summated). where dotted  traces  represent  activity
recorded from a contralateral homotopic phase reference
electrode,  and solid traces are RSA waveforms averaged
at the indicated depths of the hippocampus. Phase shifts
in the hippocampal traces are, beginning at the lop, 18°,
38°,  76°.  104°.  Calibration bar represents  100»V for
phase reference  traces and 200 u V for hippocampal
traces.
                                                                             septal
                                                                             comm
                                                                           baskets
                                                                       septal-entor
                                                                            comm
                                                                             septal
                                                                           baskets
                                                                             septal
                                                                             entor

                                                                            comm
                                                                            septal
                                                                         moss fbrs
                                                                           baskets
                                                                            septal
                                                                            comm
 Fig. 5.  The interlocking pattern of the single-layered pyramidal (CA1, CAB) and dentate granule cells (dent)
 and their intrinsic connections (mf=mossy fibers, sc=Shaeffer collaterals), are  shown schematically on the left.
 The sources of the laminated pattern ofafferents for each type of hippocampal neuron are shown schematically
 on the right (septal, commisural, etc.).

-------
NEUROMETRICS:  QUANTITATIVE
ELECTROPHYSIOLOGICAL  ANALYSIS  FOR
DIAGNOSIS  OF LEARNING DISABILITIES
AND OTHER  BRAIN DYSFUNCTIONS1
E. R. JOHN, L. S. PRICHEP, H. AHN, D. BROWN, P. EASTON, B. Z. KARMEL2, R. THATCHER,
AND A. TORO
Brain Research Laboratories and Neurophysiology Clinic, Departments of Psychiatry and
Physiology, New York Medical College, New York, NY, U.S.A.
    At the present time no quantitative criteria exist
for the  accurate  separation of learning disordered
children with an organic basis for their impairment
from children with primarily functional disturbances
or from  nonnal children. The difficulty of this differ-
ential diagnosis  can be readily ascertained by examin-
ing any  recent  volume on this topic (e.g., Wender
1971, Walzer and Wolf 1973) or by examining current
procedures for evaluation and treatment of such child-
ren. The importance  of proper diagnostic capability
for this  disorder can  be appreciated  if  one realizes
that  current estimates of its incidence  in the U.S.
population range between 5 and 15% (Wender 1971).
With early identification and subsequent monitoring
of children at risk for cognitive and behavioral prob-
lems, intervention could be initiated at the first signs
of difficulty.

    The diagnostic tools presently available to aid in
this endeavor have severe limitations.  Classifications
are often based upon vague, intuitive observations of
behavior of ambiguous origin, or upon psychometric
measures of product  which fail to make process ex-
plicit. There are too many ways to produce an inap-
propriate behavioral test response. The techniques of
classical  neurological examination are not optimal for
the analysis of subtle dysfunctions in  information
processing,  storage and  retrieval. The conventional
electroencephalogram (EEC), evaluated by traditional
methods of visual inspection, has yielded suggestive
but  inconsistent  findings of abnonnal features  in
children displaying difficulties in learning (e.g., Burks
1960, 1968; Capute  et al. 1968; Conn  and Nardini
1958; Ellingson 1954; Grunewald-Zuberbier  et  al.
1975; Klinkerfuss  1965; Pavy  and Metcalfe  1965;
Satterfield 1973; Satterfield et al. 1972; Wikler et al.
1970). More recently, studies of the sensory  average
evoked potential (EP) have suggested that such meas-
ures of brain function may reflect aspects of informa-
tion processing or cognitive functions (see reviews by
Regan  1972, John  and  Thatcher 1977).  However,
most EP  studies of learning  disabled children have
focused upon gross sensory reactivity rather than in-
dices of information processing and have further suf-
fered from the limitation of qualitative visual evalua-
tion  (e.g., Buchsbaum  and  Wender 1973, Conners
1970, Hall et al. 1976, Halliday et al. 1976, Preston
et al.  1974, Prichep et al. 1976, Saletu et al. 1973,
Satterfield 1973, Satterfield et al. 1972).

    Several conclusions may thus be drawn concern-
ing the present status of the field of assessment of
learning disabilities in children: (l)The EEC and EP,
assessed under various conditions of information pro-
cessing, potentially  provide  a  fairly  direct  insight
into  brain functions related  to sensory, perceptual
and cognitive processes likely to be relevant to learn-
ing and performance. (2) In order to take optimal
advantage of these indices, it is necessary to devise  a
technology  that  will derive precise quantitative mea-
sures of salient  features of  the EEC and  EP under
conditions that "challenge"  a wide variety of brain
functions devised to re fleet such informational proces-
ses. (3) neurobchavioral measures should be reduced
to numerical representations permitting the use of
mathematical and statistical methods of data analysis.
(4) Display and evaluation  methods  are needed to
evaluate individual data  referred to normative data
and to present  results in a clinically comprehensible
and concise way.
     A new technology known as "neurometrics" has
 been developed to meet these needs. The purpose of
'This work was supported by NSV/RANN under Grant #KRP 72-03494 (formerly Gl-34946)
"On leave from the University of Connecticut

-------
  586
                                               John et al.
   AUDIO-VISUAL-TACTILE
        STIMULATOR
        DATA TERMINAL*
             t	
      24 FIXED-GAIN
      PREAMPLIFIERS
     FULL 10/20 SYSTEM
   RECORDED MONOPOLAR
     AT ONE TIME. ALL
  MONTAGES AND BIPOLAR
   DERIVATIONS ARE CON-
     STRUCTED BY THE
        COMPUTER.
      HIGHSPEED
    ELECTROSTATIC
    MATRIX PRINTER
                                 MINICOMPUTER SYSTEM
                                 IN ACQUISITION MODE
STIMULATION PROGRAM
     LATIN SQUARE
AUTOMATIC IMPEDANCE
       TESTING
  ARTIFACT MONITOR
   ANALOG-DIGITAL
     CONVERSION
                                        II
                                   AUTOMATIC GAIN
                                      CONTROLS
                               DUAL MAGNETIC DISK

                                   PROGRAM DISK
ACQUISITION OR ANALYSIS
        PROGRAMS
                                                                   DATA
                                           DISK
                                                              PROTOCOL OF RECORDING
                                                                      SESSION
UNACCEPTABLE IMPEDANCE
       INTERRUPTS
  ARTIFACT INTERRUPTS
  STORAGE OF DIGITIZED
          DATA
                                    PROTOCOL OF
                                   AMPLIFICATION
          I
    *ONE SYSTEM CAN
     PROCESS SEVERAL
     DATA TERMINALS.
Fig.  1. Functional diagram of computerized digital electrophysiological data acquisiton and analysis system
(DEDAAS).
this paper is to provide a general overview of this tech-
nology. More detailed descriptions and results from
preliminary applications of neurometries appear else-
where (see John 1977; John and Thatcher 1977; John
et al. 1977).

Data acquisition procedures

    An automatic digital electrophysiological data
acquisition  and analysis system (DEDAAS) is repre-
sented schematically  in Fig. 1.  Data acquisition is
accomplished by a set of 24 EEC amplifiers designed
and constructed in these laboratories. These ac ampli-
fiers  have a precise fixed gain (10,000 X), a noise
level less than 2 pV, a high common mode rejection
               ratio (106 dB), a frequency response of 0.3 to 70 Hz
               (6 dB/octave roil off) with a sharp 60-Hz  filter to
               eliminate the need for a shielded room, a high input
               impedance (10 megohms) to reduce the influence of
               variation in impedance of scalp electrodes and free-
               dom from drift.

                  The output of these amplifiers goes to a PDF 11
               computer (11/45, 11/10, or 11/03) programmed to
               automatically calibrate the amplifiers and check elec-
               trode  impedances. Nineteen amplifiers are occupied
               by the International Electrode System; channel 20 is
               used for a pair of transorbital electrodes to monitor
               eye movement. The remaining channels are available
               for accelerometer or polygraph recordings as desired.

-------
Neurometrics
                                             587
The full 10/20 system is recorded simultaneously, as
monopolar derivations referenced to linked earlobes.
Data are transformed on-line into digital format and
stored  on magnetic tape, obviating the need for off-
line analog-to-digital conversion. Conventional bipolar
montages or any desired compound electrode are sub-
sequently  constructed by  computer  simulation.  We
routinely compute data from 57 derivations (i.e.,  19
monopolar, 19 coronal bipolar, and 19 saggital bipolar
pairs).

    Frequency  and voltage limits for every channel
are continuously monitored by  computer to identify
data contaminated by eye  or body movement,  or by
high electrode  impedance.  Questionable  data may
either be eliminated or marked  and stored. To assure
the adequacy of artifact rejection, these  records are
routinely inspected prior to quantitative analysis.

    Required stimulus sequences are  generated auto-
matically by computer using a stimulator and neuro-
metric  battery (NB) described below. Complete stim-
ulation protocols are included in the digital record to
permit subsequent automatic  data analysis without
operator intervention.

    The  stimulator  contains a photo-flash tube, a
source of oscillating light, an automatic slide projec-
tor, click  and pure tone (200, 500,  1000 and 3000
Hz)  sources of  specified  intensity, and  a constant-
intensity tactile stimulator. A video set and a cassette
player  provide other stimuli. Appropriate sequences
of stimuli  define all NB conditions, each condition
being a computer subroutine in DEDAAS.

Neurometric test battery

    The quantitative electrophysiological test battery,
referred to as the neurometric  battery or NB, con-
sists of EEC and EP measurements obtained under a
variety of standardized conditions. Each condition is
designed to probe maturational level and structural
integrity of the  brain, as  well as specific  aspects of
sensory,  perceptual and  cognitive  functions.  The
empirical basis  of the neurometric battery has been
reviewed elsewhere (John 1977). While  the NB  is
based on previous research, the  battery will undoubt-
edly be modified and supplemented as data accumu-
late  and the diagnostic utility of each condition is
assessed.

    Each  condition is considered  a test  item  and
yields several scores, which quantify separate features
of electrical activity (see below). Differences between
responses  on different NB conditions constitute an
additional source of information about brain function.
These  composite conditions, in  which information
obtained under one condition is evaluated relative to
 that obtained under another condition, are referred
to as "challenges." The NB includes a total of 92
conditions and challenges, which are fully described
elsewhere (John and Thatcher 1977). Administration
of the entire NB requires approximately 50 min  of
data acquisition. Actual running time may be longer
because of "time-outs" caused by artifacts.

    Normative data are available for large samples of
normal children (age 7-11) and LD children (age 7-18)
for the full set of 92 conditions and challenges (Ann,
et al.  1975; Harmony et al.  1973a, 1973b; Johnet al.
1975, 1976a, 1976b; Kaye et al.  1975;Matousek and
Petersen  1973; Otero et al.  1975a, 1975b). For other
age ranges norms are available  for only a small subset
of the NB conditions. More extensive norms are being
collected  for all conditions and challenges. Table  1
lists the set of test items that presently constitute the
NB and indicates the utility  of each item.

Quantitative indices

    These  NB conditions and challenges  provide  a
large volume of EEC and AEP data. From data record-
ed tor each electrode derivation under every condi-
tion  and  challenge, a  variety of numerical features
arc extracted,

    Under each  EEC condition,  twelve 5-sec samples
of artifact-free EEC are recorded. Numerical features
are computed from each sample separately and the
mean values and  standard  deviations  of these  are
then calculated for the full set of samples. Under each
AEP condition, the EP of every derivation is comput-
ed from 64 EPs. This computation yields the average
signal voltage  and its variance at each of 100 time
points, sampling at 10-msec intervals across a 1-sec
analysis epoch. In  addition,  a number of indices that
are intended to reflect critical features of the response
process are extracted from  these data. These derived
features are computed for  the whole analysis epoch
and for each of four latency intervals corresponding
to components or waveshape segments of special inter-
est. Computed EEC and  EP indices may be described
briefly as follows:

EEG indices

    Absolute power (microvolts squared) in  seven
frequency bands, i.e. low delta (0.5-1.5 Hz) high delta
(1.5-3.5  Hz), theta (3.5-7Hz;), alpha (7-13 Hz), low
beta (13-19 Hz), high beta (19-25 Hz), and wide band
(0.5-25 Hz).
    Relative power (% of total  power) in low delta,
high delta, theta, alpha, low beta, and high beta fre-
quency bands.
    Age-dependent quotient (ADQ): A metric reflect-
ing maturational development is obtained for major
cortical regions by calculating the ratio between the
delta  and theta energy usually observed in that head

-------
 588
                                                                    John et al.
                                 Table  1.  Test  Items in the Neurometric Battery
            Neurometrlc test Item
                            Intended purpose
EEC condition* and challenges
        I.  Eyes open, spontaneous EEC
        2.  Eyes closed, resting EEC
        3.  Eyes open minus eyes closed
        4.  Photic driving at 2.5,5,10, and 18 hertz

AER conditions and challenges
          Sensory acuity
        S.  65  lines  per  Inch,  SO  percent  trans-
           mission
        6.  27  lines  per  Inch,  SO  percent  trans-
           mission
        7.  7   lines  per  inch, 50  percent  trans-
           mission
        8.  45 db click

           Pattern perception

       9.  Large square
      10.  Small square
      II.  Large diamond
      12. Small diamond
      13.  "b"
      14.  "d"
      15.  "p"
      16.  "q"

           Prediction of temporal order

      17,  Random versus regular flash
      18,  Random versus regular click
      19.  Random versus regular tap
      20, Phasic habituation

      21. Dishabituation

      22.  Rehabttuation

          Seniory-iemary Interaction!

   23-25. Passive    Interactions   between  visual,
          auditory, and somatosensory systcmi
      26.  Flash followed by click 2SO msec
          later
      27.  Click followed by flash  250 msec
          liter
  Baseline measures
  Yields age-dependent quotient
  Kffcct of removal of visual input
  Yields  reactivity in delta,  iheta, alpha, and beta ranges when compared with baseline
    measures
  Perceived as a blank flash

  Seen as checkerboard if visual acuity Is approximately 20/20

  Seen as checkerboard unless visual acuity Is worse than 20/200

  Elicits  auditory  AER unless hearing loss  Is sufficiently severe  to  Interfere with  lan-
    guage acquisition

  Each  contributes to  an  estimate of perception of differences In geometric forms but
    preservation of shape invariance Independent of sl/.e
  Each contributes to estimates of central discrimination between shapes of letters most
    commonly reversed
 Change  in AER  waveshape  reflects diminished response  to  predictable  stimuli,  In-
   dicates recognition of repeated temporal sequence

 Reveals rate  and  amount  of suppression  of  Information  Input about  a  meaningless
   monotonous event, reflects attention and short-term memory
 Indicates whether  suppressed input is  nonetheless continuously monitored  to permit
   detection of possible change
 By  comparison with initial  phasic habituation, reveals whether suppression of mean-
   ingless Input Is facilitated by memory of previous experience
 Reveals Increase or  decrease in  response  of  brain  as  a result of simultaneous  pre-
   sentation of simple stimuli In different  sensory  modalities
 Measure of recovery cycle after visual input

 Measure of recovery cycle after auditory Input
               Figure-pound relations

   28-30. Interaction  between   meaningful  visual
          input (figure, consisting of scenes  on a
          video  screen)  and  meaningless visual,
          auditory,    or    somatosensory   input
          ground)
   31-33.  Interaction between meaningful  auditory
          Input (figure, consisting of a tape record-
          ing of a musical selection or  story) and
          meaningless visual, auditory, or somato-
          sensory input (ground)


EEC conditions and challenges
      35. Eyes open, spontaneous EEC
      36. Eyes closed, resting EEC
      37. Eyes open minus eyes closed

      38. Eyes open, beginning,  minus eyes open,
          end
      39. Eyes  closed,  beginning,  minus  eyes
          closed, end
Reflects  dynamic structuring of figure-ground relationships  which require  discrimina-
  tion between  relevant visual "signal" and Irrelevant "noise," which may  be  wither
  ipsomodal (video-visual) or cross-modal (video-auditory or vldeo-somatosensory)
Reflects  dynamic  structuring of figure-ground  relationships  requiring discrimination
  between  relevant auditory "signal"  and  Irrelevant "noise," which may be wither
  ipsimodal (music-auditory) or crosi-modal (music-vnual or muslc-somitosensory)
Replication of initial measures
Estimate of effects due to state, such as anxiety about test or fatigue due  td testing,
   versus characteristic individual features displayed across states

-------
Neurometrics
                                            589
region in a normal person the same age as the patient
and  the  amount of delta and theta energy actually
measured in the patient. If the ratio
     Ann = norma* energy '" Frequency band
            patient energy in frequency band
is approximately 1.0, the amount of slow activity in
the recording  is appropriate for a healthy person of
that age. Many brain diseases, as well as maturational
lags, are  reflected by an excess of slow activity and
ADQ values significantly less than 1.0. Progress of an
abnormal brain state, such  as might ensue from head
trauma, space-occupying lesion, cerebrovascular acci-
dent, or maturational lag,  can be followed quantita-
tively by comparing values of ADQ obtained sequen-
tially at appropriate time intervals. Effects of medica-
tion or other treatment can be similarly measured.

     Coherence between homologous pairs of deriva-
tions (phase-locked correlation) in  low delta, high
delta, theta, alpha, low beta, and high beta bands.

    Amplitude symmetry between homologous deri-
vations in low delta, high delta, theta, alpha, low beta,
and high beta bands.

     Overall waveshape symmetry between  homolo-
gous pairs as  assessed by the cross-correlation coef-
ficient for the wide-band EEC signal.

AEP indices

     The  following  indices are extracted from  EP
waveshape and variance, which constitute the initial
level ot computation.

     Signal strength. Energy represented in the EP in
four latency intervals, i.e., 40-99, 100-199,  200499,
and 500-999 msec.

     Noise. EP variance in the same four latency inter-
vals.
     Signal/noise ratio. Average signal strength across
each of the four latency intervals divided by the aver-
aged value of the noise during the corresponding inter-
val.

     Mean squared first difference, which is propor-
tional to the product of signal energy and mean squar-
ed signal frequency.

     Pairwise  energy asymmetry.  The difference  in
signal strength between homologous pairs of electrode
derivations, computed  for each of the four latency
intervals.

     Pairwise waveshape asymmetry. The difference in
waveshape between homologous pairs of electrodes,
as represented by the Pearson correlation coefficient,
computed for each of the four latency intervals.
    Peak amplitude and latency for each component
identified in the EP.

    Peak amplitude asymmetry both  absolute and
relative for each component.
    Amplitude and latency differences between cor-
responding peaks in homologous electrode pairs.

    Amplitude excursions between the peaks of suc-
cessive negative and positive components.

    t-test for  significance  of differences  between
waveshapes  recorded simultaneously from bilaterally
symmetrical derivations. The significance  of differen-
ces is  tested for all points throughout the analysis
epoch.

    F-test  for significance  of differences  between
waveshapes  recorded simultaneously from  any com-
bination of derivations.

Data reduction, display, and evaluation

AEP morphology

    The digitized EP waveshape can be  represented
as a signal  vector  in a 100-dimensional  time space,
where each dimension corresponds to signal voltage at
a particular latency point in the analysis epoch. Factor
analytic procedures can then be used to determine
the actual dimensionality of this "signal space," either
for EPs from many derivations in the same individual
or from the same derivations(s) in many individuals.
It then becomes possible to construct a more parsi-
monious description of data consisting of the linear
combination of a set of terms each defining the rela-
tive contribution (factor loading as % signal energy)
of each basic dimension (factor) to the signal vector.
These  linear equations enable great compression of
EP data. Every EPin the signal space can be represent-
ed  by a small set of factor loadings specifying the
relative contribution of the  same basic set of factors
to each of the EPs.

Ex tracted features

    The full NB yields an extremely large volume of
data, making obvious the need for compression, eval-
uation, and display techniques for  comprehension of
the mass of data available for any individual or popu-
lation. Group means and standard  deviations for nor-
mal individuals in each age range of interest, for each
quantitative index are  calculated, followed by t-tests.

    These, and other measures, can be displayed  for
individuals or populations in ways that facilitate com-
prehension.   Fig.   2   shows  an  example  of   the
"POPHIS"  display  for two graphs  from two selected
sets  of subjects. This  program plots cumulative fre-
quency curves  of both groups superimposed on  the

-------
 590
                                                           John et aL
     EXAMPLE OF POPULATION DATA COMPRESSION AND TOPOGRAPHIC DISPLAY

            DISTRIBUTION OF DIFFERENCE IN ENERGY IN 200- TO 500-msec
        LATENCY RANGE COMPARING EPs TO RANDOM VS REGULAR FLASHES

                                                   CUMULATIVE
                                                    PERCENT OF
                                                    POPULATION
                                                       Z NORMALIZED
                                                    VALUE OF DIFFERENCE
                                                         IN ENERGY
                 SAMPLE
                   SIZE
     MEAN
           F3         FZ         F4         F8
           N  65  21   N  65  22  N  65  22   N  65  22
           M  64  83   M  80  89  M  74 102   M  27  38
    STANDARD-*S   40  30  S   67 80   S  79 104   S  74 83  S  30  48
    DEVIATION  T  312 163  T  253 394   T 101 454   T 348 358  T  253 714
UNSCALED
 't'TEST
 Z-TEST
                      CZ
                      N  65  22
M  51  40   M  103  80  M 118 119  M 109 107
   77  40   S  92  81  S 130 107  S 121  90
T  219  75   T  281 215  T   7 174  T  23 143
                T5         P3          PZ        P4~
                N  65  22  N  65  22   N  65  22  N   65  22
                M 156  93  M 140  73   M 151 115  M  178  71
                S 205  94  S 158  67   S 159 124  S  207  51
                T 503  78  T [7121  91   T 274 170  T |970l  82
                   I   I
             NORMAL  LD
               DATA  DATA
                                01         02
                                N  65   22  N   65  22
                                M 238  149  M  239  197
                                S 228  128  S  224  218
                                T 580  119  T  197  253
                                                            T4
                                                            N  65  27
                                                            M  77  38
                                                            S 118  42
                                                            T 573  60
                                            N  65  22
                                            M 206  82
                                            S 309 107
                                                 	LDGROUP
                                                 	NORMAL GROUP
                                               BOXED t VALUES
                                               ARE SIGNIFICANT
                                               AT  p<.001
Fig. 2. Compression of data comparing derived features ofEP waveshapes elicited by random and regular flashes,
recorded from 19 electrode placements of the 10/20 system in a sample of 65 normal and 22 learning disabled
children. The overall display represents the topography of the electrode array viewed from above, with the front
of the head oriented toward the top of the display. Each graph compares the cumulative distribution of the selec-
ted EP features in the two different groups of children (POPHIS program).

-------
 Neurometrics
                                            591
same graph, with 19 graphs displayed in the format of
a head  for each of the three  montages (monopoiar,
saggital,  or coronal bipolar). A set of graphs  is dis-
played for each measure, and for each time epoch of
the EP.  Below each graph in  two separate columns
are displayed the group N, mean, and standard devia-
tion for the measure for  each group,  a t-test value,
and a  Z-test value (both in  the  row labeled "T").
These programs can be used  to display basic measures
as well  as difference  measures comparing bilaterally
symmetric locations  or differences between  condi-
tions that define challenges.

    It now becomes possible to implement a strategy
considered to be one of the most powerful aspects of
the neurometric technique. Each  NB index obtained
for any  individual is  subjected to Z-transformation.
The individual index is transformed from its original
units into a form representing the relative probability
of  encountering that  value for  that  index among
members of a normal reference group. It then becomes
possible  to define "abnormality" in a statistical sense
as  a set  of improbable  values  not attributable to
random  eccentric values due to the large number of
items in the measure set.

    One useful application of Z-transformed individ-
ual NB values is to construct a graphic representation
of  the  anatomical location of  improbable  indices
found across any subset of measures for an individual
or group of individuals. In  such displays, illustrated in
Fig. 3, a localized region on a head diagram represents
the anatomical position of each electrode derivation.
The density of shading in each local domain is propor-
tional to the degree of departure of represented in-
dices from the normative  reference in positive (+) or
negative (-) deviations. That is, the more deviant the
value, the more dense the entry on the head diagram.
Deviations of data within  normal limits are indicated
by  a pair of spots. This representation provides a rapid
overview of findings in any individual, providing visual
correlation between  type and anatomical locus of
abnormality.  Indices that initially reflect  disparate
types of dimensions such  as voltage, time, latency, or
coherence, are  now transformed  to the  common
metric of probability. It therefore becomes possible
to  compare or combine measures that were initially
not dimensionally comparable.

    The abnormality profile  of  any individual can
now be represented as a Z-vector in an NB-dimension-
al  probability space.  The Z-vector for an individual
whose NB indices do not deviate significantly from
expected values will not be significantly distant from
the origin of this space. However, the more deviant an
individual's NB indices, the further into the space the
corresponding Z-vector will project. The direction in
which the Z-vector points corresponds to the diagnos-
tic  definition of abnormal electrophysiologjcal activity
in the brain of that individual. Thus, abnormality is
defined quantitatively by  the length of the Z-vector
and qualitatively by the orientation of the Z-vector.
A distance matrix can now be computed between the
Z-vector representing each individual and those repre-
senting every other individual in the group being stud-
ied, yielding interindividual distances in probabilistic
terms.  The computation of such a matrix is the start-
ing point for a number of statistical pattern  recogni-
tion and cluster analysis methods that can be used for
the objective classification of groupings of data points
in  multidimensional spaces. The general field  of these
classification methods is known as "numerical taxon-
omy" (Sneath and Sokal 1973).

    The prototypic data in Fig. 3 shows five normal
and five LD children selected from  a  larger sample
and are to be considered as illustrative examples rather
than as invariable findings.

    The upper four rows of displays  represent the
distribution  of relative power  in the spontaneous
EEC,  recorded from bipolar derivations with  eyes
closed. Note the typical excess of slow delta  activity,
predominantly in  posterior head regions of the LD
subjects, usually coupled with a deficiency of alpha
and sometimes of beta activity. The  fifth row shows
that the LD subjects show significantly less change in
signal energy of the bipolar EP in the latency region
between 200  and  500 msec when  a flash delivered
randomly while the subject is watching a TV cartoon
is compared with a flash delivered randomly while the
subject looks at the defocussed TV screen. The sixth
row shows that LD subjects display significantly less
change in signal energy of the monopoiar EP in the
latency region between  200 and 500 msec when a
random flash  is compared with a regular flash. The
particular head regions displaying this less-than-expect-
ed difference when the two conditions are compared
vary from subject to subject. Nevertheless, such find-
ings show that LD children tend to  display  less sup-
pression of  P300 to an irrelevant stimulus (ground")
in the  presence of meaningful  environmental input
("figure")-  Analogously,  they also  display less of a
tendency  to distinguish between predictable and un-
predictable  events  in the environment,  reflected in
the similarity  of late positive EP components elicited
by these two different kinds of events.


    To date we have collected  neurometric data on
approximately 1000 learning disabled children and
are in  the process of submitting this data to cluster
analysis. It is our belief that a number of neurophysi-
ologically homogenous subgroups (i.e., groups of indi-
viduals with similar electrophysiologjcal profiles of
brain function), will be identified within this hetero-
geneous group of individuals who display common
behavioral symptomatology. Further, we expect that
many of these subgroups will correspond to  diagnos-
tic entities with meaningful functional andetiological

-------
 592
                                                                  John et al.
                                              NORMALS
                                                                              LD.
                                    PSYCHOMETRIC
                                        SCORE
                                                                II
                                                      PRELABELED  PRELABELED AGE
                                            ;«-i         NOHI

                                            [3l|  1129  1134 I
                                            jjl] 00802 00802 |(jj
                                                       NORMAL
                                                                            YR
SUBJIO   1154   1131]   1129   1134 I 1144
 CLASS  00800  OOsFfl  00802  00802 190802
                                                                    LD
[1142 1116   11118   1145/1110
D0809 10810  1g|06  108[59J 10BOB
                                                              MISSING







EEC
FREQUENCY-
BANDS










EP
LATENCY —
INTERVAL


DELTA
EYES CLOSED — ».. ~ 2
RELATIVE POWER1! 	 c.139
BIPOLAR DERIVATIONS 10
THETA
2
139
10
ALPHA
2
139
10
BETA
2
139
	 10
FLASH PLUS VIDEO 200 499 MS
MINUS FLASH "• " 648
— 80
SIGNAL ENERGY-*— * 10

RANDOM MINUS 200 499 MS
REGULAR FLASH*-" 576
HO
MONOPOLAR
DERIVATION * 	
                                                                      "Vo.oooi
                                                                               X
                                                                                  p Q •;•
                                                                        »*'.:•::
                                                                           •*»p-0.00001
                                      p"001

                                                                                              FRONT Of HEAD
                                                                                              LOOKING DOWN
                                                                                              FROM ABOVE
                                                           p 0.1
                                                                    p=001  p*0001
                                                                                     p-00001
 Fig. 3.  Illustrative density-coded Z-transj'ortned displays of neurometric indices extracted from various /•.'/:'G and
 EP conditions of the NB. Each column of displays represents data obtained from one subject, while each row
 represents one univariate or multivariate index. Each display represents an array of entries: each entry corresponds
 to the value of the index measured at that point on the subject's head, while the position of the entries in the array
 corresponds to the electrode locations of the JO/20 system.
implications. Preliminary results indicate that this will
be the case (John 1977, John and Thatcher 1977).

Conclusion

    While our primary concern lies with early detec-
tion and remediation of subtle  brain dysfunctions in
learning disabled children, we are also  interested  in
the application  of neurometric techniques to other
problems that would benefit from more sensitive asses-
ment of brain function (e.g.,  the problem of senility
in the elderly,  and neuropathologics sucli as tumors,
strokes and traumatic  head injuries). A description of
preliminary  applications  of  neurometrics  to other
groups  can be  found  in  John  and Thatcher (1977)
and John (1977).
                              Successful application of  this new technology
                          should have a number of important consequences. It
                          should  extend  the domain  of electrophysiological
                          assessment  into   information  processing  disorders,
                          where present methods are all but useless.  It should
                          make possible mass screening for  the early identifica-
                          tion  of victims  of such disorders, with  increased
                          opportunity  for early intervention  and  remediation.
                          Most  important,  numerical taxonomy applied to an
                          adequate  spectrum  of  neurometric  indices  might
                          make  possible the differential  diagnosis of distinct
                          etiologies within  categories of  patients now consid-
                          ered homogeneous.  Once such  differential etiologies
                          are identified, it  might be possible to identify their
                          antecedents, institute  specific  preventive  measures,
                          and  devise  individualized prescriptive   remediation
                          effective for their  treatment.

-------
COMMENTS  ON METHODS  OF  SIGNAL ANALYSIS
AND SIGNAL DETECTION1
H. WEINBERG
Brain Behaviour Laboratory, Simon Fraser University, Vancouver, BC, Canada
    It is well known that signal averaging of the EEC
is  the repetitive summing of event-locked  data. The
data usually vary in  voltage and  frequency over a
specified interval of time, either preceding or follow-
ing an event. An implicit assumption underlying use
of averaging techniques  is  that sums  of voltages at
fixed  intervals with respect  to the event  represent
those  voltage changes  that  occur  most reliably, i.e.,
consistently. However, it has always been  clear that
the sum (or average) of these voltage changes reveals
little or nothing about their  consistency. Indexes of
variability  must certainly be considered as a necessary
adjunct  to any statement  of reliability, and many
investigators do now include measures of variability
as routine qualifications of the average.

    The most common meaning of "variability" with
respect to  event related potentials (ERPs) is the varia-
bility of voltages occurring at the same instant in time
after or before repetitions of the "same" event. Varia-
bility  measures are generally used to  establish the
statistical significance  of mean changes in ERPs at
specific times after or before different events; differ-
ent events are usually, but not always, predefined
prior to the analysis. However, variability may also be
thought of as an independent variable.  Under  some
conditions, evoked potentials (EPs) are more variable.
Furthermore, variability may change  systematically
after or before events. When the event is a stimulus,
variability  of ERPs may be modified by simple repeti-
tion of the stimulus.  For example, the variability of
voltages within the intervals of N2-P2 of click-evoked
potentials seems always to be much less than the varia-
bility  of N3-P3, although P3 at the end of a contin-
gent negative variation (CNV) may be highly stable.
Another frequent observation is this: The variability
within a CNV interval  may be less near SI or near S2
than it is elsewhere. I suspect that the variability at
SI and S2 may be a function of the  amount and
nature of the information contained in these stimuli.
Patterns of variability within a CNV or EP interval
may be more significant than the mean amplitudes.
Fig. 1 shows systematic changes in variability during a
CNV interval in which a standard CNV paradigm was
used.  Fig. 1 also demonstrates that variability of a
CNV may also be differentially  distributed over the
scalp.  In  this  illustration,  variability is greater at F/
and Cz than it is at more posterior or temporal deriva-
tions.

    When considering variability of voltages occurring
at the same time after repetition  of the same event, it
is  clear that variability in the amplitude  of signals
may not be distinguishable from variability in the
latency of those signals if only  the mean  or sum is
known. This is  a fundamental limitation of averages
without concommitant indexes  of variability. This
limitation  is particularly important if the  presumed
function of averaging is to extract the real signal from
noise. This problem  leads  directly to  another  issue.
Are the characteristics of extracted signals  related to
the extraction procedures? Put  more strongly, the
question  is: Do extraction procedures in fact define
the signal?  Most people  consider extraction proce-
dures  to  be a method of separating the real signal
from noise. This problem is in fact the same encoun-
tered by all assessment procedures: Should  one set of
measures be used to validate another, and if not, what
may be used? Electrophysiologists are in a somewhat
better position than the inventors of paper and pencil
tests for they can apply different procedures to artifi-
cial signals of  known parameters. The  distinction
between  extraction  and  analytic procedres rests on
the assumption  that extraction procedures (like aver-
aging or filtering)do not constitute analysis of the sig-
nal. Fundamentally, extraction procedures cannot be
distinguished from analytic procedures. Consequently,
signal averaging can be thought of as an analytic tech-
nique in the same sense that factor analysis is consid-
ered analytic.
 The work described here was supported in part by
 the Medical Research Council of Canada.

-------
 594
                                                      Weinberg
                                     B
     EYE
     Cz
     T5
     A1
M  0,
             n
L
               1.5 sec
Fig.  1.  Changes in the standard deviation of an aver-
age CNV during  the CNV interval ± standard devia-
tions are plotted  above and below each average.
A =  trials 8-16, B = (rials 17-32 of a standard CNV
paradigm. M = markers: SI (tone) occurs at onset of
second mark (square wave); S2 (tone) occurs at onset
of third mark.  Mean reaction time and ± 1 standard
deviation  of reaction times are  shown as offset of
third mark.   All electrodes were  referenced  to   A2
(A1-A2 record between mastoids).   Eye electrode
is located on right supraorbital ridge. Note that for
all sites, with  the exception of A1-A2, variability
tends to  neck near SI, becoming  greater immedi-
ately prior to S2.  This change in variability during
the  CNV interval is more  characteristic  of  midline
than temporal electrodes.
                   A frequently used alternative to signal averaging
              is the simple method of superimposition of single
              observations or of averages. This is nothing more than
              an attempt to use the eye as a device  for estimating
              the  mean  and variability  of the  superimposed data.
              When averages are superimposed to display variability
              of the  total set of single trials  that enter into  the
              averages,  there  is an implicit assumption:  that  the
              same variability  would emerge with an equal number
              of any random selection of single trials for each aver-
              age,  an  assumption  that is probably  untrue more
              often than true. On the other hand, superimposition
              is certainly as reasonably a procedure as signal averag-
              ing,  and under some conditions may be superior. For
              example, if the activity concommitant  wih the signal,
              the  so-called noise, is discontinuous, superimposition
              allows  detection  of those  discontinuities, whereas
              averaging does not.
    The statement that signal averaging  fails to give
 specific information  about the influence  of each trial
 on the averaged signal implies more than may appear
 on (lie surface. One implication is that each individual
 (single trial) signal may  not be  an approximation to
 the real, i.e., significant,  signal. There may be no  real
 signal; there is no basis in neurophysiology for believ-
 ing that there is a  unique signal characterising  the
 effects of successive inputs.  There is good reason to
 believe  that each signal  is unique and  meaningful,
 that its difference from  other signals in  the set  ob-
 served does not simply constitute variability from the
 presumed real signal.  From this perspective, the as-
 sumption  that the time series recorded actually con-
 stitutes noise  superimposed  on a signal  comes into
 question. Put  another way, the use of signal averaging
 forces us  to assume  that (here  is indeed a signal to
 extract, that the raw data do not constitute in their
 entirety  the  signal  to be observed.  As an example,
 imagine the following: Following successive presenta-
 tions  of precisely the  same stimulus, the HRP  was a
 sine wave changing systematically in  frequency as a
 function of each new stimulus presentation.  Averag-
 ing these signals would give a new signal  that is, in a
very real sense, representative of the class of individ-
ual sine waves. The variability of each signal could be
constructed as variability of the "signal" extracted
 through averaging. There  is a basic question that must
be  faced:  Arc we justified in making the assumption
 that the signal of interest is imbedded in noise?  Hern-
ing (prc-circulated correspondence) makes this  point
succinctly by quoting M.A. Brazier who offers the ob-
servation that each event in the nervous system, in
response to  the same stimulus,  is different. This
salient fact is seldom emphasi/.cd. A  corollary of  this
is  the  argument  that  the same stimulus is  not  the
same  stimulus after  all,  since it occurs  at  different
times  from other instances of its occurrences, and is
therefore imposed on a different nervous  system. The

-------
Signal Analysis and Detection
                                            595
influence  of  the  stimulus  on  the  nervous system
would therefore define the stimulus!

    Although limitations on discriminability of the in-
fluence of single trials arc important, (here is perhaps
an even more important  constraint imposed by signal
averaging. Vaughan (1%9)  proposed the term event
related potential as a substitute  for evoked potential.
This term was adopted to describe  the constellation
of potentials that  included  CNV, P3, Nl, N2, readi-
ness potentials, evoked potentials, emitted potentials,
slow and infraslow potentials, and others. All shared
a common property:  they were  related to an event,
i.e.,  an input, an output, or a eoncomniitant change
in the environment or in the brain. The \ctmevoked
is  inadequate because it  does not  conceptually de-
scribe electrical changes like the  readiness potential
and  c'lnitted potentials. For example, what can be
said to evoke a readiness potential'?  The term related
in F.RP refers to a  temporal relationship.The focus on
temporal relationships was reinforced by the averag-
ing technique, for  the technique by its very nature al-
lows to emerge only  those signals  that have a con-
sistent temporal  relationship  to  a predefined event.
Signal averaging implies that the same electrical signal
does not occur at different latencies with respect to
the  predefined event. This  is   probably the  most
serious and  significant constraint of the technique.
It imposes a severe  limitation on the  evaluation ot
signals related  to  information  processing, which do
not result from immediate stimulus  input or precede
a response.
    Are  there alternatives to signal  averaging? It is
important  to again emphasi/e that extraction  and
analytic  techniques arc not conceptually distinct; it is
the process of analysis that defines the signal. Thus,
what  are  normally considered analytic techniques
could  be viable alternatives  to signal averaging. Mul-
tivariate  techniques like  factor analysis assume  that
each event  in the nervous system, in response to an
input, represents some variability  with  respect to the
real signal. Furthermore, multiple  interactions  are
very difficult to interpret.
     Fourier  analysis has been used  for  the analysis
 of averages and raw data. The  interpretation of fre-
 quency analysis can be ba/ardous since Fourier com-
 ponents may  result  from  signal modulation pro-
 ducing  artifactual  harmonics that  are difficult to dis-
 criminate from components  of the signal. Amplitude
 modulation  presents the same  difficulty.  However,
 the  major  limitation of frequency analysis is related
 to the  major limitation of  signal averaging,  it  gives
 very  little  information about when a signal is occur-
 ring. This is also true for signal  averaging when the
 signal is not  time-locked to an event. It is also true of
 frequency  analysis even when the signal is  time-lock-
ed to the event. For example, the analysis of a CNV
that  has  this form c=—would he the same as the one
that  had this  form«=3. A partial solution is to suc-
cessively sample epochs within the  record ot interest
and plot Fourier  transforms along a / axis represent-
ing time, i.e., to slide the epoch of analysis along the
lime domain that contains the signal of interest, in a
manner similar to that used by the recognition  index
(described  below). This  would allow the determina-
tion   of  precisely  when,  in  addition  to which,
frequency  components  within  the  lecord were shift-
ing,  whether  or  noi data  wore  lime-locked  to  an
event.  I am now  exploring this method  as a practical
technique. II an  index of when  the signal  should ap-
pear  is  available,  frequency   analysis  ot  selected
epochs could be  done for each trial, and  the Fourier
transforms  averaged. Theoretically this would lesult
iti no  more information than the Fourier  transforms
of the averaged  signal.  However, aveiaging  tends to
tiller high  frequencies  since they jitter more than
low  frequencies.  Consequently, a frequency analysis
of the average  would not reflect  the presence of high
frequencies to the same extent  as  the average  of the
single  Fourier  transforms. Digital filteiing techniques
that utili/.e Fourier  transforms, like the Wiener filter,
are  good for  discriminating  the   signal from noise
when  there is a reasonably precise estimate ot the fre-
quencies contained  in both  the signal and the noise.
A persistent problem with this technique, as pointed
out  by  Naitoh (this section) is  the  decremental et-
fccts on  high-frequency components when most ot
the  area of the signal of interest is comprised of low
frequencies. When  this  is true, the filter tends to see
high-frequency components as similar to noise.
     Development of pattern recognition  techniques
 will have great impact on our future abilities to estab-
 lish relationships between the l-T;.(i and complex be-
 haviour.  Any input that requires information pioc-
 essing must  result  in  changes throughout  the biain;
 there is  certainly much evidence to suggest this. It is
 necessary to describe  these  changes as well as to get
 an overall pieHire of the  relationships between them.
 It is particularly important to understand  how  these
 relationships  change  over  time  as  a  result  ot  ex-
 perience. Furthermore, it is extremely useful  to  be
 able to determine the degree to which .1 signal (within
 the brain) is like 01 unlike a predicted or  theoretical
 signal.

     A variety  of pattern  recognition techniques luue
 been used with the  Fli(i. I have been particularly in-
 terested  in the utility of correlational   techniques,
 such as  the recognition index (HI) illustrated  in Fig.
 2, since  they  have  the possibility of on-line applica-
 tion. Initial attempts wete to simply cross-correlate a
 predetermined  FFG pattern  (template) with the on-
 going F.HG (Fig. 2).This procedure  involves shifting
 the FEG against the template (or the template against

-------
  596
                                      Weinberg
 the EEC) a  predetermined number of increments of
 time after each  correlation.  The  correlation coef-
 ficient   (r=Exy/>/Ex2 Ey2) utilized  the  entire  dis-
 tribution of voltages within  the pattern as one vari-
 able. This did not work well since  the effect of con-
 currence in time of fast frequencies was obscured by
 slower  frequencies due to the  total greater  area oc-
 cupied  by the slower frequencies.
     The  next step  was to  divide the  template into
 segments  and correlate these segments separately,
 multiplying the correlations together for an index of
 the extent to which both of the segments in the tem-
 plate were present in the KEG (Fig. 3). This allowed a
 separation of segments  containing the fast-frequency
 and slow-frequency  components  in the EEC (partic-
 ularly the CNV).  This  procedure worked reasonably
 well; it  reduced  the number of false positives since
 a correlation coefficient of one of the segments of the
template that  was small  reduced  the resultant pro-
duct of the  coefficients. All  negative products were
ignored and plotted as zero.

    The  RI  template may be an epoch of EEC data
or a theoretical pattern. The advantage of this pro-
cedure is that a single-trial  epoch of EEC, when
designated as  the template, can be matched against
other  single-trial  epochs, giving  an  index  of the
goodness of  fit for any segment of the EEC. The pro-
cedure, therefore, does not require repetitive samples
of the same  signal for recognition, nor does it require
a temporal index of when the signal is occurring. A
disadvantage of this technique is that it gives no in-
dication  of  amplitude  differences of signals that
might have precisely the  same pattern. I have there-
fore been experimenting with  various methods  of
weighting the correlations  with  measures  of area
(fig. 4). I am  also attempting to  refine the template
definition by increasing the number of segments of a
                            A(X)B
                           C(X)D
  DATA
  TEMPLATES
                                                                JLdil
Fig. 2. Recognition index RI procedure using only two segments of the template. Note that when the segments of
the template are separated in time,  the RI is maximal (1.0) only when the data contain the same segments sepa-
rated by the same interval.  When the segments overlap, as in Cand D, recognition is maximal when both segments
overlap in a similar temporal relationship.  A, B,C, and D refer to template segments. Numbers refer to magnetic
tape blocks. Radiating lines enclose the segments of the data entering into the RI. Full scale ofRI = 1.0.

-------
Signal Analysis and Detection

template that can be independently manipulated for
determination of  the final recognition index. For ex-
ample, the evoked  potentials in  a  C'NV  may  be
much more "important" in some instances  than the
slow potential. Therefore, one may want to give them
more  weight  in  the determination of whether the
1--EG  changes observed  fit  the pal tern as  defined.
Since  Fourier transforms and correlational  analyses
are closely related, the same procedure could he used
                              0.25
                               R.I.
                                    1.5 sec
                                         597

with Fourier components, i.e., they may he itera-
tively computed as epochs of HFG are systematically
shifted  in time. However, one is again faced with the
problem of how to define the pattern!

    There are   at least two conceptually distinct
methods by which the pattern or template may be de-
fined.   It can be defined on the  basis of a decision
about what  one is  looking for, e.g., a CNV or a part-
                                                                 a,b
                        c.d
                      ^ff^^^
                      M^^^^
                                                             JJL
                   Vv^/^//^
                     ^lyA*^^
                    ^A^^^^
                    *W\^AwV^^
 Fig. 3. Use of four segments in template shown at top, without inclusion of area as a factor in the Rl. Note that
 the number of extraneous Rfs is reduced from that shown in Fig. 2 and that the first line show a trial in which
 the CNV is not obvious to visual inspection but is identified by the RL Segments are ab, cd, be, and ad. l-.ach Rl
 line refers to the trace above that line. Each EEG trace includes  one CNV trial beginning coincidental with the
 related maximal Rl on the line below each  trace. RIs shown on each line correspond to the point in time  at
 which the template is being matched to the data above that line.

-------
  598

  icular configuration of EP components. The template
  may also be defined as that pattern, whatever  it is,
  which dominates a given epoch of the  EEC. In the
  procedure I use, an epoch of EEC in which I am in-
  terested is selected and  the pattern  is  defined as a
  combination of subsets of that epoch. The template
  can consist of either a single trial or an average, or
  for that matter, some part of the spontaneous EEC.
  When selecting  the pattern in this manner, the prim-
  ary question  of interest  is the degree to which the
  pattern, as defined a priori,  is extant  in the  EEC
  searched. However, the converse question could be
                                     Weinberg

asked (it is related to the second conceptual method
for defining a template):  Are there  characteristic
patterns in the EEC  that are discoverable but not de-
fined a priori? To give an example of  the second
possible  method,  one could take an arbritrary epoch
of the  EEC and  do successive correlations of that
epoch with the spontaneous EEC. The purpose would
be to determine the goodness of fit at any particular
time within the spontaneous epoch. That epoch could
then be defined as a multiple or complex template by
dividing  it into two  epochs (corresponding to what
would be template subsets in a predefined pattern)
  ^f\fr^^^
                         Rl WITHOUT AREA,
                         FOUR TEMPLATE SEGMENTS
                         (ab, cd, be. ad)
                         Rl WITH AREA,
                         FOUR TEMPLATE SEGMENTS
                         (ab, cd, be, ad)
                                                                    0.25

                                                                    R.I.
                                                                              1.5 sec
                                                                                   C.  d
                         Rl WITHOUT AREA,
                         ONE TEMPLATE SEGMENT
                         (ad)
                                                                           TEMPLATE
                                           ilil
Fig 4  Inclusion of area as a factor in theRI. If At = area of template, Ad = area of data, and Af= area factor to
be multiplied times Rl, then forAt>Ad,Af=Ad I A t and for At
-------
Signal Analysis and Detection

completing successive correlations for each of the
template subsets. Resultant coefficients could be mul-
tiplied at each successive shift within the EEC search-
ed in the same manner in which it is done when the
template subsets are predefined. The purpose would
be  to  determine  the  point  in  time  within  the
spontaneous EEC when there is a best fit. Each of the
subsets of the original  template epoch could be sys-
tematically modified so as to  produce "all possible
patterns" that can  be defined with a given number of
subsets of the template epoch. For example, if there
were  two subsets, one subset could be  defined as 0%
of the epoch while the other is increased systematical-
ly from 0% to 100%. One of  the subsets would  be
                                                           599

                 defined as 5% of the epoch while the other  subset
                 is  systematically  increased  in increments of  5%,
                 followed by the initial subset defined as 10% and the
                 second subset  increased in increments of  5%,  etc.
                 Presumably,  the  Rls  that  indicate the  best  fit
                 throughout  the EEC record will  correspond  to the
                 dominant pattern. Unfortunately, this procedure re-
                 quires a tremendous amount 6f computer time, which
                 makes it currently impractical.
                     The RI, or some variant of it, may give informa-
                 tion  about  topographic relationships between simul-
                 taneously recorded potentials (Fig.5). For example,
   tfaffAjJ^
   ^^k//^^
                 a   b
c . d
                                                                   0.25


                                                                   R.I.
                                                                               1.5 sec
 Fig. 5. Distributional information from the RI procedure. Upper trace shows T5 and lower trace T6, coincident in
 time for the same trial, trial 6, of a standard CNV paradigm. The template is the average CNV from Cz. Note
 that T6 is more like Cz than is T5 (area factor is excluded). Using Cz as a standard, distributional differences
 in shape ofCNVs, with respect to the standard, may be described with the RJ technique.

-------
600
                                       Weinberg
one could  identify the  EEC from a single site (per-
haps  the vertex) as a "standard" against which ac-
tivity from all other sites is  compared. A question of
this sort could  then be asked:  To  what extent  is
activity recorded from the frontal cortex more similar
or less similar to vertex activity than it is to activity
recorded from the  temporal lobes? I  have done this
utilizing the averaged CNV recorded from the vertex
as the template and have shown that the frontal CNV
is more  different from  the vertex than are  CNVs
recorded from other sites.

    The  use  of pattern  recognition techniques ap-
plied  to EEC data, whether  they be variants of dis-
criminant  analysis, or the  Rl, or some other pro-
cedure,   identifies   limitations  common  to  all
measures. The   limitation  of all  methods is the in-
ability of these  methods to identify patterns within
the time domain without having  a priori knowledge
of exactly  when these patterns occurred (or are  to
occur). This limitation makes it virtually impossible
to study the  EEC related to thinking processes that
are not  time-locked to input  or output. Consequent-
ly, our understanding of the  relationship of ERPs to
complex human  thinking processes is so limited it can
only be  described as trivial. We must find methods to
go beyond  event-locked potentials if we  are to ever
have an  understanding of ERPs, and more important-
ly if we are ever to have an understanding of spon-
taneous  thinking processes.

-------
REFERENCES,  ALTERNATIVES TO  SIGNAL
AVERAGING
Anderson, T. W. The statistical analysis of time series.
    Wiley, New York, 1971.

Bland, B, H., Anderson, P. and Ganes, T. Two genera-
    tors ofhippocampal theta activity in rabbits. Brain
    Res., 1975,94:199-218.

Bloomfield, P.  Fourier analysis of time series: An in-
    troduction. Wiley, New York, 1976.

Buchsbaum,  M.  and Wender,  P.  Average  evoked
    response  in normal and minimally brain dysfunc-
    tioned children treated with amphetamine. Archs,
    gen. Psychiat, 1973, 29:764-770.

Burks, H. F.  The hyperkinetic  child. J. Exception.
    Child., 1960,27:18-26.

Burks, H.  F. Diagnostic implications of the electroen-
    cephalogram for behavior problem children. J.
    School Psychol., 1968,6:284-291.

 Burns, S. K. and Melzack, R. A method for analyzing
    variations  in  evoked responses. Electroenceph.
    clin. Neurophysiol.,  1966, 20:407-409.

 Capute, A. J.,  Niedermeyer, E. F. L. and Richardson,
    F. The electroencephalogram  in children with
    minimal brain dysfunction.  Pediatrics, 1968,41:
     1104-1114.

 Childers, D. G. and Pao, M. T. Complex demodulation
    for transient wavelet  detection and extraction.
     IEEE Trans. Audio  Electroacous., 1972, AU-20:
    295-308.

 Cohn, R. and  Nardini, J. The correlation of bilateral
    occipital slow activity  in the human EEC with
     certain disorders of behavior.  Am.  J. Psychiat.,
     1958, 115:44-54.

 Conners, C. K. Cortical visual evoked response in chil-
     dren  with learning  disorders. Psychophysiology,
     1970,7:418-428.

 Cooper,  R., Warren, W. J. and Pocock, P. V. ANEC-
    DOTE -  Advanced   NeuroElectric  Computer
    Data  Operational Tape (Export)  DECUS Pro-
    gram Library No. 12-144, 1973.
 Deecke, L, Scheid, P. and Kornhuber, H. H. Distribu-
     tion of readiness potential, pre-motion positivity,
     and motor potential of the human cerebral cortex
     preceding voluntary finger movements. Exp. Brain
     Res., 1969,7:158-168.
Dickman, K. SOUPAC program descriptions.  DCL
    Report  No.  UIUCDCS-R-72-3470-4.  Univ.  of
    Illinois, Urbana-Champaign, 1974.

Dixon, W. T. BMDP biomedical computer programs.
    Univ. of Calif., Los Angeles, 1975.

Donchin, E. A multivariate approach to the analysis
    of average evoked potentials. IEEE Trans. Bio-
    medical Engineering, 1966, 13:131-139.

Donchin, E. Data analysis  techniques  in average
    evoked potential  research.  In  E. Donchin and
    D. B.  Lindsley (Eds.), Average evoked poten-
    tials:   Methods, results,  evaluations.   NASA
    SP-191, U.S. Government Printing Office, Wash-
    ington, D.C., 1969; 199-236.

Donchin, E. Measurement in AEP  studies (opening
    remarks).  In W. C. McCallum  and J.  R.  Knott
    (Eds.),  The  responsive  brain.  Wright, Bristol,
    1976:5-9.

Donchin, E. and  Heffley, E.  Using minicomputers in
    the signal averaging laboratory. Am. Psychologist,
    1975,30:299-312.

Donchin, E. and Herning, R. I. A simulation study of
    the efficacy  of stepwise discriminant analysis in
    the  detection and comparison of event  related
    potentials.   Electroenceph. clin. Neurophysiol.,
     1975,38:51-68.

Donchin, E., Johnson, Jr., R., Herning, R. I. and  Kutas,
    M. Covariation of the magnitude of the CNV and
    P300 as a function of the subjects task. In W. C.
    McCallum and J. R. Knott (Eds.), The responsive
    brain. Wright, Bristol, 1976:76-80.

Donchin, E., Ritter, W. and McCallum, C. Cognitive
    psychophysiology: The endogenous components
    of the ERP. In E. Callaway, P. Tueting,  and S.
    Koslow (Eds.), Brain  event-related potentials in
    man. (in press).

 Donchin, E., Tueting, P.,  Ritter, W., Kutas, M. and
    Heffley, E. On the independence of the P300 and
    CNV components of the human averaged evoked
    potential.   Electroencph.  clin. Neurophysiol.,
     1975,38:449-461.

Doyle, D. J. Some comments on the use  of Wiener
     filtering for  the estimation of evoked potentials.
     Electroenceph. clin.Neurophysiol., 1975, 38:533-
     534.

-------
 602
                       Alternatives References
 Duncan-Johnson, C.  C. and Donchin, E. On quantify-
     ing surprise: The variation of event-related poten-
     tials with subjective probability. Psychophysiol-
     ogy, 1977, 14:456-467,

 Ellingson,  R. J. The incidence of EEC abnormality
     among patients with mental disorders of apparent-
     ly  nonorganic  origin:  A critical review. Am. J.
     Psychiat., 1954, 111:263-275.

 Gerbrandt, L. K. Multiplicity and overlap of move-
     ment potential components. Int. Symp. on Cere-
     bral Evoked Potentials in Man, Brussels, 1974.

 Gerbrandt, L. K., Goff, W. R. and Smith, D.  B.  Dis-
     tribution of the human average movement. Elec-
     troenceph. clin.  Neurophysiol., 1973, 34:461-
     474.

 Gerbrandt, L. K., Lawrence, J. C, Fowler, J.  R.  and
     Weyand, T. G. Multiple origins of the hippocamp-
     al  theta rhythms. Society for Neurosci., Abstr.
     No. 230, St. Louis, 1974.

 Gerbrandt, L. K., Weyand, T. G. and  Snider, R. M.
     Afferent control of hippocampal theta rhythms.
     Society for Neurosci., Abstr. No. 853, New York,
     1975.

 Glaser,  E.  M. and Ruchkin, D. S. Principles of neuro-
     biological signal  analysis. Academic Press,  New
     York,  1976.

 Green,  J.D., Maxwell, D. S., Schindler,  W.  J.  and
     Stumpf,  C.   Rabbit EEC "theta" rhythm:  Its
     anatomical  source  and  relation  to activity in
     single    neurons.   J.  Neurophysiol.,   1960,
     23:403-420.

 Grunewald-Zuberbier, E., Grunewald, G. and Rasche,
     A. Hyperactive behavior  and EEC arousal reac-
     tions in children. Electroenceph. clin. Neurophys-
    iol., 1975,38:149-159.

 Hall,  R.  A., Griffin, R.  D., Moyer, D. L, Hopkins,
     K. H. and Rappaport, M.  Evoked potential, stim-
     ulus intensity, and drug treatment in hyperkinesis.
     Psychophysiology, 1976,  13:405-418.

 Halliday, R.,  Rosenthal, J. H., Naylor,  H. and Calla-
     way, E. Averaged evoked potential predictors of
    clinical improvement  in hyperactive children
     treated with  methylphenidate:  An initial study
     and replication. Psychophysiology, 1976, 13:429-
    440.

 Harman, H. H. Modern factor analysis.  University of
    Chicago Press, Chicago, 1960.


Harmony, T., Otero, T., Ricardo, J. and  Fernandez,
    G.  Polarity  coincidence  correlation  coefficient
    and signal energy ratio of the ongoing EEC activi-
    ty.  I. Normative data. Brain Res., 1973a, 61:133-
     140.
Harmony, T., Ricardo, J., Otero, G., Fernandez, G.,
    Llorente, S. and Valdes, P. Symmetry of the vis-
    ual evoked potential in normal subjects. Electro-
    enceph. clin. NeurophysioL, 1973b, 35:237-240.

Jenkins, G. M. and Watts, D. G. Spectral analysis and
    its application. Holden-Day, San Francisco, 1969.
Jennrich, R.  I.  Stepwise  discriminant analysis.  In K.
    Enslein, A.  Ralston and H.S. Wilf, (Eds.). Statis-
    tical methods for digital  computers.  Wiley, New
    York, 1977:76-95.


John, E. R., Karmel, B. Z., Coming, W. C., Easton, P.,
    Brown,  D., Ann, H.,  John, M., Harmony,  T.,
    Prichep,  L, Toro, A., Gerson, I., Bartlett,  F.,
    Thatcher, R., Kaye, H., Valdes, P., and Schwartz,
    E., Neurometrics. Science, 1977, 196:1393-1410.


John,   E.R.,   Bartlett, F.,  Schimokochi, M. and
    Kleinman,  D.  Neural  readout  from memory. J.
    Neurophysiol., 1973,36:893-924.


John, E.  R., Karmel, B., Prichep, L. S., Ahn, H. and
    John, M. Neurometrics applied to the quantitative
    electrophysiological   measurement   of  organic
    brain dysfunction in  children.  In C.  Shagass, S.
    Gerson and A. J. Friedhoff (Eds.), Psychopathol-
    ogy and  brain dysfunction. Raven, New York,
    1977: 291-337.
John, E.  R., Ruchkin, D. S. and Vidal, J. J. Measure-
    ment of event-related potentials.  In E. Callaway,
    P. Tueting and S. Koslow (Eds.), Brain event-
    related potentials in man (in press).

John, E.  R. and Thatcher,  R. W. Functional neuro-
    science, Vol. II. Lawrence Erlbaum Associates,
    New Jersey,  1977.

Kaiser, H. F. The varimax criterion for analytic rota-
    tion  in factor analysis. Psychometrika,  1958, 23:
    187-200.

Klemm,  W.  R. Physiological and behavioral  signifi-
    cance  of  hippocampal  rhythmic slow  activity
    ("theta rhythm").Prog. NeurobioL, 1976,6:23-
    47.

Klinkerfuss, G,, Lange, P., Weinberg, W. and O'Leary,
    J.  EEG abnormalities of children with  hyperken-
    etic behavior. Neurology, 1965, 15:883-891.

Kutas, M., McCarthy, G. and Donchin, E. Augmenting
    mental chronometry: The P300 as a measure of
    stimulus  evaluation time. Science,  1977, 197:
    792-795.

Lachin, J. M. and Schachter, J. On stepwise discrimi-
    nant analysis applied to physiologic data. Psycho-
    physiology, 1974, 11:703-709.

-------
Alternatives References
                                           603
Lesevre, N., Rieger, H. and Remond, A. Le rythme
    alpha moyen et son aspect topographique: Defini-
    tion et interet. Rev. NeuroL, 1967, 117:130-136.

McCarthy, G. and Donchin, E. Brain potentials associ-
    ated with structural and functional visual match-
    ing. Neuropsychologia. (in press).

McCarthy, G. and Donchin, E. The effects of tempo-
    ral  and  event  uncertainty in  determining the
    waveforms  of  the auditory event related poten-
    tial (ERP). Psychophysiology, 1976,13:581-590.

Magnus, O. and Ponsen, L. The influence of the alpha
    rhythm in the cortical evoked response to photic
    stimulation. Electroenceph. clin. Neurophysiol,
    1965, 18:428.

Matousek,  M. and Peterson, I. Frequency analysis  of
    the EEC in normal children and adolescents.  In
    P. Kellaway and I.  Petersen (Eds.), Automation
    of  clinical  electroencephalography. Raven, New
    York,  1973:75-102.

Mulaik,  S.  A.  The foundation of factor analysis.
    McGraw-Hill, New York, 1972.

Nie, N.  H., Hull, C. H., Jenkins, J. G., Steinbrenner, K.
    and Bent, D. H. SPSS, statistical package for the
    social sciences. McGraw-Hill, New York, 1975.
 Nogawa, T., Katayama, K., Tabata, Y., Kawahara, T.
     and Ohshio, T. Visual evoked  potentials estimat-
     ed by  "Wiener filtering."  Electroenceph. clin.
     Neurophysiol., 1973a, 35:375-378.

 Nogawa,  T., Katayama, K.,  Tabata,  Y., Ohshio, T.
     and Kawahara, T. A numerical study on charac-
     teristics of the Wiener filtering for average evoked
     potentials. J.  Kansai Med. Sch.,  1973b, 25:849-
     S69.
 Nogawa, T., Katayama, K.,  Tabata,  Y., Ohshio, T.
     and Kawahara,  T. Wiener filtering and visual
     evoked potentials.  J.  Kansai  Med.  Sch.,  1973c,
     25-.S49-S69.

 Oppenheim, A. V. and Schafer, R. W.  Digital signal
     processing. Prentice-Hall, Englewood Cliffs, 1975.

 Otero, G., Harmony, T. and Ricardo, J. Polarity coin-
     cidence correlation coefficient and signal energy
     ratio of ongoing EEG activity. II. Brain tumor.
     Activitas Nerv. Sup., 1975a, 17:120-126.

 Otero, G., Harmony, T. and Ricardo, J. Polarity coin-
     cidence correlation coefficient and signal energy
     ratio of ongoing EEG activity. HI.  Cerebral vas-
     cular lesions.  Activitas Nerv.  Sup.,  1975b, 17:
     127-130.

 Pavy,  R. and Metcalfe, J. The abnormal EEG in child-
     hood communication and behavior abnormalities.
     Electroenceph.  clin.  Neurophysiol.,  1965,  19:
     414.
Peacock, Jr., S. M. Averaged "after-activity" and the
    alpha regeneration  cycle.  Electroenceph. clin.
    Neurophysiol., 1970,28:285-295.

Peters, J. F., Knott, J. R. and Hamilton, C. E. Further
    thoughts on measurement of 'the' CNV. In W. C.
    McCallum & J. R. Knott (eds.) The Responsive
    Brain, Wright, Bristol,  1976:15-19.

Pfustscheller, G. and Cooper, R.  Selective averaging
    of  the  intracerebral  click  evoked responses  in
    man: An improved method of measuring latencies
    and amplitudes. Electroenceph. clin. Neurophysi-
    ol., 1975,38:187-190.

Preston, M. S., Guthrie, J. T. and Childs, B. Visual
    evoked responses (VERs) in normal and disabled
    readers. Psychophysiology,  1974, 11:452-457.

Prichep, L.  S., Sutton,  S. and  Hakerem, G. Evoked
    potentials in hyperkinetic  and normal children
    under  certainty  and uncertainty: A placebo and
    methylphenidate study. Psychophysiology, 1976,
     13:441-451.

 Regan, D.  Evoked potentials in psychology, sensory
     physiology and clinical medicine.  Wiley,  New
     York,  1972.

 Remond, A. and Lesevre,  N. Variations in the average
     visual  evoked potential as a function of the alpha
     rhythm phase ("autostimulation"). Electroenceph
     clin. Neurophysiol., 1967, Suppl. 26:42-52.

 Remond, A., Lesevre, N,, Joseph, P., Rieger, H. and
     Lairy, G.  C. The alpha  average.  1. Methodology
     and description. Electroenceph.  clin. Neurophy-
     siol., 1969,26:245-265.

 Ritter, W., Simson, R. and Vaughan, Jr., H. G. Topo-
     graphic  analysis  of task-related cerebral poten-
     tials. Int. Symp. on Cerebral Evoked Potentials in
     Man, Brussels, 1974. (abstracts).

 Rosen, L, Silverman, N. R., Reinke, R. T. and Hellend,
     D, Videodensitometric analysis  of blood perfu-
     sion using Wiener filtering. Proc. San Diego Bio-
     med. Symp., 1975, 14:383-388.

 Ruchkin, D. S, Analysis of nonhomogeneous sequen-
     ces of  evoked  potentials.  Exp. Neurol., 1968,
     20:275-284.

 Ruchkin,  D. S. Latency characteristics and trial by
     trial variation of emitted  potentials. Int. Symp.
     on Cerebral Evoked Potentials in  Man,  1974.
     (abstracts).

 Ruchkin,  D. S. Sorting  of nonhomogeneous sets of
     evoked  potentials.  Comm.  Behav.  Biol.,  1971,
     5:383-396.

 Ruchkin, D. S. and Sutton, S. Visual evoked and emit-
     ted potentials  and  stimulus significance.  Bull.
     Psychon. Soc.,  1973, 2:144-146.

-------
604
                        Alternatives Reference!
Ruchkln, D. S. and Sutton, S. Latency characteristics
    and trial by trial variation of emitted potentials.
    In J. E. Deamedt (Ed.), Cognitive components in
    cerebral  event-related  potentials and  (elective
    attention. Progr. clin. NeurophysloL, Vol. 6.  (in
    press).

Saletu, B.,  Saletu, M. and Itil, T. M. The relationship
    between psychopathology and evoked responses
    before, during, and after psychotropic treatment.
    Biol. Psychiat.,  1973,6:45-71.

Saltzburg, B. Parameter selection and optimization in
    brain wave research. In N. Burch and H. I. Alt-
    shuler  (Eds), Behavior and brain electrical activi-
    ty. Plenum, New York,  1976:128-164.

Satterfield, J. H. EEC issues in children with minimal
    brain dysfunction. Seminars in Psychiat.,  1973,
    5:35-46.

Satterfield, J.  H., Cantwell, D.  P., Lesser, L I. and
    Podosin, R. L Physiological studies of the hyper-
    kinetic child I.  Am. J. Psychiat, 1972, 128: 418-
    1424.

Sayers, B. Me A., Beagley, H. A. and Henshall, W.  R.
    The mechanism of auditory evoked  EEC respon-
    ses. Nature, 1974, 247:481-483.

Sneath, P.H.A. and  Sokal, R.R.   Numerical taxon-
    omy; the principles  and practices of numerical
    classification.   W.H. Freeman,  San Francisco,
    1973.

Squires,  K. C. and  Donchin,  E.  Beyond averaging:
    The use of  discriminant  functions  to recognize
    event  related potentials elicited  by single  audi-
    tory  stimuli. Electroenceph.  clin. Neurophysiol.,
    1976,41:449-459.

Squires, K. C., Donchin, E., Herning, R. I. and McCar-
    thy,  G. On  the influence of task relevance and
    stimulus  probability on  event-related-potential
    components. Electroenceph. clin. Neurophysiol.,
    1977,42:1-14.

Squires, K. C., Wickens, C., Squires, N. K.  and Donchin
    E. The effect of stimulus sequence  on the wave-
    form  of  the  cortical event-related potential.
    Science, 1976,  193:1142-1146.

Tatsuoka, M. M. Multivariate analysis: Techniques  for
    educational and psychological research. Wiley,
    New York, 1971.

Tukey, J. W. Exploratory data analysis. Addison Wes-
    ley, Reading, Mass., 1977.

Ungan, P. and Basar, E. Comparison of Wiener filter-
    ing and selective averaging of evoked potentials.
    Electroenceph.   clin. Neurophysiol,, 1976, 40:
    516-520.
Vanderwolf,  C.  H., Kramis, R., Gillespie, L. A. and
    Bland, B. H. Hippocampal rhythmical slow activi-
    ty and neocortical low voltage fast activity: Rela-
    tions to behavior.  In K.  H.  Pribram and R. L
    Isaacson (Eds.), The hippocampus: A comprehen-
    sive treatise, Vol. II. Plenum, New York, 1975.

Vaughan,  Jr., H.  G.  The analysis of scalp-recorded
    brain potentials.  In R. P. Thompson and M. M.
    Patterson (Eds.), Bioelectric recording techniques,
    Part B, Electroencephalography and human brain
    potentials.  Academic, New York, 1974:157-207.

Vaughan, Jr., H. G., Costa, L. D. and Ritter, W. Topo-
    graphy of the human motor potential. Electroen-
    ceph. clin. Neurophysiol.,  1968, 25:1-10.

Walker, E., Bauer, Jr., J. A. and Grubin, M.  L. ALICE:
    A  system for manipulating and analyzing multi-
    dimensional data. Behav.  res. meth. instr.,1976,
    8:347-351.

Walter, D. 0. Aposteriori "Wiener filtering" of average
    evoked  responses.  Electroenceph.   clin.  Neuro-
    physiol., 1968, Supp. 27:59-70.

Walter, D. O. Analytical methodology since the EEC
    handbook.  In G.  Dolce  and H. Kunkel (Eds.),
    Computerized EEC  analysis.  Gustev  Fischer,
    Stuttgart, 1975:51-58.

Walter, D. 0. Digital processing of bioelectric phenom-
    ena.  Handbook  of Electroencephalography and
    clinical   Neurophysiology,  vol.  4b,  Elsevier,
    Amsterdam, 1972.

Walter, W. G. The contingent  negative variation: An
    electro-cortical sign of sensori-motor reflex associ-
    ation in man. In E. A. Asratyan (Ed.), Progress in
    brain  research, Vol.  22.  Elsevier,  Amsterdam,
    1968:364-377.
Walzer, S. and Wolff, P. H. (Eds), Cerebral dysfunction
    in children. Grune and Stratton, New York, 1973.

Weinberg, H. and Cooper, R. The Recognition Index:
    A pattern recognition technique for noisy signals.
    Electroenceph,  clin. Neurophysiol.,   1972,  33:
    608-613.


Wender,  P. H. Minimal brain dysfunction in children.
    Wiley Interscience, New York, 1971.

Wickler,  A.,  Dixon, J. F. and Parker, Jr.,  J. B. Brain
    function in problem children and controls:  Psy-
    chometric,  neurological  and electroencephalo-
    graphic  comparisons.  Am.  J. Psychiat.,  1970,
     127:634-645.

Wiener, N. Extrapolation, interpolation and smooth-
    ing of stationary time series. Wiley,  New York,
     1949.

-------
Alternatives References
                                            60S
Wiener, N. Extrapolation, interpolation, and smooth-
    ing of stationary time series.  MIT  Press, Cam-
    bridge, 1964.
Winson, J. Hippocampal theta rhythm. I. Depth pro-
    files in the curarized rat. Brain Res., 1976, 103:
    57-70.
Woody, C. D. Characterization of an adaptive filter
    for the analysis of variable latency neuroelectric
    signals. Med. Biol. Eng., 1967, 5:539-553.

Woody, C. D. and Brozek, G. Gross potential from
    facial nucleus of cat as an index of neural activity
    in response to glabella  tap. J. Neurophysiol.,
    1969,32:702-716.
Winson, J. Patterns  of hippocampal theta rhythm in
     the freely moving rat. Electroenceph. clin. Neuro-
     physiol., 1974,36:291-301.
 Woody, C. D. and Nahvi, M. Application of optimum
    linear filter theory to the detection  of cortical
    signals preceding facial movement  in cat. Hxp.
    Brain Res., 1973, 16:455-465.

-------
X. THEORETICAL APPROACHES AND MODELS
    Section Editor:

    W. Cheyne McCallum
    Burden Neurological Institute
    Bristol, England

-------
A  THEORETICAL FRAMEWORK  FOR EVOKED
POTENTIAL AND SLOW  POTENTIAL RESEARCH
W. C. McCALLUM

Burden Neurological Institute, Bristol, England

Contributors

   Dr. R. Cooper, Bristol, England
   Dr. L. Deecke, Ulm, German Federal Republic
   Dr. N. E. Loveless, Dundee, Scotland
   Dr. T. J. Marcy/nsky, Chicago, U.S.A.
   Dr. R. NBfftHnen, Helsinki, Finland
   Dr. D. Papakostopoulos, Bristol, England
   Dr. W. Ritter, New York, U.S.A.
   Dr. J. E. Skinner, Houston, Texas, U.S.A.
   Dr. H. Weinberg, Vancouver, Canada

 Introduction

   The chapter  which follows is based on several dis-
 tinct sources of information,  although it has often
 been necessary  to blur the distinctions to avoid over-
 lap and repetition. The first source was a correspond-
 ence conducted prior  to the  EPIC IV Congress  in
 which a number of the participants exchanged views
 on the necessity for a theoretical model, or models, in
 this field of research and the form that such a model
 might take.  The second  source was a series of meet-
 ings during the Congress by the subgroup concerned
 with this topic. This was followed by a plenary ses-
 sion on "Theoretical Models" at which members of
 the  subgroup presented their respective views and the
 matter  was  discussed  by Congress participants as a
 whole. Finally there was a series of written contribu-
 tions submitted either as  part of the prior correspond-
 ence or subsequently in support of various arguments
 advanced.

    To  provide  a coherent account of such diverse
 material is perhaps more than  can be expected. How-
 ever, as an  aid to  the reader, an attempt has been
 made to impose a certain minimum structure  on the
 arguments used. Some points have been summarized
 or paraphrased, but extensive use of direct quotations
 has been made where appropriate. Where contributors
 have, following the congress, reformulated their views
 and arguments  into a coherent text, these papers have
 been included  at the appropriate point  within this
 section. One paper, entitled "A biomodal slow poten-
 tial theory  of cerebral processing," originally pre-
pared by Cooper, McCallum and Papakostopoulos for
the Symposium on Cerebral Evoked Potentials held in
Brussels In 1974, was circulated prior to the Congress
as a  basis for discussion. This  is  included in full to-
gether with an account of the various comments, crit-
icisms and lines of argument which it generated.

   Because  of the range of disciplines involved, the
diversity of solutions proposed and the varied lengths
of contributions, the remainder  of  the  section has
been divided  into a  series of subsections outlined at
the end of this introduction.

The Need for a Model

   It may be asked why, after three international con-
gresses on ERPs have apparently  seen no necessity to
discuss the subject of "Theoretical  frameworks"  or
"models," there now existed a need for  such discus-
sion. Although  science  depends upon  the  steady
accumulation of the results of observation and experi-
ment, mere accumulation  is of limited value. What is
also necessary is insight to discern  relationships ex-
pressable as simple general laws. This is essentially no
more than a fundamental restatement of what science
is about. In the field  of slow potential  research the
gap  between the acquisition  and  interpretation  of
data has  steadily widened. For example, we have
spent some thirteen  years or more studying the CNV,
which was responsible for initiating this series of con-
 gresses. During  that time there has  been a steady
 accumulation of important facts, but at  the end of it
 we still have  a relatively poor  understanding of what
 this shift of  potential represents in terms of human
 function. Evoked potentials have an even longer his-
 tory of a similar kind and the Bereitschaftspotential
 and, to a slightly lesser extent, those potentials which
 accompany the execution of movements, stand  in
 much the same position.

    Largely as a result of this failure to raise our exper-
 imentation beyond  the phenomenological level, prac-
 tical applications have remained remarkably few  in
 number  for effects which were hailed as being among
 the most  significant  in human  neurophysiology  for

-------
  610

  some decades. It would  be  wrong to imply that no
  progress has been  made, but the return to be found
  from the enthusiastic pursuit of new phenomena and
  from accruing a  huge store of largely unrelated facts
  is rapidly diminishing.

     It  is both the strength and weakness of our com-
  mon field of research that it spans almost the whole
  of the neurosciences both  in the  disciplines  from
  which its investigators come and in the  directions in
  which relationships with the basic electrophysiotogi-
  cal phenomena are sought. The strength of a "tool"
  which  unites  researchers from the  fields  of  Physi-
  ology, Psychology, Anatomy, Neurology, Psychiatry,
  Biochemistry and Pharmacology is obvious. However,
  it has also the  disadvantage that it tends to encourage
  the  intensive pursuit of a variety of specialized rela-
  tionships between  the electrophysiological phenom-
  ena  and  such  diverse entities  as  cognitive states,
  anatomical   pathways, pathological processes and
  physiological mechanisms. This would in itself be no
  bad thing if it were not  for the rather poor level of
  interdisciplinary  communication that has  tended to
  exist so far. Fortunately EPIC IV itself has taken a
  large step towards crossing these communicative bar-
  riers. What is now required is an exercise in synthesis.

    To be effective it would seem that such a synthesis
  should lead to  the formulation of a theoretical frame-
  work, model (or models), which permit the integra-
  tion of our various data into a more general notion of
  brain  systems  and their  relationships to behaviour.
  The end-product should  have the  following charac-
  teristics:

    1. It should be capable of assimilating the existing
       body of scientific data.
    2. It should generate  further testable hypotheses.
    3. It should result in an increased understanding
       of the  relationships that exist, of  the systems
       and  mechanisms  which  these  relationships
       imply,  and  of the  features that differentiate
       these systems and mechanisms from others.

What Kind of a Model?

   Before attempting  to answer this  question  it is
necessary  to examine what we  mean  by the term
"model."  For some it  may imply several  pages of
flow diagrams, some may  prefer to work with  ana-
logs, some  may  think in terms  of behavioural black
boxes, some prefer to delineate physiological systems
and mechanisms while  others wish to express their
framework  in mathematical terms. Ultimately all are
attempts to impose a form upon that area of implicit
assumptions which lies behind the work of every re-
searcher. It is an area where fact shades into specula-
tion, with the result  that many of us are reluctant to
air such private  speculations in  public, being all too
                                       McCallum

 painfully aware of the gaps in both the evidence and
 the logic. However, it would seem important that we
 at least attempt to identify the common ground that
 exists in our diverse formulations. We ought also to
 be  assembling the currently  available evidence in a
 way which reveals where the major gaps in our struc-
 ture lie and perhaps enables us to state more clearly
 how the structure we are building differs from other
 structures.
    We  have  already touched upon the strengths and
 weaknesses  of  our interdisciplinary approach.  Our
 first problem becomes one of identifying our  com-
 mon ground and the  source of  the differences  that
 exist. What we have in common in this  field is a  basic
 methodology for  recording event related  electrical
 changes from  the brain, together with the body of
 electrophysiological data produced by  that method-
 ology.  Divergence  occurs because, in trying to  inte-
 grate this data  within a coherent theoretical frame-
 work,  we seem  to take a number of very different
 starting points.  While  this may be a healthy state of
 affairs  we should examine more  carefully the extent
 to which  it  facilitates or impedes our attempts at
 synthesis and perhaps  interferes with our attempts at
 communication.
   Analysis of preliminary correspondence and  dis-
 cussion, and indeed of the literature and general con-
 gress contributions,  reveals  that  there  are at least
 three distinctly different starting points.  In the re-
 mainder of this chapter we will endeavor  to outline
 these, to examine  the extent to which they can be
 reconciled with one another and then to turn to the
 data  itself and  to  take a critical look at what con-
 straints it imposes on our framework.
   "The  Behavioral  Approach"  described  by  Dr.
 Loveless provides a  starting  point.  From the per-
 spective of experimental  psychology,  Dr. Loveless
 perceives  the basic  objective as one of  integrating
 event related potential data into existing behavioural
 theory  and  relating  it  to  existing psychological
 constructs.
   The  second  approach  stems  from  researchers
whose anchors are in physiology and chemistry, rep-
resented in this section by Drs. Skinner and Marczyn-
ski,  respectively. The basic objectives are to  build
bridges  between molar  and molecular processes-to
approach macropotentials and their implications from
the unit level of activity, from functional  anatomy,
biochemistry, and neuropharmacology.

-------
Theoretical Framework
                                             611
   The third approach takes macropotentials  them-
selves as a startingpoint and looks both outwards and
inwards. This approach, as described by Dr. Papakos-
topoulos, seeks to understand the data in relation to
existing knowledge about neuro-anatomy and cellular
neurophysiology, while at  the same time  it tries to
relate the whole to systems, mechanisms and, ulti-
mately, to the direction of behaviour.

   In the last analysis the building  blocks  of our
model must  be macropotential data. We must identify
the  key  characteristics of  the data that need to be
accounted for by any theory that  is  to emerge as a
useful predictor of behaviour and a generator of test-
able hypotheses. A crude chronological division has
been chosen to deal with the phenomena themselves.
Thus the period of time which precedes voluntary
actions is considered first.  Dr. Deecke discusses the
Bereitschaftspotential or Readiness  Potential that is
associated  with this  preparatory period, and  the
slightly later group of potentials  which occur when
movement is executed. Dr. Ritter then examines the
relationship between ERPs and  consciousness, with
particular reference to those evoked potential com-
ponents which  follow the onset of an external signal
and extend over a period of three  or  four hundred
milliseconds. This is followed by  a look beyond 400
msec  by  Dr. Naatanen to  the characteristics of the
slower and sustained potential changes such as the
CNV, and some general comments  by Dr. Cooper.
The  "Bimodal  model,"  which  was primarily  -  al-
though not exclusively • based on SP data, is included
at this point together with its discussion. The section
ends with a discussion of the major issues raised.

-------
 THE BEHAVIORAL  APPROACH


 N. E. LOVELESS

 Department of Psychology, The University, Dundee, Scotland
    I share a feeling of disappointment at the relative
 absence  of theoretical discussion at recent confer-
 ences, particularly about the relationship of the CNV
 to those psychological variables to which it appears to
 be  sensitive.  This is perhaps a reaction against early
 speculation  about the "psychological correlates" of
 the  CNV, but seems to have  reached a point where
 people are unwilling to say anything at all. There is a
 danger of being too solemn about  theory. Even the
 quest for psychological correlates may have been use-
 ful in suggesting variables worth investigating. I do
 however agree with  Donchin et al. (1972)  that this
 type of theoretical interpretation, if it deserves to be
 called  theory at  all, is not likely to take us far in
 determining the functions of  ERSPs. Concepts such
 as "attention" are not only ill-defined, but also of too
 high an order.  Luria (1973), discussing the localisa-
 tion of function in the brain, remarked on the ambi-
 guity of the term "function", which may be used to
 refer either to the elementary function of a particular
 tissue, or to  a  complex  process  such  as digestion
 which  involves a number  of component systems. It
 seems equally foolish to attempt to identify a com-
 plex process  such as "attention" with a  bioelectric
 response of a  limited brain area.

   One of the factors  impeding progress in  the con-
struction  of "theoretical  models"  in  this  field
emerged   in  our  preliminary  correspondence. We
 found that we were  to some extent talking  at cross-
purposes.  We were talking about different approaches
to constructing them.  For example, I  would  epito-
mize the differences between McCallum and myself as
being that my  approach  is  one  of  physiological
psychology whereas his is one of psychological physi-
ology. The latter approach tends to take a physiologi-
cal phenomenon as primary, to set up general hypoth-
eses about its function  and then seek to test  these by
devising suitable behavioral situations.

   It is, of course, an entirely respectable strategy to
devise fresh paradigms  when  faced with a new phe-
nomenon. However, I think that we should heed what
Sutton said some years ago, namely that  pioneers in
 ERSP research tend to behave like explorers facing an
 uncharted continent. The point is  that although it is
 to be expected that a new phenomenon will excite
 fresh speculation, it  is necessary  and proper to at-
 tempt to  relate  it to what  we already  know.  The
 appropriate  level of  'psychologizing' is  surely, as
 Vaughan (1969) has suggested, to attempt to specify
 those component processes which  contribute to per-
 formance in  the task used to generate ERSPs. In this
 respect CNV research has unfortunately proceeded in
 virtual  isolation  from relevant areas of psychology.
 When  I  first entered this field, I was  astonished to
 find that although the standard CNV  paradigm was
 that of warned RT, virtually no use was being made
 of previous research on this task. It would be absurd
 to pretend that 'knowledge' in this  field amounts to a
 body of well-tested theory, but equally absurd to pro-
 ceed as if the  efforts of psychologists over  the past
 hundred years  have taught  us nothing. The RT litera-
 ture is voluminous and of very  uneven  quality, but it
 contains at least some well-tested  generalizations -
 for example, the effects of foreperiod manipulation
 which I have replicated in some of my  own research.
 While  there  is  no existing theory  which adequately
 covers all  the established facts about reaction  tasks
 (Broadbent 1971), disagreements may cover  substan-
 tial areas of agreement, and available models  have
 considerable  heuristic value.  Yet with  a few honor-
 able exceptions, CNV researchers proceed as if such
 models  do not exist. I have  found it  fruitful simply to
 replicate some well established behavioral phenomena
 and to see what happens to ERPs when  one does this.
 I  sense  from  the exciting accounts  of sequential
 effects reported elsewhere in this volume by Squires
 and Squires that others too are  following this course.

   The  situation  is not much  better  in  relation to
 classical  conditioning. Early reports on  the CNV
 made considerable use of conditioning concepts, but
 this practice  seems to have  fallen  away.  I  wonder
whether this may be due to a  mistaken interpretation
of the  relationship between conditioning  and  RT
paradigms, For instance, Rebert and  Knott  (1970)
note that  the  two paradigms are  alike in  that the

-------
The Behavioral Approach
                                             613
conditional  stimulus comes to have cue significance
similar to that of the RT warning signal, but consider
them different in that "during conditioning, the con-
ditioned response comes to precede  the unconditional
stimulus, i.e., the animal does not inhibit the required
response until onset of the second stimulus, as is true
in the reaction  time experiment". This appears  to
involve  confusion of  the  conditional and uncondi-
tional  responses. In  the  Pavlovian situation,  some
components  of  the  unconditional  response,  such  as
chewing, do  not precede  the second stimulus. Other
components, such as  salivation,  appear also  in the
conditional  response, and are  most  clearly character-
ized as  anticipatory or preparatory. In the RT situa-
tion, therefore,  the  analogue of the  conditional re-
sponse is not the reaction to the second stimulus, but
the  development of "preparatory  set" during the
foreperiod.
    I do not altogether agree with Donchin's strictures
 on the mixture of RT and conditioning terminology,
 since this may facilitate drawing upon two somewhat
 distinctive research traditions. We not only stand to
 lose whatever positive gains have come from the stud-
 y of conditioning, but risk falling into the numerous
 methodological pitfalls which have been discovered in
 that field. I am thinking particularly of our addiction
 to short inter-stimulus intervals. It may be true that
 in classical conditioning a short interval is 'optimal' in
 the sense  that it produces the most rapid condition-
 ing, but  the  short interval may not be optimal for
 purposes  of investigation. The advantages of delayed
 conditioning procedures were recognized long ago in
 Pavlov's laboratory, not least that changes due to con-
 ditioning  can be  studied without the necessity  of in-
 troducing  test trials on which the unconditional stim-
 ulus is  omitted.  This practice  was not adopted by
 Western investigators, who  have only recently come
 to realize  that the use of short ISIs  has  led to the
 confusion of several responses. Thus in eyeblink con-
 ditioning,  the long-latency closure  that  develops in
 anticipation of the UCS  must be distinguished, not
 only from voluntary blinks,  but also  from  short-
 latency responses to  the CS. These  distinctions are
 difficult to make when the ISI  is short.  Similarly, in
 autonomic and EEC  conditioning, a long-latency an-
 ticipatory response can be clearly differentiated from
 the short-latency orienting response only  when  a
 delayed conditioning procedure is used.  It is very evi-
 dent from the summary by Martin and Levey (1969)
 that this lesson has been painfully learned. Is it hubris
 or historical ignorance which leads us to suppose that
 the CNV paradigm will be exempt?  Results obtained
 by the  handful of investigators  who have used long
 ISIs suggest that  it is not. Short ISIs are  undoubtedly
 more convenient to use,  but are likely to produce a
 mix of potentials. It is  then hardly  surprising that
 terminal CNV amplitude and RT are sometimes disso-
ciated, or that it is difficult to determine CNV  func-
tion.

   Martin and Levey point out that since  the orient-
ing response, anticipatory  response, and response to
the UCS all change in the course of conditioning, it is
rather illogical to refer  to any one of them as the
conditional response. I wonder whether we have not
reached the point  where  referring to any negative po-
tential that develops during the ISI as 'the CNV is
not a positive impediment. I have in mind, for exam-
ple, the question whether the CNV is contingent on a
motor response.  I do  not doubt  that 'non-motor
tasks' produce some negative potential, but I suspect
that in  many cases it may be an orienting potential
because  the waveform  often  has a  convex  dome
shape, with no  terminal rise.  In my own research  I
have been impressed by cases in which Niave/a/fed to
observe a  clear-cut anticipatory  response, i.e., in a
signal-detection task with delayed verbal response and
in an RT  task performed  under  'sensory set',  when
muscular preparation was discouraged. Under long ISI
conditions, I have seen  a good anticipatory response
only when a vigorous motor response was required. I
am also  impressed by the  relationship between CNV
and anticipated force of movement.

   A  clear distinction   between  'motor' and 'non
-motor' tasks may,  of course, be difficult to  draw.
Nearly all tasks have some  motor component, at least
in the form of postural  adjustment or change in mus-
cle-tension. However, the  distinction  is not trivial:
one of our outstanding failures is the unresolved rela-
tionship between the CNV and Bereitschaftspotential.
McAdam et al. (1969) first pointed out the morpho-
logical similarity between the  readiness potential and
the anticipatory  potential recorded under long  ISI
conditions. Subsequent research suggests that the two
potentials are functionally  related—i.e., both are con-
cerned with motor preparation, and are similarly af-
fected by  incentive.  The  outstanding difficulty in
sorting out their relationship has been the difference
in scalp distribution. If one assumes that the conven-
tional CNV is the  sum of a Bereitschaftspotential and
a negative  frontal  wave, the difference in midline to-
pography is easily  accounted for (Gaillard 1976). The
difference  in laterality  was   more  puzzling  until
Syndulko and Lindsley (1977) suggested that it might
result from insufficient care to limit response to one
muscle group in CNV experiments. This  led to the
suggestion  that the topographical problem might be
resolved  if the  Syndulko  and Lindsley experiment
were repeated with a longer ISI. The elegant experi-
ment  by  Rohrbaugh, Syndulko and Lindsley (1976),
in fact, provides a very effective resolution.  I now
take  it as established that  the  CNV consists  of two
components which are functionally, morphologically
and topographically  distinct, and that the second of
these « a Bereitschaftspotential.

-------
 614
                                          Loveless
    In the belief that questions about the function of
 ERSPs are best approached through a process analysis
 of the CNV paradigm, I would like  to  present an
 attempt to do this.
 A  Tentative  Process Analysis  of  the  Basic
 CNV Paradigm

   While it  seems sensible to distinguish evoked po-
 tentials and  ERSPs, an adequate analysis of the CNV
 paradigm must include the functions of both evoked
 and motor  potentials. For the sake of argument, I
 assume that  the  Picton and Hillyard (1974) analysis
 of the auditory  EP is broadly true. This two-stage
 model first  posits a 'compulsory' feature analysis of
 sensory input. This first stage may include  'novelty
 detectors' that generate an  output when any feature
 changes. On  the  basis of this initial analysis, signals
 may be selected for further processing by comparison
 of current input with  memory models of expected or
 significant stimuli. In the present instance, therefore,
 these mechanisms only generate an output when an
 event is  differentiated from background activity  and
 identified as  the warning signal.

   This identification must be followed by the selec-
 tion of appropriate action, which presumably involves
 the establishment of routines for recognizing and re-
 sponding efficiently  to the imperative signal. Efficient
 response will depend  upon  the use of information
 about  the probable time of occurrence of the impera-
 tive  signal. The formation  of complex  programs of
 activity is characteristic of prefrontal cortex. This  'ex-
 ecutive'  function is, therefore, probably  associated
 with the slow negative potential which follows SI and
 appears  to   be  frontally-dominant.  This potential
 peaks in about one  second  and then decays over a-
 bout five seconds - a time-course similar to  that for
 changes in heart rate and  alpha desynchronization
 (commonly  regarded as components of the orienting
 response). Identification of the slow potential as such
 a component is supported by its appearance as a re-
 sponse to unpredictable changes in nonsignal stimuli
 and by its rather slow habituation to signal stimuli.
 The orienting response is reputed to be a frontal lobe
 phenomenon.

   An important function of the orienting response is
 to increase the probability  of detecting subsequent
 stimuli. This function appears to  be exercised through
 the control of sensitivity, which follows a timecourse
 similar to that of the  slow  potential, i.e.,  increasing
shortly after the warning signal and then declining
over a few seconds. The amplitude  of  the potential
appears to depend on the intensity and modality of
 the warning signal.  Since variation of sensitivity will
alter the rate at  which evidence  of the imperative
signal will become available in reaction tasks, it seems
 likely that this 'sensitizing'  mechanism  mediates  the
effects upon  RT of warning signal characteristics and
 of foreperiod duration over the first few seconds fol-
 lowing the warning signal.

   The 'orienting potential' is distinguished from the
 rising negative potential that develops in anticipation
 of the imperative signal when the  subject has some
 certainty about  its  time of occurrence. The most ob-
 vious illustration of the effect of temporal uncertain-
 ty is the development of a marked anticipatory po-
 tential when the foreperiod is of constant rather than
 variable duration (cf. Type A and B CNVs, McAdam
 1969). Such a potential may also be seen with a varia-
 ble-foreperiod paradigm when presentation of the im-
 perative  signal  is  probable, notably at the  longest
 foreperiod   being used. When the  foreperiod  varies
 over a short  range, the potential rises  steadily over
 this range  to  reflect the increasing  conditional prob-
 ability that response will be required. When the time
 of the imperative signal is known,  the  amplitude of
 the  potential is affected  by instructions  stressing
 speed of response through  muscular preparation. The
 anticipatory potential therefore  occurs when the sub-
 ject uses information about the  probable time of im-
 perative signal occurrence  to  produce  a brisk motor
 response. There  seems to be general theoretical agree-
 ment that  RT reflects the time necessary  for unrelia-
 ble evidence of the  occurrence of an imperative signal
 to accumulate to a  point where  a particular response
 can be selected with an acceptably  low  rate of error.
 The further time taken to  launch the response is pre-
 sumably minimized by lowering a response threshold.
 This process is presumed to be  the functional corre-
 late of the  anticipatory potential.

   Several  authors  (cf. Rebert and Tecce  1973) have
 argued against such an interpretation on the grounds
 of dissociation between  RT and CNV amplitude. A-
 part from the ambiguity of'CNV amplitude', no one
 has ever claimed that RT is determined only by effec-
 tive preparation. The observed distribution of RTs is
 generally agreed to  be a convolution of different dis-
 tributions  reflecting the different processes involved
 in the  task. Several authors (e.g., Hohle  1965) explic-
 itly analyze the distribution into components repre-
 senting preparatory set and decision time. Dissocia-
 tion has  commonly been observed  with a short con-
 stant  foreperiod  wherein temporal  certainty is high
 relative to  the  time  for which  preparation  can  be
 maintained. Under  these conditions, a practiced sub-
ject is likely to produce a fairly stable level of prepa-
 ration  and  RT  is likely to be  appreciably affected
 only on a few trials when the  subject anticipates or is
 unready. This does not mean that preparation is un-
 important  - how  else do  we explain  the persistent
 difference  between  warned and  unwarned RT? What
 it  surely means is  that under these conditions the
 proportion of trial-to-trial RT variance attributable to
 fluctuations in preparation is small  relative to  the ef-
 fect of variation in decision time.

-------
The Behavioral Approach

   This argument might be supported by evidence of
substantial correlation between RT and the latency of
late  EP components  if it  could  be confirmed that
some of  these components precede the initiation of
motor response, as Vaughan  and Ritter (1973) have
claimed.  (It might then be interesting to look at the
multiple  correlation of these latencies and anticipato-
ry potential amplitude with RT). None the less, some
components which may be related to stimulus identi-
fication occur after the subject has responded. Prac-
ticed subjects commonly report that they have react-
ed before they are aware of the stimulus. In a simple
RT task, the subject may achieve speed by reacting to
any  appreciable change in input  (subjects  trying to
produce  a fast  response  to a visual stimulus  readily
respond to an accidental noise) — i.e., the subject re-
sponds on the basis of fewer features than he needs
for  confident  identification  of the  stimulus.  In
choice RT, this finding may be related to the distinc-
tion between "fast guesses" and true discriminations
(Oilman   1966). Such  heterogeneity  of  response
would help to  explain the dominance  of decision-
time.

   I put forward this analysis more as an indication of
the sort  of model  I  would like  to see  than in the
belief that it is adequate. If this model seems too
much of a "personal ivory tower," the reason is that
the model is limited to effects obtained with long
foreperiods, although there are a  number of obvious
links with results obtained with the  conventional par-
adigm. There is a good deal of evidence  supporting a
distinction between frontal  and central potentials,
and between 'early' and 'late' CNV.  Thus, an obvious
prediction is that amphetamine affects the orienting
potential  (it would be interesting  to use non-signal
stimuli).  The 'early CNV' also seems to be particular-
ly vulnerable to distraction (see Loveless, this vol-
ume). On the other hand, distraction does not appear
                                            615

to affect RT, though the maintenance of performance
under distraction may require increased  effort. 'Ef-
fort' also  seems to be  implicated in the effect  of
sleep  deprivation on  the  'late CNV.  Pribram and
McGuiness (1975), whose  distinction between 'arous-
al'  and 'activation' seems related to  that between
orienting and anticipatory effects, have felt it neces-
sary to involve effort as a third variable. Does anyone
have a correlate for effort?

   Finally, little systematic study of individual differ-
ences in the CNV  has  been made,  despite a wide-
spread impression that there are characteristic differ-
ences in its form.  Donchin's essay in factor analysis
(1972)  tends  to confirm the distinction between
'Type A' and 'Type B' CNVs. This morphological dis-
tinction may depend,  in part at least, on the relative
amplitude  of orienting  and anticipatory  potentials,
and might  form a basis for theoretical interpretation,
perhaps related to  the work of Soviet factor analysts
(Nebilytsin 1972) who make considerable use of con-
cepts drawn from conditioning theory.

    In summary, I seek a model of what is going on in
an  individual's head when he is performing in stand-
ard behavioral paradigms. To construct such a model,
we must combine  what we know about all the event-
related potentials that occur in the task situation. We
have been urged to move  on from mere phenomenol-
ogy. The  time has come to move away from research
subgroups  clustering around discrete  physiological
phenomena and failing to communicate with one an-
other. The work based on information processing pre-
sented at  the  congress and  some of the recent re-
search on slow potentials suggest that we may now be
in  a position to give a coherent  account of at least
 one standard task.  Given that this sort of integrative
model building is possible, I would subscribe to effort
being  invested in building models of a few well cho-
sen tasks.

-------
 A NEUROPHYSIOLOGICAL MODEL  FOR REGULATION
 OF  SENSORY INPUT TO CEREBRAL CORTEX1

 I.E. SKINNER
 Neurophysiology Section, Neurology Department and Neuroscience Program, Baylor College of
 Medicine, Houston, TX, U.S.A.
   Desynchronization of the EEC, enhancement  of
 sensory evoked responses  and negative or positive
 shifts in slow potentials are electrical events in the
 human cortex that have been associated with arousal
 and  selective attention. These three types of electro-
 cortical events have been shown to be regulated  in
 animals by both the mesencephalic  reticular forma-
 tion  (MRF)  (Rossi and Zanchetti  1957) and the
 mediothalamic-frontocortical (MT-FC) system
 (Skinner and Lindsley 1973). The MRF and FC are
 known to converge upon the thalamic reticular nucle-
 us (R) (Scheibel and Scheibel 1966,1967a;Millhouse
 1969), a perithalamic structure, which presumably in-
 tegrates the  convergent inputs and  then  projects
 systematically to  the interior of thalamus  (Scheibel
 and  Scheibel 1966, Jones 1975). Evidence, reviewed
 below, suggests that  FC and MRF influences on  R
 regulate various inhibitory gates on the thalamic relay
 nuclei and, by this mechanism, continuously control
 the pattern of sensory input to cerebral cortex. I pro-
 pose that this neural mechanism, identified in the cat
 brain, mediates the processes of both general arousal
 and  selective  attention  because neurophysiological
 manipulation of the system not only alters  the three
 types of electric activities associated with these pro-
 cesses, but  also  effects the expected behavioral
 changes.

   EEC  desynchronization, modulation of slow po-
 tentials, and enhancement of sensory evoked poten-
 tials  are discussed  in the next three sections, respec-
 tively. In each case, the model will be used to explain
 the regulation  of the electrical activity, Implications
 of the model for the regulation of human-event-relat-
 ed potentials will be discussed in the final section.

 EEC Desynchronization: Reduction in Phasic
 Thalamic Inhibition

 Mechanisms of EEC Synchronization

   To understand the process of EEC desynchrontza-
tion, one must understand how synchronous activity
 Supported  by  a grant from the  National Institutes of
 Health, HL-17907.
itself occurs. Recruiting and augmenting responses
produced by 8- to 12-c/sec medial and lateral thala-
mic  stimulation, respectively, have been used as ex-
perimental  models for EEC cortical  rhythms  since
discovery by Morison  and  Dempsey  in 1942.  Both
electrically  induced responses show  systematic  in-
creases and decreases  in amplitude with  successive
stimulation in a low-frequency train, and thereby pro-
duce the fusiform  or spindle shape of spontaneously
occurring EEC  rhythms characteristically recorded
from the cortex of a quiescent or inattentive animal.
Recruiting  responses produced by medial thalamic
stimulation  were thought to be the better  model for
spontaneously occurring EEC spindles for they inter-
acted with natural waves in such a way as  to suggest
that  both rhythms shared common pathways  (Demp-
sey  and  Morison  1942).  Also, both recruiting  re-
sponses and spontaneous spindles had  primarily fron-
tal distributions in the cortex, whereas augmenting
responses did not (Morison and Dempsey 1943). Aug-
menting responses  produced by lateral thalamic stim-
ulation (i.e., sensory and  motor relay nuclei) were
found to be independent of frontally projecting spon-
taneous EEC rhythms  and later were  shown  to have
profiles of  laminar activity  in the cortex different
from those of recruiting  responses  (Spencer and
Brookhart 1961).

   Schlag and Villablanca (1967) showed that wave-
form could not serve  as  a  criterion  to  distinguish
between  recruiting and augmenting  responses. Re-
cruiting responses elicited by medial thalamic stimula-
tion  were found to have positive-negative waveforms
in the hidden region of the orbital cortex in cats, and
augmenting responses  evoked by lateral thalamic
pulses manifested their positive-negative responses in
the more accessible cortical convexity. These  authors
also pointed out that positive-negative cortical projec-
tion  fields  of  both recruiting and augmenting  re-
sponses were surrounded by large zones in which only
monophasic negative responses were recorded.

  The  best criteria for distinguishing between  re-
cruiting and  augmenting  responses  are  observed

-------
A Neurophysiological  Model

effects on behavior. Frontally project ing spontaneous
EEC rhythms are invariably associated with inatten-
tive behavior. Electric 8-c/sec stimulation of the  me-
dial thalamus completely disrupts bar-pressing behav-
ior and renders the  animal inactive. At lower intensi-
ties,  such stimuli produce a deficit in single alterna-
tion  performance (King and Skinner 1975).  Stimuli
that elicit augmenting responses have, in contrast, no
detectable effect on behavior. It is not surprising that
medial and lateral thalamic stimuli have different be-
havioral effects because the medial thalamus receives
intrinsic  inputs  from other  thalamic nuclei  and the
lateral thalamic nuclei  receive  extrinsic  projections
from  the periphery and  other  sources outside  the
thalamus. Thus  recruiting and augmenting responses
appear to be mediated  by anatomically distinct  sys-
tems that subserve different behavioral functions.  It is
important to realize that  different mechanisms exist
for different types  of cortical EEC rhythms because
false  controversies  have arisen  when EEC  synchro-
nous  activities were posited to reflect a unitary pro-
cess (see Skinner and Yingling 1977; Skinner I971a).

Role  of  the  Reticnlar Nucleus  in  Phasic
Thalamic Inhibition

   Moruzzi  and Magoun  (1949) demonstrated  that
electric  stimulation of  MRF  would  produce
behavioral arousal,  desynchronization of spontaneous
EEC, and abolition of recruiting responses elicited by
medial  thalamic stimulation.  Skinner and  Lindsley
(1967) showed that blockade of the MT-FC system at
any point would also abolish recruiting responses and
spontaneous spindles (produced  by MRF lesion) in
the thalamus us well as in the cortex. This latter result
suggests  that an extrathalamic circuit is necessary for
the geneiation of these forms of thalamic synchroni-
zation. Villablanca  and  Schlag (1968) confirmed that
blockade   of  the  pathway  between  the  medial
 thalamus and frontal granular cortex abolished spon-
 taneous  thalamic spindles in the partially  decorticate
 isole' preparation, They also found that total thalamic
 isolation resulted in a  modified  recurrence of thala-
 mic synchronization. This suggests elements entirely
within   the  thalamus  can  support  synchronous
 rhythms. It would be prudent not to identify pre-
 isolation thalamic spindles with  those recorded in the
 isolated thalamus  because (1) only the criterion of
 waveform  is employed and (2) extrathalamic  circuits
 are clearly  required in  the preisolation condition and
 not in the  latter.  How then is  one to distinguish a-
 mong the various forms of thalamic synchronization?

   Feedback from  inhibitory interneurons  upon the
  thalamocortical  cells  may be  required  to produce
  neocortical rhythms.  This hypothesis is based on  ob-
  servations (Purpura and Cohen 1962, Andersen and
  Eccles  1962)  during  synchronization  that  long-
  duration  inhibitory  postsynaptic  potentials occur
  between action potential bursting intervals of thala-
                                            617

mocortical  neurons and produce a burst-pause firing
pattern  in  thalamic cells during each cycle of rhyth-
mic  spindle  activity  (Purpura  and  Cohen  1962,
Andersen  and  Eccles  1962, Purpura  et  al.  1966,
Andersen   1966,  Andersen  and   Anderson  1968).
The  questions  that  arise  are:   (1) where  are  the
interneurons? and (2) are they the same for all types
of synchronization?  Evidence reviewed by Andersen
and  Anderson  (1968,  pt38) suggests that  interneu-
rons, at least for some types of rhythmic responses,
are  distributed locally  throughout  the  thalamus,
much like Renshaw cells are distributed locally within
or near ventral root  motoneurons  in the spinal cord.
The  evidence, however, is not conclusive.

   Recent studies have indicated that during synchro-
nisation, cells  in  the  thalamic  rcticular nucleus (R)
fire  in long high frequency bursts  when cells in near-
by thalamic nuclei are silenced, presumably by long-
duration inhibitory  postsynaptic  potentials (Massion
 1968, Massion and Rispel-Padel  1972, Schlag and
Waszak 1970,  1971,  Filion, Lamarre, and  Cordeau
 1971, Lamarre, Filion, and Cordeau 1971, Frigyesi
 1972, Frigyesi and  Schwartz 1972).  Yingling and
 Skinner (1975) showed that the envelope of the post-
 stimulus-time histogram of R-units responses to medi-
 al thalamic stimuli  were identical in wave-shape,  la-
 tency,  and duration to the  characteristic inhibitory
 post-synaptic potentials routinely observed in thala-
 mic   neurons  during  recruiting  responses. They
 also  found that lateral thalamic stimuli do not drive
 the  same  R-units that respond  to  medial  thalamic
 pulses, Thus, fairly  strong  evidence suggests Mat R is
 the  site of inhibitory interneurons  that mediate  re-
 cruiting responses. The location of inhibitory neurons
 for  augmenting responses  is unknown; the interneu-
 rons may  be locally distributed throughout the thala-
 mus or may be in regions  of R not yet explored. It is
 certain, however,  the interneurons  for augmenting
 and recruiting types of synchronous activities are not
 located in the same  cell groups.
  The Mechanism of Regulation of the Recruit-
  ing Type of Thalamic and Cortical Synchroni-
  zation

     Schlag and Villablanca (1967) showed in thalam-
  ectomized cats that 8-c/scc stimulation  in white mat-
  ter produced long latency responses in frontal cortex,
  which recruit in amplitude  similarly to cortical re-
  cruiting   responses.  Skinner  and  Lindsley (1967)
  showed  that blockade  in the pathway between MT
  and FC, i.e., in  the inferior thalamic peduncle (1TP),
  abolished  recruiting  responses  in  the  frontal lobe.
  Thus, the burst-pause pattern of activity originating
  in  MT appears to be  sufficient to evoke long-latency
  recruiting responses in  FC; anatomical  evidence sup-
  ports this physiologically inferred pathway from MT
  to FC via ITP (Scheibel and Scheibel 1967a).

-------
  618
                                         Skinner
    IIP  blockade  could also  interrupt  a descending
 projection from FC to R, a disconnection that would
 abolish  cortical and thalamic recruiting responses by
 eliminating the drive on inhibitory interneurons that
 give rise to the burst-pause pattern of activity in tha-
 lamic cells. Steriade and Wyzinski (1972) have shown
 that  FC stimulation  produces very short-latency re-
 sponses  in R units, a result consistent with a mono-
 synaptic projection from FC to R. This FC-R projec-
 tion  is  supported by anatomical evidence, which
 shows that FC axons provide innervation of the ante-
 rior neuropile region of R  (Scheibel and  Scheibel
 1967). The FC-R pathway, like the MT-FC one, also
 travels in the  ITP. Thus ITP blockade interrupts both
 inputs and outputs of FC in its extrathalamic circuit
 connecting MT to R.

   ITP blockade  prevents MT stimuli from  evoking
 unit responses in  R  and so does MRF stimulation
 (Yingling and Skinner 1975). Thus, a third alternative
 is that ITP blockade disrupts an FC-MRF-R circuit
 and has the same affect on R cells as MRF activation
 (i.e., disinhibition  of the MRF). However, since ITP
 blockade eliminates thalamic recruiting responses and
 spontaneous  spindles  in an  MRF lesioned animal
 (Skinner and Lindsley  1967), this last alternative ap-
 pears to  be ruled out and leads us to believe that the
 FC projection  to R is independent of MRF. Several
 reports, based on Golgi-stained material, trace projec-
 tions  of  the  ventral  leaf  of  the  ascending MRF
 through the subthalamic and hypothalamic fields into
 the vicinity of the rostral thalamus and R (Scheibel
 and  Scheibel  1967a, Millhouse 1969).   Thus, some
 anatomical  evidence  supports the  existence of an
 MRF-R projection that is independent  of the FC-R
 pathway.
   The regulation  of R by MRF and FC appears to
 control the pattern of thalamic inhibition that under-
 lies the recruiting  type of EEC synchronization. FC
 has a phasic excitatory affect on R cells that lasts for
 100  msec following activation  by an MT stimulus,
whereas MRF has a tonic inhibitory influence on the
same R cells that lasts for 20 to 30 sec (Yingling and
 Skinner  1975). Thus,  desynchronization of the re-
cruiting response can be produced by either inhibi-
tion ofR units by  MRF or blockade of pulsatile exci-
 tation ofR units from FC.

 Slow Potentials Also Reflect Dual Regulation
 of R by MRF and FC

 Cortical Slow Potentials

   Arduini  et  al. (1967)  first showed that a negative
 slow potential (SP) shift could be produced  on  the
 surface of the frontal cortex in a cat by stimulation
 of sensory  nerves  or the MRF. SPs and EEC desyn-
 chronization  persisted for many seconds following
 the stimulus.  The  effect of drowsiness, a
 related to a decrease in MRF activation level (Jouvet
 1967), has been shown in animals to result in a tonic
 frontocortical  slow  potential  of opposite polarity
 (i.e., surface  positive) (Caspers  1963, 1965). Thus, it
 appears that  conditions associated with an increased
 tendency  for EEC synchronization in  neocortex  re-
 sult in  SPs on the frontal lobe that are positive in
 polarity and conditions that  reduce EEC rhythmic
 activity result in negative shifts.

   Surface SPs can also be produced in FC by a warn-
 ing signal associated  with  a  meaningful  expected
 event.  Walter et al. (1964) first demonstrated  that a
 tonic negative slow potential, contingent negative var-
 iation (CNV), was produced in the frontal region of
 humans during conditioned expectancy. The CNV has
 been implicated in a cerebral process that appears to
 be  distinct from general arousal  usually associated
 with a mesencephalic  reticular mechanism. Skinner
 (1971 b) has implicated the MT-FC system of the cat
 in  the  generation of such neocortical  expectancy
 waves.  Slow  potentials, elicited by a  warning tone
 that signaled  the onset of electric shock, were evoked
 in FC and motor  cortex, but only FC responses were
 abolished  by  ITP blockade.

 Slow Potentials in R

   Extracellular positive SPs  have recently been  re-
 corded in  R following  (1) a tone reinforced by cuta-
 neous  shock and (2)  an electric  stimulus in MRF
 (Skinner and Yingling  1976).  Both types of R  re-
 sponses were found to mirror exactly those recorded
 in FC, except that the responses were positive in po-
 larity in R rather than surface-negative as observed in
 FC. By manipulating the  parameters of the stimulus
 situation,  conditioned  responses  in both  locations
 could be  modified.  Changes in the interval between
 tone and shock, introducing an unrein forced tone of
 different frequency, or massing trials with a mean of
 30  sec rather than 3 min, all modified  the responses
 by  affecting peak latency, amplitude, or other aspects
 of  the  waveform. AH modifications, no matter how
 they were created,  resulted  in mirror-image  wave-
 forms in FC and R.
   A brief stimulus to MRF evoked mirror-image re-
sponses in FC and R that had the same polarities as
and similar waveforms to those evoked by condition-
ing, but in this case the mechanism of regulation of
evoked  waveforms was found  to  be  different.  The
configuration of conditioned waveforms (i.e., peak la-
tency, amplitude, duration) was determined by para-
meters of the stimulus situation, whereas waveforms
evoked by MRF stimulation were uniformly large in
amplitude and long in duration (20 to 30 sec). Novel
stimuli or  strong cutaneous shocks also evoked similar
20- to 30-sec tonic  responses in  both FC and  R, a
                       that  these Stimuli. like  the

-------
A Neurophysiological  Model

MRF  pulse, all belong to the same class of stimuli and
are perhaps mediated by  the same cerebral mecha-
nism.
   Bilaterial cryogenic (Skinner 197 lc) ITP blockade
was found by Skinner and Yingling (1976) to abolish
the ability of the conditioned tone to evoke responses
in R or FC. Under these conditions, however, MRF
stimuli could still produce a response in R, a finding
which suggests that the conditioned response in R is
not mediated by MRF activation. Whether or not the
conditioned R response is dependent  upon  a condi-
tioned FC response is difficult to determine.  IIP
blockade invariably interrupts thalamocortical inputs
to FC as well as FC-R projections. In order  to prove
that conditioned R responses  are dependent upon FC
responses,  the latter would have  to be evoked by
some  as-yet-unknown means during ITP blockade of
FC-R fibers alone.  Since the physiological and ana-
tomical evidence of Steriade and Wyzinski  (1972),
Yingling and  Skinner  (1975), and   Scheibel and
Scheibel (1966, 1967a) convincingly  supports  a di-
rect FC-R  projection, conditioned  slow potentials in
R are probably dependent upon the occurrence of
FC responses.
Distinctions Among SPs

   Novelty or  conditioned  significance inherent in a
given  physical stimulus will result in SPs that can ap-
pear quite similar to each other. These two responses,
though, are mediated by  separate mechanisms. For
example, responses in R to novel stimuli are not  abol-
ished  following ITP blockade, whereas those elicited
by conditioned  stimuli are  (Skinner  and  Yingling
1975). This result clearly distinguishes between the
two responses even though the waveforms and polari-
ties may be identical. Identifying the underlying neu-
ral mechanism may, therefore, be necessary to distin-
guish  among some types of SP responses.

   Loveless  and  Sanford  (1974)  showed  that the
human CNV is divisible  into two components related
to conceptually  distinct cerebral processes: An  early
component that appears just after the warning signal
was attributed to an orienting reaction  (0-wave);and
a late component that manifests itself just before the
second  stimulus  was attributed to expectancy (E-
wave). Similar independent modulation of these two
components has been demonstrated recently in hu-
mans  (Rohrbough  et al., 1976) and in animals (Skin-
ner and Yingling 1977, Skinner, in preparation). By
averaging  small  blocks  of  15 trials each, the  latter
investigators were able to study the development of
frontal  cortex  SPs in  the  cat  during  acquisition of
tone-shock conditioning. After overconditioning (500
to 1000 trials), the early negative SP component that
usually follows the warning tone drops  out and leaves
unaffected the  ramp potential that just precedes the
cutaneous shock. This habituation of the early  com-
ponent is also seen in records in which the interstimu-
                                            619

lus interval is reduced from 4.0 to 1.2 sec, a situation
in  which the rise  times  of the  SP components  are
forced to accelerate (Skinner 1971b).

   Bilateral cryogenic ITP  blockade abolished both
the early and late CNV components  in FC and R
(Skinner 1971b, Skinner and Yingling 1977). This
finding  suggests that  neither of  the conditioned re-
sponses  in R is  entirely dependent upon MRF activa-
tion  since stimulation of this latter structure during
ITP blockade could still produce an SP in R (Skinner
and Yingling 1976).  A positive shift in the  baseline
occurs in R during ITP blockade, however,  and this
potential could reflect a constant process represented
by one  of the two positive  CNV components. In sup-
port  of this  position, Luria and Homskaya (1970)
have attributed the behavioral deficit following fron-
tal lobotomy in man to perpetual orienting reactions
that  can no longer be terminated or  brought under
control  by  some process that has been lost. Thus, the
positive baseline shift in R  pursuant to ITP blockade
may  represent a constant orienting process (0 wave)
that  can no longer be brought under  control by in-
puts from FC. FC input to R may normally give rise
to the expectancy process  (E wave) that always fol-
lows the orienting reaction  and perhaps suppresses it.

The Generators of Cortical SPs

   Cyclic nucleotides  have been implicated in the me-
diation  of slow postsynaptic potentials in sympathet-
ic  ganglia (McAfee and Greengard  1972, Greengard
and  Kebabian 1974, Weight et al. 1974, Libet et al.
1975, Kebabian et al. 1975). Once evoked,  these in-
tracellular  potentials  persist for  many seconds. The
pharmacology  of  the specific transmitters  involved
(Libet 1970, Greengard and Kebabian 1974), the in-
tracellular dynamics of the  specific cyclic nucleotides
activated (Greengard  and  Kebabian 1974), and the
resultant ionic effects that  produce the post-synaptic
membrane  potentials (Weight  and  Votava  1970,
Weight  and Padjen 1973, Siggins et al.  1973, Schul-
man  and Weight  1976) have been  studied in the
sympathetic  ganglion model. Tonic increases or de-
creases  in firing rates of neurons following local ion-
tophoretic  application of  various neurotransmitters
and cyclic nucleotides suggests that similar slow excit-
atory and inhibitory mechanisms exist in the cerebral
cortex (Stone et al. 1975, Phillis  et al.  1975). Subcor-
tical effects have also been  noted; local application of
acetylcholine in R will silence spontaneous firing of
units in this  structure for  many seconds (Ben-Ari et
al.  1975).  Thus, cerebral slow potentials associated
with  higher cognitive functions  may have generators
that are related to the mechanism of membrane po-
tential changes observed in  neurons of the sympathet-
ic ganglia.

   Evidence from  Skinner et al. (I978b), based on a
new cryogenic method of  neurochemical fixation in

-------
 620

 the conscious rat, showed that a close correlation ex-
 ists between shifts of cortical slow potential ampli-
 tude  and cyclic AMP (3", 5'-adenosine  monophos-
 phate) level  that follow cutaneous electric shock. An-
 other rat study (Skinner et al. 1978a) showed that
 several types of psychologic and physical stress result
 in a reduction of cyclic AMP level compared to that
 of unstressed controls. The effects of cutaneous stim-
 uli (25 V, 150 Hz, 0.5-sec duration) caused an imme-
 diate  increase in cyclic AMP in parietal cortex that
 then decreased over a 30-sec interval in parallel with
 the development of a positive SP recorded from the
 same  region  (r = 0.77, P < .01). Only SPs recorded
 from  the same  small region  of parietal cortex  in
 which the chemical measurements  were made  corre-
 lated  with the  cyclic AMP content, a result which
 suggests  that generators of the SP and cyclic AMP
 shifts  are confined to the same region. Positive SPs
 evoked by the shock did not reverse in polarity across
 the cortex,  from surface to depth, a finding which
 suggests that the SPs are generated by a closed-field
 dipole.

    Cyclic nucleotides are  thought to be intracellular
 mediators between the neurotransmitter  arriving  at
 the postsynaptic  membrane  and  the regulation  of
 ionic gates in the membrane that determines  the cel-
 lular response to the neurotransmitter (Greengard and
 Kebabian 1974). Since direct extracellular application
 of cyclic AMP  to cortical neurons by a micropipette
 causes them  to stop firing for several seconds (Stone
 et al.  1975,  Phillis et al.  1975), the  phasic  increase
 followed  by  tonic reduction of cyclic AMP described
 above suggests  that cutaneous shock  activates first a
 rapid inhibitory Input to parietal cortex neurons that
 then  becomes tonically dislnhibited over a period  of
 30 sec to drop  below the  control level. This evidence
 IE  not  conclusive because it is not known upon which
 cells the  Injected  cyclic  AMP operates or In which
 cortical  cells the cyclic  AMP increases following
 shock. All that  can be concluded  is that a  surface
 positive SP in the parietal cortex is  correlated with a
 decrease in cyclic AMP In underlying tissue.

   The coupling between SPs and underlying neuron-
al  activities is not very precise. This has been explain-
ed in part by the diffusion time required for potassi-
um ions to move between neurons and nearby gllal
cells whose  interconnected  syncytlum actually pro-
duces  the SPs (see Somjen, this volume). The slow
membrane potentials in  the  sympathetic ganglion
nerve celL seem to be produced by reduction in the
leakage currents of  sodium   and   potassium  loin
(Weight and Votava 1970, Weight and Padjcn  1973),
a finding  that implies modulation of the length con-
stant  of  membranes  much like  that  produced by
myelln on an uxon. Thus, a slow IPSP in  these cells
resulting from decreased  sodium conductance could
have the net  effect  of making some remote synaptic
EPSPs more effective  in thoir influence on a trigger
                                          Skinner

 zone in the cell and  thereby could increase the firing
 rate even though the cell is hyperpolarized.

    In  a fairly uniform structure like the thalamic re-
 ticular nucleus, however,  there does  seem to be  a
 fairly  close correspondence between the duration and
 polarity of evoked SPs and the firing of neurons, as
 illustrated in Fig.  1 and 2. In Fig. IBa negative SP is
 seen to follow each 8-c/sec  impulse  to the  medial
 thalamus (ncm) and  to  recruit in amplitude; this
 phasic  slow potential (best seen after the last pulse in
 the stimulus train, arrow) only lasts 100 msec. Fig.
 1C shows that this same thalamic train of stimuli re-
 cruits  excitation of  R units  for approximately the
 same interval as the  negative  slow potential. Fig. 2B
 shows a slow positive potential in R elicited by  a brief
 stimulus to MRF that persists for many seconds. This
 same stimulus to MRF also silences spontaneous units
  A.
             FP. (2.
C.
                                        VA
  B.
 3    V*	

 .    v—-~
      *

 7    /V*-"
           ncm
 AVf. 1. Slow  potentials and unit activity evoked In
 nucleus rettcularls tlialamt by medial thalamic stimu-
 lation. A. Filled circles: Location of units encounter-
 ed In the anterior portion of nucleus retlcularls thala-
 ml  (R) adjacent  to nucleus ventralls anterior (YA)
 that were responsive to 8 c/sec stimulation of nucleus
 centralls medlalls (nan).  Open circles: Units that
 were  not responsive  to ncm stimulation, B. Phasic
 slow potentials recorded In  the same region  of R yA
 by a DC amplifier  during ncm stimulation;  negative
 responses occur In  the Interval between the  positive-
 •polarity recruiting responses and  are best seen fol-
 lowing the last stimulus in the train farrow): calibra-
 tions 200 pV, 1 sec. C. Oscilloscope traces of units In
R VA  responding  to the first through the ninth  Im-
pulse  of an  8 c/sec stimulus train  to ncm; the trace
length  is  100  msec following each  stimulus pulse.
(Adapted from Yingling and Skinner 1975 and Skin-
ner and Yingling 1976.)

-------
A  Neurophysiological  Model
                           x^x:
 Fig. 2. Slow  potentials and unit activity evoked in
 nucleus reticularis thalami by mesencephalic reticular
 stimulation. A.  Filled circle:  Location of recording
 electrode in the anterior portion of nucleus reticularis
 thalami  (R) adjacent to nucleus vert trails anterior
 (VA) that showed a slow-potential response to a brief
 stimulus (2-6 V,  150 c/sec; 0.5 msec duration, 30
 msec train) to the  mesencephalic reticular formation
 (MRP).  Unfilled circles: Location of electrode  that
 showed  no responses to same MRF stimulation. B.
 Tonic slow potential of positive polarity recorded in
 R VA fallowing a brief stimulus to MRF; trace length
 is approximately  I min,  and  peak amplitude is 12
 mV.  C.  Oscilloscope traces of spontaneously firing
 RyA units that are inhibited by a brief stimulus to
 MRF; each trace length  is 1 sec and shows activity
 immediately before and after MRF stimulus and 10,
 20,  and 30 sec later. (Adapted  from  Yingling and
 Skinner  1975 and Skinner and Yingling 1976).

 in R for approximately  the same interval of time as
 concurrent slow potentials, a result illustrated in Fig.
 2C. Note by comparing Fig. 2B and 2C that the unit
 response latency is rapid, whereas  the onset latency
 of the slow potential is on the order of a second;
 furthermore, the peak inhibitory effect on unit activi-
 ty is earlier (i.e.,  immediate)  than the peak of the
 slow potential shift (approximately 15 sec).

    In  conclusion,  it  appears  that slow extracellular
 negative shifts are  associated with unit excitation, and
 extracellular positive responses are related to unit in-
 hibition, at least in R. Whether this same relationship
 holds  for unit activities and slow potentials in other
 parts of the brain  is not known, although some evi-
 dence indicates that multiple unit activity increases in
 the  lateral geniculate are associated with negative ex-
 tracellular slow potentials (Rcbcrt 1973). This general
 relationship does not seem to hold for parietal cortex
 in which positive SPs are associated with a reduction
 in cyclic AMP, a putative mediator of  cellular inhibi-
 tion. The cortex,  however,  contains numerous types
 of cells, and the pattern  of interactions that give rise
 to cortical SPs may be quite  complex. The etectrogen-
 esis of cerebral slow potentials associated with higher
 cognitive processes is still uncertain, but accumulating
                                            621

evidence points toward a chain of events in which
cyclic nucleotides  regulate slow postsynaptic poten-
tials in  neurons that  in turn modulate extracellular
K+ concentrations which effect responses in  the sur-
rounding glial cell syncytium to produce the recorded
closed-field potentials.

Sensory Evoked Potentials: Selective vs. Gen-
eral Regulation of Relay Nuclei

Regulation  of Cortical Evoked Potentials by
MRF and FC

   Selective attention in human subjects results in en-
hancement of cortical evoked  responses to attended
stimuli (Spong et al. 1965, Hillyard et al. 1973, Hill-
yard and Picton, this volume), whereas arousal causes
the enhancement of responses  to relevant and irrele-
vant stimuli alike (N'da'tanen  1970). In animals, MRF
stimulation causes enhancement of evoked responses
in all sense modalities independently of the precise lo-
cation  of the stimulating electrode in the structure
(Bremer and  Stoupel 1959, Dumont and Dell 1960).
These findings  suggest  that regulation of evoked po-
tentials  by the  MRF is nonspecific, as in  the process
of general arousal.  In contrast, the MT-FC-R system
regulates sensory processes selectively. For example,
various  sense modalities affected by intervention  in
the  ITP  depends upon the  precise location of the
blockade. Cryoprobes placed bilaterally in a certain
part of the ITP will enhance visual but not auditory
evoked potentials when cooled to  1ST and will en-
hance  both  when cooled  to IOT  (Skinner  and
Lindsley 1971). Alexander et al. (1976) showed that
FC stimulation  with a matrix of electrodes inhibited
spontaneous  and driven single units throughout audi-
tory cortex;  the particular  units inhibited, however,
were specific to  the point in  FC  stimulated. These
results suggest that  regulation of evoked responses by
the MT-FC-R system is specific and may be involved
in the process of selective attention.


Regulation of Sensory Evoked Responses by R

    Both ITP  blockade and MRF stimulation appear to
 enhance cortical  evoked  potentials  by facilitating
 transmission  through  thalamic relay nuclei  (Skinner
 and Lindsley  1971,  Rapasardi et al. 1974,  Bartlett
 and Doty 1974). This opening of thalamic gates may
 be effected by  two separate  mechanisms, which block
 the tonic inhibition exerted by R  on thalamic relays.
 As  evidence, stimulation of  R is found to inhibit
 transmission  of sensory information through thalamic
 relay nuclei (Skinner and Yingling  1976, Yingling and
 Skinner 1976), a finding which demonstrates that R
 is the  source of a powerful  inhibitory influence (Fig.
 3).  Evidence that R maintains a tonic output is seen
 in records where R units fire continuously (Fig. 2C).

-------
622
         STIM RLC    STIM RMQ   STIM RVBC
   VEP
   AEP
   SEP
 Fig.  3. Inhibition of evoked potentials in three major
 sense modalities by stimulation of various regions of
 nucleus reticularis thalami (R) adjacent  to different
 thalamic relay nuclei: lateral geniculate (LG),  medial
 geniculate  (MG), and ventrobasal  complex  (VBC).
 Visual evoked potentials  (VEPs) were  elicited by
 optic-tract stimulation and recorded in primary visual
 cortex. Auditory evoked potentials (AEPs) were elici-
 ted  by  stimulation of the brachium  of the inferior
 colliculus and recorded  in primary auditory  cortex.
 Somatosensory evoked potentials (SEPs) were elicited
 by stimulation of stainless steel wires sutured through
 the skin of the hind leg and recorded in the contralat-
 eral  somatosensory  cortex (far lateral anterior  sig-
 moid gyrus).  Conditioning stimuli of 250 c/sec, 20
 msec trains were delivered to either RLQ, RMG- or
 R VBC 50 msec prior to the test stimulus.  Stimulation
 1 mm on either side of these various regions of R had
 no effect on  reducing the primary  sensory evoked
 potentials.  Calibrations:  200 \t,V;4 msec (VEPs), 10
 msec (AEPs and SEPs).  (Adapted from Yingling and
 Skinner 1976).
 These spontaneous units  can  be inhibited by  MRF
 stimulation and their ability to be driven  to higher
 rates by  the MT-FC-R system can be interrupted by
 IIP blockade.  Thus, disinhibition of thalamic  relay
 nuclei could be effected by either reduced excitation
 from FC to R  or by increased inhibition from  MRF
 onR.

   Selectivity of the MT-FC  system in the regulation
 of sensory  evoked potentials implies that FC projec-
 tions through R to sensory channels maintain modal-
 ity-specific  connections. Specificity of inhibitory pro-
 jections of R into thalamic relay nuclei has been dem-
 onstrated by Yingling and Skinner (1976), who found
 that stimulation of the portion of R adjacent to a
 given  thalamic  relay nucleus  inhibited  transmission
 selectively in this afferent  pathway, without influenc-
 ing other sensory channels (Fig. 3).  Since MRF  stim-
 ulation  enhances evoked  potentials generally, this
 structure must have widely distributed axonal projec-
                                          Skinner

tions to R cells. Fig. 4 illustrates these specific and
nonspecific  projections of FC  and MRF systems to
R cells, a feature of central importance in the neuro-
physiological model discussed below.

Theoretical  Model  for  Regulation of Sensory
Input to Cerebral Cortex

   Physiological and anatomical data have been pre-
sented to support the schema  (shown in Fig. 4) by
which sensory inputs to cerebral cortex are regulated.
FC is shown to have phasic excitatory (Ep) influences
on R cells that  are specific in their projection. In
contrast, MRF has tonic inhibitory (Ij) control over
R neurons via diffuse  or nonspecific connections. R
cells project to specific thalamic relay nuclei (Th Re-
lay) and  form  a parallel  series of inhibitory gates
that control  the  transmission of  afferent activity.
The pattern of activity in R that results from its dual
regulation by FC and  MRF determines the pattern
of input from sensory receptors that reaches  respec-
tive primary projection cortices (PC). Inputs to  FC
and MRF are not specified by the  model because of
insufficient data. Both structures, however, receive in-
formation  from  external and  visceral  receptors  via
connections  with other regions of  the brain (not
shown in the diagram) including the limbic and auto-
nomic nervous systems.
   FC
Th   Th
 R   Reloy
Fig.  4, Neurophysiological model for the regulation
of sensory input to cerebral cortex. FC = frontal cor-
tex; ITP ~ inferior thalamic peduncle; ThR = thalam-
ic reticular nucleus; Th  Relay  = thalamic relay nuclei
for visual, auditory, and somatic afferent channels;
PC = primary projection cortex; MRF = mesencephal-
ic reticular formation.  White  neurons are excitatory
and black ones inhibitory. FC neurons produce phasic
excitation  (Ep)  of R  cells via specific  projections.
MRF cells produce  tonic inhibition (I-p) of R cells
and have nonspecific, widely distributed projections.

-------
A Neurophysiological Model

                    Th   Th
   FC               R   Relay
 Fig, 5. Example in which the neurophysiological mo-
 del selects auditory  input to the cerebral cortex be-
 cause of a particular pattern of activity in FC and
 MRF. + = synoptic excitation of cell; - = synoptic
 inhibition of cell; 0 = no  excitation or inhibition.
 Abbreviations same as Fig. 4.

    An example of a particular state of the model is
 shown in  Fig. 5, which  illustrates how a pattern of
 activity in FC could result  in the selective transmis-
 sion of input to the auditory cortex, as in selective
 focusing of attention to sound  stimuli. The connec-
 tions  in this model imply that selective suppression of
 irrelevant  stimuli is exerted  by the pattern of activity
 in FC; that is, FC must have information as to what
 channels to  close. This  contention  is supported by
 observation  that ITP blockade  enhances irrelevant
 evoked  responses (Skinner and  Lindsley 1967). The
 role of  FC in suppressing interfering stimulus events
 in the brain  has been suggested more  recently by
 Bartus and Levere (1977) who  showed  that the be-
 havior of monkeys with dorsolateral frontal ablations
 was guided by irrelevant  as  well as relevant stimuli in
 a discrimination  task  in  which both types of cues
 were presented sequentially.

    A  pattern of activity  in R may begin to form ini-
 tially  from a condition in which all gates are opened
 by  a  strong  stimulus to MRF.  As MRF output de-
 clines, relinquishing  priority  of inhibitory  control
 over R cells,  selective excitation  of R cells by FC may
 occur and thus close thalamic gates, one-by-one, until
 only  the  attended  channel remains open. This se-
 quence  of events in  which orienting reactions are re-
 placed by expectancy or selective  attentiveness  is,
 perhaps, reflected in the sequential  0-wave and  E-
 wave  components of SPs in  R and FC that  occur
 during  the   CNV and  related  paradigms. Another
 strong or  novel stimulus  could activate the MRF and
                                             623

cause it to exert generalized priority control over R
units and reopen the selectively closed gates. If MRF
excitation is not intense enough to drive all the gates
open, it  may simply drive  the pattern in R that is
being sculptured by the FC  into a less selective prior
condition. It  is important to realize that it is  the joint
regulation of R by FC and MRF that gives rise to the
pattern  of input to  cerebral cortex. These  two sys-
tems appear to vie constantly for control over R cells,
thus titrating the processes of general arousal against
selective attention.

   For example, King and Skinner (1975) have shown
that mild  8-c/sec MT stimulation  will disrupt single
alternation behavior, as will partial ITP blockade. In
both cases the descending pattern of activity  from FC
to R is changed, perhaps resulting in the inability of
an  animal  to control its sensory  channels in a way
that is necessary to  perform the task. In the case of
mild MT stimulation, the animal's performance  defi-
cit  can be reversed by MRF stimulation. This deficit
reversal is not the case for partial ITP blockade, how-
ever, for in this case the pattern from FC to  R is lost,
not merely superimposed upon a generalized  bias that
can be neutralized by another of opposite sign.

    This model does not pretend to solve the mystery
 concerning the  nature of the behavioral deficit that
 occurs  following frontal lobotomy, although it sug-
 gests explanations for some observations. For exam-
 ple, Konorski and Lawicka (1964) observed that fron-
 tal  preparations were easily distracted by  irrelevant
 stimuli  that impaired performance. The present mod-
 el predicts that without selective FC excitation of R,
 irrelevant sensory gates would remain open, and thus
 allow non-pertinent information  to ascend continu-
 ously to the cortex. This cortical excitation by irrele-
 vant stimuli  may  underlie the  hyperdistractibility
 characteristic of frontal animals and  lobotomized
 patients.

    A major question remains unanswered: "How does
 the pattern of activity in FC develop to set selectively
 the gates in  R that determine the pattern of sensory
 input to  the cerebral cortex?"  A definitive answer
 remains as a challenge to the future  expansion of this
 model, although some clues are  presently available
 regarding  transactions in FC. Imbert et al. (1966)
 have shown  in  "cortical shell" cat  preparations that
 FC receives  cortico-cortical inputs from virtually all
 primary and  association  cortices. Such a configura-
 tion allows for, but does not prove,  reiterative regula-
 tion of FC: FC-R-thalamic  relays-primary cortex-FC.
 Such reiterative thalamocortical circuits have always
 been presumed to  underlie EEC synchronous activi-
 ties. As stated earlier, the inherent resonant periodici-
 ty  of recruiting responses seems to depend on a delay
 in  the cortical  component  of the underlying circuit.
 Whatever is  the source of  this delay (e.g., dendritic

-------
 624
 propagation, lateral inhibition, slow PSPs), FC re-
 cruiting responses appear to have open-field dipoles
 with reversal  potentials  (Spencer  and  Brookhart
 1961). SPs in FC, however, do not appear to have the
 same type of dipole generators.  Kelly et al. (1969)
 and Arezzo et al. (1975) have shown that FC poten-
 tials in the monkey evoked by irrelevant sensory stim-
 uli have  closed-field dipoles  (i.e., no reversal poten-
 tials), a finding which suggests important bioelectric
 processes  in FC that are  distinct from those associat-
 ed  with recruiting responses. This list of FC  transac-
 tions certainly  must continue with the work in ani-
 mals by  Rowland, Fuster, Stamm,  and Marczynski
 (all  in this volume), but the incorporation of their
 data into the present model is not yet tenable. Nei-
 ther is the analytic work in parietal cortex by Mount-
 castle (1976) whose experiments  are also designed to
 backtrack from attentive  behavior into the underlying
 neocortical  mechanisms. Further  investigations into
 molecular mechanisms responsible for slow-potential
 FC shifts that  are correlated with attentive behavior
 will undoubtedly  be  the most important source of
 new information to be incorporated into the thalamic
 gating model.

   The present  model attributes a selective pattern of
 activity to FC  and a level  of activation to MRF and
 shows how these posited activities jointly regulate in-
 hibitory gates in R that control the ascent of sensory
 information through the  thalamus to  the cortex. The
 power of  this   model lies in its  ability  to  explain
 changes in three types of bioelectric activity  tradi-
 tionally associated  with the processes of arousal and
 attention in both animals and humans. EEC synchro-
 nization  in FC  may be  an  activity associated only
 with an idling state in the thalamic gating system, but
 its  use as  an investigatory probe, especially  in the
 form of recruiting responses, has been helpful in de-
 termining some of the physiological  features  of the
gating mechanism.  Any  theoretical model of atten-
tion must  deal effectively with  this phenomenon,
even if synchronous activities per se  are shown to be
of  no  particular importance.  SPs recorded  in the
gating system have confirmed the dual regulation of
 R  seen with the recruiting response  probe.  Frontal
cortex SPs appear to depend upon the integrity of a
subcortical system  that is projected centripetally via
 ITP. This projection,  however, may not be the only
 source of input that sculptures the SP pattern of ac-
 tivity in  FC, for the  frontal lobe is the recipient of
 massive cortico-cortical projections from all parts of,
 the cerebral mantle. The MRF is commonly associat-
 ed with the processes of orienting and arousal, and its
 impact upon the  thalamic  gating system at R  may
 eventually  explain how both orienting (0-wave) and
 expectancy (E-wave) components can appear in cor-
 tical SPs as demonstrated by Loveless and associates
 (this volume).   Finally,  the thalamic gating  model
 offers  a  ready explanation for  the  regulation of
                                         Skinner

sensory evoked  responses by both specific and non-
specific mechanisms.  Hillyard and Picton (this vol-
ume) have identified in  the human cortex an N100
component of  the sensory  evoked  response  that
seems to be  related to  "stimulus-set" or stimulus-
source selection and a P300 component related  to
"response-set" or stimulus relevance selection.  The
present neurophysiological model suggests a theory
of attention in which selection occurs  via active sup-
pression  of irrelevant stimuli controlled jointly by
FC and MRF interactions on the inhibitory gates in
R. These interactions determine not only what gates
are closed (source selection), but how  intensely they
are closed  by the titration of the two systems (rele-
vance selection).   These regulatory features of the
neurophysiological  model developed   with  animal
experiments appear to provide the basis for explana-
tion of the independent  modulation of the N100 and
P300  evoked  potential  components as well as  the
0-wave and E-wave components of slow potentials,
and  therefore provide a  unifying theoretical frame-
work to explain several  event  related potentials in
the human brain.

Comments on the Skinner Model

   Hillyard expressed the view  that, at present, evi-
dence from human ERP  studies of attention and ani-
mal  physiological studies of the kind presented were
so far apart  that no definite  conclusions could  be
reached as to whether a particular model, such as
Skinner's, was right or wrong. He welcomed it, how-
ever, as providing a creative framework that suggested
experiments to be undertaken. These could well give
direction to hitherto independent lines of investiga-
tion in a way  that would bring them closer together.

   One experiment suggested by the model would in-
volve showing that the  nucleus  reticularis thalami
physiologically has the capability to select or to gate
information in  a situation parallel to one in which
human psychological experimentation has shown that
good  selective  attention occurs  - i.e., a situation
where stimuli  are presented from different sources, at
a rapid rate, and with a very  high information load.
One  could,  for  example, record from two subdivi-
sions of  R, which control different selective sensory
systems,  in circumstances where the animal  was re-
quired to attend first to one source and then another.
Only then could one determine whether the animal is
actually behaving as the model predicts.

   The model also suggests other links with human
data. For instance, it suggests a place in frontal granu-
lar cortex that might be susceptible to disruption in
patients suffering from stroke or other forms of cere-
bral  damage in  the appropriate area. Such patients,
with  the equivalent of an inferior thalamic peduncle
lesion, could be  investigated for their ability to regu-
late inputs.

-------
A Neurophysiological Model
                                             625
   Unfortunately, human evoked  potential data do
not yet provide evidence for the modulation of pri-
mary cortical responses in a form that would support
Skinner's  model. Potential changes can be recorded
from the  primary cortex when a person is attending
to stimuli, at least  in the somatosensory  mode, and
possibly in the  auditory mode; however, these prima-
ry potentials do not seem to show systematic changes
related to  attention.

   Zappoli thought that the Skinner model, based pri-
marily  on  data  from experiments with cats, was not
entirely supported  by existing  evidence  from  man.
Skinner's   findings  show  that  the mediothalamic-
frontocortical system may play  an  important role in
generating and  regulating anterior frontal surface neg-
ative SPs; however, Zappoli et al. (this volume) have
observed  CNVs in  bilateral prefrontal lobotomized
patients  in whom  thalamo-frontal pathways have
been severed  by the technique of Freeman and Watts
(1947, 1966).  Regeneration must be considered im-
possible  in such cases since  autopsies have always
shown extensive retrograde degeneration of the thala-
mic dorsomedial nuclei.  Furthermore, the absence of
degeneration  of reticular and intralaminar nuclei sug-
gests that the  dorsomedial-thalamofrontal system is
not indispensable to CNV development over the fron-
tal region  in humans, as the model implies.

   These  findings, together with those of Gazzaniga
and  Hillyard  (1974) on CNVs in "split-brain"  pa-
tients, and of Marsh and  Thompson (1973) and Wein-
berg and   Papukostopoulos (1975) on the bilateral
symmetry of the CNVs  activated In  different  right-
left tusks, suggest that classic CNV activity Is a diffuse
electrical  event essentially related to a unitary, gen-
eral physiological brain process (e.g., arousal) presum-
ably mediated  by the non-specific ascending reticular
system. The  differences  In morphology and polarity
of CNVs  detectable in different brain structures are
probably  related above all to their Intrinsic anutomo-
functional characteristics and  to  the  method com-
monly utilized  In recording these potentials.

   Zappoli pointed out  that the anomalies between
Skinner's  animal data and his human data were by no
means unique  and  served to  emphasize the  caution
that should be exercised when  extrapolating from ani-
mal to man.

   Skinner replied  that  on the whole Dr. Zappoli'a
data supported rather than refuted his model. One of
the outstanding questions Is how the frontal cortex
receives Us program In order  to produce  the gain. It
certainly  receives inputs from various subcortlcal re-
gions and other cortical  areas. Evidence (Imbert et al.
 1966) from  "cortical shell"  preparations in which
only the  cortex, white matter, and its circulation are
left intact, suggests that  frontal granular cortex is the
home of the cerebral map. Stimulating every part of
the cortex and recording from every part of the cor-
tex reveals that practically everything  leads to the
frontal  cortex, including,  via cortical-cortical  path-
ways, all association and primary receiving areas.

   In man  the  inferior thalamic  peduncle  travels in
the medial ventral quandrant of the forebrain, a re-
gion normally regarded as unsafe for surgical interven-
tion.  Therefore, Skinner considered it unlikely  that
transections in Zappoli's  patients  would  have im-
pinged  upon the ITP. Skinner also argued that the
observation  of  CNV  in  prefrontal  lobotomized pa-
tients was not entirely inconsistent with his model. If
lesions  in thalamocortical pathways in  these patients
were  extensive enough  to rule  out this  source of
input, then the presence of CNVs strengthens the sug-
gestion  that frontal cortex  receives its program via
cortico-cortical pathways.

   Skinner replied  that  on the whole Dr. Zappoli's
data  tended  to expand  rather  than contradict his
neurophysiological  model.   One of the outstanding
questions is how the  frontal cortex receives its sig-
nal to produce a slow potential.  This region receives
axons from both the thalamus and the cortex. Physi-
ological evidence (Imbert  et al.  1966)  from cortical
shell  preparations in which only the  cortex,  white
matter  and circulation are left intact, suggests that
the  frontal cortex  is the  "Rome" of the  cerebral
mantle  to which "alt  roads  lead."   That is, the sen-
sory and association cortices all project, unidirection-
ally,  to the frontal  lobe.  If lesions in  thalamocorti-
cal pathways were extensive enough In Dr. Zappoli's
patients to rule out the subcortlcal source, then the
continued presence of the CNVs strengthens the sug-
gestion that the frontal cortex  receives Its SP-evoking
input   via   cortico-cortical  projections.    Perhaps,
Skinner said, an arrow should be drawn from the pro-
jection  cortices (PC) In  Pig. 4 to the frontal cortex
(FC), but this action must await experimental obser-
vation  in animals in  whom the surgical disruptions
can be completely controlled.

   Degeneration in  medlalis dorsalis thulaml does not
Indicate extensive  frontal granular cortex denerva-
tlon.   The  entire medial thalamic  nuclear complex
systematically projects to  the  frontal  lobes (Klevit
and Kuypers 1977). Partial blockade in the inferior
thalamic peduncle  or the  frontal granular  cortex on-
ly  partially  reduces  the amplitude of  cortical  or
thalamic recruiting  responses and  only partially de-
grades behavioral performance  (Skinner and  Llndsley
 1967).  In man the ITP travels in the medial ventral
quadrant of the forebrain, a region  normally regarded
as unsafe for surgical Intervention. Therefore Skinner
considered it  unlikely that transsections in Zappoli's
patients would  have  impinged upon the  ITP  where
the "bottleneck" of fibers are bundled together.

-------
 A  PARSIMONIUS MODEL  OF MAMMALIAN BRAIN
 AND EVENT RELATED SLOW  POTENTIALS1
 T.J. MARCZYNSKI

 Department of Pharmacology and Psychiatry, University of Illinois,
 Chicago, IL, U.S.A.
    During the past 15 years new ideas have been for-
 mulated concerning the basic principles that govern
 the functioning of self-organizing systems (von Foers-
 ter  1960, MacKay 1965, Uttley 1976) and the mode
 by which the information may be transmitted (Mar-
 czynski 1976, Brudno and Marczynski 1977). Models
 of  mammalian brain have  been  presented before
 (John 1967, Smythies and Adey  1970, Kornhuber
 1973, Pribram and McGuinnes  1975, Marczynski and
 Burns 1976), and many aspects of the model present-
 ed here incorporate  previous ideas, particularly those
 of Kubie (1953),  MacKay (1965),  and Maclean
 (1970). The author realizes that the model may be
 repudiated by some  investigators as an oversimplifica-
 tion; however, despite its parsimonious character, this
 model pulls together a wide spectrum of experimental
 and clinical data, including those  on event-related
 slow potentials (ERSPs) and their pathognomonic sig-
 nificance. Hence, the oversimplifications may perhaps
 be redeemed by the potential heuristic value of the
 model.

   Fig. 1  illustrates  a simple feedback control loop
 proposed by MacKay (1965). Note that hypothetical
 substrates responsible for generating the indication of
goal (!Q)  project  to a matching or comparator (C)
 system. In order to function, C must receive feedback
 in the form of sensory input generated by the organ-
 ism  during interaction with the environment in the
field of action. The brain must also contain substrates
 that can be labelled  as selector whose main function
is  to determine the general mode of motor response
by influencing the effector, i.e., all substrates respon-

1 Ed. Note: This model originally constituted the con-
cluding section of the paper by Dr. Marczynski on
'Neurochemical mechanisms in SP genesis: a summary*
which  is included  elsewhere in  this volume. Because
of its direct relevance to the topic of Neurophysiolog-
ical Models it has for convenience been incorporated
at this point, but should be read in conjunction with
the full paper.
sible for execution of movements. The effector thus
represents the motor repertoire of the organism.
   Fig. 2 illustrates a more elaborate model of infor-
 mation  flow  in the mammalian brain. This model,
 based on the theoretical servo-loop shown in Fig. 1,
 differs from  the  MacKay model in several crucial
 ways. First, anatomical and physiological details of
 well-known pathways and local neuronal circuits have
 been added to the cybernetic framework of MacKay.
 Second, the mammalian model includes two distinct
 feedback loops, whereas the cybernetic model pro-
 posed only one which does not reflect the dichotomy
 or "schizophysiology", as MacLean  put it, of the
 manner  in  which  sensory input is processed in the
 mammalian brain. The model we propose  (see also
 Marczynski and Burns 1976) is consistent both with
 the basic functional anatomy of the mammalian brain
 and with the cybernetic model of Uttley (1976), who
 used a computer model of two qualitatively different
 pathways impinging on the same substrate to provide
 mathematical and  experimental evidence for the ra-
 tionale and advantage of such an arrangement. De-
 spite the mushrooming of the association  thalamo-
 cortical  system and the cortico-cortical pathways in
 higher mammals and man, the phylogenetically older
 medial forebrain bundle (MFB) continues to perform
 a vital function in conveying septal feedback to the
 limbic system. Moreover, the selector is differentiated
 into  two basic components: one selecting a general
 mode of motor response and the other labelled as the
ascending reticular activating system (ARAS), which
 carries two functionally different projections: a cho-
 linergic and a catecholaminergic. Finally, at the tha-
 lamic and cortical level the basic recurrent inhibitory
 circuits have been included; they are  responsible for
 the phasing of neuronal activity into the alpha-like
 EEC patterns  associated with hyperpolarization  of
large populations of neurons in the thalamus and cor-
tex (cf, Andersen and Andersson 1968).

-------
Parsimonious Model of Mammalian Brain
627
  Fig. 1. Simple feedback  control loop  of MacKay (1965), depicting the flow of information responsible for
  adaptive behavior. Components include  an indication of goal {!Q), comparator (C) or evaluator of the status of
  the system resulting from selection  fS) of a particular response from the effector (E) repertoire, and an indicator
  (I) of the quality of the process under control (e.g., temperature).
                                                                      MEDIAL FOREBRAIN
                                                                      BUNDLE AFFERENT5
        INHIBITORY SYNAPSE
        INHIBITORY INTERNEURONS
        EXCITATORY SYNAPSE
    Fig. 2. Model of information flow and processing of sensory input in the mammalian brain. See text for details.

-------
  628

  Hippocampal main input-output relationships

    The basic "wiring" of the limbic system is repre-
  sented by a CA1 pyramidal cell of the hippocampus
 (matching),  whose synaptic organization is  relatively
 well established (MacLean  1970  and Gloor  et  al.
  1964). The  single cell receives  two different inputs:
  one is transmitted from adjacent entorhinal cortex via
 perforant pathways of Cajal to apical dendrites of the
 cell, and the other comes from the MFB via the septal
 pathways and impinges on basal dendrites  or  proxi-
 mal sections of apical dendrites. Since septal input is
 much closer to the soma than is cortical input, septal
 input  is much more influential than perforant input
 in modulating comparator function  (cf.  MacLean
  1970). It should be noted that  the corticofugal path-
 ways can reach the hippocampus either monosynap-
 tically, as shown in  Fig, 2, or via three synapses: den-
 tate cell dendrites, CA3 dendrites,  and Schaeffer col-
 laterals (Smythies  and  Adey  1970),  which are not
 shown for simplicity.

    Typical responses  to a volley of impulses in the
 afferent perforant  path are excitatory postsynaptic
 potentials (EPSPs) in the apical dendrites.  The result-
 ing weak depolarization is insufficient, in most in-
 stances, to fire the cell. The cell membrane, however,
 is biased, and the threshold may be reduced so that
 the cell can  be more readily fired by input  from the
 septum. Such a coincidence of two excitatory Inputs
 on the same cell may serve as a conditioning mecha-
 nism in which the effect of weak input  from  the ento-
 rhinal  cortex depends upon input  from the septum
 (cf. Gergen and MacLean 1964).

 Thalamocortical feedback

   During  a  successful  or  uniucceisful Interaction
 with the environment, sensory  Input,  both proprio-
 ceptive and  exteroceptlve (including that  from the
 cerebellum,  channeled  via  the bmchlum  con/un-
 cttvwn) is projected  to specific thalamocortical re-
 gions where the precise physical, spatial, and tempo*
 ral characteristics of each modality are analyzed, Sub-
 sequently,  the modalities converge and Interact  in
 numerous association areas of the thalamus  and cor-
 tex. Thus, for instance, the distance from  subject's
 hand to the perceived object is evaluated not only by
 "purely" visual characteristics, but also in  termi of
 muscle effort that might  bo necessary to  reach and
 lift the  object  (cf. Eccles 1966), It li  apparent that
 these synthetic judgments, in most Instances not real-
 ized by the subject, are based on previous lomesthetic
and  klnesthetlc experience, and they can be stored
and  retrieved as learned "praxlas" (cf, Kornhuber
 1971). The importance of active movements in the
development  of visual perception was  demonstrated
long ago (Held and  Hein  19S3).  Thus, the three-
dimensional visual world always has strong somesthet-
ic and kinesthetlc components that must be consider-
                                      Marczynski

 ed in trying to understand the genesis and topographi-
 cal distribution of Sps. Hence, it is not surprising that
 the CNV, P300, the large Bereitschaftspotential pre-
 ceding the Premotor Positivity (PMP) of Deecke et al.
 (1973) followed by  a  much deeper and long-lasting
 positivity, as well as  the Skilled Performance Positivi-
 ty of Papakostopoulos (this volume), and the positivi-
 ty observed by Otto et  al. (this volume) all occur over
 large areas of association cortex.

    A more recent systematic study of the cortico-
 cortical pathways (van Hoesen et al.  1972) showed
 that each primary receiving cortex  projects simulta-
 neously to two topographically and functionally dif-
 ferent association areas. For instance, visual informa-
 tion from striate cortex is projected to inferotempor-
 al  cortex and  to the parietal association area. In  the
 latter, visual information converges  with similar pro-
 jections from somesthetic and auditory cortex. In the
 infero-temporal  cortex,  visual  information  from
 striate  cortex  converges with  fronto-orbital projec-
 tions. Subsequently, integrated information  is projec-
 ted via two distinct  pathways: one  to fronto-orbital
 association cortex and the other to the ventral surface
 of the temporal lobe. The main feature of all cortical
 association areas  is that their final  projections con-
 verge on the ventral surface  of the temporal lobe. The
 entorhinal cortex, which receives the final "version"
 of the  information transforms resulting from  poly-
 modality interactions, projects via the perforant path
 of Cajal to hippocampal structures. It is therefore not
 surprising that even relatively small  lesions  involving
 the latter two links In cortical feedback to the limbic
 system  may totally disrupt adaptive  behavior in man,
 cause  hallucinations  involving  all  modalities,  and
 sometimes produce  a   syndrome   Indistinguishable
 from schizophrenia (cf.  Teuber 1972).

 Medial forebraln bundle (MFB)  feedback

    MFB pathways and their numerous  links with the
 reward-punishment system  (cf.  Olds   1964, Miller
 1961) process  and categorize  sensory input in terms
 of  the fulfilment of organism needs. These functions
 are evidenced by the fact that the rate of self-stimula-
 tion in  these pathways  and/or nuclei associated with
 them  Is closely related  to  primary drives. There are
 loci in which self-stimulation  rate is related to food
 deprivation and Inversely related  to androgen  level
 and sex  drive.  At another nearby locus, self-stimula-
 tion rate  may covary with serum androgen level and
 with food deprivation, ot It may only depend on an-
 drogen  level (cf. Olds  1961,  Brady 1961).  How-
 ever,  the view  that  level  of  motivation Is  directly
 related  to specific behavioral patterns  or  the  sum-
 total of individual drives would be a gross oversimpli-
 fication. Even  in lower mammals,  the manner  in
which  primary  drives  and  reward-punishment sub-
strates function shows  that emotional states gener-
ated in pathways Involving the hypothalamus, anteri-
or  thalamus, forebrain, and midbraln limbic system

-------
Parsimonious Model of Mammalian Brain

(e.g., Papez circuit) strongly influence the expression
of primary drives. The latter are  even further influ-
enced by previous experience (cf. Sherer 1961).

   The modulation  of primary drives becomes gradu-
ally more apparent in higher mammals, particularly in
primates and  man (cf.  Brutkowski  1965, Nauta
1971). In man the  frontal association cortex and its
reciprocal connections with other systems have all the
attributes to  function  as the main modulator of the
indication of goal (cf. Nauta  1971). Numerous clin-
ical data strongly support this view (cf. Teuber 1972).
In man the substrates that generate primary drives are
modulated  by organismic needs as well as by cultural
tradition, ethical considerations,  and philosophical
concepts transmitted from one generation to  another.
Man is the  only creature that can  walk to the gallows
with a smile  to defend his beliefs.  All these consider-
ations provided the rationale for  separating the indi-
cation of goal from MFB  pathways in the proposed
model.

   More recently,  the strategic position of frontal
association cortex was further emphasized by electro-
physiological confirmation of its commanding role in
the regulation  of the thalamic reticular nucleus  (R),
which can  be  best defined as a  pool of inhibitory
neurons responsible for gating sensory input at the
specific  thalamic  nuclei (Skinner, this section). It
appears almost certain that a specific spatio-temporal
pattern of excitatory impulses from the frontal assoc-
iation  cortex to R determines the scope of  selective
attention. Accordingly, despite the wide distribution
of the CNV  over  frontal  and postcentral areas, this
potential should have modality-dependent  compo-
nents.   Indeed, two  components of the CNV with
characteristic distributions have  been observed  dur-
ing  auditory  and  visual  discrimination  tasks  (cf.
Hillyard 1973).

Thalamocortical vs. medial forebrain  bundle
feedback

   Thalamocortical input  is extremely variable and is
designed to provide an objective analysis and a faith-
ful image of rapidly  changing  physical properties of
environmental and proprioceptive  stimuli.  Within a
 fraction of a second,  patterns of sensory input  may
dramatically   change,  and  these  changes are  most
 likely reflected in transforms of information  con-
 veyed to the limbic system  from the cortex via the
 perforant pathway of Cajal. In contrast, the  traffic of
 impulses in   MFB-septal  feedback  to hippocampal
 structures  (and other  limbic regions) is likely to re-
 flect relatively slow  changes in  the functioning of
 centers  reflecting  bodily  needs  that are protected
 against  dramatic  swings  by  numerous homeostatic
 mechanisms  designed  to  control the internal milieu
 and primary drives. The  information content of the
 final output of this system  impinging on the hippo-
 campus and  other  limbic structures most likely
                                            629

changes over a time period of hours or even days.
Hence, this  homeostatically graded  stream  of infor-
mation, which sculptures our personality and is ever
present, may be  compared to an unconditional stim-
ulus, whereas the highly variable and brisk barrages of
information  impinging on the limbic system from the
thalamocortical  system  may be  compared  to  con-
ditional stimuli.

   Kubie (1953) was the first to suggest that duality
of experience may be the basis of cognitive processes
and even memory. MacLean (1970), after reviewing
the unique  synaptic organization  of  CA1  hippo-
campal  pyramidal cells, claimed that  the "union of
internal and external experience is  as important for
memory as the combination of antigen and antibody
in developing an  enduring immunity". As already
mentioned,   mathematical   analysis  and  computer
studies of the pattern recognition capacity of units
that  simultaneously  receive a stable  and a variable
input revealed the extraordinary  power of such sys-
tems in categorizing the  information  in the variable
input (Uttley 1976). A network  of only 210 units,
called informons, and 8400  "varying" channels could
"learn" to recognize hand-printed numerals.

Hippocampal projections to the brain stem

   The  limbic  descending  pathways  may  excite or
inhibit the brainstem reticular formation (BSRF) and
ARAS (cf. Smythies and Adey 1970). The direction
of these influences seems to depend on the outcome
of the  matching  process  in  the hippocampus and
other  parts  of the  limbic system.  The  strongest
arousal influences are generated when the organism is
confronted  with a novel  barrage of sensory input.
Upon  repetitive  presentation of such stimuli, and
 their classification and subsequent anticipation habit-
 uation occurs. After full habituation, signs of active
 suppression  of  the  BSRF   and  ARAS, as  evidenced
 by high-voltage, slow-wave EEC patterns (cf. Magoun
 1964), are usually observed.

   Relevant to our holistic view of brain organization
 is the morphology of single BSRF units described by
 the  Scheibels (1967). BSRF units are characterized
 by profuse  axonal ramifications, both ascending and
 descending,  projecting  to the  thalamus,  caudate
 nucleus, hypothalamus, as  well as  to the brain stem
 motor nuclei. A substantial portion of BSRF neurons
 project rnonosynaptically   to the  neocortex.  In  all
 these  areas, each BSRF unit (including monoamin-
 ergic neurons)  is  likely  to establish  thousands  of
 synaptic contacts. Since each unit has its own charac-
 teristic domain  of influence,  each neuron  may be
 compared to a specialized  module capable of activat-
 ing and/or  inhibiting uniquely a specific set of neu-
 rons in all aforementioned areas. All these attributes
 make the BSRF system the best choice for a selector
 of gross behavioral modes in our model.

-------
 630
                                     Marczynski
   During exploratory behavior and cognitive  proc-
esses associated with selection of the most successful
behavior in  the field  of action, limbic influences on
the BSRf-selector seem to oscillate (cf. Smythies and
Adey 1970). Such influences enable the organism to
switch from one  cognitive process and  behavioral
mode to another, thus introducing maximum  flexi-
bility and the elimination of  unsuccessful behavior.
        NOR
Evidence  for such a  role of the descending limbic
pathways  in the maintenance of behavioral plasticity
is  abundant (cf. Smythies and  Adey  1970, Pribram
and McGuinnes 1975).
   If the behavioral mode is successful, i.e., provides a
match with the indication of goal, it is reinforced and
                   5 sec
 l'"ig. 3. Sensitivity of reward  contingent positive  variation  (RCPV) in the cat to  changes  in the quality  of
anticipated reward, a. On a variable interval schedule, 0.5 cc of water (W)  or milk (M) was presented after a bar
 press.  NOR,  nonrewarded bar  press; REINF, rewarded bar press. PM,  recording over posterior marginal gyrus
 (striate cortex) referred  to  medial suprasylvian gyrus, as shown in c. Lower tracing in a  and  b filtered  to
 half-amplitude response at 3 c/sec. c. Mirror reversal patterns of RCPV in bipolar recordings over PM and anterior
 ectosylvian gyrus  with reference to medial suprasylvian gyms, a phenomenon  caused by a decreasing  potential
gradient of RCl}V in the anterior lateral direction from the locus of maximum aptitude over PM. Note similarity
 of RCPV with sleep onset positive SP in b and d. (Marczynski et al. 1969.)

-------
Parsimonious Model of Mammalian Brain

perfected. Moreover, it must  be  preserved by strong
inhibitory influences from the limbic system (match-
ing)  impinging on  the  BSRF-selector  and  ARAS
immediately after a correct response. Fig. 3. shows
the sensitivity of limbic inhibitory influences to un-
expected changes in the results of a particular be-
havior.  After a bar press rewarded with 0.8 cc of
milk, cats show bursts of waxing  and waning EEC
synchronization consisting of  7 to 9 c/sec waves over
the parieto-occipital region (Clemente  et al. 1964), a
phenomenon associated with a large positive SP shift
termed  Reward  Contingent  Positive  Variation
(RCPV) whose patterns  are  virtually identical  with
those recorded during  sleep onset (Marczynski et al.
1969,  1971a). A  lesion in the basal forebrain region
encompassing the limbic descending pathways blocks
the occurrence of these  EEC phenomena (Sterman
and  Wyrwicka  1967).  Systematic studies of evoked
potentials to auditory, somesthetic, and visual stim-
uli,  showed  that the  processing  of  sensory  input
during reward-induced EEC  phenomena is virtually
identical with the processing of input during sleep
onset (cf. Marczynski 1972). Electrical stimulation of
basal forebrain areas along limbic descending path-
ways may produce  a stimulus-bound  sleep (cf,  Ster-
man and  Clemente  1962, Sterman and  Wyrwicka
1967), and  suppression of single unit activity in the
ARAS (Lineberry and Siegel 1971).
   In man during performance in a typical CNV para-
 digm, the CNV is often  abruptly  terminated  by a
 long-duration positivity (e.g., Otto  and Leifer 1973,
 Hablitz  and  Borda  1973,  Loveless  and   Sanford
 1973),  which may reflect a transient but powerful
 inhibition  of the BSRF-ARAS system. This  interpre-
 tation is also plausible for the positivity that follows
 the CNV-like potential in paradigms  that do not re-
 quire  motor  response  but  only performance  of a
 mental task (Donchin et al.  1973). The positivity
 observed  after SI  in  the CNV paradigm, including
 most P300 waves, may be triggered by a cognitive
 process  based on  matching sensory  input  with a
 "template" generated  by  the indication of goal and
 limbic  system, a  process resulting in transient but
 strong suppression  of the ARAS. This interpretation
 is compatible with  the suggestion made by Squires et
 al. (1973) and Sutton et al. (1973) that the P300, in
 most  instances,  results  from  "resolution  of un-
 certainty".
   Cooper et al. (this volume) described a large posi-
 tive SP over the vertex and occipital cortex. This SP is
 associated with detection of an  anticipated visual
 event, and it occurs prior  to motor  response. The
 duration  of approximately  1  second  indicates  that
 this SP could not have resulted from  a discharge of
 neurons and cortical "sink"  for current flow. Instead,
                                             631

this SP probably resulted from hyperpolarization of a
large population of neurons triggered by a  transient
suppression of the ARAS (see below). Convincing evi-
dence for response-induced suppression of the ARAS
and inhibitory  nature of the surface  positive SP  of
longer duration, comes from the study of Skilled Per-
formance Positivity (SPP) of Papakostopoulos et  al.
(this volume).  SPP is maximal over the vertex and
encompasses the Rolandic area. It is preceded by  an
unusually large Bereitschaftspotential and is observed
in healthy volunteers asked to perform a manipulative
task demanding precision and gradual improvement
of performance by providing real-time information
about the outcome. It is apparent that this  paradigm
strongly  engages the function of the indication  of
goal (frontal association cortex)  and matching sys-
tem.  The  most  important  aspect   of SPP  is  the
accompanying high-voltage, waxing and waning, 9- to
11 -c/sec  EEC oscillations over the  same region. In
view of  what is known about the  synchronization
process (see below), such as alpha-like burst and slow
potential could not emerge without transient but
powerful suppression of the BSRF-ARAS system.
 Hippocampal projections to the cortex
    In  order  to generate specific spatiotemporal pat-
 terns of neuronal activity and retrieve learned behav-
 ioral modes or find  solutions to pure mental tasks,
 the matching (i.e.,  categorizing) substrates must be
 promptly  converted  to a retrieval system  under the
 influence  of the indication of goal.  During this
 process the limbic system may send commands simul-
 taneously to the BSFR-selector and to wide areas of
 neo cortex via pathways described by Lorente de No
 (cf. Smythies and Adey 1970)  and marked as a bi-
 furcating  axon of  the  hippocampal pyramidal cell
 (matching) in Fig.  2. Following Kornhuber's view
 (1973), it must be emphasized that in man the limbic
 projections to  neocortex that emerge from the classi-
 cal Papez  circuit (not shown), particularly those that
 relay in dorsomedial thalamus and frontal association
 cortex and then  project to the ideational speech area
 of Wernicke, are essential in both the matching and
 retrieval processes. Hence, by projecting the relatively
 stable stream of "visceral" input to the cortex, the
 primordial categorizing and retrieval system sketched
 in Fig.  2, achieved  an enormous level of sophistica-
 tion. Correlational studies of EEC theta waves in the
 hippocampus and entorhinal cortex during learning in
 animals suggest  strong limbic influences on cortical
 transactions, a process that  is reversed after learning
 (Adey et  al. 1961). These limbic - cortical transac-
 tions  are  modulated  by nonspecific brainstem-thal-
 amocortical projections (see below). Topographically
 restricted  SPs  monitored from  the  skull almost cer-
 tainly result from these transactions.

-------
 632

 ARAS control of  thalamocortical synchroni-
 zation and positive SPs

   The  function of recurrent inhibitory circuits pres-
 ent both in specific thalamic relay nuclei (cf. Ander-
 sen and Andersson 1968) and in cortex (cf. Steriade
 and  Deschenes  1973) is inversely related to ARAS
 inhibitory  modulation of inhibitory interneurons, as
 shown in Fig. 2. Although there is still no direct evi-
 dence regarding the nature of the inhibitory transmit-
 ter involved in ARAS suppression of inhibitory inter-
 neurons, there are some indications that these ARAS
 projections may be catecholaminergic (cf. Marczyn-
 ski,  this volume). On the  other hand, it  is almost
 certain  that  the transmitter released  by  inhibitory
 interneurons   is gamma-aminobutyric  acid  (GABA)
 (cf. Krnjevic 1974).

   Emergence of alpha-like  oscillations in the  EEC
 depends on two factors, reduced  tonus of  the sup-
 pressant action  of the ARAS  on R  and  inhibitory
 interneurons,  and a  sufficient barrage  of sensory
 input to drive the recurrent inhibitory circuits (cf.
 Anderson and Andersson 1968).  If the sensory bar-
 rage  is insufficient, alpha activity can be restored in a
 stimulus-bound manner by electric stimulation of the
 lateral geniculate (Rick and Marczynski 1978). Corti-
 cofugal  activity, e.g., activation of motor cortex, also
 may  provide  electromotive  energy  to initiate  the
 function of recurrent  inhibitory circuits immediately
 after an abrupt withdrawal of ARAS influences. It
 appears that  tonic facilitory  influences of the choli-
 nergic  ARAS on  sensory  transmission in thalamic
 relay nuclei are necessary for alpha emergence over
 wide cortical  association regions, since antimuscarinic
 drugs, such as atropine or scopolamine, are known to
 block these  EEC  patterns.  After smaller doses of
 these drugs,  the bursts of alpha become  "choppy"
and are intermingled  with irregular delta waves that
most likely result from irregular "idling" of recurrent
inhibitory circuits (cf. Marczynski and  Burns  1976), a
state which can be overcome by electric stimulation
of the  lateral geniculate  (Marczynski, unpublished).


   An  important aspect of the thalamocortical phas-
 ing of neuronal activity is that only about one-third
 of the neurons actively participate in the recruitment
 of inhibitory  circuits,  as   evidenced  by  rhythmic
 sequences  of IPSPs   and post-inhibitory  discharges.
 The  rest of the neuronal population  remains hyper-
 polarized and silent (cf. Watanabe et al. 1966, Ander-
 sen and Andersson  1968,  Marczynski and  Karmos,
 this volume).  At the cortical level, hyperpolarization
 of larger pyramidal  cells results  in a positive  wave
 (Watanabe  et  al. 1966).  Similarly, hyperpolarization
 of neurons in subcortical structures is also reflected as
 a local  positive SP (cf.  Skinner,  this section; Mar-
 czynski, this volume).
                                    Marczynski

   A relatively  small change in rhythm of sequential
IPSPs affects the inhibitory process. A study of the
recovery cycle  of two sequential IPSPs triggered in
feline visual cortex by stimulation of afferent path-
ways (Watanabe et al. 1966) showed that a time inter-
val of  130 msec between the onset of two  IPSPs is
optimal to obtain  partial summation and maximum
hyperpolarization.  This  interval  corresponds  to a
7.7-c/sec high-voltage EEC pattern normally observed
in unrestrained cats during sleep onset or in a fully
alert animal after  a rewarded bar press  during the
emergence of RCPV (Marczynski et al. 1969,1971 a).
Thus, it appears  that the thalamocortical system is
"using" the most effective rhythm to produce inhibi-
tion. Data of Watanabe  et  al.  (1966) show that a
20-msec reduction or increase  in the  time  interval
between sequential  IPSPs (corresponding  to a
1.4-c/sec change from the  7.7-c/sec rhythm) would
result in approximately 40% reduction in hyperpolari-
zation.  A similar  conclusion can be drawn from a
study of the RCPV and the topographically restricted
sleep onset positive SP  in the cat induced by flash
stimuli  at  various frequencies  (Marczynski et  al.
1971a,  Marczynski and Sherry  1972). There  is  no
reason  to believe  that the basic neurophysiological
characteristics of the feline thalamocortical system
are different  from those of man. Hence, the  seem-
ingly  small  shifts  in the  power spectra of  alpha
patterns and  their "choppy"  character in schizo-
phrenic patients, coupled with  suppression of slow-
wave sleep delta patterns (cf. Itil et al. 1972), could
reflect  a substantial loss of the inhibitory capacity of
the thalamocortical system.
   Another interesting aspect of thalamic recurrent
 inhibitory circuits is that they are capable of convert-
 ing a totally patternless (i.e., "noisy") sensory input,
 devoid  of any conceivable  informational properies,
 into an effective and most  likely meaningful inhibi-
 tory process (Marczynski et  al. 1971 b; cf. Marczynski
 and Burns  1976; Marczynski and Karmos, this vol-
 ume; Rick and Marczynski  1978).  Discussion of this
 topic is,  however, beyond  the scope of this  article.
 Basic modes of operation of the thalamocorti-
 cal system; diffuse vs. localized EEC and SP
 patterns

    Despite  the  diffuse character  of cholinergjc and
 catecholaminergic projections from brainstem nuclei
 to the midbrain, forebrain, and  neocortex (cf. Hock-
 man and Bieger 1976), these ARAS components are
 essential  in generating topographically restricted  BEG
 patterns  and SPs. Recent  studies of the role of the
 nucleus reticularis thalami (R) (cf.  Skinner, this sec-
 tion)  have  clarified  the  contradictions by showing

-------
Parsimonious Model of Mammalian Brain

that specific parts of R send inhibitory axons to par-
ticular regions of sensory thalamic nuclei, thus selec-
tively gating sensory input of one or several modali-
ties. Furthermore, the spatial  and temporal patterns
in activation of  R neurons are determined  by im-
pulses coming  from fronto-orbital cortex, which  is
one  of the main  modulators  of indication of goat.
(For simplicity, inhibitory influences of the caudate
nucleus will not be discussed  and are not shown in
Fig. 2). The   contrast enhancing inhibitory back-
ground for excitatory influences of frontal cortex on
R that may result in selective attention is provided by
tonic hyperpolarizing ARAS influences (via catecho-
laminergic projections?),  impinging on R and  other
inhibitory  interneurons located outside R and in neo-
cortex  (cf. Steriade and Deschenes 1972, Andersen
and Andersson  1968).  Furthermore, the cholinergic
component of the ARAS  facilitates transmission of
sensory input at the thalamic and cortical level, and
thus further sharpens the contrast among "selected"
specific projections (cf. Marczynski,  this volume).
The interplay  between these influences may result in
several alternatives.

   Strong activation of the cholinergic and catecho-
laminergic ARAS  would  result  in hyperpolarizing
blockade of R and inhibitory interneurons, and lead
to inactivation of recurrent  inhibitory  circuits.  A
strong  facilitation of sensory input in all modalities
would  be  associated  with desynchronized EEC pat-
terns, strong  arousal, and diffuse negative SPs over
the  cortex and in specific thalamic nuclei. On the
other hand, a dc electrode in R would  show a large
positive SP (see Skinner, this section). In such a state,
the  modulation  of R by  excitatory  barrage  from
fronto-orbital cortex (Indication of goal) would not
be effective since it could not override the ARAS-
induced  inhibition.  Selective  attention would  be
blocked,  and  abstract thinking markedly Impaired.
The release of acetylcholine  from the  cortex would
be markedly Increased,  but  the  release of GAB A
would be totally  blocked  (cf. Marczynski,  this
volume).
   Mild activation  of the ARAS, with a slight pre-
 dominance  of  the  cholinergic  component,  would
 provide a background for most complex and selective
 transactions at all lavels. Mild luppreuion of R by the
 ARAS  could easily be  overriden by excitatory  in-
 fluencei from the  frontal association cortex, which
 could generate highly  specific spatio-temporal excita-
 tions of modality-specific loci  in  R. The  latter, in
 turn,  would  selectively  modulate  the  function  of
 thalamic relay nuclei. Only those thalamic nuclei or
 parts thereof that are not blocked by R would show
 Increased neuronal activity and a negative SP. Corre-
 sponding cortical projections would also show nega-
 tive  and  topographically restricted  SPs  associated
 with increased unit activity and desynchronized EEC
                                             633

patterns. The activation of cortical association projec-
tions would further determine SP topography. Recur-
rent inhibitory circuits in  the cortex and thalamus
could be  readily activated or show  only  moderate
involvement because of the suppression of inhibitory
interneurons.

   After successful performance of a motor or mental
task,  basal  forebrain  and  limbic influences  would
suppress the ARAS.  This suppression  would lead
to  vigorous  activation of  recurrent  inhibitory cir-
cuits, mainly or selectively in those thalamocortical
projections exempt from R inhibition, because only
in  those regions would there be sufficient synaptic
drive  to activate recurrent inhibitory circuits. Hence,
in  most instances,  substrates that, prior to the re-
sponse or mental problem, showed a negative SP and
desynchronized  EEC  patterns  would dramatically
shift toward  EEC synchronization of the alpha type
and a strong positive SP (e.g.,  Skilled Performance
Positivity or  RCPV in animals).  In cortical areas
which receive only weak sensory input, such as motor
cortex,  activation of pyramidal  tract  neurons  during
task performance could also  be  promptly converted,
via axon collaterals,  into vigorous activation of re-
current  inhibitory  circuits immediately  upon  task
completion.  This sequence of events  and ARAS sup-
pression would  lead  to the  emergence  of a  strong
positivity associated with EEC synchronization. The
positivity observed  after the CNV may belong to this
category.

    In  this context,  long-lasting positivity, maximal
 over the  post-central region,  observed during sus-
 tained motor response (Otto et al. 1974) Is difficult
 to explain, If sustained contraction of the  hand (most
 likely  involving some  muscles of the  forearm and
 shoulder) Is, however, associated with  massive dis-
 charge of neurons located deep In the central sulcus,
 then the deep negativity caused by depolarization  of
 neurons, apical dendrltes, and glia cells would consti-
 tute a strong sink  for current flow from surrounding
 areas that should be  reflected as positivity at the ver-
 tex. Distribution of SPs In animals supports this Inter-
 pretation (Marczynski et al, 1971a), since a strong
 positivity over the striate cortex ranging from 200 to
 400 yV measured epiduratly constitutes such a strong
 "source" for current flow  that surrounding cortex,
 e.g., the ectosylvian or the  anterior  suprasylvlan gyri,
 shows a smaller but consistent negative SP. The mech-
 anism  for  such a reversal of SP has been discussed
 elsewhere (Marczynski et al. 197 la).

  Selected aspects of ARAS influences on corti-
  cal function  and evoked  potentials:  A  new
  working hypothesis

    The basic theory  of evoked potentials (EPs)  has
  been  discus&ed  before  (Creutzfeldt  et al.  1969;
  McSherry 1973; Marczynski, this volume). As shown

-------
 634

 in Fig. 2, the interplay between excitation and inhibi-
 tion induced  by cholinergic and catecholaminergic
 projections, respectively,  is of primary importance.
 The depolarizing and profuse  cholinergic  terminals
 seem to be concentrated on cell bodies and, possibly,
 dendrites of large pyramidal cells of layers 4 and 5.
 These  cells are  endowed with  "pure" muscarinic
 receptors  (cf. Krnjevic 1974).   The distribution of
 noradrenergic  and catecholaminergic terminals is not
 yet  defined,  but it  seems logical to assume that a
 catecholaminergic input  is provided  to gaba-ergic
 inhibitory interneurons. Indirect evidence for catech-
 olaminergic modulation of the  inhibitory neurons of
 R is reviewed elsewhere (Marczynski, this volume).

   The presence  of  dopaminergic terminals  in neo-
 cortex (Thierry et al. 1973) and the potential inter-
 action of such terminals with cholinergic postsynaptic
 receptors of pyramidal cells and dendrites may be of
 primary significance in understanding pathognomonic
 aspects of EP  changes observed  in schizophrenics (cf.
 Shagass 1976). The role of the  dopaminergic system
 in the etiopathogenesis of schizophrenia is well estab-
 lished (cf. Meltzer 1976). Studies of the interaction
 between  dopaminergic and  muscarinic  cholinergic
 receptors in the autonomic ganglia (Libet, this  vol-
 ume) have shown that  dopamine dramatically en-
 hances and prolongs cholinergically mediated depolar-
ization of the postsynaptic membrane.

   Extrapolating to the cortex, one could predict that
an increased tonus of the dopaminergic system would
prolong and impair  the  recovery  of the membrane
                                     Marczynski

 potential of cholinoceptive pyramidal cells and dend-
 rites which to a  large extent determine EP patterns.
 This mechanism would account for several well-estab-
 lished  observations in schizophrenic patients, such as
 reduced amplitude recovery of somesthetic, auditory
 and visual  EPs (Shagass 1976); lower CNV amplitude'
 (Dongier 1973, McCallum and Abraham 1973) due to
 tonic depolarization  of apical dendrites; and prolong-
 ation or even lack of recovery of the CNV and the
 post-imperative  negative  variation  (PINV) (Timsit-
 Berthier 1973, Dongier et al. 1974).

   The prolongation of the spiral aftereffect (Herring-
 ton and Claridge  1965) also belongs to the same cate-
 gory  of phenomena. Callaway et al. (1965)  found
 that greater  variability in the  waveform  of  EPs in
 schizophrenics correlated with degree of thought dis-
 order, but  not with other symptoms such  as changes
 in affect or cooperativeness (Jones et al. 1966). This
 finding may also be explained by greater variability in
 the recovery of membrane potentials  "sensitized" to
 acetylcholine by an excess of dopamine or by a genet-
 ically  determined  misplacement  of dopaminergic
 terminals.

   With the consolidation of the "dopamine  theory
 of schizophrenia" (cf. Meltzer 1976),  there is a tend-
 ency to believe that  the dopaminergic projections to
 the limbic and mesolimbic structures (such as nuc. ac-
 cumbens, amygdala,  and hippocampus) are the main
culprits. This view, disregarding the neocortical mech-
 anisms, may be a  greater oversimplification than that
committed  by this  author in  sketching a  model
of the mammalian brain.
                                                      Ed, Note: A further paper relevant to the subject of
                                                      neurophysiological  models  by Dr.  Benjamin  Libet
                                                      entitled  "Slow  Postsynaptic  Responses of  Sympa-
                                                      thetic Ganglion Cells  as  Models for Slow Potential
                                                      Changes in the  Brain" appears earlier in this volume
                                                      in the Electrogenesis Section.

-------
INTEGRATIVE MODELS:  MACROPOTENTIALS  AS
A SOURCE FOR  BRAIN MODELS
DEMETRIOS PAPAKOSTOPOULOS

Burden Neurological Institute, Bristol, England

   Brain macropotentials are  products  of at  least
three factors:  the experimenter's interests, the sub-
ject's brain, and available technology. Because  tech-
nology is a matter for straightforward description, we
can, with some slight risk, omit it from discussion and
concentrate upon the remaining two factors.

   The impact of the experimenter's interests and the
way in  which he defines his variables  have a more
insidious impact upon what he records and how he
interprets  his results. As Donchin (1973) points out
"...all too often the relationship between the author's
conclusions and his data are tenuous, due to absence
of critical control conditions or to ill-chosen critical
parameters." Even earlier, Sutton (1969) presented a
"partial list of psychological terms for which evoked
potential  correlates  have been reported"  and  com-
mented "It is difficult to believe that all these find-
ings involve genuinely  different potentials. The  prob-
lem posed by a plethora of cross-cutting, poorly de-
fined concepts is not  a  trivial one. First, they  make
claim to a level of generality from which one can only
retreat. Second, they  make a  poor  foundation  for
moving forward since progress will depend less on the
broadness of the claims and more on the precise con-
trol and specification of experimental operations and
on  the precision of reasoning involved in attempts at
construct validation."

   If we are to improve upon  this rather poor per-
formance,  we  must  first examine  two  important
issues:  (1) the characteristics  of the electrophysio-
logical events with which we work and (2) the  levels
of analysis we employ  in our descriptions.

Macropotential Characteristics

   We may be guilty, in the field of human macropo-
tentials, of creating an elaborate language without
adequately  defining or  understanding the basic ele-
ments of that language. This lack of definition of
fundamental characteristics of macropotentials is per-
haps the greatest single obstacle to the formulation of
an effective theoretical model; however, the deficien-
cy has not passed entirely unnoticed.

   For example, one basic characteristic of macropo-
tentials is taken  to be their amplitude, yet  Brazier
(1964) demonstrated the ambiguity that may underly
amplitude measurement in average evoked potentials.
The cause may be either the time jitter, which occurs
across individual  trials, or the differential change in
amplitude of particular components as the  experi-
ment  proceeds. Papakostopoulos (1973) demonstrat-
ed this latter type of change in the first and second
late negative components  of the tactile EP using re-
cording  from  prefrontal cortical areas. EP  compo-
nents have  also  been  shown  by Walter (1964) and
Remond  (1964) to vary according to ongoing, intrin-
sic brain  activity at stimulus onset. Pfurtscheller and
Cooper (1975) have shown  that activities of cortical
areas  far apart as well  as local intrinsic activity can
influence the latency of individual evoked potentials
recorded  from a particular cortical area with resulting
amplitude ambiguity. In spite of these considerations,
no systematic attempt  has been made to rationalize
or define amplitude measurement.

   This lack of precision in delineating basic charac-
teristics of macropotentials is at the root  of our fail-
ure to formulate satisfactory theories about the brain
based upon macropotential data. The tendency has
been to treat macropotentials as if they were  there to
solve  problems of brain connectivity or to provide a
respectable  neurophysiological  dressing   for  vague
anthropomorphic constructs, or to justify generaliza-
tions  of principle derived from artifically created par-
ticulars.

   Some fundamental  re-thinking would at this stage
seem  advisable. We may assume that macropotentials
are a basic property of the  brain, as are weight, my-
elo-, or cyto-architectonics.  Evoked potentials can be
further characterized in terms of rise time, amplitude,
and phase or latency. We can, on the  basis  of these
parameters, begin to provide a descriptive framework
or model of the brain.

-------
 636

    Models are used by different people for different
 purposes. A common use, cited by Lachman (1960),
 is  "the  reproduction  of the theoretical prototype in
 terms of mental pictures or images."  Others regard
 this type of model building as a weak and scarcely
 reputable exercise. Yet this is not the  only function
 models can fulfill. They can be used as bases for infer-
 ence and interpretation.  The model of the servo-con-
 trol of movement (Merton  1953) is such an example.
 The known  neuroanatomical and neurophysiological
 data about the neuromuscular system were taken into
 account  in  this model of voluntary contraction. The
 proposed model suggested  indirect initiation of con-
 traction  by  activation of muscle spindles, which in
 turn activated the stretch reflex. However .more than
 a decade later, data were presented by  Valbo (1971)
 which indicate  direct activation of muscle fibers be-
 fore any change in the rate of firing of spindles. The
 original  model was modified accordingly, and a new
 servo-assistance hypothesis was  proposed (Mathews
 1972). The useful lifespan of the  new model will de-
 pend upon the extent to  which data from the experi-
 mental research it generates either support or refute it
 (Valbo 1971, Merton 1974).
   This  particular model was chosen as an example
 because  it has been used successfully in  stimulating
 and guiding research. It has led to new discoveries
 both in  animals and man over a long period of time
 and has  been applied  at both the  micro- and macro-
 levels  of the neurosciences. Another reason for the
 choice is the relatively limited experimental area from
 which the basic facts were derived. The model did not
 encompass  either  controversial  questions about the
 significance of sensory input for movement initiation
 or the complex spatiotemporal patterns of brain-body
 organization, each of which is a prerequisite of adapt-
 ive goal-directed motor activity. The model was suc-
 cessful because it was well defined and thus amenable
 to experimental verification or refutation, not  be-
 cause it was complete or because It proved correct.
   The lesson to be drawn  is that any model of the
brain, in health or disease,  that is to be based upon
macropotentials has to be preceded by a clear defini-
tion of the  characteristics of those macropotentials.
That is to say, we must first clearly specify the basic
elements of our language.
   The search for an adequate language of the brain
having its basis in macropotentials and their charac-
teristics can perhaps be discerned in the use by Walter
(1964) of such  terms as "modality  signature" and
"dispersive convergence." More recent emphasis on
waveform and distribution to interpret the functional
significance of evoked and slow potentials could be
interpreted as a step in the same direction.
                                 Papakostopoulos
     'R-1
 Fig.  1.   Superimposed  average  evoked potentials
 corticographically recorded in man from an electrode
 precen trolly  located on  the  right hemisphere (star)
 and referred to  a common  average reference elec-
 trode.  Time of stimulus presentation indicated by
 arrow.  Potentials shown following  visual (FLASH)
 auditory (CLICK), ipsilateral (R. MED. NER VE), and
 contralateral (L. MED. NER VE) median nerve elec-
 trical stimulation.
 Levels of Analysis

   The  waveform recorded with a macroelectrode is
 multidimensional in terms of both its origin and its
 description-in  terms  of origin because thousands of
 neuronal elements contribute to its generation and In
 terms of description because the macropotential can
 be approached from different levels of analysis. These
 analysis  levels  can, for convenience, be subdivided
 into local, time, and spatial.

 Local Level

   Subdural macroelectrodes chronically implanted in
 man have  revealed four  specific  characteristics of
brain electrical activity:

   1. Local polymorphism: From a particular cortical
   electrode, many or all the main  types of macropo-
   tential can  be recorded, For example,  from one
   precentral electrode  intrinsic activity, sustained,
   and evoked potentials to stimuli in various sensory
   modalities can be recorded. The latency and wave-
   form of the latter vary according to modality (Fig.
   1).
   2, Regional  reactivity: Under certain conditions
   the potentials of certain areas diminish to a point

-------
Integrative Models
                                             637
                     DISPL
  N=1«-

  20/W*
       200msec
  S/S-,
 Fig.  2.   Superimposed averaged evoked potentials
 corticographically recorded in man and following left
 index finger externally paced displacement (DISPL).
 Electrodes  located in prefrontal (upper traces), pre-
 central (middle traces), and post central (lower traces)
 cortex.

    at which they cannot be differentiated from noise,
    although the potentials  of adjacent areas  are still
    recordable (Papakostopoulos et al. 1975).
    3. Component independence; Certain components
    of an EP waveform can be localized or generalized
    according  to behavioral  conditions. A typical ex-
    ample of  this  principle has been demonstrated
    with the behavior of the P300 component under
    go and no-go circumstances (Papakostopoulos and
    Crow 197 6).
    4.  Spatial specificity:  Intrinsic  activity and EPs
    from brain locations a few millimeters apart can be
    different in frequency and waveform (cf. Freeman
    and Gerbrandt, this volume). EPs following finger
    displacement and intrinsic activities  are shown in
    Fig.  2 and 3, respectively. On both occasions each
    area has its own characteristic activities. Yet these
    activities  vary  in different  areas simultaneously
    registered.

  Time Level

    No EP occurs in  isolation. At the instant  an EP is
  recorded, other things are happening in the  brain, a
  fact which,  as Sutton (1968)  pointed  out,  is often
  ignored by experimenters. Intrinsic rhythms, for ex-
  ample, go on  continuously, as does modulation of
  membrane potentials. Even the pattern of neuronal
  connections  has  been demonstrated to be subject to
  continuous  functional modulation (Purpura  1970,
Scheibel  and Scheibel 1970, Skinner, in press, Ying-
ling and Skinner 1975). The brain is certainly no tab-
ula rasa at  the time of stimulus input. It has evolved
to its present state as the result of species-specific and
individual-specific  interaction  with  the environment.
Fig. 4 indicates schematically the influence of factors
having their origins in both nature and nurture and of
processes both overt and covert.  All these  influences
are present in the processing of each stimulus, and
their  interaction and  integration  determine the  out-
come in  response terms. Data are  now available (John
et al. 1973) that  the waveforms of macropotentials
reflect both the intrinsic significance of the present
stimulus and the influence of previous stimulus con-
figurations  which  were similar,  but not necessarily
identical.

Spatial  Level

   Discussion has  centered around an  electrode  re-
cording from a particular cortical location. The prac-
tice  of considering activity  at one location as if it
existed in  isolation can result in a form of intellectual
blindness  to  the  overall pattern of events.  Several
studies published  recently deal specifically with the
scalp distribution  of EPs in most sensory modalities
(cf.  the scalp distribution section  in  this volume).
Others  have dealt with cortico-cortical  relationships
(Papakostopoulos  and  Crow  1976) and cortico-sub-
cortical  relationships (McCallum et al.   1973, 1976).
Other evidence indicates that  during each particular
behavioral  sequence  both central structures and  pe-
ripheral  systems present  consistent  patterns  of
change.  For example, heart rate  decreases during the
foreperiod of a simple  reaction time experiment, i.e.,
during the period in  which the CNV develops (Lacey
and Lacey 1970,  Papakostopoulos 1973). The excita-
bility of the spinal monosynaptic reflex also changes
 during the same period (Papakostopoulos and Cooper
 1973, 1976, and  this volume). The total pattern of
 change in cortical, subcortical, spinal, and  peripheral
 systems appears  to  be constant for  any  particular
 stage of the preparatory process, but it  changes as the
 process  moves  towards its goal. The spatio-temporal
 description of performance can reasonably be consid-
 ered the objective index of animal and human inte-
 grated interactions with the environment, and yet the
 dynamics  of spatiotemporal processes that we are in a
 unique  position to investigate have not so far been
 studied in depth.

 Creation of a Model

    What kind of brain models  are indicated by the
 various factors discussed? What sort  of experiments
 should  we envisage? To which other disciplines and in
 what way should we look for relevant answers?

    A brain model created from macropotential princi-
 ples should take  into account (1) the areal specificity,

-------
 638
                                Papakostopoulos
                                                1	1 SEC-
                        vv*•/|^^»M^^^A^^Y"'"^V^'YA"l^^^
   PMF
                                        PKEM
                                                                           PREC
                                        POSTC
     S/B-,
                                  XI

                                                10
                                                    18   20  25  30
 /•'/£. J. Sample electrocortlcogram fright hemisphere} and corresponding power spectra from prefrontal (PREF),
 pn-motor (PRKM), precentral (PREC), and postcentral  (POSTC)  cortical areas in man. Each spectrum derived
 from 16 sec of KCoG activity; 19 spectra shown for each area.
(2) the areal plasticity, (3) the different spatial repre-
sentations of each event, and (4) the different organi-
/ation  of (lie  local and spatial relationships at every
point in  time. How, if at all, a scnsorimotor event is
going to  be represented in the  various local brain do-
mains depends on the relative weighting of the nu-
merous factors operating in the  system  at the time.
Which  factors  shape the function of a domain, and
what their value is set to at a given point in time will
depend on past, present, or anticipated future input-
output of the  domain.  Particular domains may have
many, few, or no  factors in common with other do-
mains. If there are common factors, they may vary in
unison or with varying degrees of independence in the
different  domains. The  two modes of nervous func-
tion described in the model  proposed by  Cooper,
McCalliim, and  Papakostopoulos (this section) pre-
sumably influence the setting of these factors.

   E. Roy John's "neurometric  space" (John et al.,
this  volume) is a pioneering attempt to quantify the
multifactorial  reality  of brain  macropotentials.  This
approach adheres "to the fact that sensory, perceptu-
al and cognitive processes, as well as the anatomical
and  functional integrity of the  nervous system, are
reflected in the electrical activity of the human brain
as recorded with  scalp electrodes." One can  envisage
that  psychiatric or neurological  diseases  will be de-
scribed  in the future  by their coordinates in such a
neurometric space.

-------
  Integrative Models
                                                                                                      639
                  FACTORS OF:     N
 Fig.  4. Schematic of probable factors and their inter-
 actions with processes represented in  the electrical
 activities recorded from a macroelectrode.

    One  can conceive of experiments  addressed  to
 fundamental  questions about  the  quantification  of
 each factor  in particular, about the values of many
 factors in the spatial dimension of the organism,  or
 about the principles of factor setting in time in one
 location or throughout the whole  organismic space.
 In the first instance, diversity in experimental para-
 digms is desirable. The case of CNV amplitude at the
 vertex could be an example. What, for example, is the
 maximum value that the CNV can reach at the vertex
 and  under  what circumstances? We  know  that the
 amplitude can be substantially increased by interpos-
 ing a signal between SI and S2 in such a way that its
 location in time and its  duration  cause the  subject
 continuously  to revise the nature of his ultimate re-
 sponse to S2 (Fig. 5), but is this the  optimum method
 for enhancing its value?

   When studying such factors and their relationships
 in  organismic  space, rigidity  of experimental para-
 digms is  necessary. Small changes in experimental
 procedure  can change the state of both brain and
 bodily systems (Papakostopoulos and Cooper 1973,
 1976, and this volume).

   The  question of relationships  with  other disci-
plines has practical and theoretical implications. Prac-
tically, we must know something of the electrocardio-
gram  or  plethysmogram or myogram in order to set
the characteristics of our amplifiers, or the sampling
rate of our computer, when recording brain macropo-
tentials. We should continue to bear in mind that the
activities of neurotransmitters and neuronal loops lie
beneath  our macropotcntial  configuration,  but we
should not  be restricted to verifying hypotheses of
brain  function derived from such studies. There is no
good  reason to assume that a hypothesis derived from
one restricted level  of analysis can be applied  to an-
  CLICK    TONE ON  TONE Off  FLASHES
 Fig. 5. Average vertex CNV from one subject in two
 different situations. A: Sit.  I  with CLICK as warning
 and FLASHES as  imperative signal. In half of the
 trials a tone was interposed fblack bar) and  the sub-
 ject had to withhold the response to FLASHES at the
 onset  of the tone. Note the CNV termination either
 with the  tone onset (TONE  ON) or the  motor re-
 sponses to the flashes (upper trace).
 B:  Sit. 2 was as in sit. 1 except that  the subject was
 told that if the tone terminated (TONE OFF) before
 the flashes,  then he should not keep  on  withholding
 the response, but should press to the flashes. In spite
 of the complexity of the last situation, the obtained
 CNV (upper trace) was about  twice the amplitude of
 CNV during the straightforward "press" trials (lower
 trace).


 other, particularly when the latter reflects the total
 interaction  between  organism  and environment
 Harmon (1972) makes the  point that  "...in neuro-
 physiology it seems  hopeless to accept the standard
 argument  that  one must  start at the lowest levels.
  Extrapolation in the other direction may, however,
 be somewhat easier. Knowledge at a higher level may
 elp one develop strategy for investigation at a lower
        In other  words, in  our  model-building  we
should work outward from our own phenomena. The
function of macropotential  research  should be the
 reation of questions and hypotheses that can be test-
ed at the unit, neurochemical, or psychological level
of analysis.

-------
CHARACTERISTICS OF EVENT RELATED MACROPOTENTIALS

THE ROLE  OF  THE  BEREITSCHAFTSPOTENTIAL AND
POTENTIALS ACCOMPANYING THE EXECUTION
OF  MOVEMENT
 LUDER DEECKE

 Abteilung Neurologic, Universitat Ulm, Ulm, West Germany
    My approach can hardly be called a model, but it
 may clarify motor elements of event-related macropo-
 tentials. Discussion  will  be confined to the Bereit-
 schaftspotential (BP) and  to motor aspects of the
 CNV, although I appreciate that the CNV may not be
 explainable  exclusively  in  motor terms. To me the
 exhortation  that we should now be looking beyond
 phenomenological levels of description means that
 some thought  and speculation should be addressed to
 questions such as  what  is the biological meaning of
 these slow potentials? What are the underlying mech-
 anisms and what is their functional significance?

    Kornhuber (1974)  integrated  slow potential
 changes  into a model of motor function. I do not
 wish to repeat this  theory  in detail as it is well
 known, but  I  do wish to put forward the view that
 one of  the classical  concepts of motor organization
 has to be corrected.  Despite the importance of work
 by Penfield and his colleagues (e.g., Pen field and Ras-
 mussen  1950), they over-estimated the role of the
 motor cortex. Penfield's claim  that all motor acts
 originate in motor cortex is, in our opinion, not true;
 almost all cortical areas  are capable of motor acts.
 Stimulation  of the motor  or somatosensory cortex
 yields movement, but one can elicit movements also
 from parietal areas, and eye movements from frontal
 areas with somewhat higher  stimulus strength. The
 motor cortex  is a highly  specialized  structure for
 movements that require complicated tactile analysis
 provided by  sensorimotor cortex, such as finger, lip
 and tongue movements. These movements are repre-
 sented primarily in the motor cortex, but others that
 function well without tactile analysis, e.g., eye move-
 ments, are not  represented there.

   Let us start with the electrogenesis of slow poten-
 tials, make  the crucial  assumption  that CNV and
Bereitschaftspotential are negative shifts of the cor-
tical d-c potential, and then  consider which general
influences  cause shifts of the cortical d-c potential.
There are many such influences, ranging from meta-
bolic  and physiological  to  psychological  factors.
Hyperoxia produces a negative shift; hypercapnia, a
positive shift (Caspers and Speckmann 1972). Increas-
ing wakefulness leads to a negative shift; falling asleep
is  accompanied by a positive shift; and arousal reac-
tion in the EEC is associated  with a marked negative
cortical d-c shift (Caspers 1963,1965). In freely mov-
ing animals, negative shifts occur during orienting re-
sponses to certain stimuli (Arduini 1957). Pharmacol-
ogical  studies show  that narcotics, sedatives, and
GABA cause  a  positive shift, while amphetamines
cause  a negative shift (O'Leary and Goldring 1964).
On the basis of these observations, we may conclude
that negative  d-c shifts are  usually associated with
activation and positive shifts are usually associated
with inactivation or inhibition. This view is further
supported  by intracellular data  from Caspers and
Speckmann (1972) who found that negative shifts of
steady potentials are associated with increased EPSP
rates of cortical neurons and that positive shifts are
accompanied by decreased EPSP rates.

   This evidence suggests that cortical d-c potential
shifts  provide a fairly reliable indicator of cortical
state.  Both CNV and Bereitschaftspotential represent
shifts  of the cortical steady potential. If we use them
as  indicators in the above sense, we can infer which
cortical areas  are active in relation to  a particular
goal. I regard both Bereitschaftspotential and CNV as
early preparatory processes, and I infer that they re-
present a facilitatory cortical process taking place in
the dense dendritic network of upper cortical layers. I
also believe that this facilitatory process is  specific

-------
                                                                                                     641
and highly selective to the extent  that one  can  tell
from the topography  of these potentials where  the
action is or where it will be in the intended task.

   Recent experimental  evidence from  the  Bristol
group (McCallum et al. 1976), from Rebert  (1972),
and from Skinner and Lindsley (1973) suggests that
the early preparatory process  of  CNV and BP is not
exclusively a cortical phenomenon, but a thalamocor-
tlcal phenomenon.  It becomes possible to envisage a
thalamocortical  focus of activity  that shifts over  the
cortex depending upon the site of action. This thal-
amocortical focus of activity is constantly changing-
i.e., it projects to speech centers during speaking, to
the right parietal region with constructive or spatial
tasks, to the visual area during attentive visual percep-
tion, and to  the frontal areas with decision.  In order
to  record this focus of activity with present methods,
we  tend  to  keep our experiments  as constant  and
stereotyped as possible with the result that this focus
is froxen in one position. I think we can assume that
this early preparatory process also occurs preceding
thoughts, intentions, ideas, and other cognitive or vo-
litional  acts, which unfortunately as yet  cannot be
 transformed to a trigger pulse. I  was speaking of the
 early preparatory process of the CNV and the readi-
 ness potential in the same sense and indeed it is my
 opinion  that  they represent the  same  underlying
 mechanism. The CNV differs from the BP in (1) the
 experimental situation producing  it, (2)  waveform,
 (3) topography, (4) relative bilateral symmetry, and
 (5) possible  independence  from  motor  acts  (cf.
 Deccke et al. 1976, plllf.). These differences are of
 a  merely operational  nature  and represent modifica-
 tions  of the same underlying early preparatory  pro-
 cess,

    The constantly changing thalamocortical focus of
 activity  associated with  these early preparatory pro-
 cesses may  be called a type of arousal, alertness, or
attention, but the  term should not be taken to imply
that such processes are nonspecific; on the contrary,
they are highly selective. Even such basic processes as
regional cerebral blood flow are highly selective (Ris-
berg et  al. 1975). When the subject performed a ver-
bal test, regional blood flow was increased in the left
hemisphere; when  he performed a spatial test, it was
increased in  the right hemisphere. Thus, even basic
vascular and  nutritional processes are to  some extent
selective and many kinds of dynamic changes going
on in the living brain are more specific than we think.
Experiments with the emphasis on topographical eval-
uation  of slow  potentials are  needed because  slow
potentials  reveal, more effectively than faster poten-
tials, which brain regions are active in a given experi-
mental situation.

    In conclusion, the potentials preceding a voluntary
act, such as  a finger movement, that must be taken
account of by any theoretical model arc essentially of
three kinds: (1) an early preparatory  process  - the
Bereitschaftspotential  - recordable as a slow negative
shift of the d-c potential, (2) a faster potential - pre-
motion positivity  (PMP) described by Deecke et al.
(1969), and (3) the motor potential (MP) proper. The
latter potential is  a unilateral potential that precedes
unilateral  finger movement (but not eye movement)
with a  mean onset  time of 60-50 msec prior to the
first EMG activity. It  should, however, be noted that
other workers, notably Vaughan et  al.  (1968), have
used the  term "motor potential"  to  refer to the
whole  complex of BP, PMP, and MP. The MP, as we
define  it, is a  negative-going potential recorded over
 the motor  cortex  contralateral to movement and
 most  probably reflects  cortical activity  associated
 with the  ongoing pyramidal  tract volley (cf.  Evarts
 1966).  The temporal relationship  of  the MP and
 movement led us to postulate a cortico-cerebello-cor-
 tical loop preceding voluntary  movement (Deecke et
 al. 1973, 1976),  a  feature confirmed in monkey ex-
 periments by Thach (1975).

-------
                                                                                                     641
and highly selective to the extent  that one  can  tell
from the topography  of these potentials where  the
action is or where it will be in the intended task.

   Recent experimental  evidence from  the  Bristol
group (McCallum et al. 1976), from Rebert  (1972),
and from Skinner and Lindsley (1973) suggests that
the early preparatory process  of  CNV and BP is not
exclusively a cortical phenomenon, but a thalamocor-
tlcal phenomenon.  It becomes possible to envisage a
thalamocortical  focus of activity  that shifts over  the
cortex depending upon the site of action. This thal-
amocortical focus of activity is constantly changing-
i.e., it projects to speech centers during speaking, to
the right parietal region with constructive or spatial
tasks, to the visual area during attentive visual percep-
tion, and to  the frontal areas with decision.  In order
to  record this focus of activity with present methods,
we  tend  to  keep our experiments  as constant  and
stereotyped as possible with the result that this focus
is froxen in one position. I think we can assume that
this early preparatory process also occurs preceding
thoughts, intentions, ideas, and other cognitive or vo-
litional  acts, which unfortunately as yet  cannot be
 transformed to a trigger pulse. I  was speaking of the
 early preparatory process of the CNV and the readi-
 ness potential in the same sense and indeed it is my
 opinion  that  they represent the  same  underlying
 mechanism. The CNV differs from the BP in (1) the
 experimental situation producing  it, (2)  waveform,
 (3) topography, (4) relative bilateral symmetry, and
 (5) possible  independence  from  motor  acts  (cf.
 Deccke et al. 1976, plllf.). These differences are of
 a  merely operational  nature  and represent modifica-
 tions  of the same underlying early preparatory  pro-
 cess,

    The constantly changing thalamocortical focus of
 activity  associated with  these early preparatory pro-
 cesses may  be called a type of arousal, alertness, or
attention, but the  term should not be taken to imply
that such processes are nonspecific; on the contrary,
they are highly selective. Even such basic processes as
regional cerebral blood flow are highly selective (Ris-
berg et  al. 1975). When the subject performed a ver-
bal test, regional blood flow was increased in the left
hemisphere; when  he performed a spatial test, it was
increased in  the right hemisphere. Thus, even basic
vascular and  nutritional processes are to  some extent
selective and many kinds of dynamic changes going
on in the living brain are more specific than we think.
Experiments with the emphasis on topographical eval-
uation  of slow  potentials are  needed because  slow
potentials  reveal, more effectively than faster poten-
tials, which brain regions are active in a given experi-
mental situation.

    In conclusion, the potentials preceding a voluntary
act, such as  a finger movement, that must be taken
account of by any theoretical model arc essentially of
three kinds: (1) an early preparatory  process  - the
Bereitschaftspotential  - recordable as a slow negative
shift of the d-c potential, (2) a faster potential - pre-
motion positivity  (PMP) described by Deecke et al.
(1969), and (3) the motor potential (MP) proper. The
latter potential is  a unilateral potential that precedes
unilateral  finger movement (but not eye movement)
with a  mean onset  time of 60-50 msec prior to the
first EMG activity. It  should, however, be noted that
other workers, notably Vaughan et  al.  (1968), have
used the  term "motor potential"  to  refer to the
whole  complex of BP, PMP, and MP. The MP, as we
define  it, is a  negative-going potential recorded over
 the motor  cortex  contralateral to movement and
 most  probably reflects  cortical activity  associated
 with the  ongoing pyramidal  tract volley (cf.  Evarts
 1966).  The temporal relationship  of  the MP and
 movement led us to postulate a cortico-cerebello-cor-
 tical loop preceding voluntary  movement (Deecke et
 al. 1973, 1976),  a  feature confirmed in monkey ex-
 periments by Thach (1975).

-------
 THE PLACE OF CONSCIOUSNESS IN BRAIN  RESEARCH1
 WALTER RITTER

 Department of Psychology, Lehman College, City University of New York,
 and Department of Neuroscience, Albert Einstein College of Medicine,
 Bronx, NY, USA.
   The question as to the kind of model or models we
 should use brings to my mind the broader question of
 how  the  purpose  of brain research is best defined.
 Indeed, the adequacy of models is ultimately judged
 with  respect to how well they aid in achieving speci-
 fied goals. In this sense, there is not just one kind of
 model. The models (this section) of Deecke, Skinner,
 and  the Bristol group, for example, may each be ap-
 propriate  to the  questions asked by the respective
 authors.

   Research is commonly classified into categories of
 practical or clinical utility, on one hand, and intellec-
 tual understanding, on the other. Models discussed in
 this paper are of the  latter kind. In  broad terms, the
 central theoretical question of  brain research is  to
 understand how the  brain does what it does. This
 includes many things, such as regulation of the endo-
 crine  system, water balance, body temperature, and
 breathing. The  focus of this conference, however, is
 on  psychological  processes of the  brain. I wish  to
 propose   that  intellectual  understanding  of the
 psychological processes of the brain is of two kinds.

   First,  how does the  brain  produce and regulate
 psychological processes  that occur outside of aware-
 ness?  We  are  not conscious,  for example, of how
 memory works. We have no subjective experience of
 the manner  in  which   memories  are constructed,
 stored, organized, and retrieved. Tricks can be learned
 to insure that certain material is adequately memori-
zed (e.g., by  rehearsal)  and to retrieve certain items
(e.g., remembering a person's name by  thinking about
places and events associated with that person). In this
way, crude notions can be developed as to how mem-
ory is established and organized, but these notions are
inferences and not direct experiences of memory pro-
cesses. We also have no  awareness of the mechanisms
of perception, coherent thought, emotions or even
      work  was supported by  USPHS Grant  HD
 10804 to the author.
 the simplest voluntary motor acts. Although we have
 a profound ignorance of most of the  psychological
 processes that underlie the mind, we nevertheless can
 use these nonconscious processes, and  usually in ef-
 fortless ways. We  can make memory yield relevant
 portions of its contents, direct our thoughts to partic-
 ular topics, or walk into the kitchen to make a cup of
 coffee. The system  is somewhat analogous to a person
 who can drive a car and yet grasp virtually nothing of
 the means by which automobiles work. The theoreti-
 cal  significance of understanding  the  nonconscious
 psychological  processes  lies  in  comprehending  the
 role they play in forming the contents of conscious-
 ness.

   The second kind of understanding of psychological
 brain processes concerns those events of which we are
 conscious.  The initial question here is how the brain
 produces consciousness, i.e.,  how biophysical events
 are transformed  into personal experience. It was,  I
 think, the sense that discovery of the significance of
 the mesencephalic reticular  formation opened  the
 door  to understanding how the  brain produced con-
 sciousness that gave the conference on Brain Mech-
anisms and Consciousness (Adrian et al. 1954) a spe-
 cial ambiance  of excitement  and achievement. Most
 brain  scientists, however, limit their thinking to how
 the brain produces piecemeal, static contents of con-
sciousness -i.e., how we 'see lines,  geometric shapes,
or colors; hear individual sounds; or have complicated
percepts, such as a person's face. But the phenomena
of consciousness go considerably beyond  that. There
is a flow and continuity of the contents of conscious-
ness,  so admirably  described  by William  James
(1890). When  we  are with a friend, it is not just a
conglomeration of sensations  or percepts that occurs:
what we experience is that person. Taken in its broad-
est  scope, the contents of consciousness  constitute,
for  each of us, reality as we know it. Through a com-
plex sequence of transformations and  constructions
(Neisser 1967), the  brain produces the world we (con-
sciously) live in, our personal identities, and the very

-------
Consciousness in Brain Research

goals of comprehending  how the brain does these
things.

   Some readers may wonder why two kinds of un-
derstanding have been proposed, since understanding
nonconscious processes which provide the contents of
consciousness might provide all the information need-
ed. There are two reasons. (1) It may be assumed that
nonconscious processes do not have conscious proper-
ties. (This is in contrast to Freud's view that the "un-
conscious" has  conscious properties, such as wishes,
aims, the ability to make judgments as to what should
and should not enter consciousness, and the construc-
tion of strategies).  (2) Conscious events have causal
efficacy in brain activity of a different kind than non-
 conscious  processes, because of their different proper-
 ties (cf. Sperry 1969, 1970). Whatever the processes
 are  that produce the  experience of anger, for exam-
 ple, subsequent psychological events (both conscious
 and nonconscious) are activated and shaped by con-
 scious  properties of that anger.  In  other words, the
 flow of consciousness constitutes the reality in terms
 of which we think, feel and behave.  Accordingly, the
 reality we experience  affects the brain activities relat-
 ed to thinking, feeling, and behavior.

  How  should relevant models be evaluated?

    Models should not be judged  by the volume  of
  research they stimulate. The models of Titchener and
  Hull,  for example, produced vast quantities of data,
  most  of which currently reside on dusty shelves.  In
  recent years there  has  been  a major shift in experi-
  mental psychology to information processing models
  of cognitive psychology. Since  these models  play a
  significant role in the search for event-related-poten-
  tial concomitants of psychological processes, it is im-
  portant  to judge whether these models, and the  re-
  search they prompt, will also one day find their way
  to dusty  shelves. It is, of course, difficult to estimate
  the stability of particular models because of the cur-
  rent relative lack of knowledge concerning psycholog-
  ical processes and related brain events.

    Since  the position taken in this paper is that the
  ultimate  goals of brain  research  concerned with
  psychological processes  pertain  to understanding the
  physiological  mechanisms associated with conscious
  experience (cf. John and Schwartz, 1978),  I  pro-
  pose that one essential way to evaluate relevant mod-
  els is  the degree to which they relate, at one point or
  another,  to our own personal experience. This means
  that introspection must play an important role in pro-
  viding basic notions of what it is we are attempting to
  explain,  as I  know  of no other way in which it is
  possible  to know  what consciousness is or whether
  given lines of research are in any way related to con-
  scious experience. The  failure of Titchener's efforts
  was not that he used introspection, but that the kinds
                                             643

of specialized introspections his subjects were trained
to perform were not related in any meaningful way to
everyday conscious experience or to what conscious-
ness is about: the construction of personal reality. As
William James  (1890) put it, "Introspective observa-
tion is what we have to rely on first and foremost and
always." For a more  recent, incisive  statement of a
similar  position,   see  Bertrand  Russell  (1948,
pp43-53). Using personal experience as a frame of
reference, then, some  comments will first be made on
information processing models and  then on event-re-
lated potentials.

Information processing models

   The revival of interest in selective attention in the
early models of Broadbent (1958) and  other investi-
gators brought to experimental psychology theoreti-
cal considerations  that could readily be interpreted in
terms of and related to experience. That we generally
are  conscious of only a limited portion of the envi-
ronment, and usually some coherent portion, squares
well with introspection. Investigations of the process-
es underlying  selective attention, therefore, are im-
mediately relevant to consciousness because such pro-
cesses play  an  important role in forming the contents
of consciousness.  The work of Moray (1970), which
showed  that unattended material  was not likely to
enter long-term memory, whereas certain unattended
 words (such as the subject's name) could  capture at-
 tention, was significant in that it  spoke to the rele-
 vance of consciousness to memory  and to the  fact
 that unattended stimuli are processed  to  a consider-
 able extent even  though they may never enter con-
 sciousness. The key, as Sperling's  (1960) work sug-
 gested, is that if attention is focused on a given subset
 of  sensory input  in such a way that it becomes con-
 scious, then that input can influence other psycholog-
 ical processes, such as thinking, feeling, memory and
 behavior, whereas sensory input which is  not attend-
 ed lasts  a few seconds at most and has no further
 influence.

     The models and experiments that grew out of the
 information  processing approach  were  a breath  of
 fresh air after the dreary days  of Behaviorism, where
 conscious  experience was excluded as  a relevant con-
 cern. Thus, where the purpose is to unravel the man-
  ner in which consciousness is produced,  the early
 work of  Broadbent  and others  provides relevant
  models.

     A more recent  model  of information processing
  views attention as a capacity that can be "allocated"
  among various psychological processes (e.g., Kahne-
  man 1973). Broadbent conceived of  the mechanism
  of selective attention as a filter which, by admitting
  certain sensory inputs for further processing and  ex-
  cluding others,  prevented  overload  of the system.
  While both models share  the assumption of limited

-------
  644
                                            Ritter
 capacity,  the  allocation  model views attention  as a
 flexible process that operates at various states of in-
 formation processing. Thus, attention can be directed
 to perception,  storage in or  retrieval from memory,
 imagination, or motor activities; and, attention can be
 allocated to more than one psychological process at
 the same time (up to some limit of capacity), depend-
 ing on the amount  of attention required for  each
 operation. It is certainly possible to read a book and
 monitor crying from an infant in a nearby bedroom,
 but it is  not  likely that a book could be  read while
 performing a  vigilance task that required detection of
 minute changes in ongoing stimuli. Furthermore, in
 circumstances where complex activities are overlearn-
 ed, such  as walking or driving a car, it is possible to
 simultaneously  engage in  another complex  process,
 such as conversation. These modifications in  the con-
 cepts of attention also fit with introspection. Notice,
 however, an important change with respect to the
 fate of stimuli  that do not reach awareness. In the
 earlier model, stimuli that do not reach consciousness
 decay in  a brief period of time and have no  further
 influence on other psychological processes. In the al-
 location model, stimuli that do not  reach  conscious-
 ness can influence behavior, in certain circumstances,
 as in driving a car; however, both models can include
 the  idea that stimuli that do not reach awareness do
 not  enter  memory. Surely  each of  us  has  had the
 experience of doing something, such as driving a car
 or crossing a series of heavily  trafficked streets, with-
 out  paying (conscious) attention, and then suddenly
 become aware  that for some  time we had not been
 paying attention. In these circumstances, we usually
 do not have any recollection of the streets  crossed or
 the road traveled. Thus, current experiments examin-
 ing the manner  and degree in  which  attention can be
 allocated among various psychological processes com-
 mend  themselves for our purposes because they are
 related to conscious experience.

 Event-Related Potentials (ERPs)

   With  the information models  just described, in
 combination with personal experience, certain tenta-
 tive conclusions can be  drawn  about the  relationship
 between  ERPs and consciousness.  First, some ERP
 components can be related to  conscious  processes,
 whereas others can not. In auditory ERPs,  for exam-
 ple,  all components  from  brainstem  potentials to
 N100 and P200 may be necessary  for conscious per-
 ception of an auditory stimulus, but none  Is sufficient
 for  conscious  perception to occur. That is, a  subject
engrossed  in reading  or asleep, may  be  unaware of
 tones or clicks  which nevertheless elicit  these com-
ponents.  On the other hand, P300  seems to occur
only when  a subject is aware of the eliciting stimulus.

  The work of  Hillyard and Picton (In press) is perti-
nent, The demonstration that N100 is larger to stimu-
li in attended channels than in simultaneously  pre-
 sented unattended channels suggests two things. First,
 subjects seem  to be able  to consciously direct certain
 nonconscious  psychological processes,  reflected  by
 the enhancement  of N100, to select  stimuli on the
 basis of simple physical characteristics. I say "direct"
 because it seems  unlikely that N100 enhancement
 would  occur unless a subject consciously intended to
 listen selectively to one channel over  other channels
 and  maintained that intention during stimulus pre-
 sentation. If a subject's mind wandered, for example,
 and  for a period of time he stopped selectively  at-
 tending to the  stimuli, NIOO enhancement would pro-
 bably disappear and NIOO amplitude would be equal
 for stimuli of all channels. Second, a  prediction can
 be made about the subject's attention on the basis of
 the relative amplitude of NIOO. Imagine telling a sub-
 ject to count  infrequent  stimulus changes in one of
 two  channels, but not to reveal which channel  he
 intended to monitor. The  relative amplitude of NIOO
 should  permit  the experimenter to  infer which chan-
 nel had been attended, and to test the inference  by
 subsequent questioning of the subject.

    In terms of the so-called "cocktail party phenome-
 non," NIOO findings suggest that subjects clearly hear
 stimuli in  an  attended channel, while stimuli in  an
 unattended  channel  are  perceived as  background
 noise, analogous to the background babble of voices
 we  experience  at  cocktail parties.  Unfortunately,
 subjective reports have not been obtained (or report-
 ed) in these experiments.  I am  in full agreement with
 Tecce's remarks (this section) about obtaining subjec-
 tive reports in appropriate ERP  experiments.


    The circumstance that P300 is  elicited by infre-
 quent changes in an attended channel and not in  an
 unattended channel, suggests that  stimuli  in  the  at-
 tended channel are heard clearly enough so that infre-
 quent, small changes in the attended channel are con-
 sciously perceived, whereas changes  in unattended
 channels are not.  It  is not  plausible  to argue  that
 changes in both channels are equally well perceived  or
 that changes in an attended channel elicited P300 be-
 cause they are  task-related (are being counted) where-
 as  changes in other channels are not task-related. If
 only one channel of stimuli is presented and subject!
 are instructed to press a key whenever  they detect an
 infrequent, small change in stimulation, changes  that
 are  not detected  do  not  elicit P300  (Ritter  and
 Vaughan 1969), Presumably subjects are not aware  of
 undetected changes. Subjective reports are  of no use
 in this situation, since subjects cannot be expected  to
 report changes of which they are unaware. Further-
 more, when one set of stimuli is delivered while  sub-
jects are engrossed in reading, P300 Is not elicited by
small, infrequent changes (Squires etal. 1975). While
it seems probable that subjects were not conscious of
the changes  in this experiment, subjective  reports
were not obtained.  Finally, if subjects are reading or

-------
Consciousness in Brain Research

otherwise ignoring stimuli,  changes obtrusive enough
to capture a subject's attention do elicit P300 (Ritter
et al.  1968,  Roth et al. 1976). In  the Ritter et  al.
study, subjective reports confirmed this  conclusion.
Thus,  the weight of evidence suggests  that P300 oc-
curs only when a subject is conscious of the eliciting
stimulus (or, to be more precise, when  a subject per-
ceives  that a change in stimulation has  occurred).
Similar considerations apply to the slow wave report-
ed by  Squires et al. (1975)  and  the frontal P300 asso-
ciated  with complex, novel stimuli (Courchesne et al.
1975).

   The bimodal model  of Cooper and associates in
this  section can readily be related to  the allocation
model of attention. Note that Cooper et al. in evalua-
ting this model  make joint reference to "the experi-
mental evidence and everyday  experience." It is dif-
ficult  to think of the CNV occurring without the con-
scious  participation of the subject. The  suggestion
that the CNV is no longer  elicited in warned, simple
RT  tasks when  the  task is so  overlearned that con-
scious participation is no longer required to perform
the  task, is an intriguing way of interpreting the data.
It is  not  clear whether relevant subjective reports
were obtained from the subjects. If not, it would be
desirable to do so since, as mentioned before, we can
know  when  we have not  been paying attention to
some routinized activity.

   In  terms of the allocation model, subjects should
have spare capacity when CNVs cease occurring if the
bimodal hypothesis  is correct. This could be tested,
for  example,  by giving subjects an  appropriate con-
comitant task at various points during  the RT experi-
ment. The prediction is that performance on the sec-
ond task will improve in conjunction with stabilized
RT  performance, which is  associated with absence of
the  CNV.


 Conclusions

    Information processing models appear to be useful
 in ERP research because such models can be  related
 to  personal  experience and certain ERP components
 can be associated with conscious  events. There are
 several ways in which  ERP components may  be re-
                                             645

lated  to  consciousness: they may be influenced by
consciousness, have an influence on consciousness, re-
flect consciousness itself in some way, or  represent
some  combination of these  possibilities. N100 en-
hancement to stimuli in attended channels, for exam-
ple, presumably occurs because of conscious inten-
tions  of the subject to attend to a given channel. In
this case,  N100 enhancement demonstrates an influ-
ence of conscious processes on an ERP component.
On the other hand, N100 enhancement may influence
consciousness by playing a role in bringing stimuli of
the attended channel  to  the center of conscious a-
wareness and thereby increasing their clarity.

   In  terms of the bimodal model, the CNV in the
"scopeutic" mode, may be thought of  as being influ-
enced by  consciousness or perhaps even being a con-
comitant  of consciousness. That the CNV occurs first
and then  engages the subject "scopeutically" in the
task does not seem  reasonable. Perception of the
warning stimulus is more  likely to activate conscious
processes which direct preparatory events, in which
case the CNV could be related to the conscious pro-
cesses or to the mechanisms that  underlie prepara-
tion.  In  the "categoric" mode, the warning stimulus
presumably triggers the appropriate preparatory pro-
cesses directly, without the intervention of conscious
processes. Thus, no CNV is elicited.

   The  P300  is more ambiguous  than N100 or the
CNV in  that all of the possibilities mentioned above
could apply. P300 could affect the contents or quali-
ty of consciousness if it is assumed that  P300 occurs
prior to awareness of the eliciting stimulus. P300 may
also be a correlate of conscious  awareness (Posner
 1975, Simson et al. 1976). Finally, P300 could reflect
the effect of  conscious awareness on nonconscious
psychological processes pertaining to preparation for
or strategies related  to future events (Donchin et al.
 1973, Picton and Hillyard 1974).  Selection between
the  various possibilities  (or some  combination of
them) awaits clarification of the functional signifi-
cance of P300.

   In summary, it seems  fair to  conclude  that ERP
research  will contribute meaningfully  to our under-
standing of conscious and nonconscious psychological
processes of the mind.

-------
  SIGNIFICANCE  OF  SLOW POTENTIAL SHIFTS
  IN  ANTICIPATION OF AND  DURING TASK
  PERFORMANCE1 2


  R. NA'A'TANEN
  Department of Psychology, University of Helsinki, Finland
    In discussions on the functional significance of
  slow negative shifts such as the CNV, one crucial as-
  pect has received  little  comment—namely, that on
  many individual trials even with high-level  task per-
  formance, no definite CNV can  be observed in scalp
  recordings.  Some  relevant data from a study by
  Gaillard and Naatanen (1973) are shown in Fig,  1.
  Although the amplitudes of averaged CNVswere rela-
  tively large, many single trials,  even with fast RTs,
  in a visual  two-choice RT task were associated with
  no negative  SP during the S1-S2 interval in  the EEC
  record.  Although  people in the field may be aware
  of such data, the implications for the functional role
  of these  negative task-related shifts in the behaving
  organism have largely been  ignored.  It is of crucial
  importance to our theorizing on the nature  and role
  of SP shifts  such as the CNV always to keep in mind
  that their existence — inferred  from  scalp record-
  ings - does not appear to be a necessary concomitant
  of preparation or performance, even under conditions
  associated with well-developed averaged CNVs.

    First let us consider the basic  assertion that there
 really are trials with high level performance and no
 CNV. It could be argued that the CNV is not always
 observable on single  trials for the same reason as for
 evoked potentials (EPs),  namely  that  they  are  ob-
 scured by ongoing EEC activity. This argument is val-
 id  for EPs,  but does not apply  to slow potentials
Supported by the Academy of Finland.
      present discussion does not involve the slow
negative frontal component (e.g., Weerts and  Lang
1973, Loveless and Sanford 1974, and Gaillard 1976)
and more  specific processes such as Bereitschaftspo-
tentials (cf. Deecke, this section) and those related to
language production (e.g., Grozinger et al. 1976).
 of average amplitude because the dominant EEC fre-
 quencies  are  of a similar order of magnitude  as
 those of most EP components, but not of SPs^. The
 occurrence of spontaneous SPs, with a duration  of
 one second or  more,  should be clearly observable
 at  the scalp if present  on any given individual trial.
 The only reason for a prominent slow negative poten-
 tial to remain undetected in scalp recordings would
 be its temporal coincidence with a spontaneous posi-
 tive SP of similar size, duration and distribution, such
 that the two events cancel each other. However, we
 find far too few examples of such positive SPs in our
 primary records  to account for the lack of a negative
 shift in the EEC record  on so many response  trials.
 Hence we  may conclude  that CNV absence on indi-
 vidual  trials is not merely the  result of masking by
 other potentials. Trials  really occur in which no CNV
 is generated in those same structures which, on the
 average, produce a well-developed CNV4. This  infer-
 ence cannot, of  course, be made with regard to EPs.

    What, then, are the implications if task-related SPs
 are not necessarily present during anticipation of, and
 preparation for,  performance? This  observation sug-
 gests that the processes giving rise to the generation
      frequency of most spontaneous SP activity ap-
pears  to be much slower (see Gaillard et al., in prepa-
ration) than that of task-related SPs recorded during
inter-stimulus intervals  or foreperiods of 1 second.

4If the "ceiling hypothesis" of CNV  amplitude is
valid, then some of the intra-trial variability  of SP
amplitude  can, of course.be attributed to this factor.
However, this hypothesis could by no means totally
account for CNV absence on individual trials. For
example, in  trials 20 and 27 of Fig. 1, the baseline
immediately after the  foreperiod  exceeds  the CNV
in negativity suggesting that a ceiling on negativity
has not been in effect.

-------
Significance of SP Shifts
   17-32
   17

   18

   19

   20

   21

   22

   23

   24
                         RT
                                         3CK 5
                                       median
'^^TV    299

(^^W^^^SA^    324
  F&. •/.  fpp*'' trace: Vertex CNV averaged over the
  I7th-32nd trial in a 2-choice RT task with an inter-
  stimulus interval of 1 sec. Data from the 2nd series of
  the 2nd session  of one subject. Lower traces: Single
  vertex records of the upper trace (from Galllard and
  NUStUnen, 1973; reproduced with the kind permission
  of the North-Holland Publishing Company).

  of negative shifts in the human brain  are temporari-
  ly coincident with, but not causally related to or con-
  tingent upon, the physiological and psychophysiologj-
                                             647

cal processes necessary for a certain pattern of behav-
ior. Hence  SPs-at least as recorded from the scalp-
would  not  be  necessary to performance in the same
compelling sense  as, say, the activation of motoneu-
rons is for movement. Many other CNS processes
share the same property as suggested above  for SPs.
For  example,  EEC activation does not seem to be a
necessary condition for good task performance (e.g.,
Dureman and Edstrom 1964). On the other hand,
even a sleeping animal can have a highly activated
cortex under certain  drug  conditions (Lacey 1967).
In my evaluation  of the inverted-U hypothesis of acti-
vation and performance (Naat'anen 1973), I pointed
out  that the'level of activation'-within certain wide
limits—has  not yet been shown to be  an important
determinant of performance. Lacey (1967) also cites
numerous dissociations between behavioral and physi-
ological events (e.g.,  Bradley  1958, Dureman and
Edstr&m 1964,  Feldman  and  Waller  1962, Malmo
 1966). These observations lead to the inevitable con-
clusion that much of present-day psychophysiological
knowledge is  at  the level of temporal coincidences
rather  than causal relationships. Wikler (1952)  ob-
served  similar dissociations between sleep states and
 EEC in atropinized cats. He  concluded that  "sponta-
neous electrical activity of the cerebral cortex reflects
 the  activity of neuronal  systems  which, in part at
 least, are independent of those neuronal systems that
 subserve behavior in general" (p.265).
                                                           Feldman and Waller (1962) provide additional im-
                                                        portant evidence. Cats with nearly complete bilateral
                                                        lesions  of  the  posterior hypothalamus were somno-
                                                        lent and  unresponsive to sensory stimuli, required
                                                        tube  feeding,  showed no spontaneous movements,
                                                        and could  not  be behaviorally  aroused. Cortical de-
                                                        sychronization, however, was easily produced by peri-
                                                        pheral sensory  stimulation or by stimulation of the
                                                        midbrain  reticular formation, even in  the complete
                                                        absence of behavioral arousal. The converse dissocia-
                                                        tion-a  behaviorally aroused cat with a* "sleeping"
                                                        cortex-was produced by bilateral lesions in the mid-
                                                        brain reticular formation.
                                            What then is the  functional significance of CNV-
                                         like SPs in relation to overt behavior? To take a spe-
                                         cific example, Tecce (1972) demonstrated that  CNV
                                         correlates  with  attention, but we must not be tempt-
                                         ed to infer that the relationship is more than a correl-
                                         ative  one. We  have already  seen  that  task  perfor-
                                         mance-which is frequently used as an index of atten-
                                         tion-can  occur in the absence of a negative SP. Pre-
                                         sent evidence does not permit an unequivocal deter-
                                         mination  of the  psychological or  behavioral dimen-
                                         sion of performance  which is reflected in scalp  nega-
                                         tivity. There  is, however, some reason to think that
                                         conscious  effort (as exemplified in situations of stress
                                         or "trying harder") might be the key.

-------
  648
                                        NSMtanen
      193-208
     289*304
     385-400
                                                S1           S2          S1
     481-4%
     577-592
     673-668
     769-784
     865-880
     961-976
                                        RTs



                                        173


                                        UO



                                        138



                                        U?



                                        133



                                        1S6



                                        160



                                        US



                                        U6



                                        137


                                        US
              IduV
 Fig, 2,  The EP elicited by SI, the CNV and the median RT as a function of time for one subject in a simple RT
 task with an inter-stimulus interval of 1 sec and inter-trial interval of 5 sec. Derivations from left to right: vertex
 (Cz), frontal (Fz), and temporal (T3); negativity upwards (from  Ntitttttnen and Gaillard, 1974; reproduced with
 the kind permission of the North-Holland Publishing Company).
   This interpretation seems to be consistent with the
division of performance  into "scopeutic" and "cate-
goric" modes as suggested by Cooper et al. (this sec-
tion). The former implies conscious, nonreflexive per-
formance, associated with large  negative task-related
SPs. The categoric mode pertains to more automatic,
less conscious behavior. This change would take place
during the course of learning and would be associated
with gradually attenuating SPs5. Cooper et al. suggest
s Distraction  and  overload were  also mentioned as
factors causing SP reduction.
that the execution of performance in such circum-
stances is transferred to separate, second-order, semi-
automatic systems in the brain.

   According to these authors, individual trials show-
ing little or no negative SPs  should exist during the
later parts of extended experimental sessions involv-
ing simple repetitive tasks. The authors' data showing
habituation  of CNV during a prolonged simple-RT
experiment (Cooper et al. 1976) suggest this to be the
case. However, NSStanen and Gaillard (1974) failed
to observe any progressive CNV habituation in a sim-
ple-RT experiment with one-thousand trials (Fig. 2).
The explanation might  be, as  suggested by McCallum
(personal communication), that the frequently (every

-------
Significance of SP Shifts

10th  trial) given RT feedback  prevented the transi-
tion from the scopeutic  to categoric mode, i.e., the
task required  continuous  conscious or "scopeutic"
processing.

   It  is possible  that during automatic, less conscious,
performance the SP generators giving  rise  to  scalp
negativity might be  deactivated and  that preparation
for performance at the cerebral level takes place only
on the basis of a relatively small, absolutely neces-
sary, cerebral cell population whose  summed activity
is not enough to produce large negative fields.  Large
task-related negative SPs might reflect motivational
concomitants  of behavior  and  performance which is
not yet automatized.(However, this maybe only one
among many performance dimensions or concomitant
processes reflected by large task-related negative SPs.)
Conscious effort and stress have  already been men-
 tioned as  such processes. The  role of factors  of
 this kind in performance  situations has been exten-
 sively  discussed elsewhere (Naatanen  1973, 1974).
 Such processes, which are  distinct from performance
 per se, have their own  physiological activation pat-
 terns that may  overlap and interfere with those phys-
 iological  activation  patterns  forming  the  necessary
 physiological  substrates  of various kinds of perform-
 ances  and  may even cause performance to deterio-
 rate.  As stated above, the necessary level of cerebral
 activity in many performances (even those of high
 standard) may  involve only a relatively small cell pop-
 ulation  whose  activation does  not induce  large
 SPs  such as "CNVs."   These  large potentials, when
 coincident with performance,  would signify the oper-
 ation  of the "extra factors"  discussed above.  This
 would explain  why complex and high level perform-
 ance might be achieved  without any notable nega-
 tive SP.
                                              649

    In addition to the extensive CNV literature which
 supports the concept of CNV as reflecting mainly
• motivational factors (motivation, attention, arousal,
 effort, interest, significance, relevance), research find-
 ings  on the  relationship between performance and
 negative SPs seem  to  fit  particularly well with the
 present hypothesis:  namely,  correlations  between
 performance and CNV measures have generally been
 rather low,  even non-existent (e.g., McCallum  1973,
 Naatanen  and Gaillard  1974).  The relationship be-
 tween the level of conscious effort (and similar fac-
 tors)  and performance is probably just as  tenuous.
 Many  other psychophysiological measures than CNV
 have  shown similar relationships with performance
 (see review by Naatanen 1973).
     To  sum up, it appears that SPs such as the CNV
  mainly reflect conscious effort to increase the level of
  performance,  as well as other motivational and emo-
  tional aspects of the situation, to a much greater ex-
  tent than they  reflect performance processes them-
  selves.  Undeniably, task-specific SPs, such as those
  associated   with  language   production  (Grozinger
  et al. 1976) also exist, but they usually seem to be of
  small size and cannot, consequently, account for the
  large CNVs often  observed.  Since motivational and
  emotional processes - above a certain level - probably
  have no consistent relationship with performance,
  this view would explain the  inconsistent relationship
  observed  between  scalp-recorded  negative  SPs and
  performance  —  one of the  most confusing  issues in
  the field.  This hypothesis  also provides an explana-
  tion of the great variance and even absence of single-
  trial   CNVs  under   conditions   of   excellent
  performance.

-------
SOME GENERAL  CONSIDERATIONS  IN  FORMULATING
ELECTROPHYSIOLOGICAL  BRAIN MODELS
 R.  COOPER

 Burden Neurological Institute, Bristol, England

    Both in everyday life and in scientific endeavor the
 interpretation of new information is influenced  by
 previous  knowledge.  Thus  attempts  to determine
 whether there is life on Mars are based on experi-
 ments that determine  the presence of lower life forms
 on earth. Completely new life forms  would be diffi-
 cult to detect because of the difficulty of testing the
 large number of possibilities open.

   Similarly, new electrophysiological data from the
 brain are interpreted in terms of our past knowledge
 of and experience with electrical phenomena, dating
 from publications such as, "On the electricity excited
 by the mere contact of conducting substances of dif-
 ferent kinds." (Letter from Mr. Alexander Volta.  . .
 Read June 26,1800).  Within frameworks of this kind
 we attempt to describe the propagation of the nerve
 action potential along an isolated fiber in terms  of
 well-known physical phenomena such as current, volt-
 age,  capacity, and resistance. More complex systems
 are often  studied by breaking them down into com-
 ponent  parts and then synthesizing the behavior  of
 the whole; however, above a certain level of complex-
 ity, particularly when there is multiway interaction
 between  the components, this technique has limited
 value  and different methods must be sought. At one
 extreme  many psychologists deal  with  the problem
 by treating the individual as a black box, i.e., measur-
 ing only the inputs and the behavioral outputs of the
 black box, and are content to make inferences about
 the mechanisms within the box. At another extreme
 many animal  physiologists isolate themselves and
 their preparations, and try to understand the mech-
anisms by changing one variable  at a time.  Yet an-
other, essentially pragmatic approach has to be taken
by the psychiatrist who must make far-reaching deci-
 sions  on a minimum  of information  obtained from
 sources of very varied reliability.

   Somewhere in between these extremes we find the
 psychological approach described by Loveless and the
 neurophysiological approach  described  by Skinner.
 Despite our claims  to represent  a multidisciplinary
approach, the amount of real integration of overlap-
ping concepts is minimal. This is not meant as a criti-
cism of these excellent sections, but to point out that
most  of us make little attempt to cross interdiscipli-
nary boundaries and to understand  what other disci-
plines are trying to say. The situation arises  largely
because advances in modern science and technology
make it necessary  to  acquire a  mass  of detailed
knowledge before one can contribute at more than a
trivial level in another discipline. Furthermore, we al-
most certainly choose our own speciality because it is
the type of work we prefer.

   Nevertheless, what seems to be lacking at the pre-
sent time  are  general models of brain systems  that
describe the  way they operate and interact without
being too much concerned about  what  each  indivi-
dual nerve cell is doing. In many ways we are  in a
similar position to politicians or economists trying to
understand,  for instance, the role  of multinational
corporations in maintaining the stability of monetary
systems. The operational characteristics of the various
parts have to be determined, but there is little need to
know how these are generated from individuals in the
corporations. Physicists and chemists  have  come
through this  stage by "discovering" the gas laws, grav-
itation,  the periodic table, and so on. New informa-
tion is assimilated into such global theories until data
arrive which  do not fit.  Then the theories have to be
examined for error and changed just  as Newtonian
mechanics had to give  way to Einstein's theory of
relativity. Cyberneticists would probably claim that
this is what  they are already endeavoring to do, but
so far these specialists seem to be too detached from
reality to establish an effective  rapport with brain
scientists.

   Earlier in  this section Papakostopoulos  suggested
that we should in our model building start  with the
macropotentials,  which  constitute our  common
ground, and  work outward. In the paper that follows
we have attempted  to do this, at least for slow poten-
tial changes.

-------
BIMODAL  SLOW POTENTIAL THEORY  OF CEREBRAL
PROCESSING
 R. COOPER, W.C. McCALLUM, AND D. PAPAKOSTOPOULOS
 Burden Neurological Institute, Bristol, England

   More than 13 years have passed since the first re-
 cordings in man of slow potential changes, such as the
 contingent negative variation (CNV) and the Bereit-
 schaftspotential (BP).  During this time, much experi-
 mentation has  been  carried  out  in  laboratories
 throughout the world, and the evidence to date has
 confirmed the  relevance  of such slow changes to  a
 range of behavioural states requiring planned decision
 or action.

   The relationship, if any, of the slow potentials to
 the  more transient  evoked potentials and to the in-
 trinsic  activity of the brain has not yet been ade-
 quately investigated.  A unified theory that includes
 all aspects of brain  electrical change from steady po-
 tentials to unit activity is no doubt desirable, but gaps
 of knowledge occurring at many levels prevent its for-
 mulation. For example, the site and characteristics of
 the generators  of macropotentials - intrinsic or
 evoked - are largely unknown, and the mechanisms
 of rhythmicity and synchronization in various parts
 of the  brain are not yet determined.  Nevertheless, it
 is relevant to ask in what respect the different types
 of activity can be regarded as independent from one
 another and whether they can be attributed to dif-
 ferent  aspects of cerebral processing.

    With the present state of knowledge, it seems that
 there is sufficient dissociation between the character-
 istic changes of the evoked potentials and the event-
 related slow potentials (ERSPs) for them to have dif-
 ferent  functions.  It would also appear that the
 ERSPs, such as the CNV, are quite different in char-
 acter from spontaneous infraslow activity of the type
 reported by Aladjalova and others. Thus a case can be
 made for an independent function to be attributed to
 ERSPs. Can this function be determined with the pre-
 sent state of knowledge?

    In considering this question, the following factors
  warrant consideration. ERSPs are widespread in cor-
  tical and subcortical structures. They are usually ne-
  gative on the scalp, but  can be positive in deep struc-
  tures. They are time-related to an event in the exter-
nal  (or internal) world. They have a  characteristic
waveform consisting of a slow rise and rapid fall that
could be indicative of the  function of the system.
Waveforms of this kind are encountered in many and
varied physical systems. In general, such characteris-
tics tend to be found in integrative systems where the
buildup leads to a point at which there is collapse or
breakdown, e.g., the buildup of charge in a thunder-
cloud giving rise to lightning discharge. Such systems
are often used when the energy  supply is limited and
time has to be utilised to accumulate sufficient power
— a photographer's flash is a good example. The prin-
ciple is a familiar one in many fields of human activ-
ity from insurance to  the building-up of armaments.
The basic characteristic is accumulative storage  of the
commodity in a form that  can be expended quickly.

   When  considering ERSPs in  this light, we  might
expect some correlation between the quality or quan-
tity of output and the level of the CNV, and indeed
the literature abounds with attempts to establish such
relationships. In   some instances  they  have  been
found, as in the case of CNV amplitude and anticipat-
ed  energy  output, but  in other circumstances they
have proved elusive, as in the case of CNV and reac-
tion time. Generally speaking, it has become evident
that simple output measures are not the most reliable
indicators of internal state. This is not necessarily to
say that energy is not being built up in some way, but
rather that its  deployment may be with respect to
decision-making processes  rather  than to processes
that manifest themselves in muscular action. Perhaps
attention should be directed more to the information-
-organizing characteristics  of the system rather than
 its integrative, energy-storing characteristics.

    Analogs of information systems are not so abun-
 dant as those for energy systems, but early warning
 radar could serve as an example. After an initial con-
 tact, there presumably  follow various stages of alert-
 ness during which additional  information flows  in
 from various sources leading, for example, to identifi-
 cation of the target as friendly and the sudden stand-
 ing down  of all  systems.  Here the CNV would  be

-------
  652

  analogous  to the  buildup  of information as various
  possibilities are checked out. This analogy would also
  be consistent with the theory of Skinner (this vol-
  ume)  in that  if various  selective gates were  being
  opened (or closed) one by one as systems were check-
  ed out, the total current flow would rise in the man-
  ner  of the CNV, rapidly in the beginning and slowly
  afterwards. This is analogous to  the conductance  of
  an increasing number of parallel  resistors, i.e., resist-
  ance (1/RH1/RO+0/R2)+(1/R3)... .  Any one item
  of information  in the interstimulus interval could
  cause a stand-down of the system. This would be con-
  sistent with  the data  of Papakostopoulos (this sec-
 tion, Fig. 5), who showed that when a third stimu-
 lus, meaning 'don't press', was inserted during the in-
 terval  between the warning  and imperative stimuli
 of d classic  CNV  paradigm, the CNV  immediately
 collapsed to baseline.

    Attempts  to relate slow changes directly to exist-
 ing psychological constructs have been manifold, but
 none lias proved entirely satisfactory or acceptable to
 the majority of those working in this field. Inspection
 of the  constructs with which correlations have been
 demonstrated reveals a substantial overlap in defini-
 tions and usage. For example, attention, motivation,
 expectancy,  subjective  probability,  conation, and
 orientation  lead one to a recognizable but ill-defined
 area of human cognitive function.

    Any subject in a slow potential experiment recog-
 nizes these  psychological  elements, and indeed his
 subjective awareness might well be considered a key
 factor in  the situation, albeit  one that is almost im-
 possible to define or quantify on an objective scientif-
 ic  basis.  There  is also an  alternative state in which
 these elements are  not  operating in the same  way,
 which  is  characterized by  a  reduced and modified
 awareness and in which a more  automatic mode of
 functioning seems  to predominate. In the definition
 of these two states of the system - or modes of ac-
 tion if one prefers to stress their dynamic aspects - a
 need  arises  for a terminology  which is not contami-
 nated by  the  many overtones of the present range of
 psychological constructs.  Low came close to solving
 this problem when he proposed that the term 'cona-
 live'  might  appropriately be substituted for 'contin-
 gent*  in the CNV context.  But  even this revival of
 a word  which first appeared In the literature several
 hundred years ago does not succeed in embracing one
 or  two  of the critical elements which we now know
 to  be intrinsic to the mode of action in which the
 slow potentials develop.  Reluctantly one is therefore
 obliged  to look to a neologism which can be the sub-
ject of operational definition.  But first it is necessary
 to  delineate  more  precisely   the  mode  of  action
 concerned.

    It  is  proposed that, when the organism has to deal
 with  a  sequence  of two or more  events, a mode of
                                    Cooper et al.

 action of the nervous system is adopted in which the
 subject enters  a  selective state  of dynamic involve-
 ment with the  environment. This state is directed to-
 wards the execution of a planned action which can be
 an overt motor response, the inhibition  of a  motor
 response, or a decision. It  may develop rapidly or
 slowly and is sustained, subject to an upper time lim-
 it, until  the point  of decision or action. During its
 course there exists a continuous  interaction with  the
 environment such that  modification  of  the end re-
 sponse is at any time possible. The system on which
 the state depends has  a limited processing capacity
 and can thus be subject to overload when the quanti-
 ty or  complexity of incoming information exceeds
 certain limits. To express the main elements of the
 mode of action described, namely involvement of the
 subject, continuous interaction with the environment
 and  an  end result, we  propose  to  use the word
 'scopeutic'.*

    This term is used to express a selective  state of
 involvement with a particular set of circumstances
 which  have taken over  the  central processor for a
 given  period of time and which  require at their end
 point some action or decision.


    Physiologically, each occurrence of processing in
 the scopeutic mode is reflected  by a  particular pat-
 tern of activity throughout the nervous system. The
 central feature  of this pattern is  the development of
 slow potentials. This is accompanied by other physio-
 logical features ranging from changes in the excitabili-
 ty of spinal monosynaptic reflexes to changes in auto-
 nomic activity as shown by reduction of heart rate.

   These  physiological changes can be  reduced by in-
 creased task complexity, which leads to overload, or
 by distraction.  In these conditions some aspect of
 performance is usually impaired.  However, reduction
 can also occur in conditions where performance is not
 impaired,  for example  when  repeated presentation
 has resulted in overlearning.

   The notion  that  repeated practice of complex
 skills and overlearning of simple skills can result in
 the execution of those skills being transferred to sep-
 arate,  second order, semi-automatic systems  in the
*From  'scope', which in turn is derived  from the
Greek  ckorroa,  a mark  for  shooting at or aim, and
from the  Italian scopo, aim or purpose,  Among the
meanings  listed for 'scope' (O.E.D.) are:  goal, some-
thing aimed at  or desired,  something which  one
wishes to effect or attain, an end in view, an object,
purpose, aim. The distance to which the mind reach-
es in its working or purpose, reach or range of activi-
ty, extent of view, outlook or survey.

-------
Bimodal Theory

brain is not a new one. There has long been a distinc-
tion between states of involvement about which the
individual is aware and those  in which he performs a
task in an automatic and 'unthinking' manner, but
with the ability to  maintain  a good level of perfor-
mance. What is new in the situation is the suggestion
that such states are associated with the reduction or
absence of the slow potential changes described.

   To distinguish this mode of action from the previ-
ously outlined  scopeutic mode, it is proposed to use
the term  'categoric'. The word has been chosen be-
cause it reflects  both the conceptual categorization
that takes place and  the relative inflexibility of this
mode of action once established. As used in the con-
text of this  paper, it is taken to refer  to states  in
which  actions (and possibly  decisions) are executed
largely  automatically  and mechanically according  to
some predetermined and pre-established response pro-
cedures. Establishment  of this preset pattern of re-
sponse  is usually the result of repeated previous oc-
currences in  the scopeutic mode of a particular con-
figuration of stimuli and responses, which  become
firmly associated. This procedure has been referred  to
as overlearning in the sense that association  has been
carried well  beyond  the stage necessary for simple
learning or  the basic acquisition  of a skill  to take
place.

   The  predetermined  procedure  for   responding
adopted in the categoric mode, together  with the ab-
sence  of slow potential changes,  persists as long  as
there is no  unexpected change of environmental cir-
cumstances.  Relevant interaction  with the  environ-
ment  is maintained  to  the  extent  that unexpected
change will result in reversion to the scopeutic mode.
The essential feature of the categoric mode is that
performance executed within this mode  is not  focal;
it is secondary or ancillary to some other  activity that
is taking place at the same  time. This other activity
may be internal (mental) or  external (motor). Exam-
ples of categoric mode activity are the  motor skills
performed during driving, typing, and violin-playing.
During the performance of such actions, control can
be returned to the  scopeutic mode either voluntarily
or because the situation so demands. When such a
transfer takes place, it should be accompanied by the
 reappearance of slow  potential  changes associated
 with the task.

    Both  experimental evidence and everyday experi-
 ence  suggest  that  the two modes of  action, the
 scopeutic and the  categoric, are not mutually  exclu-
 sive. Thus, one  may be performing in the  categoric
 mode  a  routine, overlearned skill,  such as changing
 gear during car driving, while at the same time making
 quite complex decisions in the scopeutic mode in re-
 sponse to current traffic conditions or selection  of
 route. It would seem that the overall cerebral systems
                                             653

concerned have a finite capacity and that this capaci-
ty is shared between  the  two modes of action. The
scopeutic mode is relatively demanding on this limit-
ed capacity but is flexible in its responses to input,
whereas the  categoric mode is  probably  more eco-
nomical in its demands,  but economy is achieved at
the expense of flexibility. Division of the total capaci-
ty between the two  modes is constantly  changing.
Even when a task has been overlearned and has passed
from the scopeutic to the categoric mode, it appears
that "periodic sampling"  continues to take place in
the scopeutic mode,  perhaps as a  monitoring func-
tion. Some of the trial-to-trial variability encountered
in CNV experimentation might well be accounted for
on this basis.

   One of the clearest demonstrations of the interac-
tion of the two modes of processing is to be seen in
CNV experiments in which a simple, fixed-foreperiod
reaction-time task is extended for a large number of
trials over  a long period. During the early stages of
such  an experiment, CNV amplitude is  fairly con-
stant,  while reaction time tends to fall slightly. There-
after,  as the  task becomes boring and is responded to
automatically by the subject, CNV values  diminish,
but reaction-time values remain relatively constant. If
 any novel  element, which re-engages the subject's in-
 terest  and involvement, enters the situation, CNV am-
 plitudes increase once more. Reaction times may well
become longer for a time in such circumstances.

    These  stages  reflect  the  transition  from  the
 scopeutic  mode to the categoric  mode  as the  task
 becomes boring and back to the scopeutic mode  as it
 once  more engages the subject. In an extended  vigi-
 lance  task, in  which involvement remains high, slow
 potential changes persist to the end, the  subject re-
 maining in the scopeutic mode. If errors are made,
 they are found to be associated with the  failure of
 slow  potentials  to  develop. Experimental evidence
 also already exists for reductions in CNV amplitudes
 in conditions of task complexity  and when  distrac-
 tion is present. A  further experimental finding  that
 could be explained by the bimodal theory is the in-
 creased trial-to-trial variability of CNVs in children.
 The suggestion would be that children have a lower
 ability to  maintain the scopeutic  mode than adults.
 The additional  finding that CNVs can be recorded at
 very low ages if an appropriate and engaging experi-
 mental situation can  be found might suggest  that
 what  is being engaged is the scopeutic mode of ac-
 tion.

    It  has  been postulated that  the balance between
 scopeutic  and categoric  modes  of processing is con-
 stantly changing. There is some indication that if this
 balance shifts significantly in the direction of the cat-
 egoric, that is to say that routine, automatic tasks are
 occupying the major part of the processors capacity,
 there will be  a tendency for the individual to drift

-------
 654

 towards  a  state of sleep. It would seem that some
 minimum level of scopeutic processing is necessary to
 the maintenance of an active waking state.

    Although prominence has tended to be given to
 the occurrence or nonoccurrence of slow potential
 changes as indicators of the existence of individual
 states influencing  behaviour, such changes are only
 one sign of their existence. The involvement of other
 physiological factors, such as the changing excitabili-
 ty  of spinal   monosynaptic  reflexes and  cardiac
 changes,  have  already been  mentioned. Pupillary re-
 sponses and blood  pressure may also be involved, and
 yet more factors doubtless remain to be discovered. It
 would not be  surprising, for example, to find  bio-
 chemical correlates of these  states, perhaps even to
 find that each  mode of action was mediated by its
 own chemical transmitter or transmitters.

    Dissatisfaction  with  a purely phenomenological
 level of approach to brain slow potential changes has
 led  us to try to construct a theoretical framework,
 based  largely   on  existing  experimental evidence,
 which might enhance the general understanding of
 some brain  systems and their modes of operation. No
 theory of this kind could be acceptable unless it gen-
 erated further testable hypotheses. It is suggested that
 the  present  theory fulfills this condition and offers
 scope  for  the  construction  of further experimental
 paradigms both to  confirm the existence of the two
 modes and to investigate the conditions under which
 transition from one mode to the other takes place.

 Summary

   Slow  potential changes such as the CNV and BP
 have been related to a range of behavioural states in
 which  planned  actions or  decisions are  required.
 There is  reason to suppose that their presence reflects
 a  particular  mode  of brain function. It is proposed
 that this mode  of  action, termed 'scopeutic', repre-
 sents an aspect of dynamic involvement with the envi-
 ronment  that is selective, has a high degree of flexibil-
 ity,  and  is  sustained until the point of decision or
 action. It is in some sense related to cognitive  capaci-
 ty and  as such can be subject to overload.

   One method of avoiding overload of the system as
a whole utilizes  a secondary mode of action to which
 integrated elements  of central processing can be trans-
 ferred when, for example, skills have been overlearn-
ed and the input-output procedures stereotyped. This
 mode of action, which is less flexible  than the first,
 has been termed 'categoric'. It is relatively automatic
 in character  and is associated with a lower level of
subjective awareness. It is also associated with  the
absence or reduction of slow potential changes.

   The evidence suggests that these two modes of ac-
 tion can  co-exist and that together they share  the
                                    Cooper et  al.

 cerebral capacity for information  processing. It is
 likely that they can be defined not only by slow po-
 tential changes but by a whole range of physiological
 factors,  only a  few of which have so far been deter-
 mined.

 Comments on the bimodal model

   Loveless:  I have already expressed my general ob-
 jection to  this model, namely that I seek a different
 starting  point. I require a model  of what is going on
 in performance of an RT task, or a vigilance task, or
 another  standard psychological situation, rather than
 a model based on a particular physiological phenome-
 non.

   As implied in my foregoing analysis  of the basic
 CNV paradigm, I am dissatisfied with the reliance of
 the bimodal  theory on 'CNV amplitude' measured in
 a short foreperiod.  If there is indeed a distinct orient-
 ing potential, it is surely this that is likely  to habitu-
 ate with practice, as Weerts and Lang (1973) suggest,
 and  to be dishabituated by any novel element. There
 may not be habituation of the anticipatory potential
 (it  was  still increasing over the Weerts  and Lang
 series). To be persuaded about the dissociation be-
 tween RT and anticipatory potential, I would require
 that the foreperiod used had been  long enough (at
 least 5 seconds) to separate  the orienting and antici-
 patory potentials.


   Since the  model refers to slow potentials in gener-
 al, and since I suspect the anticipatory potential is a
 Bereitschaftspotential,  I should like to know whether
 the latter decreases with practice. By implication, the
 response should become automatic with practice,
 although an  appropriate criterion might be difficult
 to find.

   I  also find it difficult  to  relate the model to my
 findings  in signal detection  and  sensory set experi-
 ments. In both cases, there was a distinct  end point
 (or "scopes"), but  a  slow  potential conspicuously
 failed to  develop. The signal detection series was long,
so that one might appeal to practice and  boredom,
but attempts were made to  prevent this. This seems
much  less  plausible  for  the  sensory set case, since
sensory set  is notoriously difficult  to  maintain, so
that loss  of concentration would almost certainly pro-
duce a reversion to motor set.
   Naattinen:  The model is  well presented and sys-
tematizes much existing knowledge. Many analogies
presented were stimulating. However, I think that the
formulation would gain from a more detailed discus-
sion on the phenomenon of overlearning. One reason
slow potentials are attenuated during the course of a
session  during which the performance level remains
stabilized might be that the  pattern of performance

-------
Bimodal Theory

(or activation) becomes more economical. Much per-
ceptual-motor and other learning seems in essence to
be gradual inactivation of irrelevant mechanisms, and
this might have counterparts or reflections in central
phenomena such as slow potentials.

   A further concept  that could  be particularly rele-
vant to the bimodal view is the  concept of "effort"
(Naatanen 1973, 1976). While I can offer no direct
evidence from my own data that  slow potentials tend
to be absent in error situations, I share the view that
there is  practically no correlation between RT and
CNV amplitude, although  some  exceptionally  slow
RTs might be associated with a zero CNV.

   With regard  to the hypothesis that  one method of
avoiding overload of the system as a whole involves a
shift to a secondary (categoric)  mode of  processing
(associated with a relative  absence of CNV), I think
the shift to the secondary mode of action does not
occur for the sake of avoiding overload, but occurs to
avoid unnecessary  effort, this avoidance being made
possible by, and indicating, learning. The result then
is that we are  further  from the overload threshold
than before such learning had taken place.

   Some further points and data relevant to the bi-
modal view generally, and in particular to the notion
of a categoric mode, can be found in our experiments
dealing with RT in relation to expectancy and prepa-
ration  (Naatanen  and  Gaillard   1972, Gaillard and
Naatanen 1973). The "purpose" of the expectancy
phenomena has also been discussed  (Naa'tSnen and
Summala 1976) in  relation to road-user behavior. In
this last study, we found substantial evidence for the
'categoric' mode of action and have described many
 forms of expectancy phenomena as semi-automatic or
automatic processes  that  do  not require 'conscious
 control', but instead have direct access to the deter-
 mination of behavior.
    Weinberg: It seems clear from strictly behavioral
 observations that a distinction can be made between
 scopeutic and categoric activity. The illustrations of
 piano- and  violin-playing are good, for they press
 home the point that behavior can occur in a highly
 programmed manner that appears to be independent
 of immediate stimulus input, including proprioceptive
 input. I think it should be pointed out that behavior
 of the categoric kind can be not only highly complex
 motor  movements,  but  what  may ordinarily  be
 thought of as "cognitive function."

    In the original formulation of the bimodal theory,
 it was not clear whether scopeutic and categoric pro-
 cesses were  dichotomous, i.e., whether an individual
 was either in one state or the other (or possibly in
 both), or whether cerebral processing in these terms
 was to  be  thought of as a continuum, the ends of
                                             655

which were scopeutic and categoric. Almost all litera-
ture on simultaneous and successive processing deals
with situations in which both events being processed
are presumed to be in the scopeutic mode. In this
situation, the classic findings have been a "psycholog-
ical  refractoriness," in which  processing is succes-
sive-i.e., there is  a necessary  interval between the
two inputs for equal performance to  occur to both.

   I am glad to note that the current reformulation of
the theory clarifies the point that the  two modes can
exist in parallel. I think that the experimental data of
Kahneman (1973) suggest that  two tasks, a scopeutic
and categoric task, can  occur simultaneously, but that
there is a limited  capacity for what Kahneman calls
effort, which can be allocated to both tasks differen-
tially,  depending upon the payoff. In other words,
the  CNV  may  be thought of as an index of the
amount  of effort being invested in  one or several
tasks,  whether they be scopeutic or  categoric. This
explanation seems to fit better with my own disposi-
tion to identify CNV  amplitude as a variable that is
inversely related to the amount of effort invested in
the task.

    When the task requires discrimination in SI, the
 amount  of effort is presumably greater near SI than
 near S2, especially in long inter-stimulus  intervals.
 Normally, when SI and S2 are nondiscriminative, one
 might expect most of  the effort to occur after SI; S2
 simply triggers occurrence  of the response, the nature
 of which is determined shortly  after SI. If the data of
 Loveless and Sanford (1974) are considered, there are
 at least  two other interacting  factors that influence
 the CNV: an orienting resonse to SI and a prepara-
 tory motor set (the latter may  be related to the readi-
 ness potential). An orienting response would presum-
 ably be greater for a  scopeutic than for a categoric
 task, especially when  SI is a  difficult discriminative
 stimulus, and therefore, one would predict a greater
 amplitude of the  orienting response followed by a
 rapid  decline attributable to information  processing
 between SI and S2, followed  by an  increased ampli-
 tude just prior to S2, if no discriminative information
 is contained in S2. This seems to be what happens
 and brings into question  a positive  relationship be-
 tween scopeutic processing and CNV  amplitude.

    In Vancouver, we  have recently been looking into
 the question of how CNV amplitude  can be influ-
 enced by distracting  stimuli. From preliminary data,
 it appears that reduced CNV amplitude occurs prima-
 rily when distracting  stimuli can be confused with SI
 (what we prefer to call "confusion" rather than "dis-
 traction"); when what would be distracting stimuli, as
 defined by other  measures, occur either in the inter-
 trial interval, the interstimulus interval, or both, they
 have  very little influence on CNV amplitude. Our cur-
 rent data suggest that in a difficult discriminative task

-------
 656

 (when the CNV amplitude is in fact small) confusion,
 and  for  that matter  distraction without confusion,
 have less of an effect than they would in a simple
 highly  practiced task (categoric task). One interpreta-
 tion  is that, during a  scopeutic task, attention is riv-
 eted  to SI  and makes distraction or confusion more
 difficult. Distraction,  for  me,  means simply the de-
 gree  to which stimuli, other than  those involved in
 the primary task, are processed during the interval
 when processing  with respect  to  the  primary task
 must occur. If  there is no processing  of extraneous
 stimuli, then there is no  distraction.  Therefore, ac-
 cording to the hypothesis, if two scopeutic tasks were
 being performed simultaneously, the amplitude of the
 CNV should be greater than the amplitude for each
 individually, or  at least not less than  the amplitude
 associated with each task done separately. But in fact
 two scopeutic tasks done  concurrently lead to  the
 greatest interference.  When  a  categoric task  is in-
 volved, there should be a smaller effect of a concur-
 rent  scopeutic  task than  there would be if one
 scopeutic task was concurrent with a categoric task.
 This would be an interesting hypothesis to test.

   McCallwn: I would  like  to acknowledge the help-
 ful views  of  Dr. Weinberg,  which contributed to  the
 reformulation of our model to make explicit that pro-
 cessing in the two  modes could occur simultaneously,
 or at least could overlap one another in  time.

   It  is  interesting that both Naatanen  and Weinberg
 have developed their own views in terms of "effort.'
 This currently popular  concept  has  been put forward
 by Pribram and  McGuiness (1975), together with a-
 rousal and activation, as one of the basic attentional
control processes. I am  sure  that it deserves considera-
tion in  the construction of a theoretical framework,
but personally I find it  more fruitful to  think in terms
of the available capacity of the system rather than the
direction of effort.  Nevertheless, I accept  that  the
two concepts are to some extent complementary.
                                   Cooper et al.

   One aspect of the  bimodal theory commented
 upon by Naatanen and Weinberg, either explicitly or
 implicitly, was in its attempt to account for some of
 the changes that occur in CNV amplitudes in terms of
 overload.  This approach is consistent with the general
 notion of a limited capacity system seeking to achieve
 the most economical use of that capacity. However, I
 agree that more thought should be addressed to the
 question of whether reductions in CNV amplitude as
 a result of overload are to be subsumed as instances
 of transition to  the categoric mode or whether they
 require a separate form of interpretation in their own
 right. Kahneman's experiment, cited  by  Weinberg,
 provides a good example of this problem. The ab-
 sence of concomitant SP data makes it difficult to
 comment  more precisely on  the mechanisms opera-
 ting in this particular situation, but in general terms,
 one would expect  subtasks in a discriminative situa-
tion to be dealt with in  the scopeutic mode, with the
additional possibility that overload might occur if the
 discrimination were difficult. It is clear, however, that
simple performance measures cannot be relied upon
as indicators of central state. The dissociations found
between RT and CNV amplitude  (Rebert and Tecce
 1973, McCallum and Papakostopoulos 1973) and the
comments of Naatanen  in this volume leave little
doubt on this point.


   A theoretical formulation of the kind attempted in
 the bimodal model is  open to a number of criticisms.
 It starts from the data, uses these to infer something
 about the systems and  modes of action underlying
 their production, and then uses this inferential frame-
 work to predict or comment upon the  data that will
 be obtained in given sets of circumstances. Because of
 this it has been accused  - although with little justifi-
 cation - of being  circular  in its arguments. On the
 whole, the existing formulation would seem to fulfill
 two  of the basic  criteria  for scientific theorizing,
namely those of being descriptive and predictive.

-------
GENERAL DISCUSSION OF THEORETICAL ISSUES
RELATED TO EVOKED AND  SLOW POTENTIAL
CHANGES
    Weinberg: It seems to me that two distinct points
 of view have developed. One is that different roads
 lead to the same  place. There is a fundamental brain
 system and pattern of organization, but there are dif-
 ferent ways of looking  at it. Essentially, all we have
 to do to solve the problems is to communicate. The
 other point of view, which seems to me to be reflect-
 ed  in  Papakostopoulos' approach,  is  that  each  re-
 searcher in this field defines the system he is talking
 about in his own way on the basis of his own particu-
 lar  methods and data and that  it may be naive to
 assume  that  all these  data and all  these parallel
 courses will converge on one place.

    Papakostopoulos: I would, however, add the hope
 that because we are a common species we will some-
 how be able  to  communicate in the  future. In the
 meantime, let  us make  a start with the data  in our
 individual laboratories and collect these data in such a
 way that we cover the total individual.

    Cooper: I don't think we can ever do that, because
 one can never know when and  where  to  stop. One
 must be selective or  the procedure  becomes unman-
 ageable. We are already producing a large volume of
 data; what we need  now is to integrate this into a
 system. What impressed me about Skinner's approach
 was that he talked  not  so much about unit potentials
 as  about  intercommunication. We  have limited re-
 sources and are limited in many other ways, so it
 becomes imperative that we select  the relevant vari-
  ables to study. What the contributions to this volume
  show is that there  are relevant variables available for
  study.

    Donchln: I don't have an answer about what kind
  of system we are seeking, but I think that the parable
  of the blind man and the elephant is relevant to this
  discussion. There is no  way of guaranteeing that each
  one of us who  designs an elephant in the end will
  produce the same picture of an elephant. My elephant
  looks very different from Skinner's elephant because
  mine has no  gates.  My elephant does hundreds of
  things in parallel and does not  do things serially. It
  processes Information in many different ways, and I
  am only  able to measure limited aspects of this pro-
cessing at given points in time. If I measured them in
some  way  and  at  another point in  time, I would
doubtless obtain a very different answer. On  the
other  hand, if a particular elephant happens to bear a
resemblance to a cat and is in an anaesthetized state,
being  measured  in certain ways, then we find it has
gates  that can be turned on and off, and the problem
becomes one of the finding the switches.

   The point is that, when we are dealing with human
 beings interacting to complex tasks, it seems to me an
 almost hopeless endeavor to attempt to create models
 that  describe  the system  in terms  of  gates being
 switched on and off. It is very convenient to  say that
 we should  all come up with the same system, but the
 problem is that at the moment we are not talking the
 same language.

   Skinner: The problem remains, are we looking at
 the same elephant? The  axiomatic assumption is that
 a biological mechanism is producing  the phenomena,
 and  I cannot accept, as Donchin suggests,  that the
 phenomena are independent  of the biological system.
 Cooper pointed out, in order  to draw our scientific
 conclusions, we have to keep as many as possible of
 the variables constant while we look at a particular
 variable. The system is  so complex  that there is no
 guarantee  that we are ever keeping  all  the  variables
 constant except the one we looking at, and this raises
 obvious problems for  the application of scientific
 method. Nevertheless, this does not mean that we will
 ultimately fail to arrive at  the  same picture of an
 elephant. What I am suggesting is that instead of one
 of us concentrating on the trunk and the other on the
 tail,  we both concentrate on the trunk.

    Rebert; There are two  general methods of theory
 building, the constructionist and reductionist, which
 are  characterized respectively by  relating  observed
 phenomena  to  concepts  on more  general  or  more
 particular levels of discourse. One may "explain" the
 CNV, for example, either as a reflection of a set of
 hypothetical  psychological  states  (e.g., arousal, at-
 tention, Information processing - the constructionist
 approach, or in  terms of more particular, observable
  physiological processes - the reductionist approach. I

-------
  658
  am somewhat surprised that the general trend of the
  present discussion has been to place the emphasis on
  reductionist explanations,  as we have for the last 10
  years almost totally ignored that dimension and have
  concentrated  on  constructionist theories.  It  is  my
  feeling that this reversal has occurred to some extent
  because  of the frustrating nature  of our previous
  endeavors to make sense  of ERPs  in terms of their
  psychological  correlates.   Although  we  have  now
  obtained a massive data base concerning the minutiae
  of experimental variations  that will promote system-
  atic  modification  of ERPs, that data base seems not
  to have  brought  us  closer to  understanding the
  psychological  correlates of ERPs,  but to have con-
  fused us further. Ritter's  earlier presentation on  P3
  phenomena  seems to confirm that view.  It may be
  argued with some  justification  that we have come a
  long way in applying concepts from signal detection
 theory to electrophysiology, but it seems that this has
 been primarily a way of systematizing treatment of
 the data  rather than explaining it. As an explanatory
 framework, signal detection theory was applied in the
 earliest reports of P3 phenomenon and fairly early in
 the CNV literature. How much advance can be said to
 have  occurred as a consequence?


    The most fruitful developments in ERP analysis
  appear  to have derived   from  a  semi-reductionist
  approach in the systematic dissection of ERP  com-
  plexes,  relating specific  components to  particular
 experimental  manipulations  without   necessarily
 attributing specific  psychological  processes to the
 components. Essentially, this constitutes defining the
 orthogonal dimensions of the total  waveform, and is
 primarily descriptive.

    One of the  major stumbling blocks  to relating
 ERPs to psychological processes is that we see in the
 scalp-recorded event  such  a minute proportion  of
 brain activity, and we have no idea where the ERP we
 observe lies  in some supposed chain of events. Per-
 haps, too, we  have tried too hard to "freeze our fo-
 cus," as  Deecke put it, and have  missed the  true
 dynamic nature  of brain functioning. We know that
 there are  billions  of neuronal  transactions taking
 place  each second  and  the patterns of transactions
 vary  from instant to instant. It seems a truism that
 any particular psychological state is  associated with a
 complex and dynamic pattern of activity that we are
 unable to capture adequately at  the  scalp. That prob-
 lem is,  of course, what prompts  multiple  channel
 recording  and single trial analyses; yet even with such
 techniques we are constrained to a very limited view
in our human studies.

   Unfortunately, little  intracerebral  recording has
been  directed  to the delineation of patterning of
activity  among  intracerebral sites,  especially  with
 respect to the CNV paradigm, and only a crude esti-
                              General Discussion

  mate of intracerebral patterning in that situation has
  been developed. Roy John, however, is moving elec-
  trophysiological analyses in the right direction with a
  multivariate approach, and Skinner has provided us
  with some specifics about  a thalamocortical system
  that may be involved in mediating the CNV. Deecke
  has described  that system  as  an  everchanging thal-
  amcortical  focus having control over activation and
  inhibition  of specialized cortical  foci such as speech
  and visuoconstructive centers.

     Nevertheless, a thalamocortical model presents too
  simplistic a view to account for the widespread intra-
  cerebral system involved in  the inter-stimulus interval
  of the cued RT task, as  evidenced by SP recordings in
  thalamic and brainstem  nonspecific nuclei, basal gang-
  lia, hypothalamus, and paleocortex (cf. Rebert 1972).
  In addition, general systems  may be crudely delimited
  on the basis of SP polarity,  and such findings suggest
  that nonspecific and rhinencephalic  regions are in a
  dynamic reciprocal balance during development of
  the cortical  CNV.  We  often  forget, too, that the
  cortex is under inhibitory  control of other regions,
  notably the caudate nucleus, and  I have suggested'
  that the positive caudate SP observed in association
  with the CNV  may represent inhibition of the cau-
  date, which in turn releases  the cortex from caudate
  inhibition,  giving rise to a  surface-negative SP. This
 appears to  occur in concert with  thalamocortical
 excitatory processes.

    Skinner  places emphasis on the  role  of frontal
 granular cortex as the "Rome of the cerebral map."
 In the cat, which has a minute area of frontal granular
 cortex, it would be very difficult  without very fine
 mapping to conclude that frontal SPs were maximally
 generally in granular cortex. In monkeys performing
 in the CNV paradigm, the major focus of SP activity
 is in agranular promoter cortex, with a second major
 focus in the motor strip. This has been  a  consistent
 finding in several investigations that have mapped the
 monkey's CNV (McSherry and  Borda 1973, Rebert
 1972),  and  I believe  the same has  been found  to be
 true for humans, despite very early indications of a
 far  frontal  focus.  It  is unlikely,  therefore, that a
 frontal  granular cortex system plays  a major role in
 the preparatory or "waiting"  process, at least as speci-
 fically defined by the CNV paradigm.

   My main point is that attempts to relate the CNV
 to particular psychological constructs were frustrating
 in part  because of an inability  to observe the  total
 CNV system. However, so many experimental manip-
 ulations produce CNV variation that the CNV process
 must be a general one - arousal, if you will. What is
 meant  by arousal in  this context is  that a surface-
negative  SP  (probably)  represents excitation  of a
region of cortical tissue.  It is the total distribution of
excitatory and inhibitory events in the  total brain

-------
General Discussion

that constitutes the  neurophysiological correlate of
momentary psychological states, and it is to the end
of observing as much of that total brain, in as dynam-
ic a form as possible, that we must continue to move.

   Donchin: I cannot accept that the constructionist
approach, as  Rebert has labelled it, has come  to a
dead end. It seems to me that in the past year or so
there has been  a remarkable integration of behavioral
findings. Models have been developed (e.g., Hillyard
et  al.  1978)  that explain the  relationship between
psychological  variables and  macropotentials  in a
much more convincing way and face up to the defini-
tional  problem of the  psychological variables them-
selves. There has not been the opportunity to develop
the expectancy model  in detail, but I am impressed
by  the fact that we can operationally define,  with
considerable precision, what we need to measure in
order to predict an  enormous percentage  of the vari-
ance  in  P300  (e.g. Squires, Wickens, Squires and
Donchin, this volume). Assumptions on which such
predictions are made derive from outside the immedi-
ate evoked-potential research area -  from literature
on information processing  and reaction time - and
describe the amplitude  of P300 in meaningful psycho-
logical terms. Consequently, I see such models as by
no means  defunct,  but as being in  generally  good
shape.

    It is worth remarking that in this Fourth Congress
much  of the discussion has centered around motor
behavior  and  cognitive situations other than  those
involved in the CNV. The reason is, I think, that it is
much easier  to control and measure motor behavior
than define the properties of the behavior associated
with more cognitive preparatory operations.

    I would like, however, to pinpoint one issue that
has proved confusing in the design and interpretation
 of CNV experiments and has led to some discrediting
 of attempts to interpret the CNV. I would suggest
 that what has discouraged some  CNV research has
 been  the very low  correlation found between CNV
 and reaction time. Consider as an analogy an experi-
 ment  that tries to relate the depression of an accelera-
 tor of a car with the speed of the car. If one very
 carefully measured how  much the  accelerator  is
 depressed and how fast the  car is  going over the
 whole lifetime of the  car, one would probably find a
 very  low  correlation. To  conclude  from this that
 there is no relationship between the depression  of the
 accelerator and the speed of the car would, of course,
 be foolish. The reason is that such a conclusion does
 not take into account  the terrain over which the car is
 traveling.

    In the case of the CNV,  we tend to make the
  assumption that it represents excitation of 1he cortex
  in the sense used by Deecke.and that such excitation
                                            659

should be accompanied by faster performance. How-
ever, it might well be that the cortical system is mak-
ing an adjustment so as to maintain a steady level of
performance, given everything else that is going on in
the system. This hypothesis would explain some of
the difficulties in interpreting the CNV, although I
have not yet come up with any good experiments for
testing  the hypothesis.  I  raise  the  issue because I
think we  should not dismiss the CNV merely because
we have misled ourselves into looking for correlations
in circumstances where we  really should never have
expected  them.

   McCallum: I would  add that one of the  reasons
many of  us  persist with CNV research is  exactly
because it does  not show  a  simple, neat, and  tidy
correlation with reaction  time. If it did, we would
probably find that the vast existing body of literature
on RT, referred to by Loveless, would provide most
of the answers. What is so fascinating about the CNV
- and possibly other event-related potentials - is that
it apparently provides  an insight into  psychological
states about which simple performance measures are
singularly uninformative.


    Tecce: I assume that eventually we  all want our
 scientific models to apply to human behavior. I think
 the notion of amphetamine inducing surface negativ-
 ity is a little too simple, and I think the notion of one
 specific  arousal being  a comprehensive explanatory
 construct, at least in humans, is destined to  failure. I
 am  reminded  of an  experiment (Tecce and  Cole
 1974) in which amphetamine produced a paradoxical
 double-effect. One group showed high arousal, as
 measured by elevated  heart rate, but this was accom-
 panied by a quiescent cognitive state. Another group
 showed high arousal accompanied by cognitive surges,
 i.e.,  a flood of associations extraneous to the current
 experimental situation. With cognitive surges, CNV
 decreased; with  no cognitive  surge, CNV  did not
 decrease. This suggested to me that one has to make a
 distinction  between nonspecific arousal  and direc-
 tional properties that accompany nonspecific arousal.
 As Hebb (1955) pointed out,  high levels of arousal
 are  accompanied by  clear disturbances in steering
 functions, and therefore  I think this two-process, or
 bimodel, theory has to be kept in mind.

    My second  point is  that we are destined to failure
  in any  theory about human behavior unless we take
  subjective reports into consideration. This is one tool
  we  have  with humans that we do  not  have with
  animals. In our experiments, we were only able to
  find out about cognitive surges by asking the subjects.
   An additional argument in favor of the use of subjec-
  tive report is that, with individual differences running
   rampant in most of  our experiments, particularly
   with regard to  the use of strategies, it is important to
   ask the subject what he did during that experimental

-------
  660

  run. This provides a valid and important source of
  information relating to fluctuations in the amplitudes
  of CNV and P300 that otherwise would not be avail-
                                        McCallum

 able. I make a plea, therefore, for the use of subjec-
 tive  reports and the analysis of subgroups in terms of
 individual differences.
 CONCLUDING  REMARKS
 W. C. McCALLUM

 Burden Neurological Institute, Bristol,
 England

   We never  expected that some  neat and agreed
 model would emerge from  the Congress. Indeed, in
 preliminary discussions, one sensed a distinct reluc-
 tance  of individuals to formulate  even a tentative
 model of their own working hypotheses. Perhaps,
 therefore, the most  fruitful feature to  emerge has
 been the  ultimate willingness of a substantial number
 of researchers to  commit  themselves on paper to
 some of  the  basic  principles  and  hypotheses that
 guide their own research and  to examine critically
 and constructively the problems of overcoming inter-
 disciplinary barriers  of knowledge and  communica-
 tion. This is no small achievement, but it also serves
 to emphasize  the  distance  remaining  to overcome
 those  barriers  and  attain a true pooling and integra-
 tion of effort  toward the common goal of providing
 an adequate description and understanding  of the
 systems and mechanisms of the brain as they affect
 human behavior.

   Dr.  Loveless made the  plea that, in designing ex-
 periments to study  the  relationship of physiological
 variables   to various  aspects  of  behavioral  perfor-
 mance, we should  not ignore the extensive store of
behavioral data available. We would do well, for ex-
ample, to consult the considerable body of psycho-
logical  literature on  reaction-time before embarking
on complex RT paradigms In pursuit of CNV changes.
 However,  we are under no obligation to accept uncrit-
 ically  all the behavioral Interpretations of such data,
 or the plethora of constructs that psychology has pro-
 duced in the form of Intervening variables, as the only
 effective  behavioral anchor for our physiological find-
 ings.

   We  should subject to close scrutiny constructs
 such as arousal, attention, motivation and orientation
 before attempting1 to use them to account for ob-
 served physiological changes. I suspect that closer in-
 spection may lead to the rejection of  some as too
 loose and ill-defined. Arousal is a good example: It
 did not figure prominently in the psychological litera-
 ture until  after the discovery in the 1950s of that
 range of physiological phenomena associated with the
 excitatory  influences of the  brainstem reticular for-
 mation. It  became inextricably linked with the paral-
 lel concept of activation, the two terms being now
 virtually synonymous, and was further linked physio-
 logically with rather imprecisely defined patterns of
 autonomic  activity, having  a  very dubious unity.
 Psychologists  using  the  term somewhat naively  as-
 sume themselves to be talking about a well-establish-
 ed physiological state, whereas physiologists using the
 term equally naively assume that the diverse range of
 physiological changes they are describing are in some
 way  united by  a common  behavioral state  that
 psychologists  are competent to define and measure
 on the basis of independent criteria. The net result is
 a circularity in explanations offered, which does little
 to advance  our understanding of the brain systems
 and processes concerned.

   Similar objections might be raised to  the concept
of orientation. While  I  have  sympathy for  the argu-
ment  that we should not perpetuate notions of 'the
CNV' as an  invariant unitary phenomenon, I see It as
no solution to say that there are two components
rather than one, the first being an orienting response
and the second a Bereitschafts potential.  Even on the
most charitable definition of the orienting response,
the facts do not appear to support  this contention!
Furthermore, whatever is  taking place in terms of a
negative shift prior to the Imperative S2 stimulus, It is
in  several clear respects  independent from negative
changes taking place prior to  the execution  of a vol.
untary action.  It is important to  study  the  time

-------
Concluding Remarks

course of slow potential changes and variations in the
cortical distributions of these changes, but arguments
about whether there are one, two, three, or ten com-
ponents present in the  interstimulus interval between
SI and S2 are in the end likely to be sterile, implying,
as they  do, a fixed number of discrete and localized
generators. A more flexible  view would be that cor-
tical slow potential changes reflect a depolarizing pro-
cess  in a complex mosaic of neuronal domains, acti-
vated to a major  extent by  subcortical mechanisms.
Exactly which  elements  of this mosaic are active
during the various stages of signal processing, associa-
tion, and  preparation  for decision or action seems
likely to depend  on numerous factors  such  as con-
text, meaning, time relations, presence or absence of
 competing signals, instructions, set, and possibly even
 the level of activity in other systems, be they auto-
 nomic, endocrine, or  those  central mechanisms con-
 cerned with posture and motor control.
    The  mechanisms underlying the molar  levels  of
 change studied through event-related potentials need
 to be understood. The contributions of those investi-
 gating these mechanisms at the cellular level is vital to
 this understanding, particularly when  the lower cen-
 ters studied appear to perform important integrating
 and regulating functions as outlined by Skinner.  At
 another  neurophysiological level, it may be  that the
 systems  being investigated can be differentiated  on
 the basis of neurotransmitters that mediate them and
 on other aspects of CNS biochemistry. It is encour-
 aging, therefore, to note  an increasing biochemical
                                              661

and neuropharmacological interest leading to real pro-
gress in this direction.
   Ultimately, the one factor that unifies the diverse
lines of research and  theoretical positions outlined
here is the common use of event-related brain macro-
potentials  as a  research  tool.  The  reasons  that
emerged  piecemeal  throughout  the  various discus-
sions  deserve a brief reiteration. First, these macro-
potentials  enable us to study  the  brains of intact
human beings in relatively normal situations  with
minimal interference.  Secondly, in  a systematic, al-
though as yet poorly understood way, these macropo-
tentials seem to reflect the operation  of brain systems
and mechanisms concerned with the  highest levels of
cognitive   processes involved in  man's transactions
with his environment. Finally,  they  offer a possible
unit for the objective  measurement  of functions and
states previously only  describable in  subjective, men-
talistic terms. It would be foolish to suggest that they
solve at a  stroke all the problems of mind-brain rela-
tionships, but they open up possibilities, which  be-
cause of the constraints on recording, the dependence
on animals,  and the complexities of interaction, have
led many, working exclusively at  the  unit level of
 neurophysiology, to despair of achieving. For the be-
 havioral  scientist, brain  macropotentials may well
 have  a role to play in redefining and strengthening
 some of the looser psychological constructs.  In  the
 light of these hopes and prospects,  the need to con-
 struct an  adequate theoretical  framework for the in-
 terpretation  of brain  macropotentials and  for  the
 guidance of future research in this field can hardly be
 overemphasized.

-------
THEORETICAL  MODELS  REFERENCES
 Adey, W.R., Dunlop, C.W. and Hendrix, C.E. Com-
      puter  techniques in correlation and spectral
      analyses of cerebral slow waves during discrimi-
      native behavior.  Exp.   Neurol.,  1961,
      3:501-524.

 Adrian, E.D., Bremer, F., Jasper, H.H. and Delafre-
      snaye, J.F. (Eds.).  Brain Mechanisms and Con-
      sciousness. Oxford, Blackwell, 1954.

 Alexander, S.E., Newman, J.D. and Symmes, D. Con-
      vergence of  prefrontal and  acoustic inputs
      upon neurons in the superior temporal gyrus of
      the awake squirrel monkey. Brain Res., 1976,
      116:334-338.

 Andersen, P. Rhythmic 10/sec activity in the thala-
      mus.  In D.P.  Purpura  and  M.D. Yahr (Eds.),
      The  Thalamus.  Columbia  Univ.  Press,  New
      York, 1966:143-152.

 Andersen, P. and Andersson, S.A. Physiological Basis
      of  the  Alpha  Rhythm.   Appleton-Century-
      Crofts, New York, 1968.

 Andersen, P. and Eccles, J.C. Inhibitory phasing of
      neuronal  discharge.  Nature  (Lond.),  1962,
      196:645-647.

 Arduini, A., Mancia,  M.  and Mechelse, K.  Slow
      potential changes in the cerebral cortex  by
      sensory and reticular stimulation.  Arch. Ital.
      Biol., 1957,95:127-138.

 Arezzo,  J., Pickoff,  A. and Vaughan,  H.G. The
      sources and intracerebral distribution of audi-
      tory   evoked   potentials in  the alert rhesus
      monkey. Brain Res., 1975,90:57-73.

 Bartlett, J.R. and Doty, R.W. Influence  of mesence-
      phalic stimulation on unit activity in the striate
      cortex  of squirrel monkeys. J. Neurophysiol.,
      1974,37:642-652.

 Bartus, R.T. and Levere, T.E. Frontal decortication in
      rhesus  monkeys:  A  test of the interference
      hypothesis. Brain Res., 1977, 119:233-248.

Ben-Ari, Y., Konazawa, I. and Kelly, J.S. Exclusively
      inhibitory  action  of  iontophoretic acetylcho-
      line on single  neurones  of feline  thalamus.
      Nature (Lond.), 1976, 259:327-329.

Bradley, P.B. The central action  of certain drugs in
      relation to the reticular  formation of the brain.
      In H.H. Jasper et al. (Eds.), Reticular Forma-
      tion of of the  Brain. Little,  Brown, Boston,
      1958:123-149.

Brady, J-v- Motivational-emotional factors and intra-
      cranial self-stimulation. In  D.E. Sheer (Ed.),
      Electrical Stimulation of the Brain. The Hogg
      Foundation Research Series. Univ.  of  Tex.
      Press, Austin,  1961:413-430.
 Brazier, M.A,  Evoked responses recorded from the
      depths of the human brain, Ann, N Y  Acad
      Sci., 1964, 112:33-59,

 Bremer,  F. et  Stoupel,  N. Facilitation et inhibition
      des  potentials evoques corticaux dans 1'eveil
      cerebral. Arch. int. Physiol. Biochim., 1959
      67:240-275.

 Broadbent, D.E.  Decision  and  Stress.  Pergamon,
      London, 1>71.

 Broadbent, D.E.  Perception and Communication.
      Pergamon, New York, 1958.

 Brudno,  S. and Marczynski, T.J.  Temporal patterns,
      their distribution and redundancy in trains of
      spontaneous neuronal spike intervals  of the
      feline hippocampus studied with a non-para-
      metric technique.  Brain Res.,  1977, 25:65-89.

 Brutkowski,  S. Functions  of prefrontal cortex in
      animals. Physiol. Rev., 1965, 45:721-746.

 Callaway,  E.,  Jones, R.T. and  Layne, R.S.  Evoked
      response and segmental set of  schizophrenia.
      Arch. gen. Psychiat., 1965, 12:83-89.

 Caspers,  H. Relations of steady potential shifts in the
      cortex to the wakefulness-sleep spectrum. In
      M.A.B.  Brazier (Ed.), Brain Function, vol.  I:
      Cortical  Excitability and Steady  Potentials.
      Univ. of Calif. Press. Berkeley, 1963:117-123!

Caspers, H. Shifts of the cortical steady potential dur-
     ing various stages of sleep.  In M. Jouvet (Ed.),
     Aspects  Anatomo-fonctionnels  de  la  Physio-
     logic du Sommeil. Colloques Internationaux du
     Centre National  de la Recherche  Scientifique
     No. 127. Paris, 1965:213-229.

Caspers,  H. and Speckmann,  E.J. Cerebral p02  and
      pH:  Changes during convulsive  activity  and
      their significance for spontaneous arrest of sei-
      zures. Epilepsia, 1972, 13:699-725.

Clemente,  D.C., Sterman, M.B. and Wyrwicka, W.
      Post-reinforcement EEC synchronization  dur-
      ing alimentary behavior.  Electroenceph. clin
      Neurophysiol., 1964, 16:355-365.

Cooper,  R., McCallum,  W.C. and Papakostopoulus,
      D.A. Bimodal slow potential theory of cerebral
      processing. In I.E. Desmedt (Ed.), Cognitive
      Components in Cerebral  Event-related Poten-
      tials and Selective  Attention. Progr. clin. Neu-
      rophysiol., Vol. 6, Karger, Basel, in press.

Courchesne,  E.,  Hillyard, S.A.  and Galambos, R.
      Stimulus novelty,  task relevance and the visual
      evoked potential in man. Electroenceph. clin.
      Neurophysiol., 1975,39:131-143.

-------
Theoretical Models References

Creutzfeldt, O.D., Rosina, A., Ito, M. and Probst, W.
      Visual evoked response of single cells and of
      the EEC in primary visual area of the cat. J.
      Neurophysiol., 1969,32:127-139.

Deecke, L., Becker, W., Grozinger, B., Scheid, P. and
      Kornhuber, H.H. Human  brain potentials pre-
      ceding  voluntary limb movements.  Elec-
      troenceph.  clin.  Neurophysiol., 1973, Suppl.
      33:87-94.

 Deecke,  L., Grozinger, B. and Kornhuber, H.H. Vol-
      untary  finger movement  in  man:  Cerebral
      potentials and theory. Biol. Cybernetics,  1976,
      23:99-119.

 Deecke,  L., Scheid, P. and Kornhuber, H.H. Distribu-
      tion  of  readiness  potential, pre-motion  posi-
      tivity, and motor potential of the human cere-
      bral cortex  preceding voluntary finger move-
      ments. Exp. Brain Res., 1969, 7:158-168.

 Dempsey, E.W. and Morison, R.S. The interaction of
      certain spontaneous  and  induced  cortical
      potentials.  Amer.  J.  Physiol.,  1942,
       135:301-308.

 Donchin,  E. Methodological issues in CNV research.
      Electroenceph.  clin.  Neurophysiol.,  1973,
      Suppl. 33:3-17.

 Donchin, E., Gerbrandt, L.K., Leifer, L. and Tucker,
       L. Is the contingent negative variation contin-
      gent on a  motor response? Psychophysiology,
       1972,9:178-188.

 Donchin,  E., Gerbrandt, L., Leifer, L. and Tucker,
       L.R. Contingent negative variations and  motor
       response. Electroenceph. clin. Neurophysiol.,
       1978, Suppl. 33:187-190.

 Donchin, E., Kubovy, M., Johnson, R. and Herning, I.
       Graded  changes  in evoked  response (P300)
       amplitude as a function of cognitive activity.
       Percept. Psychophys. 1973, 14:319-324.

 Donchin,  E., Otto, D., Gerbrandt, L.K. and Pribram,
       K. Studies in the physiology of CNV. Electro-
       enceph.  clin.  Neurophysiol.,  1973,   Supp.
       33:257-261.

 Dongier, M., Dubrovsky, B. and Garcia-Rill, E. Slow
       cerebral potentials in psychiatry, Can. psychiat.
       Assoc. J., 1974, 19:177-183.

 Dongier, M. Clinical applications of the CNV: A re-
       view. Electroenceph. clin. Neurophysiol., 1973,
       Suppl. 33:309-315.

 Dumont, S. et  Dell, P. Facilitation reticulaire des
       mechanismes visuels corticaux. Electroenceph.
       clin. Neurophysiol., I960, 12:769-796.

 Dureman, R.A. and Edstrom,  R. EEG  and time per-
       ception. 22nd Rep.,  Dept. of Psych., Univ. of
       Uppsala, Sweden, Dec., 1964.
                                            663

Eccles, J.C. Conscious experience  and memory. In
      J.C. Eccles (Ed.), Brain and Conscious Experi-
      ence.  Springer  Verlag,  New  York,
      1966:314-338.

Evarts, E.V. Pyramidal tract activity associated with a
      conditioned hand movement in the monkey. J.
      Neurophysiol., 1966, 29:1011-1027.

Feldman,  S.M. and  Waller, H.J. Dissociation of elec-
      trocortical  activation  and behavioral  arousal.
      Nature, 1962, 196:1320-1322.

Filion, M., Lamarre, Y. and Cordeau, J.P. Neuronal
      discharges of the ventrolateral nucleus of the
      thalamus during sleep and wakefulness in the
      cat. II. Evoked activity. Exp. Brain Res., 1971,
      12:499-508.

Freeman, W. and Watts, J.W.   Retrograde degenera-
      tion of the thalamus following prefrental lobot-
      omy.  J.Comp. Neurol., 1947,86:65-93.

Freeman, W. and Watts, J.W.   The thalamic projec-
      tion to the frontal lobe.  In  Association  for re-
      search in  nervous and  mental disease:   The
      frontal  lobes,  Hafner,   New York,  1966:
      200-209.
 Frigyesi, T.L. Intracellular recordings from neurons in
       dorsolateral thalamic reticular nucleus  during
       capsular,  basal  ganglia, and mid line  thalamic
       stimulation. Brain Res., 1972, 48:157-172.

 Frigyesi,  T.L.  and Schwartz,  R. Cortical control of
       thalamic sensorimotor relay activities in the cat
       and the squirrel monkey. In T.L. Frigyesi, E.
       Rinvik, and M.D.  Yahr (Eds.), Corticothalamic
       Projections and Sensorimotor Activities. Raven
       Press, New York, 1972:161-195.

 Gaillard, A.W.K. Effects of warning-signal modality
       on  the  contingent negative variation  (CNV).
       Biol. Psychol., 1976,4:139-154.

 Gaillard,  A.W.  and  Naatanen,   R.  Slow  potential
       changes and choice reaction time as a function
       of  interstimulus  interval.  Acta  Psychologica,
       1973,37:173-186.

 Gaillard, A.W., Naatanen, R. and Mlntysalo, S.The
      effect of background slow potentials on CNV. (in
      preparation).

  Gazzaniga, M.S. and Hillyard, S.A. Attention mech-
       anisms  following brain bisection. In S. Kom-
       blum (Ed.).  Attention and Performance, Vol.
       4, Academic Press, N.Y., 1974:221-238.

  Gergen, J.A. and  MacLean, P.O. The limbic system:
       Photic activation of limbic cortical areas in the
       squirrel  monkey. Ann.  N.Y. Acad. Sci.,  1964,
        117:69-87.

  Gloor, P., Vera, C.L. and Sperti, L. Electrophysiologi-
       cal studies  of hippocampal  neurons.  Electro-
       enceph. clin. Neurophysiol., 1964, 17:353-370.

-------
664
                Theoretical Models References
Greengard, P. and Kebabian, J.W. Role of cyclic AMP
      in  synaptic  transmission  in the  mammalian
      peripheral nervous system. Fed.  Proc., 1974,
      33:1059-1067.

GrOzinger, B., Kornhuber, H. and Kriebel, J.EEG in-
    vestigation of hemispheric asymmetries preceding
    speech. In W.C. McCallum and J.R. Knott (Eds.),
    The  Responsive  Brain. Wright,  Bristol, 1976,
    103-107.

Hablitz, J.J. and  Borda, R.P. The effect of Dalmane
      (Flurazepam hydrochloride) on the contingent
      negative variation.  Electroenceph.  clin.
      Neurophysiol., 1973,Suppl. 33:318-320.

Harmon,  L.D.  Neural subsystems:  An  interpretive
      summary. In P.O. Schmitt (Ed.) The Neuro-
      sciences; Second Study  Program.  Rockefeller
      Univ. Press, New York, 1970:486-494.

Hebb, D.O. Drives and  the CJN.S. (conceptual nerv-
      ous  system). Psychol. Rev., 1955,62:243-254.

Held, H. and Hein, A. Movement-produced stimula-
      tion  in the  development of visually guided be-
      havior.  J,  comp.  Physiol.  Psychol.,  1963,
      56:872-876.

Herrington, R.N. and Claridge, G.S. Sedation thresh-
      old and  Archimedes' spiral after-effect  in early
      psychosis. J. psychiat. Res., 1965, 3:159-170.

Hillyard, S.A.  The CNV and human behavior:  A re-
      view. Electroenceph. clin. Neurophysiol., 1973,
      Suppl. 33:161-171.

Hillyard, S.A., Hink, R.F., Schwent, V.L. and Picton,
      T.W. Electric signs of selective attention in the
      human brain. Science, 1973, 182:177-180.

Hillyard, S.A.  and Picton, T.W, Event-related poten-
      tials  and selective information processing in
      man. In J,E. Desmedt (Ed.), Cognitive Compo-
      nents in  Cerebral  Event-related Potentials and
      Selective Attention. Prog.  clin. Neurophysiol,,
      Vol.  6, Karger, Basel, in press.

Hillyard, S.A., Picton, T.W. and Regan, D.  Sensa-
      tion, perception, and attention: Analysis using
      ERPs.  In E. Callaway, P. fueling, and S.H.
      Koslow (Eds.).  Event Related Brain Potentials
      in  Man.    Academic  Press,  1978:223-322.

Hockman, C. and Bieger, D. Chemical Transmission in
      the Mammalian Central Nervous System, Univ.
      Park Press, Baltimore, 1976.

Hohle  R.H. Inferred components of reaction times as
      functions  of  foreperiod  duration. J.  Exp.
      Psychol., 1965,69:382-386.

Imbert, M., Bignall,  K.E.  and  Buser, P.  Neocortical
      interconnections in  the cat. J,  Neurophysiol.,
      1966,29:382-395.

Itil, T.M., Saletu, B. and  Davis,  S. EEC findings in
      chronic  schizophrenics based on  digital  com-
      puter period analysis and analog power spectra.
      Biol. Psychiat., 1972, 5:1-13.

James, W. The Principles of Psychology. Holt, New
      York, 1890.

John, E.R. Mechanisms of Memory. Academic Press
      New York, 1967:468.
John, E.R., Bartlett, R., Shimokochi,  M. and Klein-
      man, D. Neural readout from memory. J. Neu-
      rophysiol., 1973,36:893-924.

John, E.R. and Schwartz, E.L. The neurophysiology
      of information processing and cognition. Ann.
      Rev. Psychol., 1978,29:1-29.
Jones, E.G. Some  aspects of the organization of the
      thalamic reticular complex. J. comp. Neurol.,
      1975, 162:285-308.

Jones, R.T., Blacker, K.H. and Callaway, E. Percep-
      tual  disfunction  in schizophrenia:  A clinical
      and  auditory response  findings.  Amer,  J.
      Psychiat., 1966, 123:639-645.

Jouvet,  M. Neurophysiology of the states of sleep. In
      G,C. Quarton, T,  Melnechuk and P.O. Schmitt
      (Eds.), The Neurosciences: A Study Program.
      Rockefeller Univ.  Press,   New  York,
      1967:529-576.

Kahneman, D. Attention and Effort. Prentice  Hall,
      Englewood Cliffs,  1973.

Kebabian,  J.W.,  Bloom, F.E.,  Steiner,  A.L. and
      Greengard, P. Neurotransmitters increase cyclic
      nucleotides  in postganglionic  neurons: Im-
      munocytochemical demonstration. Science,
      1975, 190:157-159.

Kelly, Jr., D.L., Goldring, S.  and O'Leary, J.L. Aver-
      aged evoked somatosensory responses from ex-
      posed cortex of  man. Arch. Neurol.,  1965,
      13:1-9.

Kievit, J. and Kuypers,  H.G.J.M. Organization of the
      thalamo-cortical connections to the frontal lobe
      in the Rhesus monkey.  Exp. Brain Res., 1977,
      29:299-322.

King, G.,  and Skinner, J.E.  Mesencephalic reticular
      activation reverses the disruptive  effects of low
      frequency medial  thalamic stimulation  on de-
      layed alternation  behavior.  Neuiosci. abstr.,
      1975,1:493.

Konorski,  J. and Lawicka, W. Analysis of errors by
      prefrontal  animals on the  delayed-response
      test.  In J.M. Warren and K. Akert (Eds.), The
      Frontal  Granular  Cortex  and  Behavior.
      McGraw-Hill, New York, 1964:271-294.

Kornhuber, H.H.  Cerebral  cortex, cerebellum and
      basal ganglia: An  introduction to their motor
      functions. In P.O. Schmitt (Ed.), The  Neuro-
      sciences.  III. Study  Program.  MIT  Press,
      Cambridge-London, 1974:267-280.

-------
Theoretical Models References
                                           665
 Komhuber, H.H. Motor functions of cerebellum and
      basal ganglia:  The  cerebello-cortical saccadic
      (ballistic) clock, cerebellonuclear hold  regula-
      tor and the basal ganglia ramp (voluntary speed
      smooth  movement)  generator.  Kybernetik,
       1971,8:157-162.

 Kornhuber, H.H.  Neural control of input into  long
      term memory: Limbic system  and amnestic
      syndrome in man. In H.P. Zippel (Ed.), Mem-
      ory and Transfer of Information. Plenum Press,
      New York, 1973:1-22.

 Krnjevic, K. Chemical nature of synaptic transmission
       in  vertebrates. Physiol.  Revs.,  1974,
       54:418-540.

 Kubie, L.S. Some implications for psychoanalysis of
       modern concepts of the organization of the
       brain. Psychoanal. Quart., 1953, 22:21-68.

 Lacey, J.I. Somatic response patterning and stress:
       Some revisions  of activation  theory. In  M.H.
       Appley and R. Trumbull (Eds.), Psychological
       Stress:   Issues in Research.  Appleton,  New
       York, 1967:14-44.

 Lacey, J.I. and Lacey, B.C. Some autonomic-central
       nervous system interrelationships. In P.  Black
       (Ed.),  Physiological  Correlates  of  Emotion.
       Academic Press, New York, 1970:205-228.

 Lachman,  R. The  model in  theory  construction.
       Psychol. Rev., 1960, 67:113-29.

 Lamarre, Y., Filion, M. and  Cordeau, J.P. Neuronal
       discharges  of the ventrolateral nucleus of  the
       thalamus during sleep  and wakefulness in  the
       cat. I.  Spontaneous activity.  Exp.  Brain Res.,
       1971,  12:480-498.

 Libet, B. Generation  of slow inhibitory and excita-
       tory postsynaptic potentials. Fed. Proc.,  1970,
       29:1945-1956.

 Libet,  B., Kobayaski, H, and  Tanaka, T. Synaptic
       coupling in the production and storage of a
       neuronal memory trace. Nature (Lond.),  1975,
       258:155-157.

 Lineberry, C.G. and Siegel, J. EEC synchronization,
       behavioral Inhibition,  and mesencephalic unit
       effects produced by  stimulation of orbital cor-
       tex, basal forebrain and caudate nucleus. Brain
       Res., 1971,34:143.161.

  Loveless, N.E. and Sanford, A.J. Slow potential cor-
       relates of preparatory set. Biol. Psychol., 1974,
        1:303-314.

  Loveless, N.E. and Sanford, A.J. The CSV  baseline:
        Consideration of internal consistency of data.
        Electroenceph. clin.  Neurophyslol.,  1973,
        Suppl. 33:19-23,

  Luria,   A.R.  The Working  Brain. Penguin,  London,
         1973.
Luria, A.R. and Homskaya, ED.  Frontal lobes and
     the  regulation  of arousal processes.  In  D.I.
     Mostofsky  (Ed.),  Attention:   Contemporary
     Theory  and  Analysis.    Appleton-Century-
     Crofts, New York, 1970:303-330.
Me Adam, D.W., Knott, J.R. and Robert, C.S. Cortical
      slow potential change in man related to inter-
      stimulus interval and to pre-trial prediction of
      interstimulus interval. Psychophysiology, 1969,
      5:349-358.

McAdam, D.W. and  Whitaker, HA.  Language pro-
     duction:  electroencephalographic localization
     in  the normal  human  brain.  Science,  1971,
      172:499-502.

McAfee,  D.A. and   Greengard,  P.  Adenosine 3',
      5 '-monophosphate:  Electrophysiological evi-
      dence  for a  role  in  synaptic  transmission.
      Science, 1972,178:310-312.

McCallum, W.C. and Abraham, P. The contingent neg-
      ative  variation  in  psychosis.  Electroenceph.
      clin. Neurophysiol., 1973, Suppl. 33:329-335.

McCallum, W.C. and Papakostopoulos, D. The CNV
      and  reaction time in situations  of increasing
      complexity.     Electroenceph. Clin.   Neuro-
      physiol., 1973, Suppl. 33:179-185.

McCallum, W.C., Papakostopoulos,  D. and Griffith,
      H.B.   Distribution of CNV  and  other slow
      potential changes  in human brainstem struc-
      tures. In W.C. McCallum and J.R. Knott (Eds.),
      The   Responsive  Brain.   Wright,   Bristol,
      1976:205-210.

 MacKay, D.M.  Cerebral  organization  and the con-
       scious  control of action. In J.C. Eccles (Ed.),
       Brain  and  Conscious Experience.  Springer
       Verlag, New York, 1965: 422-445.

 MacLean, P.D. The  triune brain,  emotion, and scien-
       tific  bias. In P.O. Schmitt (Ed.), The Neuro-
       sciences:  Second Study Program. Rockefeller
       Univ. Press, New York, 1970:336-349.

 McSherry, J.W. Physiological origins of steady poten-
       tials:  A  review. Electroenceph. clin.  Neuro-
       physiol., 1973, Suppl. 33:53-61.

 McSherry, J.W. and Borda, R.P. The intracortical dis-
       tribution of the CNV in rhesus monkeys. Elec-
       troenceph.  clin. Neurophysiol,,  Suppl. 33,
       1973:69-74.

 Magoun, H.W. The Waking Brain. Charles C. Thomas,
       Springfield, 1964:188.

 Malmo, R.B. Cognitive factors in impairment: A neu-
       ropsychological  study  of  divided set. J. Exp.
       Psychol., 1966,71:184-199.

 Marczynski,  T.J. Electrophysiological correlates  of
       positive  reinforcement:   Post-reinforcement

-------
        synchronization,  modulation of sensory input,
        and steady potentials. In A.G. Karczmar and
        J.C. Eccles (Eds.), Brain and Human Behavior.
        Springer Verlag, New York, 1972:303-321.


  Marczynski, T.J. Pattern recognition in sequences of
        neuronal spike intervals; a new non-parametric
        approach  and its implications. Materia med.
        Polona,  1976,8:97-107.

  Marczynski, T.J., York, J.L., Hackett, J.T. Steady
        potential correlates of positive reinforcement:
        Reward  contingent positive variation (RCPV).
        Science, 1969:301-304.

  Marczynski, T.J. and Bums, L.L. Reward  contingent
        positive  variation  (RCPV)  and post-reinforce-
        ment  EEC synchronization (PRS) in the  cat:
        Physiological aspects, the  effect of  morphine,
        and LSD-25, and a new interpretation of cho-
        linergic  mechanisms.  Gen. Pharmacol.,  1976,
        7:211-220.

  Marczynski,  T.J.  and Sherry, C.J. Induction of sleep
        and reward contingent positive  variation by
        photic stimuli in  the  cat.  Exp. Neurol., 1972,
        34:103-114.

  Marczynski, T.J., York, J.L., Rick, J.H., Allen, S.L.
       and Sherry, C.J. Steady potential  correlates of
       positive reinforcement  and sleep onset in cat:
        Reward contingent positive  variation (RCPV).
        Brain Res., 1971,26:305-332.

  Marczynski,  T.J., Hackett,  J.T.,  Sherry,  C.J.  and
       Allen, S.L. Diffuse light input and  quality of
       reward determine  the occurrence of reward
       contingent  positive variation (RCPV) in cat.
       Brain Res.,  1971b, 28:57-70.

 Marsh, G.R.  and  Thompson, L.W.  Effect of verbal
       and   non-verbal psychological  set  on  hemi-
       spheric asymmetries in the CNV.  Electroen-
       ceph,   Clin.   Neurophysiol.    1973,  Suppl.
       33:195-200.

 Martin, I. and Levey, A.B. The Genesis of the Classi-
       cal Conditioned Response. Pergamon, London,
       1969.

 Massion, J.  Etude  d'une  Structure Matrice  Thalami-
      que, le Noyau Ventrolateral et de sa Regulation
      par les Afferences Sensorielles. Theses de  Doc-
      torat d'Estudes Science Naturelles,  Faculte des
      Science de Paris, 1968.

Massion, J. and  Rispal-Padel, L. Differential control
      of motor  cortex and sensory areas on ventro-
      lateral  nucleus  of the thalarnus.  In T.L.  Frig-
      yesi, E. Rinvik and M.D. Yahr  (Eds.), Corti-
      cothalamic Projections and Sensorimotor Ac-
      tivities. Raven Press, New York, 1972:357-374.

Mathews, P.B.C. Mammalian  Muscle Receptors and
      their Central Actions. Arnold, London, 1972.
                  Theoretical Models References

 Meltzer, H.Y. and Stahl, S.M. The dopamine hypothe-
       sis of schizophrenia:  A review.  Schizophrenia
       Bull., 1976,2:19-76.

 Merton, P.A. The  properties of  the  human muscle
       servo. Brain Res., 1974, 71:475-478.

 Merton, P.A.  Speculations  on the servo-control of
       movement. In  J.L. Malcolm and J.A.B.  Gray
       (Eds.),  The Spinal Cord.  Ciba Foundation
       Symp.,Churchill, London, 1953:247-255.

 Miller, N.E. Learning and performance motivated by
       direct  stimulation of the brain.  In D.E. Sheer
       (Ed.),  Electrical Stimulation of  the Brain. The
       Univ. of Tex. Press, Austin, 1961:387-396.

 Millhouse, O.E.  A Golgi study  of the descending
       medial forebrain  bundle.  Brain Res   1969
       15:341-363.

 Moray, N. Attention: Selective Processes in Vision
       and Hearing. Academic Press, New York, 1970,

 Morison, R.S. and  Dempsey, E.W. A study of thala-
       mocortical relations. Amer.  J. Physiol., 1942
       135:281-292.

Morison, R.S.  and Dempsey, E.W.  Mechanism  of
      thalamocortical  augmentation and repetition
      Amer. J. Physiol., 1943, 138:297-308.

 Moruzzi, G.  and Magoun, H.W. Brainstem reticular
       formation and activation of the  EEG. Electro-
       enceph. clin. Neurophysiol., 1949, 1:455-473.

 Mountcastle,  V.B. The world around us. Neural com-
       mand functions for selective attention. Neuro-
       sci. Res. Prog. Bull., 1974, 14:1-47.

 Mountcastle,  V.B., Lynch,  J.C.,  Georgopoulos, A.,
       Sakata, H.  and  Acuna, C. Posterior parietal as-
       sociation cortex of the monkey: Command
       functions for operations within  extrapersonal
      space. J. Neurophysiol., 1975, 38:871-908.

NaStanen, R. Components of physiological activation
      in  psychophysiological performance  and per-
      sonality research.  Paper presented at Int.Conf.
      on Temperament and Personality,Warsaw 1976

Naatanen, R.  Evoked potential, EEG, and slow poten-
      tial correlates  of  selective  attention. Acta
      psychol. (Amst.), 1970,33:178-192.

Naatanen, R. Inverted U-shaped relationship between
      activation and performance.  A critical review
      In S.  Kornblum (ed.). Attention and Perfor-
      mance  IV.  Academic Press,   New  York
      1973:155-174.

Naatanen, R.  and Gaillard, A.W. The relationship be-
      tween  the  contingent  negative  variation  and
      reaction-time under  prolonged   experimental
      conditions.  Biol. Psychol., 1974, 2:277-291.

-------
Theoretical Models References

Naatanen, R. and Summala, H. Road-User Behavior
      and Traffic Accidents. American Elsevier, New
      York, 1976.

 Nauta,  W.J.H.  The problem of the  frontal  lobe: A
      reinterpretation. J.  psychiat.  Res.,  1971,
      8:167-187.

 Nebylitsin,  V.D.  Fundamental  Properties  of  the
      Human  Nervous  System.  Plenum,  London,
      1971.

 Neisser, I), Cognitive Psychology. Appleton-Century-
      Crofts, New York, 1967.

 Olds, J,  Differential effects of drives  and drugs on
      self-stimulation at different brain sites. In D.E.
      Sheer (Ed.),  Electrical Stimulation  of  the
      Brain.  The  Univ.   of Tex.  Press,  Austin,
       1961:350-366.

 Olds, J. and Olds, M.E. The mechanisms of voluntary
      ' behavior. In R.G. Heath (Ed.), The  Role of
       Pleasure in Behavior. Hoeber Med. Div., Harper
       & Row, New York, 1964:23-53.

  O'Leary, J.L. and Goldring, S. D-C potentials of the
        brain. Physiol. Rev.  1964,44:91-125.

  Oilman, R.T. Fast guesses in choice reaction time.
        Psychon. Sci., 1966, 6:155-156.

  Otto,  D.A. and Benignus, V.A. Slow positive shifts
        during  sustained motor activity in humans.
        Electroenceph. clin.  Neurophysiol.,  1975,
        38.-542P.

  Otto,  D.A.,  Benignus,  V.A.,  Ryan, L.J. and Leifer,
        L.J. Slow potential  components of stimulus, re-
        sponse and  preparatory process in man.  In J.
        Desmedt (Ed.), Attention, Voluntary Contrac-
        tion  and  Event-related  Cerebral Potentials.
        Progr. clin. Neurophysiol., 1977, 1:211-230.

  Otto,  D.A. and Leifer, L.J. The effect of modifying
        response and performance feedback parameters
        on the  CNV in humans.  Electroenceph. clin.
        Neurophysiol.,  1973, Suppl. 33:29-37.

  Papakostopoulos,  D. CNV and autonomic  function.
        Electroenceph. clin.  Neurophysiol,,  1973,
        Suppl. 33:269-280.

  Papakostopoulos, D.  Discussion. CNV and human be-
        haviour. Electroenceph,  clin.  Neurophysiol.,
        1973, Suppl. 33:243-244.

  Papakostopoulos,  D.  and Cooper, R. Brain, spinal
        cord and autonomic changes before, during and
        after a planned motor action in man. In V/.C.
        McCallum and J.R. Knott (Eds.), The Respon-
        sive Brain. Wright, Bristol, 1976:114-119.

  Papakostopoulos, D. and Cooper, R.  The contingent
        negative variation  and the excitability of the
        spinal monosynaptic reflex.  J.  Neurol. Neuro-
        surg. Psychiatry, 1973, 36:1003-1010.
                                            667

Papakostopoulos, D.  Cooper,  R.  and  Crow, H.J.
      Inhibition of cortical evoked  potentials and
      sensation by  self-initiated movement in man.
      Nature (Lond.), 1975, 258:321-324.

Papakostopoulos, D.  and Crow,  H.J.  The cortical
      somatosensory  evoked  responses  and  their
      changes  with  stimulus,  sensory or motor con-
      text.  In  J. Desmedt (ed.) Prog, clin. Neuro-
      physiol., Vol. 7, in press.

Penfield, W. and Rasmussen, T. The Cerebral Cortex
      of Man. Macmillan, NY, 1950.

Pfurtscheller, G. and  Cooper, R. Selective averaging
      of the intracerebral click evoked responses in
      man; An improved method of measuring laten-
      cies and  amplitudes. Electroenceph. clin. Neu-
      rophysiol., 1975, 38:187-190.

Phillis, J.W., Kostapoulos, G.K, and Limacher, J.J. A
      potent depressant action  of adenine derivatives
      on cerebral cortical neurons. Europ, J, Pharma-
      col., 1975, 30:125-129.

Picton, T.W.  and  Hillyard,  S.A. Human auditory
      evoked  potentials: II.  Effects  of attention.
      Electroenceph.  clin.   Neurophysiol.,  1974,
      36:191-199.

 Posner, M.I. Psychobiology of attention. In M.S. Gaz-
       zaniga and C. Blakemore (Eds.), Handbook of
       Psychobiology.  Academic  Press,  New York,
       1975:441-480.

 Pribram,  K.H. and McGuinness, D. Arousal, activa-
       tion, and effort in the  control of  attention.
       Psychol. Rev.,  1975,82:116-149.

 Proctor,  L.D., Knighton, R.S.,  Noshay,  W.C.  and
       Costello, R.T. (Eds.).  Reticular Formation of
       the Brain. Little, Brown, Boston, 1958.

 Purpura, D.P. Operations and processes in thalamic
       and synaptically related neural subsystems. In
       P.O. Schmitt (Ed.), The  Neurosciences. Second
       Study  Program. Rockefeller  Univ.  Press,
       1970:458-470.

 Purpura, D.P. and Cohen, B.  Intracellular recording
       from thalamic neurons during recruiting re-
       sponses. J.  Neurophysiol., 1962, 25:621-635.

 Purpura,  D.P., McMurtry,  J.G.  and  Maekawa, K.
       Synpatic events in ventrolateral thalamic  neu-
       rons during  suppression of recruiting responses
       by brainstem reticular stimulation. Brain Res.,
        1966,1:63-76.

  Rapasardi, S.C., Wilson, P.O. and Alvarez, F.L. Retic-
       ular modulation  of evoked potentials at the
       cortex and lateral geniculate nucleus of the un-
       anesthetized squirrel  monkey. Exp.  Neurol.,
        1974,44:282-294.

  Rebert, C.S. Cortical and subcortical slow potentials
       in the  monkey's brain during  a preparatory

-------
  668
                 Theoretical Models References
      interval.  Electroenceph. clin.  Neurophysiol.,
      1972,33:389-402.

 Rebert,  C.S.  Slow potential correlates of neuronal
      population responses in the cat's lateral genicu-
      late  nucleus.  Electroenceph.  clin.  Neuro-
      physiol., 1973,35:511-515.

 Rebert, C.S. and Knott, J.R. The vertex non-specific
      evoked  potential and latency of contingent
      negative variation. Electroenceph. clin. Neuro-
      physiol., 1970,24:561-565.

 Rebert, C.S. and Tecce, J.J. A summary of CNV and
      reaction  time.  Electroenceph.  clin.  Neuro-
      physiol., 1973, Suppl. 33:173-178.

 Remond, A.  Level  of organization of  evoked  re-
      sponses in man. Ann.  N.Y.  Acad.  Sci., 1964,
      112:143-159.

 Rick, J.H. and Marczynski, T.J.  Electrical activation
      of visual pathways  substitutes for tonic light
      input  in triggering EEC correlates  of food re-
      ward during conditioned behavior in cats. Brain
      Res. 1978, 154:105-118.

 Risberg,  J.,  Halsey, J.H., Wills, E.L. and Wilson, E.M.
      Hemispheric specialization  in  normal  man
      studies  by bilateral measurements of  the  re-
      gional cerebral blood  flow. A study  with the
      133-Xe  inhalation  technique.   Brain,   1975,
      98:511-524.

 Ritter, W. and Vaughan,  Jr., H.G.  Averaged  evoked
      responses  in vigilance and  discrimination:  A
      reassessment. Science, 1969,164:326-328.

 Ritter, W., Vaughan, Jr., H.G. and Costa, L. Orienting
      and habituation to auditory stimuli: A study of
      short  term changes  in  average  evoked  re-
      sponses.  Electroenceph. clin.  Neurophysiol.,
      1968,25:550-556.

 Rohrbaugh, J.W., Syndulko, K. and Lindsley, D.B.
      Brain  wave components of the contingent neg-
      ative  variation  in  humans.    Science,  1976,
      191:1055-1057.

 Rossi, G.F.  and Zanchetti, A. The brain stem retic-
      ular  formation:    Anatomy  and  physiology.
      Arch.itil.Blol., 1957,95:199435.

 Roth, W.T., Ford, J.M. and Kopell, B.S.  Effects of
      stimulus probability  and task  relevance  on
      event-related  potentials.    Psychophysiology,
      1976,13:311-317.

Ruiiell, B. Human Knowledge: Its Scope and Limits.
      Schuiter, New York, 1948.
Scheibel, M.E. and Scheibel, A.B. The organization of
      the nucleus reticularii thalami: A Golgi study.
      Brain  Rei., 1966, 1:43-62.

 Scheibel, M.E. and Scheibel, A,B. Structural organiza-
      tion of nonspecific thalamic nuclei and their
      projection  toward cortex,  Brain Res.,1967a,
      6:60-94.

 Scheibel, M.E. and Scheibel, A.B.  Anatomical basis
      of attention  mechanisms in vertebrate brains.
      In  G.C.  Quarton, T.  Melnechuk and  P.O.
      Schmitt  (Eds.), The  Neurosciences.   Rocke-
      feller Univ. Press, New York,  1967b:577-602.

 Scheibel, M.E. and Scheibel, A.B. Elementary  proc-
     esses  in  selected  thalamic  and  cortical  sub-
     systems  -  the  structural substrates.  In  P.O.
     Schmitt (Ed.), the Neurosciences. Second Study
     Program. Rockefeller Univ. Press, 1970:443-457.

 Schlag, J. and Villablanca, J. Cortical incremental re-
      sponses  to thalamic  stimulation.  Brain  Res.,
      1967,6:119-142.

 Schlag, J. and Waszak, M. Characteristics of unit re-
      sponses  in nucleus reticularis  thalami. Brain
      Res., 1970,21:286-288.

 Schlag, J. and Waszak, M. Electrophysiological prop-
      erties of  units of the thalamic reticular  com-
      plex. Exp. Neurol., 1971, 32:79-97.

 Schulman, J.A. and Weight, F.F. Synaptic transmis-
      sion:  Long-lasting  potentiation  by  a  post-
      synaptic  mechanism.  Science,  1976,
      194:1437-1439.

 Shagass, C. An electrophysiological  view of schizo-
      phrenia. Biol. Psychiat, 1976, 11:3-30.

 Sheer, D.E. Emotional  facilitation in learning situa-
      tions with subcortical  stimulation. In  D.B.
      Sheer (Ed.) Electrical  stimulation of the brain.
      Univ. of Tex.  Press, Austin, 1961:431-464.

 Siggins,  G.R.,  Battenberg,  E.F.,  Hoffer,  B.J. and
      Bloom,  F.E.  Noradrenergic  stimulation  of
      cyclic adenosine  monophosphate  in rat  Pur-
      kinje neurons: An immunocytochemical study.
      Science, 1973,179:585-588.

Simson,  R., Vaughan, Jr., H.G. and Ritter, W. The
      scalp  topography of potentials associated  with
      missing  visual and  auditory stimuli.  Electro-
      enceph. clin.  Neurophysiol., 1976, 40:33-47.

 Skinner, J.E.   Abolition of several forms of cortical
      synchronization during blockade in the inferior
      thalamic  peduncle. Electroenceph.  din. Neuro-
      physiol., 197 la, 31:211-221.


Skinner, I.E.   Abolition of a conditioned, surface-
      negative,  cortical potential  during cryogenic
      blockade  of  the  nonspecific  thalamocortlcal
      system.    Electroenceph.  clin. Neurophysiol.,
      1971b, 31:197-209.

Skinner, I.E.  Neurosclence: A Laboratory Manual.
      Saunders, Philadelphia, 197 Ic.

-------
Theoretical Models References
                                            669
Skinner, J.E. and Lindsley, D.B. Electrophysiological
      and behavioral effects of blockade of the non-
      specific thalamocortical  system.  Brain Res.,
      1967,6:95-118.

Skinner, J.E. and  Lindsley, D.B. Enhancement  of
      visual  and  auditory evoked potentials during
      blockade  of the nonspecific  thalamocortical
      system.  Electroenceph.  clin.  Neurophysiol.,
      1971,31:1-6.

Skinner, J.E. and  Lindsley,  D.B. The nonspecific
      mediothalamic-fronto-cortical  system:  Its
      influence  on electrocortical activity  and be-
      haviour.  In  K.H.  Pribram and  A.  R. Luria
      (Eds.), Psychophysiology of the Frontal Lobes.
      Academic Press,   New  York-London,
      1973:185-234.

Skinner, I.E., Reed, J., Welch, K.M.A. and Nell, J.
      Cutaneous  shock procedures correlated shifts
      in slow  potential  amplitude  and  cyclic  3',
      5'-adenosine  monophosphate  levels  in   the
      parietal cortex of the  conscious rat.  J. Neuro-
      chem. 1978a, 30:699-704.

Skinner, J.E., Welch, K.M.A., Reed, J., and Nell, J.
      Psychological stress  reduces  cyclic  3',  5'-
      adenosine  monophosphate  levels  in  the  cere-
      bral  cortex  of  conscious rats, as determined
      by a  new  cryogenic  method  of rapid tissue
      fixation.  J.  Neurochem., 1978b,  30:691-698.

 Skinner, J.E. and Yingling, C.D. Regulation of slow
      potential shifts in nucleus reticularis thalami by
      the  mesencephalic  reticular formation and  the
      frontal granular cortex.  Electroenceph. clin.
      Neurophysiol., 1976,40:288-296.

Skinner, J.E. and Yingling, C.D, Central gating mech-
      anisms that regulate  event-related  potentials
      and  behavior. A neural model  for attention. In
      J.E. Desmedt (Ed.), Prog.  Clin. Neurophysiol.
      Karger, Basel, 1977, 1:30-69.

Smythies, J.R. and Adey, W.R. Brain Mechanisms and
      Behavior. Academic Press, New York, 1970.

 Spencer, W.A. and Brookhart, J.M. Electrical patterns
       of augmenting and recruiting waves in depths
       of  sensorimotor  cortex  of  cat.  J.  Neuro-
       physiol.,  1961,24:26-49.

 Sperling, G. The information available in brief visual
       presentations. Psychol. Mon.,  1960, 74 (Whole
       No. 498).

 Sperry, R.W. A modified concept of  consciousness.
       Psychol. Rev., 1969, 76:532-536.

 Sperry, R.W.  An objective  approach  to subjective
       experience: Further explanation of a hypothe-
       sis.  Psychol. Rev., 1970,77:585-590.

 Spong, P.,  Haider, M. and Lindsley,  D.B. Selective
       attentiveness and cortical  evoked responses to
     visual  and  auditory  stimuli. Science, 1965,
     148:395-397.

Squires, K.C., Hillyard, S.A. and Lindsay, P.H. Vertex
     potentials evoked  during auditory signal detec-
     tion:  Relation  to decision  criteria. Percept.
     Psychophys. 1973, 14:265-272.

Squires, N.K., Squires, K.C. and Hillyard, S.A. Two
     varieties of long-latency positive  waves evoked
      by  unpredictable  auditory  stimuli.  Electro-
      enceph. clin. Neurophysiol., 1975, 38:387-401.

Steriade, M. and Deschenes, M. Cortical interneurons
      during sleep and  waking in  freely moving pri-
      mates. Brain Res., 1973, 50:192-199.

Steriade, M. and Wyzinski, P. Cortically elicited activ-
      ities in thalamic reticularis neurons. Brain Res.,
      1972,42:514-520.

Sterman,  M.B.  and Wyrwicka, W.  EEC correlates of
      sleep:  Evidence  for  separate  forebrain  sub-
      strates. Brain Res., 1967,6:143-163.

Stone,  T.W., Taylor, D.A.  and Bloom,  F.E.  Cyclic
      AMP  and  cyclic  GMP may mediate opposite
      neuronal responses in the rat  cerebral cortex
      Science, 1975, 187:845-847.

Sutton, S. The specification of psychological variables
      in the average evoked potential experiment. In
      E.  Donchin and D.B. Lindsley (Eds.), Average
      Evoked Potentials. NASA SP-17 I.Washington
      D.C., 1969:237-297.

Syndulko, K. and Lindsley, D.B. Motor and sensory
      determinants of cortical slow potential shifts in
      man.  In J. Desmedt (ed.) Progr. clin.  Neuro-
      physiol., Karger, Basel, 1977, 1:97-131.

Tecce,  J.J.  Contingent  negative variation  (CNV) and
      psychological processes in man.  Psychol. Bull.
      1972,77:73-108.

Tecce,  J.J.  and  Cole, J.O. Amphetamine effects in
      man: Paradoxical drowsiness and lowered elec-
      trical  brain  activity (CNV).  Science,   1974
      185:451-453.

Teuber, H.L. Prospects  for research on schizophrenia:
      Effects of focal  brain lesions.  Neurosci  Res
      Prog. Bull., 1972, 10:381-384.

Thach,  W.T. Timing of activity in cerebellar dentate
      nucleus  and  cerebral  motor  cortex  during
      prompt volitional movement. Brain Res., 1975,
      88:233-241.

Thierry, A.M., Stinus, L., Blanc, G. and Glowinaki, J.
      Some evidence for the existence of dopaminer-
      gic neurons in the rat cortex. Brain  Res., 1973,
      50:230.

Timsit-Berthier, M., Delaunoy, J., Koninckx,  N. and
      Rousseau,  J.C.  Slow potential  changes  in
      psychiatry. I.  Contingent  negative variation.

-------
  670
                 Theoretical Models References
       Electroenceph.  clin. Neurophysiol.,  1973,
       35:355-361.

 Uttley,  A.M.  Simulation  studies of learning  in an
       informon  network.  Brain   Res.,  1976,
       102:37-53.

 Valbo, A.B. Muscle spindle  response at the onset of
       isometric voluntary  contractions in man. Time
       difference between  fusimotor and  skele-
       tomotor  effects.  J. Physiol. (Lond.),  1971,
       218:405-431.

 VanHoesen, G.W., Pandya, D.N. and Butters, N. Cor-
       tical afferents to the entorhinal cortex of the
       rhesus   monkey.   Science,   1972,
       175:1471-1473.

 Vaughan,  H.G.  The relationship  of brain activity to
       scalp recordings of event-related potentials. In
       E.  Donchin and D.B. Lindsley (Eds.), Average
       Evoked Potentials. NASA SP-171, Washington,
       D.C., 1969:213-241.

 Vaughan,  Jr., H.G., Costa, L.D.,  Ritter, W. Topogra-
       phy of the human  motor potential. Electro-
       enceph. clin. Neurophysiol., 1968, 25:1-10.

 Vaughan, H.G. and Ritter, W. Physiologic approaches
       to  the analysis of attention and performance.
       In S. Kornblum (Ed.), Attention and Perform-
       ance  IV.  Academic  Press,  New  York,
       1973:129-154.

 Velasco, M. and Lindsley, D.B. Role of orbital cortex
       in regulation of thalamocortical electrical  activ-
       ity. Science, 1965, 149:1375-1377.

 Velasco, M., Skinner, J.E.,  Asaro, K.D. and Lindsley,
       D.B. Thalamo-cortical systems regulating  spin-
       dle bursts and recruiting responses. I. Effect of
       cortical ablations. Electroenceph. clin.  Neuro-
       physiol., 1968,25:463-470.

 Villablanca, J. and Schlag, J. Cortical control of thala-
      mic  spindle  waves.  Exp.  Neurol.,  1968,
      20:432-442.

 VonFoerster, H. On self-organizing systems and  their
      environment. In  M.  Youits and  S. Cameron
      (Eds.),  Self-organizing  Systems. Pergamon
      Press, Oxford, 1960:35-50.

Walter, W. Grey. The convergence and interaction of
      visual, auditory, and tactile responses in human
       nonspecific cortex. Ann. N.Y.  Acad.  Sci
       1964,112:320-361.

 Walter, W.G., Cooper, R., Aldridge, V.J., McCallum,
       W,C. and Winter, A.L. Contingent negative vari-
       ation:  An electric sign of sensory-motor associ-
       ation  and expectancy  in the human  brain
       Nature (Lond.),  1964, 203:380-384.

 Watanabe, S., Konishi, M. and Creutzfeldt, O.D. Post-
       synaptic potentials in the cat's visual cortex
       following  electrical stimulation  of  afferent
       pathways.  Exp.  Brain Res., 1966, 1:272-283.

 Weerts, T.C.  and Lang, P.J. The effects of eye fixa-
       tion and stimulus and response location on the
       contingent negative  variation  (CNV). Biol
       Psychol., 1973, 1:1-19.

 Weight, F.F.  and Padjen, A. Slow synaptic inhibition:
       Evidence for  synaptic inactivation of sodium
      conductance  in sympathetic  ganglion  cells
      Brain Res.,  1973,55:219-224.

 Weight, F.F., Petzold,  G.  and Greengard, P. Guano-
      sine 3 ', 5 -monophosphate in sympathetic gang-
      lia:  Increase associated with synaptic transmis-
      sion. Science,  1974, 186:942-944.

 Weight, F.F. and Votava, J. Slow synaptic excitation
      in synaptic ganglion cells:  Evidence for synap-
      tic  inactivation  of  potassium  conductance.
      Science, 1970, 170:755-758.

 Weinberg, H. and Papakostopoulos, D.  The frontal
      CNV:  its dissimarity to CNV's recorded from
      other sites.  Electroenceph. clin. Neurophysiol.,
      1975,39:21-28.

 Wikler, A. Pharmacologic dissociation of behavior and
      EEC  "sleep  patterns"  in  dogs:  Morphine,
      N-allylnormorphine, and atropine. Proc. Soc.
      Exp. Biol. Med., 1952, 79:261-265.

Yingling, C.D. and Skinner, J.E.  Regulation of unit
      activity  in  nucleus  reticularis thalami by  the
      mesencephalic  reticular  formation  and  the
      frontal  granular  cortex.   Electroenceph. clin.
      Neurophysiol.,  1975,39:635-642.

Yingling, C.D. and Skinner, I.E.  Selective regulation
      of thalamic sensory  relay nuclei by nucleus re-
      ticularis thalami. Electroenceph. clin. Neuro-
      physiol., 1976,41:475-482.

-------
671
                                            TECHNICAL REPORT DATA
                                   (Please read Instructions on the reverse before completing)
  REPORT NO.
     EPA-600/9-77-043
                                                                          3. RECIPIENT'S ACCESSION»NO.
  TITLE AND SUBTITLE
   MULTIDISCIPLINARY PERSPECTIVES IN EVENT-RELATED BRAIN
   POTENTIAL RESEARCH
                                         5. REPORT DATE
                                           December 1978
                                         6. PERFORMING ORGANIZATION CODE
  AUTHORIS)
   David A. Otto, Editor
                                                                          8. PERFORMING ORGANIZATION REPORT NO.
  "PERFORMING ORGANIZATION NAME AND ADDRESS
   Health Effects Research Laboratory
   U.S. Environmental Protection Agency
   Research Triangle Park, NC 27711
                                         10. PROGRAM ELEMENT NO.
                                           1AA817
                                         11. CONTRACT/GRANT NO.

                                           R803494-02
 "SPONSORING AGENCY NAME AND ADDRESS
   Health Effects Research Laboratory
   Office of Research and Development
   TJJS. Environmental Protection Agency
   Research Triangle Park, NC 27711
                                         13, TYPE OF REPORT AND PERIOD COVERED
                                           in-house
                                         14. SPONSORING AGENCY CODE
                                           EPA 600/11
  6. SUPPLEMENTARY NOTES
    This volume is the Proceedings of the Fourth International Congress on Event-Related Potentials of the Brain (EPIC-IV)
   held in Hendersonville, North Carolina in April 1976.  It contains 118 manuscripts including critical reviews and data reports
   in the following areas of ERP research:  1) electrogenesis and neurochemistry, 2) motor control, 3) information processing,
   4) language, 5) developmental disorders, 6) psychopathology, 7) environmental toxicology, 8) scalp distribution, 9) alter-
   natives to signal averaging, and 10) theoretical models. The environmental section assesses the progress of neurobehavioral
   research on the health effects of environmental toxicants, evaluates the utility of evoked potential research techniques in
   environmental toxicology, and encourages neurobehavioral research in problems of environmental concern.
    Many issues concerning the neurochemical substrate  and functional significance of ERPs are discussed with the objective
   of defining multidisciplinary approaches for resolution. The volume provides useful reviews and models to assimilate the
   growing body of ERP data from anatomy, physiology, pharmacology, psychology, linguistics, toxicology, neurology,
   psychiatry, and allied clinical sciences.  The proceedings will serve, therefore, as a planning document as well as a progress
   report in the field of ERP research.
 17.
                                         KEY WORDS AND DOCUMENT ANALYSIS
                        DESCRIPTORS
                          b.lOENTIFIERS/OPEN ENDED TERMS
                                    COSATI Field/Group
   CNS function
   electroencephalography
   electrogenesis
   evoked potentials
   neuroanatomy
   neurolinguistics
   neuropharmacology
neurophysiology
neuropsychology
neurotoxicology
psychopathology
psychophysiology
contingent negative variation
developmental disorders
event-related brain potentials
information processing
motor-related potentials
P300
05 J
06P
                N STATEMENT

     RELEASE TO PUBLIC
                           8. SECURITY CLASS (ThisReport)
                             UNCLASSIFIED
                                 21. NO. OF PAGES
                                    698
                                                            20. SECURITY CLASS (TMipage)
                                                               UNCLASSIFIED
                                                            22. PRICE
  If A form 2220-1 (9-73)

-------