-------
ABSTRACT (continued)
The results of the present study using 22 healthy men did not show a
statistically significant effect of CO exposure to 100 ppm for four hours
on either tracking or monitoring. The criterion for expen'mentwise
significance was a = 0.05 which was divided equally between the two kinds
of performance so that for each overall significance test, a = 0.0?5. The
The planned analysis was based on (a) reanalysis and power analysis of
Putz's original data and (b) pilot data from our laboratory. For tracking,
the test of CO effects of interest (the CO x hour interaction) had p =
0.035. For monitoring the appropriate test of CO x hour effect yielded p
> 0.39.
In the present study (a) observed trends were in the same direction
as those of Putz et al. (b) results approached statistical significance
criterion and (c) several inadvertent methodological changes from Putz
et al. apparently occurred. Due to these considerations and the findings
of Putz et al. (1976) and Putz (1979), it may be tentatively concluded
that (a) tracking may be sensitive to impairment by CO exposure (b)
monitoring does not appear to be affected by CO exposure and (c) important
variables in research on the effects of CO exposure on tracking appear to
be the level of subject training and the task difficulty. It is important
to note that this publication does not claim that the results of Putz et
al. have been replicated. Much, however, was learned from this study in
terms of the stability of tracking behavior for further quantitative CO
research planning.
-------
AD -
EFFECTS OF LOW LEVEL CARBON MONOXIDE (CO) ON TRACKING AND MONITORING.
AN ATTEMPT TO REPLICATE THE FINDINGS OF PUTZ ET AL. (1976)
PROTOCOL 2, FINAL REPORT
Vernon A. Benignus*, Keith E. Muller ^, Curtis N. Barton^ and James D.
JANUARY 1985
Supported by
U.S. ARMY MEDICAL RESEARCH AND DEVELOPMENT COMMAND
Fort Detrick, Frederick, MD 21701-5012
Project Order 1812
lU.S. Environmental Protection Agency and
Department of Psychology
University of North Carolina, Chapel Hill
Chapel Hill, NC 27514
^Department of Biostati sties
University of North Carolina, Chapel Hill
Chapel Hill, NC 27514
•^Department Psychology
University of North Carolina, Chapel Hill
Chapel Hill, NC 27514
Project Officer: Major John Kelly
Health Effects Research Division
U.S. ARMY MEDICAL BIOENGINEERING RESEARCH AND DEVELOPMENT LABORATORY
Fort Detrick, Frederick, MD 21701-5010
Approved for public release;
distribution unlimited
The findings in this report are not to be construed as
an official Department of the Army position unless so
designated by other authorized documents.
-------
TABLE OF CONTENTS
Page
ACKNOWLEDGEMENTS 1
EXECUTIVE SUMMARY 2
LIST OF FIGURES 4
LIST OF TABLES 5
TABLE OF ABBREVIATIONS 6
GLOSSARY 7
INTRODUCTION R
METHOD 10
Experimental Design 10
Subjects 10
Tasks 11
Tracking Task 11
Monitoring Task 12
Auditory Task 13
Hood Li ghti ng 13
Reversal Test 13
Procedure..., .„ 14
Differences From Putz et al. (1976) 15
Counterbalanced orders IB
Double Blind Design 18
Stati sti cs 18
COHb 19
RESULTS 20
COHb 20
Tracking (confirmatory) 20
Monitoring (confirmatory) 20
Reversal (Expl oratory) 21
DISCUSSION 22
Tracking (confirmatory) 22
Moni tori ng (confi rmatory) 23
Reversal (exploratory) 23
General Concl usi ons 24
Tracking 24
Moni tori ng 25
Reversal 25
Conclusions 26
REFERENCES 27
-------
ACKNOWLEDGEMENTS
The authors would like to express their gratitude and appreciation
to the following:
1. Lynne Clapp and Martin Case for the task of testing, preparing and
debriefing subjects,
2. EPA Clinical Research Branch for their aid and support of this project,
3. Josephine Nichols and Linda Mack for their conscientious and careful prepa-
ration of this manuscript.
-------
EXECUTIVE SUMMARY
Carbon Monoxide (CO) is the byproduct of incomplete combustion. As
such it occurs in many environments ranging from outdoor air to industrial
and military. The neural and behavioral effects of CO exposure are
poorly understood but have been reported to affect the variables measured
in this study: hand-eye coordination (in the form of compensatory tracking)
and visual monitoring of peripheral events.
This is the final report of a protocol designed to replicate the
finding of Putz et al. (1976) who have published much of the original
work in this area. The objective was to help determine if tracking and
perhaps monitoring are reliably affected by CO exposures in other, un-
related experiments and laboratories than those of Putz et al. If so,
then those behavioral dependent variables could be used in further
dose response studies of CO effects.
Twenty-two healthy young men were exposed to either 100 ppm CO or
to ambient air for four hours while they were tested on a tracking and
monitoring task. Tracking consisted of attempting to keep a spot on a
cathode ray tube (CRT) centered by manipulating a pressure sensitive
joystick. Monitoring was tested by having subjects detect the occurrence
of an unusually bright flash of light in a string of regularly flashing
red lights. Tracking and monitoring were performed simultaneously.
The results of the present study did not show a statistically signifi-
cant effect of CO exposure to 100 ppm for four hours on either tracking
or monitoring. The criterion for expenmentwise significance was a =
0.05, divided equally between the two kinds of performance so that for
-------
each overall significance test, a = 0.025. The planned analysis was
based on (a) reanalysis and power analysis of Putz's original data and
(b) pilot data from our own laboratory. For tracking, the test of CO
effects (the CO x hour interaction) had p = 0.035. For monitoring the
test of CO x hour effect yielded p > 0.39.
In the present study (a) observed trends were in the same direction
as those of Putz et a!. (b) the results approached statistical significance
criterion and (c) several inadvertent methodological changes from Putz et
al. apparently occurred. Due to these considerations and the findings of
Putz et al. (1976) and Putz (1979), it may be tentatively concluded that
(a) tracking may be sensitive to impairment by CO exposure (b) monitoring
does not appear to be affected by CO exposure and (c) important variables
in research on the effects of CO exposure on tracking appear to be the
level of subject training and the task difficulty. It is important to
note that this publication does not claim that the results of Putz et al.
have been replicated. Much, however, was learned from this study in
terms of the stability of tracking behavior for further quantitative CO
research planning.
Conclusions. From the above discussion and in consideration of findings
by Putz et al. (1976) and Putz 1979, the following conclusions seem appro-
priate.
(a) tracking may be sensitive to impairment by CO exposure.
(b) monitoring may not be affected by CO exposure.
(c) important variables in research on the effects of CO exposure
on tracking appear to be the level of subject training and the
task difficulty.
-------
LIST OF FIGURES
Figure No. Figure Heading Page
1 v Sketch of tracking/monitoring
apparatus. 34
2 Sketch of CRT screen and monitoring
lights as seen by the subject through
the viewing port. 35
3 Sketch of the inside of the viewing
hood. 36
4 Temporal order of events during each
hour of the experiment and during
training. 37
5 Temporal order of events during
tracking/monitoring task. 3R
6 Performance of tracking task, fast
FF, for data from present study and
from Putz et al. (1976). 39
7 Performance of tracking task, slow FF,
for data from present study and from
Putz et al. (1976). 40
8 Performance of monitoring task for
data from present study and from Putz
et al. (1976). 41
9 Performance of the tracking task
during reversal. 42
10 Performance on the monitoring task
during reversal. 43
-------
LIST OF TABLES
Table No. Title Page
1 Daily Schedule of Events 28
2 Differences Between Present 29
Study and That of Putz et al
(1976)
3 Mean, SO and Range of COHb, in 30
Percent, for Exposed and Control
Groups Before and After Exposure
Period
4 Significance Tests (a = 0.025) for 31
Tracking Behavior Using Gieser-Green-
house Adjusted Degrees of Freedom
5 Significance Tests (a = 0.025) for 32
Monitoring Task Using Gieser-Greenhouse
Corrected Degrees of Freedom
6 Means of Tracking Error in cm and of Re- 33
action Time in MS for CO Exposed and
Control Groups During Reversal Learning
(Exploratory Analysis) +_ 1SD.
-------
TABLE OF ABBREVIATIONS
CNS central nervous system
COHb carboxyhemoglobin, a measure of the level of carbon monoxide
in the blood
CRT cathode ray tube, a video screen
dB decibel, a measure of sound intensity
FF forcing function. A signal or voltage which drives the CRT
spot away from the target
Hz cycles per second, a measure of frequency
MANOVA multivariate analysis of variance, a statistical analysis
technique
n number of subjects
p probability
ppm parts per million, used to express the level of CO in the air
-------
GLOSSARY
a priori
A-weight
Bonferroni
Correction
double blind
evoked potential
forcing function (FF)
joystick
post hoc
power
pseudorandom
SAS
single blind
threshold
type I error
An experimental hypothesis made before conducting
the experiment
The weighting of a sound by a frequency weighting net-
work which approximates human loudness of sounds of
different frequency compositions
A statistical technique employed to adjust the
calculated p values for the number of statistical
tests conducted to arrive at a more accurate estimate.
This reduces the probability of committing a type
I error for the experiment as a whole (Kirk, 1968)
An experimental strategy in which neither the subject
nor the experimenter in contact with the subject is
aware of the exposure condition. This strategy is used
of minimize experimenter and subject bias
Electrical activity in the central nervous system
occasioned by a sensory stimulus
A time varying voltage used to drive a spot on a CRT
A lever similar to that used on video games or in
remote control system to manipulate some event
Here refers to analyses which do not satisfy criteria
of a-priori hypotheses
Chances of (correctly) detecting a true effect
Random except for certain constraints, e.g. no strings
of repeated numbers longer than some fixed number
A statistical analysis package
An experimental paradigm in which the experimenter but
not the subject is aware of the experimental condition
In psychophysics the arbitrarily defined point on the
physical continuum above which perception of a stimulus
is regarded as reliably occurring, frequently defined
as > 50% detection.
In hypothesis testing, falsely rejecting the null
hypothesis and concluding that a significant dif-
ference was due to the experimental manipulation
-------
INTRODUCTION
Experiments on the effects of carbon monoxide (CO) on neural and
behavioral variables in man have frequently not been replicable (Benignus
et al., 1983; Laties and Merigan, 1979). Such nonreplicability could be
due to (a) low level exposures which are near threshold for effects (b)
unreliable performance of human subjects on the tasks chosen for study
(c) changes in procedure in replication studies with respect to the
original studies (d) lack of double blind design or (e) type 1 errors
which are reported, in contrast to nonsignificant findings which are
typically not reported (Rosenthal, 1979).
Putz et al. (1976, 1979), a possible exception to the above, per-
formed studies of the effects of CO upon (a) tracking (b) monitoring and
(c) auditory evoked potentials. The tracking task involved keeping a moving
spot centered on a CRT by manipulating a joystick. The monitoring task
was performed simultaneously with tracking and consisted of detecting
bright flashes of light interspersed among a series of dimmer flashes.
The subject's task was to detect the bright flashes. The auditory evoked
potential task consisted of listening to a series of tones presented
serially and pressing a button when a target tone was 1 KHz. Nontarget
tones were either higher or lower in frequency.
Subjects were exposed to either 70 ppm CO, 35 ppm CO or room air for
4 hours. In one study an independent group of subjects served at each ex-
posure level (Putz et al., 1976). In the other study each subject served
in all exposure levels (Putz et al., 1979). The 70 ppm group reached
about 5 percent COHb at the end of the 4 hour exposure.
-------
In both studies the 70 ppm group showed statistically significant
decrements in tracking accuracy. The amount of decrement increased as a
function of time of exposure. Latency of detection on the light flash
task also was significantly increased by 70 ppm CO exposure in a time-
on-task related fashion. No effect of 35 ppm CO exposure was observed.
The evoked potential was not affected by any CO level, at any time.
The studies discussed above have a high credibility. They constitute
a study which was independently replicated, although by the same investiga-
tor. This pair of studies appears to be the only neurobehavioral work
in the extant literature which can boast this quality. The behaviors
which were affected have high face validity as measures of important
nonlaboratory tasks in which humans engage, such as driving and machine
operation.
A replication of the above studies is urgently needed because if an in-
dependent laboratory can show similar results then the results will have
more credibility in the scientific community. These results are among
the lowest CO exposure levels reported to have produced neurobehavioral
decrements in healthy young subjects. For these reasons a replication
should be undertaken.
The present study was undertaken to attempt to replicate the findings
of Putz et al. (1976), since this report gives more methodological detail
and since raw data are published there. Much effort was expended in
exactly replicating procedure and equipment.
-------
METHOD
EXPERIMENTAL DESIGN
A complete list of raw data was published by Putz et al. (1976).
These data were reanalyzed by the present investigators and the results
verified those of Putz et al. Analyses were also performed to determine
the power (Muller and Peterson, 19R4) of the significance tests used by
Putz et al. The p-values were smaller and the estimated powers were
greater when only 0 and 70 ppm groups were used. For the same number of
subjects (per exposure level), a = 0.05, assuming the same size of
effect and the same variance as Putz et al., the CO x TIME effect would be
tested with power approaching 1.0, if only room air and 70 ppm groups
were used. The same test for the light monitoring behaviors would have
a power of 0.72. Thus for a total of 20 subjects, using room air and 70
ppm CO exposure, the probability of nonreplication was considered to be
very small provided that the findings are reliable.
It was planned, therefore, to test two groups of subjects, control and
CO exposed, using approximately 10 subjects per group according to the
Putz et al. paradigm and experimental design. It was decided to collect
data from only the tracking and monitoring tasks since the auditory
evoked potential results were nonsignificant. Subjects were, however,
tested in the tone judgement (evoked potential) task since the performance
of the task may have influenced performance on other tasks.
SUBJECTS
Subjects were 24 healthy males. One subject was eliminated from the
exposed and one from the control group because they failed to perform the
10
-------
yrs (SD = +_ 2.83, range = 19-31 yrs). Subjects were paid $5.00 per hour to
participate. All subjects had been given examinations by a physician and
had made normal scores +_l S.D. on the Minneapolis Multiphasic Personality
Inventory.
TASKS
Tracking Task. The tracking task consisted of keeping a moving spot
centered on a cathode ray tube (CRT) by manipulation of a pressure sensi-
tive, nonmoving joystick. The CRT was a 20.3 x 25.4 cm (8" x 10") green
phosphor oscilloscope with the 25.4 cm dimension on the vertical axis.
The oscilloscope was equipped with a viewing hood and joystick mount as
shown in Figure 1. During task performance the subject was seated on an
adjustable chair in front of the viewing hood. The screen was viewed by
the subject by positioning his head in the rubber viewing port, thus
fixing the position and distance from eye to screen.
The view of the screen as seen by the subject is shown schematically
in Figure 2. In the center of the screen was a circular fixed target con-
sisting of a white ring, glued to the CRT screen. The inside (clear) circle
of the target ring had a diameter of 4 mm and the outer target diameter
had a diameter of 12 mm. A spot on the CRT moved up and down (vertically
only) as driven by a forcing function (FF) generated by a microcomputer.
The subject's task was to keep the spot centered in the stationary target.
The subject exerted control over spot position by pushing forward or back-
ward on the joystick.
The FF in this task consisted of half sinusoids of one of four amplitudes
and either positive (up) or negative polarity. Zero FF left the spot centered
in the target ring. Polarity and amplitude of the FF were independently and
11
-------
pseudorandomly selected for each half cycle. The amplitudes which were used
produced +_ 12.7, +_ 6.35, +_ 4.763 and +_ 3.175 cm of deflection. There
were two speeds of the FF, 4 sec/half cycle and 8 sec/half cycle (the
"fast" and "slow" speeds). The fast and slow FFs were alternately pre-
sented in blocks of trials during the testing as will he discussed later.
The joystick was a Measurement Systems PN 436 pressure sensitive
control. The joystick did not move in response to pressure. The length of
the joystick handle was about 12 cm overall with a wooden ball at the end
of 4.5 cm diameter. The handle of the joystick was mounted on a wooden
platform below the CRT. It was calibrated such that full scale consisted
of a push or pull of 2.4 Kg (5.3 Ibs) for +_ 12.7 cm deflection. Error
scores for tracking were expressed in mean absolute deviation in cm of
the spot from the target (averaged over time).
Monitoring Task. The monitoring task consisted of monitoring the
brightness of two red lights on either side of the CRT as shown in Figure
2. One or the other of these lights was on for 1.5 sec and both were
then off for .74 sec. The on/off periods were repeated continuously for
the duration of the tracking task. Left and right lights were selected
pseudorandomly.
Within any block of trials, the light level to which the red lights
were illuminated was either "dim" or "bright". The subject was required
to press a hand-held switch when a bright light level occurred on either
left or right light. On alternate blocks the overall level of illumina-
tion was switched between two levels. During the overall bright condition
-------
the nonsignal intensity was .38 lux and the signal intensity was 1.5 lux.
During the overall dim condition the nonsignal intensity was .04 lux and
the signal intensity was .17 lux. Intensity was measured by placing a
photometer sensor 6.5 cm in front of the lights. Signals and non signals
were pseudorandomly interspersed. Among 25 light stimuli 9 were signals.
Auditory Task. A third task was performed by subjects in an alternating
fashion with the tracking/monitoring task. This (auditory) task was the same
as the one used by Putz et al. to elicit evoked potentials. In the present
study the task was performed by the subjects and electrodes were attached
to vertex and linked mastoids but no data were actually collected. The
only rationale for including this task was that it might affect performance
of the other tasks.
The auditory task consisted of judging tones sounded over a loudspeaker.
Each tone had a duration of 1.5 sec and a loudness of approximately 80 dbA
(SPL). The time between tones was .75 sec. The three tone frequencies were
400 Hz, 1000 Hz and 2000 Hz. The subject's task was to press a handswitch
only when the 1 KHz tone signal was sounded.
Hood lighting. Figure 3 shows a schematic drawing of the inside of
the viewing hood. A diffused green light was used to illuminate the screen
so as to reduce contrast effects. The intensity of the light was adjusted
to .22 lux when the photometer sensor was located 6.5 CM from the CRT. In
order to avoid glare from reflected light, the light source was vertically
polarized while a horizontal polarizing filter was placed over the viewing
port.
Reversal Test. After all of the replication data was collected and
just after the last tracking/monitoring task, another tracking/monitoring
task was substituted. This last task was included for exploratory work
13
-------
only (Muller et al. 1984). The results were not to be included in the
first study's a prion hypotheses. The reversal test might become a part
of a formal experiment later, however, should the exploratory results
prove useful.
This last exploratory task was the same as the other tracking/
monitoring tasks except that the responses which were required were re-
versed. Pushing on the joystick moved the spot downward rather than
upward. The monitoring target signals were made dim rather than bright.
The subject was instructed as to what was to be expected via intercom,
just before the reversal task was to begin. This corresponds to the
reversal learning paradigm, performance of which is frequently impaired
by injury or drugs.
PROCEDURE
The schedule of events for a typical day in this experiment is shown
in Table 1. This was the same schedule and same times of day as used by
Putz et al. (1976). Once informed consent and training was completed,
the schedule followed an hourly cycle of performance.
Figure 4 shows the schedule of events within each hour of performance
testing. The tracking/monitoring task was performed first for 32 min.
Immediately following, the subject performed the auditory task for 14 min.
This performance was followed by a rest period of 11 min, during which the
subject was permitted to read, after which the subject briefly exited from
the chamber to have an end-tidal alveolar air sample collected. The
cycle then repeated itself during the subsequent hour. Again, this was
exactly the procedure used by Putz et al. (1976).
The temporal order of events within tracking/monitoring task was
quite complex as shown in Figure 5. Performance was divided into 56 sec
14
-------
trials. Two trials form a block and there were 8 blocks in the 32 min
performance period. Blocks were separated by 118 sec. pauses. Trials
were separated from each other (within blocks) by a 10 sec pause.
Monitoring behavior alternated in a blockwise fashion between the
overall bright and dim conditions. Tracking FF speed alternated between
fast and slow on a two-block cycle. Thus, each block had a bright or dim
(monitoring) condition and a fast or slow (tracking) condition. Within a
32 min performance period, each combination of conditions occurred twice.
Differences from Putz et al. Despite major efforts to make this re-
plication exact, several differences from Putz et a!. were found to be
necessary or extremely convenient. The differences are summarized in
Table 2.
The two forcing functions used by Putz, et al. had speeds of 7 cycles/
min and 4 cycles/min for fast and slow conditions. Presumably this means
that the fast function had a speed of 4.28 sec. per half cycle (8.57 sec
per cycle). This would have necessitated stopping the FF during a nonzero
crossing since each trial was only 50 sec long. Putz was unable to shed
any light on this problem (personal communication). The only speed which
was close to the Putz et al. FF and still resulted in an integer number
of cycles was a 4 sec per half cycle FF which went through 14 half cycles
in 56 sec. A slow FF which also came out in integer half cycles was 8
sec per half cycle, which completed 7 half cycles in 56 sec. Thus the
trial was made 56 sec long (rather than 50 sec) and the fast and
slow FFs were 4 and 8 sec/half cycle (instead of 4.28 and 7.5 sec/half
cycle). The interblock pause was adjusted to 10R sec (rather than 120
sec) to keep task length constant. These are relatively minor changes.
15
-------
The CRT size used by Putz et al. was 25.4 cm square (10" x in").
The screen available for the present study was ?4.5 cm high by 20.3 cm
wide (8" x 10"). Since no horizontal spot movement was used, the stimulus
lights were mounted so as to have the same horizontal visual angle from
center as those used by Putz et al. A different hood design (square vs.
long tapered) was also used. Since the visual effect was the same, this
should not matter.
The CRT hood shown in Figure 1 was illuminated to an intensity
of .22 lux, measured at 6.5 cm from the screen. Putz et al. did not specify
the distance of the photometer sensor but in the same paragraph specified the
distance for measurement of the signal light intensity. It was assumed it
was the same distance for the CRT screen. Putz was unable to resolve the
problem (personal communication).
Putz et al. stated that in 22 light stimuli, 8 were signals. They
also stated that lights on/off cycles were 2.25 sec long (1.5 sec on/.75
sec off). Thus 22 stimuli would take 48.5 sec. In order to provide
stimuli for the 56 sec trial length, stimuli in the present study were 1.5
sec long but the off time was .74 sec. In this case, 25 stimuli exactly
occupied the 56 sec. trial. Of the 25 stimuli, 9 were signals. The
proportion of signals was therefore .36 whereas in Putz et al. it was
.364. These are, again, probably trivial differences.
A major difference between the present study and that of Putz et al.
was that they tested two subjects at a time on two tracking/monitoring
devices in the same chamber. In the present study, only one subject was
studied at a time in a visually and acoustically isolated chamber. In
the Putz et al. study, the two subjects were tested alternately, one
during the 1?0 sec rest time of the other. The possible social contact
16
-------
between subjects, even without transmission of information about perfor-
mance, could have affected baseline values. It is not reasonable to
assume that results should be differentially affected for CO and air
exposed subjects, however. The reasons for testing only one subject at
a time are (a) resource limitations in terms of computer speed/time (b)
logistic complications in terms of technician time per experiment, thus
increasing possible confusion related errors and (c) the general principle
that for vigilance-like tasks, isolated subjects show clearer and earlier
decrements. If the baseline performance of the subjects in the present
report should not differ from those of Putz et al., it would be additionally
reasonable to suppose that the social effect was not important for these
tasks.
Random event scheduling was apparently done with random number genera-
tors by Putz et al. in their online control program. Certain highly
desirable distribution characteristics do not usually obtain in such pro-
grams, e.g. exactly equal number of signals per trial, exactly balanced
amplitudes in the FF, absence of runs of specifiable length, local sta-
tionarity, etc. For these reasons, random events were scheduled by ex-
haustively sampling from finite populations of playing cards and then
entering these fixed schedules into the controlling computer program.
All event schedules were checked for distribution similarity, short run
lengths and balanced first order conditional probability distributions.
The schedules for FF amplitudes and polarity and for the monitoring
events repeated after eight two-min blocks of performance. Considering
the number of events, the length of the trials and the pauses between, it
is very doubtful that the subjects could detect a pattern.
17
-------
As discussed in the introduction, Putz et al. exposed their subjects
to 70 ppm CO. Subjects were exposed to 100 ppm CO in the present study.
Counterbalanced Orders. Figure 5 implies that four possible combina-
tions of FF speed and monitoring brightness exist, depending on which
pair the subject is given first. In order to eliminate possible difficulty
order differences due to starting combinations, the starting speeds and
brightness were pseudorandomly selected for each subject. Since complete
counterbalancing would require 12 subjects per group, this strategy was
not possible. The same number of each starting combination were used
in CO exposed as in air exposed groups.
Double Blind Design. This study was entirely double blind. CO
level was determined by a computer stored pseudorandom schedule. No one
knew a priori what that day's subject was to be exposed to unless they
looked at the computer code, which was cumbersome to read. Everyone who
had any contact with the subject was strictly enjoined not to attempt to
break the blind. CO level was monitored by one of the senior investigators
who had no contact with the subject after the study began. Thus that
senior investigator remained blind until after his last subject contact.
Statistics. Two dependent variables were analyzed: (a) mean absolute
deviation for the tracking task and (b) response time for the peripheral
light task. Univariate repeated measures analysis (Kirk, 1968) was
used for both. Each analysis is detailed below.
For the tracking task variable, a CO by Difficulty by Hour factorial
design (2x2x4) was used. CO is a "between" factor and the other two
are "within" factors. Levels were 0 and 100 ppm CO, low and high frequency
for difficulty, and 1st, 2nd, 3rd, 4th hour.
18
-------
For the peripheral light task, CO and Hour (2 x 4) were the factors,
with levels as above. A factorial model was tested, again with CO "between"
and Hour "within".
For both analyses, stepdown tests (Kirk, 1968) were to be done as separate
tests for each hour. If the three-way interaction were involved in tracking,
second level stepdowns were to be CO within difficulty. Three differences
exist with Putz1 analyses: (a) the dropping of the 35 ppm CO level (b) the
use of a Geisser-Greenhouse correction for F tests and (c) a Bonferroni
correction to test each overall test at a =0.025 =0.05/2. Stepdowns,
if needed, were to be tested at o =0.025/k, with k = number of step-
downs within the family. These choices were made after using Putz's data
for exploratory analysis.
COHb. Blood was drawn, as shown in Table 1, before and after exposure.
At each drawing, two three-mi vacutainers were used. COHb values were measured
in triplicate immediately after blood was drawn for each set of samples
by use of an Instrumentation Laboratories model 282 CO oximeter. Triplicate
values were averaged to provide a final pre and post COHb value. Alveolar
air samples were not analyzed because of the need to keep the number of
dependent variables in this study at a minmum. Comparison of blood and
air values would have constituted additional hypothesis tests and therefore
required further division of a. This would have reduced the power of
the hypothesis tests and thus required more subjects in the study.
19
-------
RESULTS
COHb.
Table 3 shows the results of the COHb analyses. These were well within
expected limits.
TRACKING (CONFIRMATORY).
The results of the significance tests for tracking behavior are
shown in Table 4. No CO related test was significant by the Bonferroni
adjusted criterion of a =0.025. The CO x Hour interaction, which was
significant in the study by Putz et al. (1976) had a value of p = 0.035
in this study. Figure 6 shows the mean tracking errors for the fast FF
for both the data reported by Putz et al. and the data collected in the
present study over the four hours of behavior and exposure. Figure 7
shows the corresponding data for the slow FF. In both figures it is seen
that data from the present study tend to follow in the same direction of
those of Putz et al., i.e. the CO exposed group made a higher mean track-
ing error than the control group. This difference was statistically
significant for the data from Putz et al. but not for data from the
present study. The mean squared error (MSE) for the tracking data for
the test of CO x Hr interaction from Putz et al. was 1.5 whereas for the
present study the same MSE was 4.6.
MONITORING (CONFIRMATORY).
Table 5 shows the significance tests for the peripheral light montoring
task. Here, no test was even close to statistically significant. Figure
8 shows the mean reaction times over the four hours of performance and ex-
posure for both the data of Putz et al. and the data from the present study.
For the data from Putz etal., MSE was 775 and for the present study MSE=14?3,
-------
REVERSAL (EXPLORATORY).
Exploratory analyses showed that the reversal behavior was extremely
non-Gaussian in its distribution. For exploratory purposes the mean tracking
scores and mean reaction times in the CO and control groups were tested
using the Wilcoxon Rank-Sum Test. There were no significant differences
between control and CO groups on tracking, Z = -1.02, p > 0.31 or on
reaction time, Z = 0.60, p > 0.55. Table 6 shows the mean scores during
reversal training.
In order to explore the question of performance improvement (learning)
a plot was made of the tracking score over the eight two-min reversal blocks
(Figure 9) and of the reaction times for the same blocks (Figure 10).
21
-------
DISCUSSION
TRACKING (CONFIRMATORY).
Figures 6 and 7 demonstrate that at the start of the experiment, the
subjects from the present study performed with nearly identical level of
error to the subjects of Putz et al. (1Q76). This was true for both fast
and slow FFs. Especially for the fast FF, the CO exposed group from Putz
et a!. increased their mean error of tracking more than for the subjects
in the present study. This results in a greater divergence between groups
for Putz et al. than for the present study. By the end of the four-hour
exposure period the between group difference for the fast FF in the Putz
et al. data was approximately three times as great as for the data from
the present study.
Inspection of the control group values in Figures 6 and 7 reveals that
subjects in the present study improved the mean level of performance over
the four hour period by a greater amount than did the subjects of Putz et al
This fact, together with the higher MSE for present study, suggests that
subjects in the present experiment were not as well trained as those of
Putz et al.
If it is the case that subjects for the present study were not trained
as well as those of Putz et al., then it is also implied that the tracking
task for the present study was in some way easier than that of Putz et al.
If their subjects were better trained but performed on a par with more
poorly trained subjects, then their task must have been harder to perform.
It is not clear from the parameters of the task that any difference
22
-------
should exist, however. The above comments ignore the possibility that
all of the differences in results could be explained by the fact that
a different pool of subjects was used in the two studies.
The possible reasons for the nonsignificant finding in this study
(relative to that of Putz) are multifold. The task for the present study
might have been slightly easier while the subjects might have been more
poorly trained. The MSE for the present study was larger while the CO
exposed and control groups diverged less. A Bonferroni corrected a
criterion was used and the degrees of freedom were Geisser-Greenhouse
corrected in the present study while the Putz et al. study used no such
corrections. The present study would have yielded p = 0.0274 without
correction. The power in the Putz study approached 1.00, however, even
with corrections when the original data was reanalyzed in the planning
stage of the present study. The increased variability and the reduced
effect size jointly reduced the power from that which was expected. It
is presumed that the small effects size reflects subject or task differences
while the large error variance reflects less well trained subjects.
Monitoring (CONFIRMATORY)
Figure 8 reveals that the mean reaction times for the data from the
present studies were substantially lower than for the data from Putz et al.
This implies, again, that the task in the present study was easier for some
unknown reason. Task parameters were carefully measured. Minor differences
do not provide a clue to the reasons for the different level of difficulty.
REVERSAL (EXPLORATORY)
The reversal task was included as a possible measure of the effect of
CO on learning (relearning). It appears that most of the improvement in
23
-------
tracking occurs during the first two-min block of performance after which
little systematic change occurs. Even after eight blocks of reversal
practice the level of performance in neither group was as good as it had
been during the regular performance (compare Figure 9 to Figure 6).
The non Gaussian distribution of the reversal data resulted from what
appeared to be behavioral lapses of individual subjects from the learned
reversal to the previous polarity of responding. This is conjecture
because the time functions of the tracking lever response was not recorded
and so could not be inspected. The hypothesis of lapses to the previous
polarity of responding seems justified by the occasional but nonsystematic
occurrence of large scores.
It would seem from inspection of Figure 9 that the CO group performed
consistently worse than the control. High variability and small differences,
however, prevented the difference from becoming significant. While the CO
group consistently performed more poorly during reversal (as it also did during
previous regular performance) there was a much less clear difference in the
rate at which performance improved. This might have been due to the brief
improvement during block 1 and the general lack of further learning.
The reaction times during reversal were unsystematic (see Figure 10) and
continued to be much higher than in the previous performance (compare Figures
10 and 8). The only conclusion one can draw from this observation is that the
monitoring task was poorly relearned.
GENERAL CONCLUSIONS.
Tracking. One of the major objectives of this work was to determine
whether compensatory tracking was sensitive to CO effects. It appears,
despite the nonsignificant effects observed, that the confidence in this
24
-------
measure has increased, if the task can be properly designed. Due to the
higher variance, smaller mean differences, and methodological differences
from the research and data published by Putz et al. (1976) the a priori
hypothesis tested in the present study was not significant. The mean
trends for tracking scores were apparently in the expected direction.
While none of the above considerations, or all of them collectively, can
be used to claim that the results of Putz et al. have been replicated,
the findings of Putz et al. (1976) and Putz (1979) combined with these
results to suggest that tracking is sensitive to degradation by CO exposure.
Monitoring. Much of what has been said about tracking can also be
said about monitoring. With respect to Putz et al. the trends in this
study were in the same direction but not significant. There were more
extreme deviations from the data of Putz et al. in the present study,
compared to tracking e.g. the task was much easier in the present study.
It was also true that in the data reported by Putz et al. the CO effects
upon reaction time were weaker than the CO effects upon tracking.
Reversal. The reversal was apparently so disruptive as to have pre-
vented complete relearning in the time allowed. The evidence for this state-
ment is that (a) not much relearning was demonstrated in either task except
on the first block of trials (b) the variance was high and the distribu-
tion skewed by apparent frequent behavioral lapses and (c) the performance
never recovered to prereversal levels. If reversal learning is to become
25
-------
a useful measure of CO effect in this context, either a longer relearning
period will have to be used or some way of reducing the difficulty will
have to be devised.
Conclusions. From the above discussion and in consideration of find-
ings by Putz et al. (1976) and Putz (1979), the following conclusions
seem appropriate.
(a) tracking may be sensitive to impairment by CO exposure
(b) monitoring may not be affected by CO exposure
(c) important variables in research on the effects of CO exposure on
tracking appear to be the level of subject training and the task
difficulty.
26
-------
REFERENCES
Benignus, V., K. Muller, C. Barton, J. Prah and L. Benignus. 1983. Neuro-
behavioral Effects of carbon monoxide (CO) exposure in humans, Protocol
1, Final Report, U.S. Army Medical Research and Development Command,
Report No. Ad-A133891, Fort Detrick, MD.
Kirk, R. 1968. Experimental Design: Procedures for the Behavioral Sciences.
Belmont Calif. Brooks/Cole.
Laties, V. and W. Merigan 1979. Behavioral effects of carbon monoxide on
animals and man. Ann. Rev. Pharmacol. Toxicol. 19: 357-392.
Muller, K., C. Barton and V. Benignus 1984. Recommendations for appro-
priate statistical practice in toxicologic experiments. Neurotoxicol.
5: 113-126.
Muller, K. and B. Peterson 1984. Practical methods for computing power in
testing the multivariate general linear hypothesis. Comp. Stat. Data
Anal. 2: 143-158.
Putz, V.R., Johnson, B.L. and Setzer, J.V. 1976. Effects of CO on vigilance
performance. U.S. Dept. of Health Education and Welfare, National
Institute for Occupational Safety and Health, Cincinnati, Ohio.
Putz, V.R., Johnson, B.L. and Setzer, J.V. 1979. A comparative study of the
effects of carbon monoxide and methylene chloride on human performance.
J. of Environ. Pathol. and Toxicol. 2: 97-112.
Rosenthal, R. 1979. The "file drawer problem" and tolerance for null results,
Psycho! . Bull. 86: 638-641.
27
-------
TABLE 1. Daily Schedule of Events
Time
Event
Cum. Time
10:00 AM
10:15
10:30
10:35
11:35
11:45
12:17
12:31
12:45
1:45
2:45
3:45
4:00
Prescreened subject arrives.
Sign consent forms
Blood and breath samples collected
Enter chamber for training*
Training completed - apply electrodes
Begin tracking/vigilance task
0
15 min.
30 min.
35 min.
1:35 min.
1:45
1
Begin evoked potential (tone) task
Begin rest - lunch - breath sample
Begin 2nd hour
Begin 3rd hour
Begin 4th hour
Exit chamber - blood sample - debrief
Subject leaves
2:17
2:31
2:45
3:45
4:45
5:45
6:00
*Training hour is typical of regular test hour except that verbal feedback is
used.
-------
TABLE 2. DIFFERENCES* BETWEEN PRESENT STUDY AND THAT OF PUTZ ET AL. (1976)
Value in Value in
Parameter Present Study Putz et al .
Fast FF speed
Slow FF
Trial length
Interblock
CRT screen size
Hood illumination
Monitoring stimulus ISI
No. of monitoring stimuli per trial
Monitoring signal/nonsignal ratio
Number of subjects per chamber
Window in chamber
Randomization of events
CO level **
4 sec/half cycle
8 "
56 sec.
108 sec.
25.4 cm. vert., 20.3 horiz.
.22 lux at 6.5 cm
.74 sec.
25
.36
1
No
fixed schedules
100 ppm
4.28 sec/half cycle
7.5
50 sec.
120 sec.
25.4 cm square
unclear
.75 sec.
22
.364
2
Yes
computer controlled
70 ppm
*Rationale for these differences is discussed in text.
**100 ppm exposure was inadvertantly selected in the present study. A 30 ppm
difference is unlikely to affect results and should if anything increase effects
of CO.
-------
TABLE 3. MEAN, SD AND RANGE OF COHb, IN PERCENT, FOR EXPOSED AND CONTROL
GROUPS, BEFORE AND AFTER EXPOSURE TO 100 PPM CO FOR FOUR HOURS
Before Exposure
Period
After Exposure
Period
Control Group
Exposed Group
SD =+ 0.39
M = 1.42
range = 0.9 - 2.32
SD =+0.18
M = 1.36
range = 1.07 - 1.57
SD=+ 0.24
M = 1.22
range = 0.87 - 1.55
SD=+ 0.49
M = 8.24
range = 7.57 - 9.03
sn
-------
TABLE 4. SIGNIFICANCE TESTS (a = 0.025) FOR TRACKING BEHAVIOR USING
GIESER-GREENHOUSE ADJUSTED DEGREES OF FREEDOM
Effect
CO
Speed (of tracking
stimulus)
Hour
CO x Speed
CO x Hour
Speed x Hour
CO x Speed x Hour
F*
0.66
346.56
0.79
0.96
3.27
1.75
0.39
p**
0.43
0.0
0.49
0.34
0.035
0.18
0.72
dfi.dfp***
1,20
1,20
2.6, 51.3
1,20
2.6, 51.3
2.4, 47.8
2.4, 47.8
*F = Computed value of significance test (Fisher's F)
**
p = Probability value represented by the particular value of F
***df] and dfp are the degrees of freedom for the hypothesis and for the
for error term, respectively, in the F test tables.
31
-------
TABLE 5. SIGNIFICANCE TESTS (a = 0.025) FOR MONITORING TASK USING
GIESSER-GREENHOUSE CORRECTED DEGREES OF FREEDOM.
Effect
CO
Hour
CO x Hour
F*
0.29
1.11
0.96
P*
0.60
0.34
0.39
dfN, dfD*
1,20
1.9, 38.5
1.9, 38.5
* See TABLE 4 for explanations
-------
TABLE 6. MEANS OF TRACKING ERROR IN CM AND OF REACTION TIME IN MS FOR
CO EXPOSED (TOO PPM FOR FOUR HOURS) AND CONTROL GROUPS DURING
REVERSAL LEARNING (EXPLORATORY ANALYSIS), + 1 SD.
CONTROL
CO EXPOSED
TRACKING
REACTION TIME
1.10
865.4
+_ 0.28
+_ 169.1
1.45
849.2
+_ 0.73
+ 107.2
33
-------
CHAMBER WALLS
VIEWING
PORT
OSCILLOSCOPE
Figure 1. Sketch of tracking / monitoring apparatus,
-------
20 3cm
UJ
_l
0
2
CO
o
00
E
o
If)
(M
O
O
A2
A! = LEFT VIGILANCE LIGHT
A2 = RIGHT VIGILANCE LIGHT
B = TRACKING TARGET (STATIONARY)
C = CRT MOVING SPOT
Figure 2. Sketch of CRT screen and monitoring lights as seen by the
subject through the viewing port.
-------
UH
B = VERTICALLY POLARIZED
FILTER
C = HORIZONTALLY POLARIZED
FILTER
D = VIEWING PORT
E = VIGILANCE LIGHT
F = CRT
G = SUBJECT'S PUPIL POSITION
32cm-
Figure 3. Sketch of the inside of the viewing hood.
-------
OJ
-J
TRACKING/MONITORING TASK
AUDITORY TASK
REST
ALVEOLAR AIR SAMPLE
I
0
10
20
30
TIME (min)
40
50
60
Figure 4. Temporal order of events during each hour of the experiment
and during training.
-------
U)
00
TRACK (FAST)
TRACK (SLOW)
MON (BRIGHT)
MON (DIM)
PAUSE
PRINT DATA
TIME (mm)
1 ••
1
1
1 1
1
1 1
•H
i
1
1
1
I
1 I
1
1
••
1
D 2 4 6 8 10 12 14 1
*- | x « i x
ili Q
0 ^
si
m _i
CO
rJ
BLOCK
CD
.OW/BRI
CO
^
O
O
CO
Q
FAST/
<5j-
BLOCK
O
AST/BRI
u.
1
6 1
in
^
U
O
m
I
1
1
1
BH
1
1
1
1
1 1
1
1
-
8 20 22 24 26 28 30 3
I I Is- 1 i
Q
SLOW/
(O
BLOCK
O
.OW/BRI
CO
^
O
g
GO
Q
FAST/
00
BLOCK
O
AST/BRI
LL.
Figure 5. Temporal order of events during tracking/monitoring task.
Four schedules exist depending upon the particular combinations
of tracking speed and vigilance brightness which are selected
for the first trial. Pause durations: Short = 10 sec.,
Long = 118 sec. Trial length = 56 sec. Printout occurs after
every two trials. Fast forcing function = 4 sec./half wave,
slow forcing function = 8 sec./half wave.
-------
O
z
*
U
1.3 r-
1.2 -
£ n'1
U
DC
O
DC
DC
LU
1.0
0.9
0.8
PUTZ, ET AL.
• • CONTROL
O O CO (70nnm)
PRESENT STUDY
A -± CONTROL
A -A CO (lOOppm)
0.0
2 3
HOURS OF EXPOSURE
Figure 6. Performance of tracking task, fast FF, for data from prestent study
and from Putz et al. (1976). Tracking error is mean absolute
deviation of spot from center screen in cm.
39
-------
PUTZ, ET AL.
0.75 -
cc
O
EC
EC
Ul
O
2
*
O
<
CC
h-
0.70
0.65
0.60
0.55
O
•• CONTROL
•O CO (70ppm)
PRESENT STUDY
A A CONTROL
A A CO (lOOppm)
0.0
HOURS OF EXPOSURE
Figure 7. Performance of tracking task, slow FF, for data from present
study and from Putz et al. (1976). Tracking error is mean
absolute deviation of spot from center screen in cm.
-------
850
825
800
_ 775
V)
LLJ
2 750
O
(-
O
<
HI
c 725
700
675
PUTZ, ET AL.
• • CONTROL
-O CO (70ppm)
PRESENT STUDY
A A CONTROL
A A CO (lOOnnm)
-A—-
"-A
A
o
\
2 3
HOURS OF EXPOSURE
Figure 8. Performance of monitoring task for data from present study and
from Putz et al. (1976).
-------
DC
O
CC
DC
UJ
(D
V.X
u
DC
f-
2.5
2.0
1.5
1.0
0.5
- A
\
\
\
\
\
CO EXPOSED
I
I
I
I
3456
BLOCKS OF TWO MIN PERFORMANCE
8
Figure 9. Performance on the tracking task during reversal. Tracking
error is mean absolute deviation of spot from center screen
in cm.
-------
925
900
"w"
£
UJ
875
U
2 850
cc
825
I
3456
BLOCKS OF TWO MIN PERFORMANCE
8
Figure 10. Performance on the monitoring task during reversal.
-------
DOCUMENT DISTRIBUTION LIST
No. of Copies
5 US Army Medical Res & Dev. Command"
Attn: S6RD-RMS
Fort Detrick
Frederick, MD 21701
12 Defense Technical Information Center (DTIC)
Attn: DTIC-DDA
Cameron Station
Alexandria, VA 22314
1 Dean
School of Medicine
Uniformed Services University of the
Health Sciences
4301 Jones Bridge Road
Bethesda, MD 20014
1 Commandant
Academy of Health Sciences, US Army
Attn: AHS-CDM
Fort Sam Houston, TX 78234
2 Librarian
US Army Medical Bioengineering
Research and Development Laboratory
Attn: SGRD-UBZ-I
Fort Detrick
Frederick, MD 21701
1 Chief
USAEHA Regional Division, West
Fitzsimmons AMC
Aurora, CO 80045
1 Chief
USAEHA Regional Division, North
Ft. George, G. Meade, MD 20755
1 Chief,
USAEHA Regional Division, South
Bldg. 180
Ft McPherson, GA 30330
44
-------
1 Commander
US Army Health Services Command
Attn: HSPA-P
Fr. Sam Houston, TX 78234
1 Commandant
Academy of Health Sciences
United States Army
Attn: Chief, Environmental Quality Branch
Preventive Medicine Division (HSHA-IPM)
Fr. Sam Houston, TX 78234
1 Commander
US Army Material Development
and Readiness Command
Attn: DRCSG
5001 Eisenhower Avenue
Alexandria, VA 22333
1 Commander
US Army Environmental Hygiene Agency
Attn: HSHV-0
Aberdeen Proving Grounds, MD 21010
1 Commander
US Army Environmental Hygiene Agency
Attn: Librarian, HSHD-AD-L
Aberdeen Proving Ground, T1D 21010
1 Commander
10th Medical Laboratory
Attn: AEMML-PM
APO, New York 09180
1 Commander
US Army Medical Command-Korea
Attn: EAMC-H-L
APO San Francisco &(.?[:!
1 Commander
7th Medical Command
Attn: AEMPS-PM
APO New York 09102
17 Commander
US Army Medical Bioengineering
Research and Development Laboratory
Ft. Detrick, Fredrick, MD 21701
45
-------
10 US Environmental Protection Agency
Neurotoxicology Division
Research triangle Park, NC 27711
1 US Environmental Protection Agency
Library
Reasearch Triangle Park, NC 27711
46
-------