EPA 550/9-78-100
AMRL-TR-78-4
THE ABILITY OF
MILDLY HEARING-IMPAIRED INDIVIDUALS
TO DISCRIMINATE SPEECH IN NOISE
JANUARY 1978
JOINT EPA/USAF STUDY
I
55
O
\
S3ZZ
\
UJ
(3
U.S. Environmental Protection Agency
Washington, D.C. 20460
Aerospace Medical Research Laboratory
Wright-Patterson Air Force Base, Ohio
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
SUBJECT: Report on "The Ability of Mildly Hearing-Impaired
Individuals to Discriminate Speech in Noise"
FROM:
TO:
DATE: April 7, 1978
Rudolph M. Marrazzo
Scientific Assistant, ONAC
Regional Noise Enforcement Officials
The attached report represents a major effort by Alice Suter while she
was here at ONAC. Her research, conducted jointly with the Biodynamics
and Bioengineering Division of the Aerospace Medical Research Laboratory
at Wright-Patterson Air Force Base, was aimed at assessing the current
standard for determination of hearing handicap. The study demonstrates
that most present hearing protection standards are based on a definition
of hearing "handicap", which classifies as "normal", persons who are in
fact functionally handicapped. Specifically, it evaluated the American
Academy of Ophthalmology and Otolaryngology (AAOO) standard of 26 dB,
averaged over the audiometric frequencies of 500, 1000, and 2000 Hz, as
the point above which hearing handicap occurs. The adequacy of this
standard has been questioned recently because this method of computing
hearing handicap is widely used in many states for compensation statutes,
and in establishing damage-risk criteria as well as occupational noise
standards.
In assessing the ability of mildly hearing-impaired individuals to under-
stand speech in everyday listening conditions (i.e., with a background
of moderate-level noise), it was found that frequencies above 2000 Hz
should be included in the techniques for measurement of hearing impair-
ment, although at present they are not. It was recommended that the use
of the term "speech frequencies" be defined to include all of the audible
frequencies of the speech spectrum, not just 500, 1000 and 2000 Hz.
In relation to this, the "26 dB fence", or the point of beginning hearing
handicap as established by the AAOO, averaged only over the frequencies
500, 1000 and 2000 Hz, was shown to be far too high.
It was recommended that this fence be lowered, based on the higher fre-
quencies, and to include more people with a real handicap who are not
now recognized as impaired at all.
It is hoped that this research will prompt not only the AAOO but responsi-
ble legislators and standards setters to define more accurately the point
of beginning hearing handicap. Such a recommendation has great importance
especially for us here in the noise control office, since our regulations
and standards must be based on accurate and timely appraisals of the actual
health effects of noise exposure.
Enclosure
EPA Form 1320-6 (Rev. 6-72)
-------
THE ABILITY OF
MILDLY HEARING-IMPAIRED INDIVIDUALS
TO DISCRIMINATE SPEECH IN NOISE
by
Alice H. Suter
Investigation Conducted and Report Prepared
under the Joint Sponsorship of
U.S. Environmental Protection Agency
Washington, D.C.
Aerospace Medical Research Laboratory
Wright-Patterson Air Force Base, Ohio
January 1978
-------
SUMMARY
The purpose of the investigation was to explore the relationship
between hearing level at various audiometric frequencies and speech
discrimination in different noise backgrounds. The study was designed
specifically to test the American Academy of Ophthalmology and Otolaryn-
gology's (AAOO) selection of a 26-dB average of 500, 1000, and 2000
Hz, as the point above which hearing handicap occurs. The AAOO method
for computing hearing handicap has lately been brought into question
for two primary reasons: that the 26-dB fence is too high, and for
the exclusion of frequencies above 2000 Hz. The present study, there-
fore, attempted to see if there were differences among individuals
whose hearing was at or better than the low fence, and if so, what
factors caused or affected the differences.
In designing the study, the following experimental questions were
posed:
1. What is the relationship between average hearing level at 500,
1000, and 2000 Hz and speech discrimination scores in noise
for individuals whose average hearing levels are at or better
than the AAOO low fence?
2. Is the relationship dependent upon speech-to-noise ratio?
3. Is the relationship between average hearing level and speech
discrimination scores differently described by different
speech materials?
4. Which combination of audiometric frequencies best predicts
speech discrimination scores?
Forty-eight subjects were tested with two types of speech materials:
the University of Maryland Test #1, which employs simple, "everyday"
sentences, and the Modified Rhyme Test, a closed-set test of rhyming
monosyllables. Speech stimuli were presented at 60 dBA measured at the
listener's ear. The noise stimulus, a babble of twelve voices, was
presented at levels of 60 to 66 dBA. Subjects were divided into three
groups according to their average hearing levels at 500, 1000, and 2000
Hz. Group I had normal hearing at all frequencies, Group II had mean
hearing levels of 13.4 dB in the mid-frequencies, and Group III, mean
hearing levels of 24.7 dB in the mid-frequencies. Both groups II and
III had considerable amounts of loss in the high frequencies, which is
typical of noise-induced hearing loss. Subjects listened to both speech
materials in a quiet condition and in three levels of background noise.
Group scores were compared in a three-factor analysis of variance, and
correlations between audiometric frequencies and individual discrimina-
tion scores were performed.
-------
Results of the investigation provided answers to the above ques-
tions as follows:
1. Significant differences were found in mean speech discrimina-
tion scores among all of the three groups, showing that
within the "normal" area under the 26-dB fence, there was
considerable variation in the ability to discriminate speech
in noise.
2. The relationship between average hearing level and speech
discrimination scores proved to be dependent upon speech-to-
noise ratio. The discrimination scores of all three groups
decreased as the speech-to-noise ratio became lower, and the
differences between groups increased.
3. Mean scores were similar for the two kinds of materials in
the two intermediate noise conditions, but not in quiet, or
in the most difficult noise condition. However, the two
materials were equally effective at delineating differences
among the three groups in the various conditions.
4. Correlational tests revealed that frequency combinations
that included frequencies above 2000 Hz were significantly
better predictors of speech discrimination scores than the
combination of 500, 1000, and 2000 Hz.
11
-------
PREFACE
According to a Public Health Survey* in 1962, 8.4% of the U.S. popu-
lation, or nowadays approximately 18 million people have hearing levels
of 25 dB or greater at the average audiometric frequencies of 500, 1000
and 2000 Hz. This hearing level is at the point of beginning hearing
handicap, as it is defined by the medical profession today. Many more
individuals have hearing levels that are less severe, but nevertheless
complain of difficulties in understanding speech, especially in a back-
ground of noise. Many of these 18 million or more individuals have
suffered their hearing losses as a result of exposure to noise.
On the authority of the Noise Control Act of 1972 the Environmental
Protection Agency is charged with conducting research on the effects of
noise. The resulting information is to be used for developing and re-
fining criteria, which in turn is used for setting standards and regula-
tions, advising other Federal agencies, giving technical assistance to
local communities, and educating the general public, all for the general
purpose of protecting the public against the adverse effects of noise.
The research described in this report was undertaken to assess the
functional abilities of individuals with hearing levels of approximately
25 dB in the mid-frequencies, as well as those with hearing levels less
severe. By doing so, the Agency would obtain information on the masking
effects of certain levels of environmental noise on a substantial portion
of the population. More specifically, the research was designed to test
the adequacy of the 25-dB mid-frequency demarcation point as the beginning
of hearing handicap. The results would, therefore, be of interest to
those who develop rules or guidelines for medico-legal purposes that are
based upon a definition of hearing handicap. They would also be of in-
terest to physicians, audiologists, public health specialists, and Federal
agency personnel who have traditionally used the 25-dB, mid-frequency
hearing level to differentiate between "normal" and "impaired" hearing.
The research was carried out in the Biodynamics and Bioengineering
Division of the Aerospace Medical Research Laboratory at Wright-Patterson
Air Force Base, Ohio, in support of Project 7231, Work Unit 03.** It was
conducted by Ms. Alice Suter of the Environmental Protection Agency's
Office of Noise Abatement, under the supervision of Dr. H. E. von Gierke
of the Air Force, under the auspices of an Inter-agency Agreement between
the two agencies. In addition to Dr. von Gierke the following individuals
served as reviewers of this report:
Dr. William Burns Dr. Karl Kryter Mr. Karl Pearsons
Laleham-on-Thames Stanford Research Institute Bolt Beranek and Newman, Inc.
England Menlo Park, California Canoga Park, California
* U.S. Department of Health, Education and Welfare, Public Health Service,
Hearing levels of adults by age and sex, United States 1960-1962.
** The voluntary informed consent of the subjects used in this research was
obtained as required by AFR 80-33.
111
-------
TABLE OF CONTENTS
Chapter Page
I. INTRODUCTION n
Statement of the Problem 3
II. HISTORY AND SURVEY OF THE LITERATURE 5
Historical Efforts to Assess Hearing Handicap 5
Support for the AAOO Rule 8
Criticism of the AAOO Rule 9
Other Methods of Assessing Hearing Handicap 10
Veterans Administration Method 10
Social Adequacy Index 11
Self Evaluation 11
Experimental Attempts to Characterize Hearing 12
Impairment or Handicap
Tests in Quiet 12
Tests in Noise 12
Other Distortions 14
Rationale for the Present Study 15
III. PROCEDURES 17
Subject Selection 17
Speech Materials 23
UM Test #1 23
MRT 24
Masking Noise 25
Relationship of Speech and Noise Stimuli 25
Speech Level 25
Speech-to-Noise Ratio 26
Sound Field 27
Listening Mode 27
Instrumentation 28
Equipment 28
Measurement 28
Calibration 30
Hearing Protective Devices 30
Method of Presentation 31
Analysis of the Data 32
Corrections for Guessing 33
Three-Factor Design 33
IV. RESULTS 35
Analysis of Variance 35
Correlational Tests 39
Further Ad Hoc Comparisons 43
V. DISCUSSION 46
Statistical Analysis 46
Analysis of Variance 46
IV
-------
TABLE OF CONTENTS (Cont)
Chapter
Other Tests on the Data
Comparisons of Results with Other Studies
Differences Between Groups
Effect of Speech-to-Noise Ratio
Similarity of Speech Materials
Comparison with Other Correlational Studies
Monaurality
The Speech Frequencies
Fences
VI. SUMMARY AND CONCLUSIONS
Summary
Conclusions
Recommendations
Appendix A Pilot Experiments
Instrumentation
Pilot Study #1
Pilot Study #2
Pilot Study #3
Appendix B Consent of Volunteer
Appendix C Instructions to Subjects (MRT)
Appendix D Speech Materials
Modified Rhyme Test
University of Maryland Test #1
Lists B through J
Appendix E Raw- Data
Subjects' Hearing Levels and Discrimination Scores
References
Page
47
48
48
49
50
50
53
53
54
58
58
60
60
62
62
62
62
64
68
69
70
70
71
80
83
v
-------
LIST OF TABLES
Table Page
1. Means, standard deviations, and ranges of better-ear 20
hearing levels in dB for each audiometric frequency.
2. Means, standard deviations, and ranges of better-ear 21
hearing levels in dB for various combinations of
frequencies.
3. Age and educational levels of the three groups. 22
4. Attenuation values in dB for hearing protectors 31
as reported by U.S. Air Force (1973).
5. Mean discrimination scores in percent correct for 34
MRT before and after corrections for guessing, and
for UM Test #1.
6. Mean scores and standard deviations in percent 36
correct responses.
7. Summary of analysis of variance to determine effects 37
of hearing loss, speech-to-noise ratio, and speech
materials on speech discrimination scores.
8. Tests on significant interactions of hearing loss 38
with speech-to-noise ratio, and speech materials with
speech-to-noise ratio.
9. Correlational matrix among six combinations of audio- 42
metric frequencies and discrimination scores in
various speech-to-noise ratios.
10. Matrix displaying significance of differences among 44
correlations of various frequency combinations and
speech discrimiation scores averaged over the four
experimental conditions.
A-l. Pilot Study #1. Mean scores and standard deviations 63
in percent correct of ten normal-hearing subjects for
various single and combined lists of UM Test |1.
A-2. Pilot Study #2. Mean scores and standard deviations 65
in percent correct on the Modified Rhyme Test of
thirteen normal-hearing subjects in monaural and
binaural listening modes.
A-3. Pilot Study #3. Individual scores on MRT and UM 66
Test #1 in a variety of speech-to-noise ratios.
VI
-------
LIST OF FIGURES
Figure Page
1 Mean better-ear hearing levels and ranges of the three 19
experimental groups.
2 Block diagram of equipment. 29
3 Mean percent correct responses of the three groups as a 40
function of speech-to-noise ratio. Scores are averaged
across the two speech materials.
4 Mean percent correct response in the two speech materials 41
as a function of speech-to-noise ratio. Scores are
averaged across the three groups.
5 Mean speech discrimination scores in percent correct as 45
a function of speech-to-noise ratio. Groups II and III
have been combined and partitioned according to whether
subjects have more or less hearing loss than the median
(47 dB) at the average of 2000, 3000, and 4000 Hz.
Group Y = < 47 dB, Group Z = > 47 dB.
VII
-------
CHAPTER I
INTRODUCTION
In today's complex society, hearing and understanding speech is by
far the most important function of the hearing mechanism. Warning
signals and the sounds of nature are also important to the human
listener, depending upon individual circumstance and preference. But
it is generally agreed that when loss of hearing becomes a handicap, it
is because an individual can no longer adequately hear or understand
speech.
Hearing loss from exposure to noise has been recognized through the
ages, especially in conjunction with inetalworking and gunpowder (Burns,
1973). It became a pervasive occupational condition with the advent of
the industrial revolution. Ibday, there are approximately 14 million
American workers in the production industries, 70% of whom are exposed
to noise levels of 85 dBA and above (BBN, 1974). Most of these indivi-
duals will incur some amount of hearing loss, however small in some
cases, if they remain in their noisy jobs over a working lifetime (EPA,
1974).
Before any significant attempts were made to prevent occupational
hearing loss, the condition was recognized as job-related and therefore
compensable under workmen's compensation laws. In the early part of
the century only "acoustic trauma" was compensated, but in the 1950"s
noise induced hearing loss of gradual origin was recognized as a com-
pensable occupational disease (Newby, 1964; Ginnold, 1974). Since com-
pensation preceded prevention, it is not very surprising that compensa-
tion formulas found their way into damage-risk criteria, and thus the
concepts of compensation and prevention became confused (Suter and von
Gierke, 1975; von Gierke, 1975). The risk of hearing loss from noise
was stated in terms of the percentage of workers expected to incur
compensable losses when exposed to certain levels of noise for certain
amounts of time (Rudmose, 1957).
The terms disability, handicap, and impairment were used almost
interchangeably until 1965, at which time the American Academy of Oph-
thalmology and Otolaryngology (AAOO) decided to make a long-needed
distinction (Davis, 1965).
1. Disability: actual or presumed inability to remain employed
at full wages.
2. Impairment: a deviation or change for the worse in either
structure or function, usually outside of the range of normal.
-------
3. Handicap: the disadvantage imposed by an impairment sufficient
to affect one's personal efficiency in the activities of daily
living.
That these terms are still confused is evident in the 1974 proposed stan-
dard for noise exposure of the U.S. Department of Labor's Occupational
Safety and Health Administration (OSHA). The proposed OSHA standard uses
the same criterion for impairment that the AAOO has used for handicap,
(OSHA, 1975).
In the ensuing controversy over the proposed OSHA standard new ques-
tions about occupational hearing loss have been raised. It has become
clear that there are legal and administrative aspects of the issue on the
one hand, and scientific aspects on the other. Examples of the former are
the differentiation between prevention and compensation, and the decisions.
as to the amount of hearing loss that should be prevented, and the amount
of compensation that should be awarded. Examples of the latter are the
estimation of hearing loss resulting from noise exposure, and the deci-
sion as to how much hearing loss is a handicap. This last point should
be decided primarily on the basis of hearing loss for speech.
Currently, the most widely used method for dealing with hearing loss,
both for compensation and preventive purposes, is the AAOO method
(Lierle, 1959). This method designates lower and upper cutoff points or
"fences" at 26 dB and 93 dB (converted to the ANSI 1969 standard for audio-
metric zero), respectively, for the averaged frequencies 500, 1000, and
2000 Hz with 1-1/2% handicap assigned to each decibel of hearing level
between those points.
The AAOO rule and its rationale have been criticized recently on
a variety of grounds: (a) that the rule is inappropriate for preventive
criteria, (b) that the "low fence" is too high, (c) that the rule dis-
counts the value of high-frequency hearing, (d) that it fails to take
the noisy aspect of day-to-day living into account, and (e) that it is
not based on sufficient scientific evidence (Kryter, 1963, 1973, and
1975; Niemeyer, 1967; Kuzniarz, 1973; EPA, 1974; von Gierke, 1975; Meyer,
1975; Thomas, 1975; UK-Member Body, 1975). The reason for the persisting
confusion between preventive and compensation criteria is not entirely
clear, except, as it was pointed out above, that the early emphasis on
industrial hearing conservation was to prevent compensable hearing loss.
Some other nations (such as the United Kingdom) have made the distinc-
tion, even though the AAOO's definition of beginning handicap has found
its way into the International Organization for Standardization's
"Assessment of occupational noise exposure for hearing conservation
purposes" (ISO, 1971). This confusion between prevention and compensa-
tion has been brought to the attention of OSHA by the Environmental
Protection Agency (EPA, 1975), and by Meyer (1975), Thomas (1975), and
von Gierke (1975) in formal testimony during OSHA-sponsored public
hearings.
The height of the 26 dB-fence has been criticized by Kryter (1973).
Based on calculations using the Articulation Index, Kryter maintained
that an individual with an average hearing level of 26 dB (at 500, 1000,
-------
and 2000 Hz) can correctly repeat only 90% of sentences at a normal con-
versational level and 50% of monosyllabic words at a "weak" conversational
level. The calculations assume a quiet background, and a distance of one
meter between talker and listener. Kryter recommended a low fence
of 15 dB for the frequencies 500, 1000, and 2000 Hz.
With respect to the frequencies used in predicting hearing handicap,
the National Institute of Occupational Safety and Health (NIOSH, 1972)
has dropped 500 Hz and substituted 3000 Hz in its criteria for the assess-
ment of occupational hearing impairment, thereby affirming the importance
of high frequency hearing. The decision was based largely on the research
of Harris, Haines, and Meyers (1960), Kryter, Williams, and Green (1962),
and Niemeyer (1967). (While no distinction is mentioned, the NIOSH
criteria pertain to preventive rather than compensation criteria.)
Kryter, Williams, and Green (1962), Niemeyer (1967), and Kuzniarz
(1973) among others, have demonstrated the importance of high frequency
hearing in conditions of background noise. Since some amount of back-
ground noise is very common in everyday listening conditions, these inves-
tigators have proposed that frequencies higher than 2000 Hz be included
in the averaging process for the prediction of speech discrimination scores.
Statements issued at the time of the AAOO's deliberations (DeForest
and Lierle, 1955; and Lierle 1959) defined hearing for speech as "the
ability to identify spoken words or sentences under average everyday con-
ditions of normal living ..." but did not specify the noise conditions
under which such sentence or ward identifications could occur, or the
degree of correct identification that was considered acceptable. Although
one assumes that the selection of a formula must have been guided by some
kind of scientific evidence, such evidence is not apparent. According to
the AAOO's Subcommittee on Noise (Lierle, 1959): "These principles are
based on current medical opinion."
Statement of the Problem
The present study was undertaken with the intent of examining the
AAOO low fence, both in terms of the 26-dB cutoff level and the use of
the simple, unweighted average of 500, 1000, and 2000 Hz. The object
was to see whether individuals whose hearing impairments are at this
criterion or better are indeed "not handicapped." In other words, does
the AAOO method implicitly classify some individuals as normal who yet
have considerable difficulty in understanding speech?
In order to investigate this issue subjects were selected whose
hearing in the mid-frequencies (500, 1000, and 2000 Hz) fell in the
vicinity of or better than the AAOO low fence. Any amount of loss above
2000 Hz was permitted in the experimental subjects, since these losses
are not considered in the AAOO formula. The effect of these losses was
central to the study. Speech discrimination scores were obtained for
sentence and monosyllabic material in several different levels of back-
ground noise in order to assess the influence of increasingly difficult
listening conditions.
-------
A dramatic cutoff in hearing level between little and great diffi-
culty in speech discrimination could not be expected, just as the 26-dB
fence cannot be viewed as a magical turning point. Loss of hearing is a
complex phenomenon, with great possibilities for subtle physiological
and behavioral differences. Statements of pure-tone acuity as measured
by standard clinical techniques cannot fully describe it. However,
pure-tone thresholds are a simple, objective method for measuring hearing
level, and they continue to be used for medico-legal purposes despite
certain shortcomings. Therefore, in this experiment pure-tone thresholds
have been related to speech discrimination scores under a variety of
conditions, and some attempt has been made to describe the functional
abilities of individuals according to their pure-tone thresholds.
In summary, this study investigated the adequacy with which the
AAOO rule predicts the point of beginning handicap through (a) application
of the fence at 26 dB, and (b) exclusion of frequencies above 2000 Hz. To
this end the following specific questions were posed:
1. What is the relationship between average hearing level at 500,
1000, and 2000 Hz and speech discrimination scores in noise for
individuals whose average hearing levels are at or better than
the AAOO low fence?
2. Is the relationship dependent upon speech-to-noise ratio?
3. Is the relationship between average hearing level and speech
discrimination scores differently described by different
speech materials?
4. Which combination of audiometric frequencies best predicts
speech discrimination scores?
-------
HISTORY AND SURVEY OF THE LITERATURE
Historical Efforts to Assess Hearing Handicap
The early experimental work on the intelligibility of speech was not
done to assess hearing handicap, but rather to study and improve communi-
cation systems. A vast body of knowledge was developed by such well-known
investigators as Davis, Fletcher, French, Steinberg, and Stevens. This
information was utilized in the early attempts to assess hearing handicap
and still is used today. These early investigators spoke of "hearing loss
for speech" since the distinctions between disability, impairment, and
handicap had not yet been made. Fletcher (1929) developed the first well-
known method for assessing hearing loss for speech, based on loudness and
intelligibility data. The entire audible range from 0 dB to 120 dB (re
ASA 1951 audiometric zero) for the averaged frequencies 500, 1000, and
2000 Hz was divided into degrees of loss with a slope of 0.83% loss per
dB. The simple (unweighted) average of 500, 1000, and 2000 Hz was chosen
because it correlated well with results of unamplified live voice and
phonograph record tests using digits. The slope was later modified to
yield an even 0.8% per dB and hence the time-honored "Fletcher Point-Eight
Rule." According to Davis, the rule was not meant for purposes of compen-
sation because there was no threshold of handicap and the ceiling was
virtually unattainable (Davis, 1970).
In 1939, the American Medical Association's (AMA) House of Dele-
gates of the Council on Physical Therapy approved a recommendation for
standardizing tests and preparing a method for estimating percentage
loss of hearing. The result was a "Tentative Standard Procedure for
Evaluating the Percentage of Useful Hearing Loss in Medicolegal Cases"
(Carter, 1942). The proposed standard ascribed percentage values of
hearing loss to octave intervals between the frequencies 256 and 4096 Hz,
and to 10-dB intervals from 10 dB to approximately 95 dB (re ASA 1951
audiometric zero). One plotted an audiogram, connected the hearing
levels at the various frequencies, selected the percentage values imme-
diately above the connected lines and summed them. Binaural hearing loss
was computed by weighting the better ear seven times the value of the
poorer ear. The standard was endorsed by the AMA in 1942.
This first AMA method was prepared by Bunch, Fowler, and Sabine
(as reported by Fowler, 1947) and was derived almost entirely from a
method developed by Sabine (1942). Sabine's method was in turn based
upon studies conducted by Knudsen (1923) and Fletcher (1929), that des-
cribed the number of discriminate units for both pitch and loudness
within the speech range. It was also based on the work of Steinberg
and Gardner (1940) who measured the abilities of both normal-hearing
and hearing-impaired persons to understand speech at various supra-
threshold levels. In addition it was influenced to some extent by a
method developed by Fowler (1942) which relied on Fowler's own clinical
-------
experience as an otologist, and on a combination of the experimental
data on frequency filtering by Fletcher, Steinberg, and others at Bell
Laboratories (Fletcher, 1929).
The 1942 method was revised in 1947 (Carter, 1947) and later became
known as the Fowler-Sabine or AMA method. The new method more closely
resembled Fowler's 1942 formula in that percentage values were ascribed
to discrete frequencies rather than to octave intervals, and the fre-
quency weighting was uniform at all hearing levels, namely 15%, 30%, 40%,
and 15% for 500, 1000, 2000, and 4000 Hz, respectively. In the old AMA
method, frequency weighting varied according to intensity.
Although the new formula was somewhat simpler than the 1942 version,
it was still not very popular, and many physicians preferred the old
Fletcher Point-Eight Rule. According to Davis (1973), "it was essential
to simplify the more accurate but complicated Fowler-Sabine scale in
order to gain acceptance. Otologists just wouldn't use the complicated
table " In addition to complaints about the complexity of the AMA
method, it was felt that the weightings ascribed to the different frequen-
cies were somewhat arbitrary (Davis, 1971). Another objection was that
although the formula worked fairly well with conductive losses, it was
not appropriate for individuals with sensori-neural losses that were pre-
dominantly in the high frequencies (De Forest and Lierle, 1955). Conse-
quently, another method was proposed in 1959 (Lierle, 1959), which was
adopted by the AMA in 1961. It was prepared by the Subcommittee on
Noise of the American Academy of Ophthalmology and Otolaryngology, and
it became known as the AAOO method or rule.
The AAOO method used the simple average of 500, 1000, and 2000
Hz with a "low fence" or normal cutoff at 15 dB (ASA; or 26 dB re ANSI
audiometric zero) and a "high fence" or total loss cutoff at 82 dB (ASA;
93 dB ANSI), with 1-1/2% impairment (or later handicap) for each dB be-
tween the two fences. The better ear was given five times the weight of
the poorer ear. This method was purportedly based on the ability of the
hearing-impaired individuals to hear speech. In the words of the Subcom-
mittee (Lierle, 1959):
Ideally, hearing impairment should be evaluated in terms
of ability to hear everyday speech under everyday condi-
tions... The ability to hear sentences and repeat them
correctly in a quiet environment is taken as satisfactory
evidence of correct hearing for everyday speech. Because
of present limitations of speech audiometry, the hearing
level for speech should be estimated from measurements
with a pure tone audiometer. For this purpose, the Sub-
committee recommends the simple average of 500, 1000 and
2000 cps... If the average hearing level at 500, 1000
and 2000 cps is 15 dB [ASA] or less, usually no impairment
exists in the ability to hear everyday speech under every-
day conditions.
According to Davis (1973), the Subcommittee determined, on the basis of
clinical evidence, that an average hearing level at 500, 1000, and 2000
-------
Hz of approximately 16 dB (ASA) was the point at which individuals begin
to have difficulty hearing sentences in quiet and seek medical help for
their hearing problems.
GLorig and his colleagues (see Glorig and Baughn, 1973) conducted a
self-assessment poll in conjunction with the Wisconsin State Fair hearing
survey of 1954. Responses to the question "Is your hearing good, fair,
or poor?" were correlated with median hearing levels. Individuals who
reported that their hearing was "good" had median hearing levels of 10
to 16 dB (ASA; averaged over 500, 1000, and 2000 Hz). However, Glorig
and Baughn point out that the responses were clearly age-dependent, with
younger people rating themselves by more stringent criteria than the
older people.
The Subcommittee's report (Lierle, 1959) made no mention of experi-
mental data in support of its decision to change the Fowler-Sabine
method, although there were studies conducted in the decade between the
AMA and AAOO rules that favored the simple average of 500, 1000, and 2000
Hz as being the most important indicator of "hearing for speech." Car-
hart (1946) found that the AMA method and the simple average of the
three frequencies 512, 1024, and 2048 Hz correlated equally well with
speech reception thresholds for bi-syllabic words in a variety of hearing-
impaired cases. Although the AMA method was found to be a slightly
better predictor of speech reception for subjects with marked high tone
losses, there was less variability associated with the 3-frequency
method. For this reason and for practical considerations, Carhart
favored the 3-frequency method.
As mentioned above, otologists had continued to use Fletcher's
Point-Eight rule, mainly because of its simplicity. As a result of
additional studies, Fletcher first reaffirmed the simple 3-frequency
average, and then proposed a more complex method employing weighted
frequencies from 250 to 8000 Hz. The former method (Fletcher, 1950),
based entirely on loudness calculations, was validated against three sets
of speech reception data. These data consisted of the averaged speech
reception thresholds (level of 50% correct responses) for digits, spon-
dees, phonetically balanced (PB) words, and sentences, transmitted by a
variety of means (phonograph records, calling directly through the air,
and voice attenuated by an audiometer). Fletcher found that while the
average of 500, 1000, and 2000 Hz was a good predictor of speech recep-
tion for relatively flat losses, the average of the best two of these
three frequencies was more often correct, especially for sloping losses,
and hence the widely used "Fletcher Average" or "Two-Frequency Average"
that still is in clinical use today.
Fletcher's next method of assessing hearing loss for speech (Fletcher
and Gait, 1950; and Fletcher, 1952) was based on calculations of the
Articulation Index. This formula employed the following weights: 250 Hz
= 0.04, 500 Hz - 0.13, 1000 Hz = 0.23, 2000 Hz = 0.30, 4000 Hz = 0.25,
8000 Hz = 0.05. The study's purpose was to match transmission systems to
hearing-impaired ears. It does not seem to have influenced current think-
ing on the assessment of hearing handicap as much as his 1950 formula.
-------
Two other investigations are likely to have given impetus to the
AAOO method. Harris, Raines, and Myers (1956) compared six different
pure-tone methods for predicting "hearing loss for speech" as measured
by thresholds of intelligibility (50£ correct) for PB words. _ The authors
concluded that the best predictor was a combination of a multiple regres-
sion method (including 500 to 6000 Hz) and the 3-freguency (500, 1000,
and 2000 Hz) average. After an error in calculation was pointed out, the
authors concluded that the simple average of 500, 1000, and 2000 Hz was
the most satisfactory method (Harris et al, Erratum, 1956).
In a later study, Quiggle, Glorig, Delk, and Summerfield (1957)
showed that the average of 500, 1000, and 1500 Hz was the best predictor
of hearing loss for speech as measured by spondee word thresholds. The
authors simplified the method by substituting 2000 for 1500 Hz (since
1500 Hz is rarely tested), but cautioned that spondee words were not
"fully representative of the speech sounds a man must 'understand' to
communicate verbally."
Although the psychoacoustical and statistical techniques used in
these attempts at characterizing hearing handicap may have been fairly
sophisticated, the actual assessment of hearing for speech was incom-
plete, and often primitive by today's clinical standards. Such tech-
niques as calling words and digits through the air give little quanti-
fiable information on an individual's capacity to discriminate speech
sounds. In addition, threshold measures (a 50% criterion), for PB's
as used by Harris et al (1956), or spondees as used by Carhart (1946)
and Quiggle et al (1957), give little information on one's ability to
understand speech in a variety of everyday conditions. All of the above
experiments were presumably carried out in quiet since none mentions a
noise background.
Support for the AAOO Rule
Some investigators have continued to support the AAOO rule since
its acceptance by the AMA in 1961. Davis (1971) stated, "The AAOO rule
now enjoys considerable legal prestige by its incorporation into many
rules or even State laws relating to compensation for hearing handicap,
from whatever the cause." It has been enthusiastically supported by
Glorig (see Discussion of Part I and Summing-up in Robinson, 1971;
Glorig and Baughn, 1973). Glorig and Baughn (1973) cited three studies
in defense of the use of 500, 1000, and 2000 Hz. The first, by Ward,
Fleer, and Glorig (1962), showed relatively small differences for
speech discrimination in quiet between subjects with sloping hearing
losses above 2000 Hz and those with normal hearing. In the second,
Myers and Angermeier (1972) tested speech discrimination in subjects
with a variety of hearing losses. The authors found large amounts of
scatter when the audiometric average of 500, 1000, and 2000 Hz, and
discrete frequencies of 2000 Hz and 3000 Hz were related to speech dis-
crimination scores in quiet and noise, and they concluded that no pure-
tone audiometric index could explain the variance among their individual
listeners. The third study, by Murry and Lacroix (1972) investigated
speech discrimination in noise of individuals with hearing losses
-------
above 2000 Hz. Ihe authors found that none of the pure tone predictors
tested (average of 500, 1000, and 2000 Hz; 1000, 2000, and 3000 Hz; or
3000 Hz alone) correlated well with speech discrimination scores, but
that the average of 1000, 2000, and 3000 Hz correlated slightly better
than the other two. They also found that individuals with hearing losses
above 2000 Hz scored about 5 to 10 percentage points more poorly than
normal-hearing subjects on the discrimination test used (the Modified
Rhyme Hearing Test). In short, these studies lend no support to the
500, 1000, and 2000 Hz formula, contrary to the implication of Glorig
and Baughn.
Criticism of the AAOO Rule
At present there is considerable criticism of the AAOO rule in the
scientific community, especially as it applies to preventive criteria.
The Department of Labor held public hearings on the proposed standard
for occupational exposure to noise in 1975, which provided a forum for
this kind of criticism. The most frequent complaint was that the simple
average of 500, 1000, and 2000 Hz penalizes persons with noise-induced
hearing losses by giving equal weight to these three frequencies, and by
ignoring frequencies above 2000 Hz, even though noise-exposed individuals
sustain most of their loss in the higher frequencies (von Gierke, 1975;
Kryter, 1975; Thomas, 1975). In the criteria for an occupational noise
exposure standard developed by NIOSH (1972) the 500, 1000, and 2000 Hz
average is rejected in favor of 1000, 2000, and 3000 Hz because of the
importance of higher frequencies for understanding speech in everyday
conditions. Similarly, the British Standard for Assessment of Occupa-
tional Noise Exposure (BS 5330) has recently changed to an average of
1000, 2000, and 3000 Hz, and these same frequencies are incorporated
into the U.K.-Member Body's recommendation for revision of the ISO 1999
(1975) . The importance of the higher frequencies for understanding
speech has been frequently reported in the literature.
Kryter (1963 and 1973) has maintained that the 26-dB (ANSI) fence
is too high, especially in a background of noise. From calculations
based on the Articulation Index, he predicted that an individual with
a 26-dB hearing loss at 500, 1000, and 2000 Hz can correctly repeat only
90% of sentences at a normal conversational level (55 dB long-term rms),
and 50% of monosyllabic words at a weak conversational level (50 dB
long-term rms). The calculations assume a quiet background and a dis-
tance of one meter between talker and listener. Kryter advocated a 15-dB
fence if the frequencies 500, 1000, and 2000 Hz are used. Even Davis
(1971), who has been a strong supporter, admitted that the AAOO rule
may be harsh at the low fence. He has also stated that it is not appro-
priate for "steep" audiograms (Davis, 1970), although he has continued
to defend the rule over the years.
The AAOO rule has also been criticized on the grounds that the
criterion for understanding speech, namely hearing sentences and repeat-
ing them correctly in a quiet environment, is not representative of real
life. Kryter (1973) summarized this point as follows:
-------
10
It is not obvious why the noise-deafened ear should not be
considered impaired nor the individual handicapped when
losing the ability the normal person has to understand:
(a) individual words, the unexpected message, the unfamil-
iar name, or the important telephone number; (b) the weaker-
than-normal intensity speech that can occur because the
talker drops his voice level, or the distance between the
talker and listener is greater than one meter or so and
the talker is using a normal conversational level of effort
for the quiet; or (c) speech in the presence of everyday
noise, at a party or conference when several people are
talking, etc.
Although undistorted speech in quiet is used in many experimental
and in most clinical evaluations, Harris (1965) has pointed out that
this condition is characteristic of not more than half of our everyday
listening conditions. Niemeyer (1967) and Kuzniarz (1973) have also
stressed the universality of background noise and its detrimental effects
on individuals with high frequency hearing loss.
Other Methods of Assessing Hearing Handicap
Veterans Mministration Method
Although pure-tone audiometry has been the primary method used for
medico-legal and damage-risk purposes, speech audiometry has been used
for some time in certain systems. The Veterans Administration (VA) has
rated hearing impairment for many years either by a combination of speech
reception threshold (SRT) for spondee words and speech discrimination
scores for PB words, or by pure-tone thresholds. Normal limits of hear-
ing are defined by an SRT of less than 26 dB (ANSI), a discrimination
score of higher than 92%, and pure-tone thresholds of better than 40 dB
(ANSI) for all audiometric frequencies from 250 - 4000 Hz and better than
25 dB (ANSI) for at least four frequencies, (VA, 1976). Hearing losses
that exceed these amounts are not necessarily eligible for compensation,
even though they are no longer considered normal. Compensation is usually
awarded on the basis of speech audiometry, but in those cases where only
pure tone thresholds are used it can be awarded for any of the following
conditions (VA, 1976): hearing loss in both ears of 50 dB or greater
in any of the frequencies 500, 1000, or 2000 Hz; average hearing level
in both ears of 38 dB or greater at 500, 1000, and 2000 Hz; hearing
loss in one ear of 75 dB or greater in any of the frequencies 500, 1000,
or 2000 Hz; or average hearing level in one ear of 500, 1000, and 2000
Hz of 58 dB or greater (all values re ANSI zero).
-------
11
Social Adequacy Index
The Social Adequacy Index (SAI) was developed as a tool for pre-
dicting improvement by a hearing aid or by the fenestration operation,
and for assessing social adequacy for medicolegal purposes. The SAI was
based on the speech discrimination scores, using Harvard PB words, of
normal and hearing-impaired listeners that would be predicted at three
different listening levels: faint speech at 55 dB (sound pressure level),
average speech at 70 dB, and loud speech at 85 dB. The SAI represented
the average of discrimination scores over the three conditions.
Davis (1948) developed a chart where the SAI could be estimated from
just two audiometric tests, the speech reception threshold in dB, and the
discrimination score at about 35 dB above SRT. Difficulty in social
situations was thought to begin at an SAI of 67, when about two-thirds of
the PB monosyllables would be understood for an average of the three
conditions. This point occurred between an SRT of 28 dB (converted to
ANSI zero for speech audiometry) if the discrimination score was perfect,
and a discrimination score of 70% if the SRT was 2 dB (ANSI) or better.
The threshold of social adequacy, the point at which one could barely
"get by" was an SAI of 33. This point was judged to occur between a
speech reception threshold of 46 dB (converted to ANSI zero for speech
audiometry) if the discrimination score was perfect, and a discrimination
score of 35% if the SRT was 2 dB or better. Although the scheme was a
logical one, Davis (1970) offers two possible reasons why it did not work
very well in practice. One was the fact that speech discrimination tests
were not considered as accurate as tests for speech reception threshold.
The other was an admitted lack of knowledge about the relationship
between the understanding of connected speech, and its component parts
(sound frequencies, phonemes, and syllables).
Self Evaluation
There have been many attempts to assess hearing handicap through
questionnaires or self-evaluation techniques. Notable examples are
studies by Nett, Doerfler, and Matthews (1959), High, Fairbanks, and
Glorig (1964), and Atherley and Noble (1971). The investigators expressed
reservations with pure-tone audiometry as an adequate predictor of handi-
cap since two people with an identical pure-tone impairment will often
suffer different degrees of handicap (High et al, 1964; Atherley and
Noble, 1971). These methods, however, have failed to produce quantita-
tive results. Too many variables besides hearing loss are involved.
The answers appear to be dependent upon age (Glorig and Baughn, 1973;
Merluzzi and Hinchcliffe, 1973), on occupation (Simonton and Hedgecock,
1953), and on a number of other factors as brought out by Nett et al
(1959). Although the study by Nett et al was an ambitious attempt to
assess handicap by self-evaluation, the authors failed to recommend a
usable verbal model based on the resulting data, and the study was never
published.
-------
12
Experimental Attempts to Characterize Hearing Impairment or Handicap
Tests in Quiet
The relationship between audiometric frequency and the understanding
of speech in hearing-impaired listeners has been studied for over thirty
years. Most of the earlier tests were conducted in quiet (Utley, 1944;
Fletcher, 1950; Quiggle, Glorig, Delk, and Summerfield, 1957; Mullins and
Bangs, 1957; and Ward, Fleer, and Glorig, 1962). The work of Fletcher,
Quiggle et al, and Ward et al has been discussed above.
Utley (1944) compared the hearing threshold levels of severely hard-
of-hearing children, and their losses as computed by five different pure-
tone methods of assessment, to discrimination scores for vowels and con-
sonants. She found that all five pure-tone methods correlated well with
discrimination scores, as did hearing levels at 512, 1024, and 2048 Hz.
However, the study is not applicable to mildly hearing-impaired individuals
since many of the children had no residual hearing at 2048 Hz, and even
fewer had hearing above that point. Consequently, the author did not at-
tempt to correlate thresholds above 2048 Hz with discrimination scores.
Mullins and Bangs (1957) compared pure-tone thresholds of hearing-
impaired veterans to speech discrimination scores on Harvard PB word lists.
Of the pure-tone thresholds 250, 500, 1000, 2000, 3000, and 4000 Hz, the
investigators found that the frequencies 2000 and 3000 Hz were the best
predictors of speech discrimination scores. They also found that the
steepness of the slope between 500, 1000, and 2000 Hz, which they called
the "index of inferred masking" for these three frequencies, yielded the
highest correlation with discrimination ability of any of their measures,
although none of the correlations was high enough to achieve statistical
significance.
Quist-Hanssen and Steen (1960) tested three pure-tone methods of
assessment against scores for monosyllables, disyllables, digits, and
"context" speech (in Norwegian), in subjects with noise-induced hearing
loss. The authors found that all three methods overestimated speech
reception scores. Although the averaged frequencies 500, 1000, and 2000
Hz came closer to predicting speech scores than the other two methods,
the authors concluded that speech audiometry should be performed in these
cases.
Tests in .Noise
Daring the 1960 's and 1970 's investigators began to examine the ef-
fects of background noise, presumably because noise is characteristic
of many everyday listening conditions. Most of these investigations
showed that good hearing at and above 2000 Hz was necessary to overcome
the adverse effects of masking noise (Kryter, Williams, and Green, 1962;
Ross, Huntington, Newby- and Dixon, 1965; Acton, 1970; Elkins, 1971;
Lindeman, 1971; Aniansson, 1973; Kuzniarz, 1973; and Dickman, 1974).
-------
13
One of the first, and largest of these studies was that of Kryter, Wil-
liams, and Green (1962), which correlated audiometric thresholds with
speech discrimination scores for the Harvard sentences and monosyllables.
The authors grouped a large population of sensori-neural hearing loss
subjects according to the frequency at which the hearing loss began (for
example 1000 Hz and above, or 2000 Hz and above, etc.). They found
that the frequencies 2000, 3000, and 4000 Hz were the best indicators
of speech discrimination in their experimental conditions and concluded
that formulas that do not take into account frequencies above 2000 Hz are
not appropriate predictors for the understanding of speech in "realistic
acoustic environments."
Ross, Huntington, Newby, and Dixon (1965) also studied speech dis-
crimination in a background of noise in an attempt to relate discrimi-
nation scores to a combination of an "exogenous distortion" (noise) and
a variety of "endogenous distortions" (abnormal difference limen func-
tions, reduced linear range, and audiometric configuration). The authors
wished to test the hypothesis that suprathreshold distortions contribute
to speech discrimination problems over and above the amount that would
be explainable by frequency filtering. Although the multiple distortion
hypothesis was not validated, the study did indicate the value of the
higher audiometric frequencies (2000 and 4000 Hz), for speech discrim-
ination in quiet. But these two frequencies did not predict the rela-
tive discrimination shift (the differences between scores in quiet and
scores in noise) as well as 500 Hz.
A study of speech discrimination by Elkins (1971) showed that hear-
ing-impaired subjects performed predictably more poorly than normal-
hearing subjects, both in quiet and in noise, except for the most diffi-
cult noise condition, where the mean difference between groups was much
smaller than expected. Like Ross et al, Elkins found that the frequen-
cies 2000, 3000, and 4000 Hz were good predictors of speech discrimination
in quiet, but significant correlations between audiometric threshold and
speech discrimination in moderate levels of noise were lacking. Both
Elkins and Ross et a'l presented the speech material at a level of 40 dB
above each subject's SRT, with the masking stimulus at a fixed level in
relation to the speech. This technique may help to explain the smaller
than expected differences between normal and hearing-impaired groups in
noise. It may also help to explain the lack of clear cut results in the
studies by Myers and Angermeyer (1972) and Murry and Lacroix (1972).
Presenting the stimuli at a level of 40 dB above SRT is usually done
in an attempt to find the asymptote of a subject's performance-intensity
(P-I) curve, or "PB max." While this procedure is useful for evaluating
potential performance under amplification, it is not the most appropri-
ate method of describing the performance of hearing-impaired individuals
in everyday, unamplified conditions. In everyday conditions, people are
forced to listen at various levels along the P-I curve, not just at the
point of maximum performance. If the investigators had presented the
stimuli at a fixed level for all subjects, greater differences between
normal and hearing-impaired groups would most likely have occurred. For
example, a presentation level of 40 dB above audiometric zero would most
-------
14
likely produce a 100% correct response from a normal listener. A presen-
tation level of 40 dB above an SRT of 24 dB in a hypothetical case of
hearing impairment, (which would be 64 dB above audiometric zero) would
also be likely to produce a discrimination score of 100%. If, however,
the speech were presented at the same level to both listeners (+40 in one
case and +16 in the other) the discrimination scores might be quite
different.
Studies by Acton (1970), Lindeman (1971), Aniansson (1973), Kuzniarz
(1973), and Dickman (1974) presented both the speech and noise at fixed
levels regardless of the individual subjects' hearing acuity. Each of
these investigators used subjects with predominantly high-frequency hear-
ing losses who would be considered either normal or only mildly impaired
when assessed by the 500, 1000, 2000 Hz average, with the possible excep-
tion of Lindeman's group, whose audiograms were not reported. Acton stud-
ied discrimination of monosyllables in a background of pink noise that
had been filtered so as to roll off 6 dB per octave in the higher fre-
quencies. He found that subjects whose losses included 2000 Hz were de-
cidedly more impaired than those whose losses began above that point.
Lindeman compared audiometric thresholds of persons with noise-induced
hearing loss to their ability to identify monosyllables (Dutch) in back-
grounds of "cocktail party" noise. He found that in a speech-to-noise
ratio of +10 (speech 10 dB above noise), the most important audiometric
frequency was 2000 Hz. Aniansson evaluated Swedish PB monosyllables in
a background of traffic noise plus, in some conditions, competing radio
voice and live voice. He concluded that the frequencies 3000 and 4000 Hz
are just as important as 500 and 1000 Hz when estimating speech discri-
mination in an "everyday milieu." Kuzniarz studied the effect of white
and predominantly low frequency noise on Polish monosyllable and sentence
discrimination. Dickman used the Central Institute for the Deaf (CID)
"everyday" sentences in three different noise backgrounds. Both of
these latter investigators found significant differences in scores
between normal-hearing and hearing-impaired individuals. They also con-
cluded that hearing acuity above 2000 Hz is critical for speech dis-
crimination in noisy conditions.
Other Distortions
Since noise is not the only distortion to which everyday speech is
subject, some investigators chose to examine other distortions, such as
poor articulation, speeded speech, and filtering. A study of sentence
intelligibility for speeded speech led Harris, Haines, and Myers (1960)
to reconsider the value of frequencies higher than 2000 Hz in hearing-
impaired listeners. The authors concluded that near-normal hearing at
3000 Hz is essential for sentence intelligibility if the speech is
speeded. Harris (1965) hypothesized that distortions which may have
little adverse effect when experienced singly can produce a much more
serious effect when combined. This theory prompted him to examine a
number of speech distortions including atypical accents, interruptions,
reverberation, and three degrees of speeding. Correlations with pure-
tone audiograms caused him to recommend an average of 1000, 2000, and
3000 Hz as the best predictor of everyday speech, which he estimated is
-------
15
distorted by sources other than noise about 50% of the time. Some of
the previously mentioned experimental work also employed some forms of
speech distortion as well as noise. Kryter et al (1962) introduced low-
pass filtering (attenuating the speech above 2000 Hz), Acton (1970) and
Aniansson (1973) introduced mild reverberation, and as mentioned above,
Aniansson's study included competing speech in some conditions.
Rationale for the Present Study
Although the majority of experimental evidence points to the impor-
tance of high frequency hearing for understanding speech, further research
was needed specifically to test certain conditions. As Davis (1973)
suggested, the AAOO assumptions needed to be validated in an experiment
utilizing "everyday" speech in an "everyday" noise background. Subjects
were needed whose hearing levels were in the vicinity of or better than
the 26-dB low fence.
Most of the researchers mentioned earlier in this chapter did
not present the speech material in "everyday" environments. The earlier
experiments were conducted in quiet (Carhart, 1946; Harris et al, 1956;
Mullins and Bangs, 1957; Quiggle et al, 1957). Later, most of the
studies that employed noise presented the stimuli under earphones rather
than in the sound field (Kryter et al, 1962; Ross et al, 1965; Elkins,
1971; Lindeman, 1971; Myers and Angermeyer, 1972; Murry and Lacroix,
1972; and Dickman, 1974). Conditions were somewhat more lifelike in the
experiments of Harris (1965) and Aniansson (1973) where the stimuli were
recorded in the sound field and then presented to the subjects through
earphones. However, Harris' reverberation time of five seconds was
considerably greater than that of most rooms, which according to Nabelek
and Pickett (1974a) is more likely to be one second or less. Of all the
speech discrimination studies of mildly-impaired persons discussed above,
only Acton (1970) and Kuzniarz (1973) presented the speech and noise
material in a live sound field, with reverberation times of 0.5 and 0.2
second, respectively. Even those experiments could have been somewhat
closer to "everyday" conditions if sentences in addition to or instead
of monosyllables had been used.
The need for lifelike experimental conditions in assessing hearing
handicap was stressed by Webster, Davis, and Ward (1965) in their
comments on Harris1 experiment (1965). The authors suggested that the
distortions that Harris had chosen were not "everyday" speech distortions:
electronic chopping with a 50% duty cycle and 8 interruptions per
second, reverberation time of 5 seconds, and speedups of 250 to 345 words
per minute. Moreover, they criticized the study for not using noise as a
listening condition, and for presenting the stimuli at 40 dB above
speech reception threshold. Instead, they recommended that hearing
handicap be assessed under the following conditions:
1. "everyday patients" with high-frequency hearing losses
2. "everyday noise," low-frequency masking
-------
16
3. "everyday talkers" such as those used by Harris
4. "everyday distortions," 1 to 2 seconds reverberation time, and
a background of meaningful babble
5. "everyday listening levels," not 40 dB above each patient's
threshold.
The present experiment was designed to satisfy the above conditions
as closely as possible. The experimental conditions will be discussed in
detail in the next chapter. Under these everyday environmental conditions,
the relationship between hearing levels and speech discrimination scores
was explored. Also investigated was the dependency of this relationship
on increasing levels of background noise, the ability of two different
speech materials (sentences and rhyme words) to describe the above rela-
tionships, and the ability of various audiometric frequency combinations
to predict speech discrimination scores.
-------
CHAPTER III
PROCEDURES
This chapter will address the selection of subjects and the setting
of experimental conditions. Those conditions include speech materials,
masking noise, the relationship between the speech and noise stimuli,
sound field, and mode of listening (monaural as opposed to binaural).
Instrumention will also be discussed, as well as the method of stimulus
presentation, and the selected methods of data analysis.
Two types of speech materials, a closed set test of monosyllabic
words—the Modified Rhyme Test (MET)—and an open set test of sentences—
the University of Maryland Test #1 (UM Test #1)—were presented to forty-
eight subjects. The subjects were divided into three equal groups ac-
cording to their better-ear average hearing level for the frequencies 500,
1000, and 2000 Hz. Mean average hearing levels were as follows: Group
I 2.5 dB, Group II 13.4 dB, and Group III 24.7 dB. The tests were con-
ducted in a mildly reverberant room (T = 0.625 sec.), in four different
noise conditions with the speech stimuli delivered at a fixed level of
60 dBA measured at the subject's ear. The speech-to-noise ratios were
slightly different for the two different speech materials. The speech-
to-noise ratios for the MRT were quiet (Q), 0, -3, and -6 dB; and for the
UM Test #1 were Q, -1, -3, and -5 dB, the minus designation indicating
that the noise level was higher than the speech level. Experimental
conditions were determined on the basis of three pilot studies which are
discussed in Appendix A.
Subject Selection
Subjects were recruited through the clinical and screening facili-
ties at Wright-Patterson Air Force Base and the Springfield Air National
Guard in Ohio, by advertisements in local newspapers, and by word-of-
mouth. Both normal-hearing and hearing-impaired subjects were recruited.
Testing was accomplished in one visit, and subjects were paid $15. Those
with hearing losses were tested by bone conduction so that those with con-
ductive losses could be eliminated. Also, those with losses too severe or
too mild to meet the hearing loss criteria were eliminated. Subjects
were screened informally in order to eliminate those with problems of
articulation or dialect, which might interfere with scoring procedures on
the sentence discrimination task where the subject responded orally-
Air-conduction and bone-conduction testing was performed with a Grason-
Stadler 1701 audiometer with TDH 49 earphones, in a room that met the
ANSI (1960) specifications for audiometric test rooms.
Subjects who qualified were classified into one of three groups accord-
ing to the following criteria:
17
-------
18
Group I, the normal or control group, could have better-ear hearing
levels in the averaged mid-frequencies 500, 1000, and 2000 Hz no worse
than 8 dB (ANSI, 1969) with no frequency from 250-6000 Hz worse than 20 dB.
Group II could have better-ear average hearing levels in the mid-
frequencies from 10 to 18 dB.
Group III could have better-ear average hearing levels in the mid-
frequencies from 20 to 28 dB.
No restrictions were placed on presence, absence, or amount of high-
frequency hearing loss in groups II and III, but since most of the sub-
jects had noise-induced hearing losses, there were considerable amounts
of loss in the 3000 to 6000 Hz range.
Figure 1 shows mean better-ear hearing levels and ranges for the
frequencies 250 through 6000 Hz for the three groups. Table I shows
standard deviations and ranges of hearing level for each audiometric fre-
quency for the three groups. Table II shows the average hearing levels,
standard deviations, and ranges for a number of possible frequency combi-
nations. As a result of the study's design, inter-subject variability
was fairly low in the low and middle frequencies, but considerable
variability was allowed in the higher frequencies.
Table III shows means, standard deviations, and ranges for the fac-
tors of age and educational level. A total of 48 subjects were'selected,
16 in each group. No attempt was made to control for sex since it has
been shown that there is no difference in speech perception between male
and female listeners (Silverstein, Bilger, Hanley, and Steer, 1953).
Subjects were required to be within the ages of 18 and 56, and to
have at least an 8th grade education so as to have acquired basic lan-
guage skills. The upper age limit was imposed in order to minimize the
contribution of presbycusis. Investigators have found a decline in
speech discrimination as a function of aging, despite controlling for
actual hearing level, (a summary of these studies is found in CHABA,
1977). The effect appears to be greater with more difficult speech mate-
rial (Goetzinger, Proud, Dirks and Embrey, 1961). Feldman and Reger (1967)
found that discrimination scores for phonetically balanced (PB) words
decreased by approximately 5% per decade after age 50 in a population
that was not controlled for hearing acuity, (which decreased about 10 dB
per decade in the mid-frequencies and about 15 dB per decade at 4000 Hz).
(PB words are more difficult than the materials selected for the present
study). Blumenfeld, Bergman, and Millner (1969), using the Fairbanks
Rhyme Test, found that correlations of scores with age were much higher
with subjects over age 60 than for those below that age. Bergman (1971)
tested mildly hearing-impaired subjects in quiet, with the "CHABA"
sentences (the same material that was selected for the present study),
and found that there was little degradation in scores until age 80.
The effect of aging also appears to be greater when the speech is
presented in a background of noise than in quiet (Jerger, 1973). Mayer
-------
Frequency in Hz
19
§
CD
V)
z
<
CO
•o
_c
1
0>
2
o
V)
£
.c
o>
n
03
10
20 H
30 H
40 H
50 H
60 H
70 H
80 H
90 H
100-^
250
500
1000
2000
3000
4000
6000
Ranges:
Group I
Group II —
Group III •••
Figure 1. Mean better-ear hearing levels and ranges of the three
experimental groups.
-------
Table I. Means, standard deviations, and ranges of
better-ear hearing levels in dB for each audiometric frequency.
250
500
Mean Hearing Levels (dB)
1000 2000 3000
4000
6000 Hz
Group I
Group II
Group III
1.9
7.8
11.3
3.4
9.7
13.4
0.6
11.3
19.4
3.4
19.0
41.3
6.3
47.2
57.8
9.4
58.8
63.1
10.9
55.6
67.8
250
500
Standard Deviations (dB)
1000
2000
3000
4000
6000 Hz
Group I
Group II
Group III
5.1
5.5
6.7
4.4
4.3
5.4
3.1
3.4
4.8
5.7
9.0
13.4
8.5
17.7
16.7
6.0
15.6
20.1
8.0
18.7
20.3
Ranges (dB)
Group I
Group II
Group III
250
-5 to 10
0 to 20
0 to 25
500
-5 to 10
0 to 15
5 to 20
1000
-5 to 5
5 to 15
10 to 25
2000
-5 to 15
10 to 35
25 to 70
3000
-5 to 20
25 to 85
25 to 90
4000
0 to 20
30 to 95
35 to 100
6000 Hz
-5 to 20
20 to 90
35 to 100
N>
O
-------
Table II. Means, standards deviations, and ranges of better-
ear hearing levels in dB for various combinations of frequencies.
Mean Average Hearing Levels (dB)
Group I
Group II
Group III
Group I
Group II
Group III
500,1000,
2000 Hz
2.5
13.4
24.7
500,1000,
2000 Hz
3.4
2.9
3.1
500,1000,
2000, 3000 Hz
3.4
22.0
33.0
Standard
500,1000,
2000, 3000 Hz
4.4
5.6
5.7
1000,2000,
3000 Hz
3.4
25.9
39.6
Deviations
1000,2000,
3000 Hz
4.9
7.2
8.2
1000,2000,
4000 Hz
4.4
29.6
41.2
(dB)
1000,2000,
4000 Hz
3.8
5.0
8.4
2000,3000,
4000 Hz
6.6
41.6
54.1
2000,3000,
4000 Hz
5.7
10.3
14.3
3000,4000,
6000 Hz
8.8
54.0
62.9
3000,4000,
6000 Hz
6.4
15.1
17.6
Ranges (dB)
Group I
Group II
Group III
500,1000,
2000 Hz
-3 to 8
10 to 18
20 to 28
500,1000,
2000, 3000 Hz
-4 to 11
14 to 30
21 to 43
1000,2000,
3000 Hz
-3 to 13
15 to 35
25 to 55
1000,2000,
4000 Hz
0 to 13
20 to 37
28 to 55
2000,3000,
4000 Hz
-2 to 18
25 to 63
30 to 80
3000,4000,
6000 Hz
-2 to 18
27 to 90
37 to 93
-------
22
Table III. Age and educational level of the three groups.
N
Age (years)
Mean
SD
Range
Educational level
(Years)
Mean SD Range
Group
Group
Group
I
II
III
16
16
16
37.8
45.3
50.7
8.9
7.8
5.4
21
27
38
to
to
to
55
54
56
13.6
12.6
11.6
2.2
3.1
2.1
12
9
9
to 18
to 20
to 17
-------
23
(1975) found decreases in performance as both age and noise level in-
creased. Inspection of Table III shows that age levels of the three
groups in the present experiment were not equivalent. There was slightly
more than a decade between the ages of groups I and III. However, it
was expected that the contribution of aging would be minimal since the
speech materials selected were not as difficult as PB words, an age
limit was set in the mid-fifties, and there was some attempt to match
the groups for aging.
Speech Materials
Both sentence and monosyllabic materials were used in order to pro-
vide added means of comparison with other research, and so that the two
types of materials could be compared in similar test conditions. One of
these was a standardized form of the Revised Central Institute for the
Deaf (RCID) Sentences. The recorded form of these sentences is known as
the University of Maryland Test #1 (UM Test #1) (Elkins, Causey, Beck,
Brewer, and de Moll, 1975). The other was a test of monosyllabic discrim-
ination, the Modified Rhyme Test (MRT). Speech materials are presented in
Appendix C.
UM Test #1
Sentence material was selected because of its close relationship
to connected discourse (Giolas and Epstein, 1963; Giolas, 1966), and
because it is the form of speech material that the AAOO rule was sup-
posedly based on. Webster (1969) has recommended the use of sentences
for finding out how everyday speech will be heard, and Davis (1973) has
recommended the use of these particular (CID) sentences for validating
the AAOO assumptions.
The original CID sentences were developed by researchers at the
Central Institute for the Deaf in response to criteria set forth by the
Committee on Hearing and Bio-Acoustics (CHABA) of the National Research
Council. CHABA"s criteria were that the sentences should closely
resemble "everyday" speech in such parameters as vocabulary, sentence
length, grammatical structure, and redundancy (Silverman and Hirsh,
1955). The sentences that constitute the UM Test #1 were modified from
the original lists in order to achieve homogeneity of sentence length
by Harris, Haines, Kelsey, and Clack, (1961) to form the Revised CID sen-
tences, (RCID).
The UM Test #1 is a tape recording of the ten RCID lists of ten
sentences each, recorded by a male talker with general American speech.
The test was standardized on 100 normal-hearing subjects and 55 subjects.
with various degrees of sensori-neural hearing loss (Elkins et al, 1975).
The investigators found a fairly steep performance-intensity function,
resembling the curve for spondee words, and that the slope was steeper for
hearing-impaired than for normal-hearing subjects. The UM Test #1 was
originally recorded in the anechoic chamber of the University of Maryland's
-------
24
Biocoramunications Laboratory. The tape consists of ten lists of ten
sentences, each about seven words in length. Each list contains 50 key
words for which correct responses are scored. Each sentence is preceded
by the carrier phrase "Number ". The subjects respond by repeating
the whole sentence (minus the carrier phrase), or as much of it as is
perceived.
The tape of the UM Test #1 was re-recorded in order to juxtapose
desired lists, and a speech-shaped noise was recorded at the beginning of
the tape in order to serve as a calibration signal. A pilot study was
then conducted to examine the equivalency of the lists and to check for a
learning effect. (See discussion of Pilot Study #1 in Appendix A.)
MRT
The Modified Rhyme Test is an outgrowth of the Fairbanks Rhyme
Test (1958), which was modified by House, Williams, Hecker, and Kryter
(1965) in order to provide a multiple choice, closed-set test for the
purpose of assessing speech communication systems. It was modified
very slightly once more, recorded by three talkers, and mixed with three
levels of speech-shaped background noise in order to produce what is
known as the Modified Rhyme Hearing Test (MRHT) (Kreul, Nixon, Kryter,
Bell, Lang, and Schubert, 1968). The resulting test was intended pri-
marily for clinical purposes. According to Kreul et al it should be
"capable of rank-ordering patients according to their ability to dis-
criminate speech under 'everyday' listening conditions." Aside from
a few word changes in order to eliminate some objectionable words and
to reduce word redundancy, the lists are essentially the same as those
of House et al (1965).
There are several reasons why the MRT was selected. First, material
less redundant than the CID sentences was considered useful. Kryter
(1973) points out that monosyllabic material is sometimes characteristic
of everyday speech (the unfamiliar name, the important telephone number,
the unexpected message, etc.). Also, the MRT is noted for ease of
administration and scoring (Kryter and Whitman, 1965; Kreul et al, 1968),
it can be used with untrained listeners, and is reported as providing
fairly stable scores (House et al, 1965; Kreul et al, 1968; Nabelek and
Pickett, 1974a and 1974b). Additional advantages of the MRT are that it is
a closed-set task, eliminating problems of vocabulary familiarity, and
that the subject needs only to circle the selected response item. The
write-down feature was desirable since the subjects' responses to the
other speech materials were oral. Each form contains 50 sets of 6 words
each, one of which acts as a test word and the other 5 as foils. Thus,
there are 6 different test lists for each form.
For this experiment the Kreul tapes (MRHT) were not selected because
the background noise, which had been mixed with the speech signal, would
have prevented the use of any other level or type of masking stimulus.
Instead, eight lists of the Kreul version of the MRT were recorded in
quiet, by a male talker with general American speech and good voice
control. The recording procedures and carrier phrase were the same as
those described by Kreul et al (1968). The lists were recorded in a
sound-treated room, on pricTiTon quality tape, using a Sony TC-850 tape
-------
25
recorder with a Sony electret condenser microphone ECM-270. Each test
word was imbedded in the carrier phrase, "You will mark the ,
please." Each item was given six seconds: three seconds for the utter-
ance and three seconds' pause. Timing was monitored by an oscilloscope
trace that was recycled every six seconds, and voice level was monitored
by a VU meter. Seven practice sessions were conducted before the final
recordings were made.
All eight lists were employed in a pilot experiment on normal-hear-
ing subjects using a fixed speech-to-noise ratio. The four lists with
the closest means and least variability were selected for the final ex-
periment. (See discussion of Pilot Study #2 in Appendix A.) The tape
was then copied to include only the four selected lists and one practice
list. A speech-shaped noise was recorded at the beginning of the tape
to be used as a calibration signal.
Masking Noise
The masking noise used was a "babble" of twelve talkers. A tape
consisting of sound-on-sound recordings of six voices, three male and
three female, producing a babble of twelve voices, was supplied by Karl
Pearsons of Bolt, Beranek, and Newman, Inc. This tape was rerecorded so
as to provide a continuous 20-minute segment of babble, and a speech-shaped
calibration signal was applied to the beginning of the tape. The range of
level fluctuation of the babble was a maximum of 2 dB.
Babble was selected as a masker because of its lifelike quality
(subjects reported that it sounded like a party), and because of its
speech-like spectrum. A mixture of many voices was first used by Miller
(1947) who concluded that "the best place to hide a voice is among other
voices." Babble has been used as a masker also by Nabelek and Pickett
(1974a and 1974b), and by Miner and Canhauer (1976).
Relationship of Speech and Noise Stimuli
Speech Level
A review of the literature concerning "everyday" speech levels
reveals that the frequently-cited levels of 65 to 70 dB (unweighted
long-term rms level measured one meter from the talker) are not neces-
sarily typical of conversational speech. Gardner (1966) found that the
level of conversational speech in a "free-space" room was approximately 50
dB (B-weighted) and in a quiet office, 58 dBB. When subjects were asked
to read prepared text in the same conditions, their voice levels were 6 to
8 dB higher. Pearsons, Bennett, and Fidell (1976) found a similar pheno-
menon when subjects were asked to recite a memorized passage. Voice
levels at one meter in anechoic conditions were an average of 52 dBA for
"casual" (conversational) effort and 57 dBA for "normal" effort (reciting
prepared material). These levels would be approximately 3 dB higher when
-------
26
measured on the linear scale of a sound level meter (Kryter, 1970). It
appears, therefore, that conversational levels in quiet are approximately
52 to 60 dB (long-term rms level), depending on the acoustics of the room
and the conditions of the conversation.
Studies by Kryter (1946), Korn (1954), Pickett (1958), Webster and
Klumpp (1962), and Gardner (1966) show that individuals automatically
raise their voices as background noise levels increase. Kbrn postulates
an increase in speech level of 0.38 dB for every 1-dB increase in noise.
Webster and Klumpp found a 0.7 dB/dB increase, while both Kryter and
Pickett estimate a 0.3 dB/dB increase. Pearsons et al (1976) also studied
voice levels in noise, and made measurements in field as well as laboratory
conditions. Speech levels were measured at the listener's ear, which was
usually about one meter from the talker. The authors found average speech
levels of 57 dBA in urban homes and 55 dBA in suburban homes, with ambient
noise levels of 48 dBA and 41 dBA, respectively. They also found speech
levels of 61 dBA in department stores, 73 dBA in trains, and 77 dBA in
airplanes. By comparing speech levels with background noise levels the
authors concluded that individuals raise their voices about 0.6 dBA for
every 1-dBA increase in noise, from 48 dBA up to about 67 dBA, at which
point talkers and listeners move closer than one meter in order to maintain
intelligibility. Accordingly, the following levels of speech in noise
could be expected at the listener's ear:
Speech dBA 55 57 63 69
Noise dBA 41 48 58 68
These relationships could become less favorable if communicating dis-
tances were greater than one meter, if individuals failed to move closer
together at higher levels, or if the talker failed to raise his or her
voice.
A speech level of 60 dBA was selected for this study to reflect a
slightly raised conversational voice. This level could commonly occur
outdoors, inside department stores or inside urban homes (Pearsons et al,
1976).
Speech-to-Noise Ratio
The relationships between speech and noise in this study were
determined mainly by the location of the performance-intensity func-
tions of certain normal-hearing and hearing-impaired subjects (see discus-
sion of Pilot Study #3 in Appendix A). Speech-to-noise ratios were
selected so that the hearing-impaired subjects achieved discrimination
scores between approximately 20% and 80%. The normal-hearing pilot sub-
jects had lesser amounts of difficulty for the same conditions. Slightly
different speech-to-noise ratios were employed for the two materials since
the performance-intensity function of the UM Test #1 has been reported to
be somewhat steeper (Elkins et al, 1975) than that of the MET (House et
al, 1965). As a result of the pilot work, the speech-to-noise ratios
selected were 0, -3, and -6 dB for the MET and -1, -3, and -5 dB for
the UM Test #1. A quiet condition for each material was also included in
-------
27
order to provide a basis for comparison. (The residual sound level of
the unoccupied room was 19 dBA, so the speech-to-noise ratio for the quiet
condition was approximately +40 dB.)
These speech-to-noise ratios are not always typical of "everyday"
speech communication. They are not implausible, however, since any of
the conditions listed above could occur (conversationalists more than one
meter apart, one party fails to raise his voice, etc.). Negative speech-
to-noise ratios do occur on aircraft (Pearsons et al, 1976) and undoubtedly
occur in other noisy situations such as cocktail parties, city streets,
and factories.
Sound Field
The speech materials were presented in a mildly reverberant test
space. The reverberation time of the room was 0.625 second, as measured
with pink noise, (see section on measurement). This reverberation time
was not unlike everyday conditions. Nabelek and Pickett (1974a and 1974b)
noted that reverberation times of 0.5 to 1.0 second are typical of small-
to-medium sized rooms with hard-wearing, non-porous surfaces (such as
school rooms or cafeterias). Reverberation times used in their studies
were 0.3 and 0.6 second. In other speech intelligibility studies Harris
(1965) used a 5-second reverberation time, Millin (1968) used 0.45 sec.,
Acton (1970) used 0.3 second, MacKeith and Coles (1971) used 0.35 second,
Aniansson (1973) used 0.5 second, and Kuznairz (1973) used 0.2 second.
Nabelek and Pickett (1974b), and Bullock (1967), report that hearing-
impaired listeners are more sensitive to increased reverberation than
normal listeners. Also, subjective reports of hearing-impaired individuals
indicate an adverse effect of reverberation on speech discriminiation.
Listening Mode
Although binaural listening more closely resembles real life condi-
tions, a monaural, better ear condition was used for this study. Exper-
imental evidence shows that normal-hearing and even hearing-impaired
individuals experience a binaural advantage in noise, especially in
reverberant conditions (Moncur and Dirks, 1967; MacKeith and Coles,
1971; Nabelek and Pickett, 1974a and 1974b). Therefore, in assessing the
speech discrimination abilities of hearing-impaired subjects, there was
the danger that binaural hearing would influence discrimination scores to
an unknown and uncontrollable extent.
For this reason a pilot study was conducted on normal-hearing sub-
jects in the proposed experimental conditions. The rationale was that if
no binaural advantage were evident for normal listeners, it would be safe
to conclude that the hearing-impaired listeners, whose poorer ears would
contribute even less than if they had binaurally symmetrical hearing,
would not be affected. However, the results showed a binaural advantage
of 13% for normal listeners, and therefore the monaural condition was
selected (see discussion of Pilot Study #2). In order to occlude the
non-test ear, a combination of an earplug and a monaural earmuff was
utilized.
-------
28
Instrumentation
Equipment
All experimental tests were conducted in a double-walled Industrial
Acoustics Go. Model 1205-A sound-treated room that met the requirements
for background noise in audiometer rooms established by ANSI (1960).
Dimensions of the room were 9'10" wide by 9'2" long by 6'6" high. It
had been paneled throughout with formica-surfaced particle board in
order to create a mildly reverberant space.
Inside the test room subjects were seated in the middle of the room,
facing a corner. A two-way window was located to the subjects' right
so that they and the investigator could see each other easily. A Grason-
Stadler Model 162-4 loudspeaker was located on the floor in the corner,
the base of which was approximately 80 inches from the subject's head.
A KLH Model Six loudspeaker was located directly above the Grason-Stadler
speaker approximately 65 inches from the subject's head. The location
of each was marked with masking tape to assure a constant position
throughout the experiment.
Signal generating and monitoring equipment was located outside the
test room. The recorded noise and speech materials were each reproduced
on Kudelski Nagra Type D3 tape recorders, attenuated by Hewlett-Packard
350 C attenuators, and amplified by SWTP 207/A power amplifiers. The
speech signal was fed into the KLH loudspeaker and the babble into
the Grason-Stadler loudspeaker. A Balentine Model 320 true rms voltmeter
was situated so that it could be connected to the output of the tape
recorder, attenuator, or amplifier of either the speech or noise sys-
tem in order to monitor the voltages.
Hearing threshold tests were performed with a Grason-Stadler
Model 1701 diagnostic audiometer with TDH-49 earphones in MX/41 AR
cushions. Oral responses to the UM Test #1 were picked up by a Shure
lavalier-type microphone, routed through the Grason-Stadler audiometer,
and recorded by a Sony TC-850 tape recorder. Tney were also monitored
through a small loudspeaker in the Grason-Stadler audiometer. Both of
these systems provided adequate intelligibility of the subjects' re-
sponses, even at the lowest speech-to-noise ratio. A simplified block
diagram of the experimental equipment is shown in Figure 2.
Measurement
Pink noise was used to measure the reverberation of the test roan.
It was generated by a Hewlett-Packard 8057A precision noise generator,
and measured with a Bruel and Kjaer 4145 condenser microphone, and
a Bruel and Kjaer Type 2305 Level Recorder. The reverberation time of
the room was 0.625 second, measured according to the ANSI, Sl.l (1960)
definition: "The reverberation time of a room is the time that would
be required for the mean-square sound pressure level therein, originally
in a steady state, to decrease 60 dB after the source is stopped."
-------
29
NAGRATYPE 3
TAPE RECORDER
NAGRATYPE 3
TAPE RECORDER
H-P 350C
ATTENUATOR
H-P 350C
ATTENUATOR
SWTP
207/A
AMPLIFIER
SONY TC-850
TAPE RECORDER
SWTP
207/A
AMPLIFIER
G-S 1701
AUDIOMETER
BAND K
MEASURING
AMPLIFIER
SHURE
MICROPHONE
BAND K
MICROPHONE
O
O
Figure 2. Block diagram of equipment.
-------
30
Calibration signals were necessary to determine the level of the
speech and noise stimuli. In order to insure stable calibration signals
in the reverberant test room speech-shaped noise, generated by the Grason-
Stadler Model 1701 audiometer, was recorded onto the beginning of each
tape. It was then necessary to determine the level of the speech and
babble in relation to the calibration signals. The speech and noise,
including the calibration signals, were measured by an integrating sound
level meter or Noise Average Meter manufactured by Computer Engineering
Ltd. Since the instrument's display time interval was not fast enough to
measure each of the short speech segments, the durations of silence
between each stimulus segment were timed with a stop watch, and durations
of speech were calculated. Timing of the MET was almost exactly three
seconds of speech and three of silence. The UM Test #1 lists consisted of
20-22% speech. The Noise Average Meter's microphone was placed in the
sound field at the location of the listener's ear, and the A-weighted
sound level was then measured for the entire duration of each list.
Corrections to the observed levels were made according to the duration
of the speech. For example, 7 dB was added when speech accounted for
20% of the total, 6.7 dB when speech accounted for 21%, and 6.5 dB
when speech accounted for 22%. These timing procedures were unnecessary
in measuring the levels of the babble and the calibration noises because
the continuous seaments were sufficiently long for the Noise Average Meter
to handle.
Calibration
Acoustic calibration of both the speech and noise systems was per-
formed before and after testing each subject and at the end of each day.
For this purpose a Bruel and Kjaer 4145 condenser microphone was placed in
a position corresponding to the center of the subject's head. In order
to assure that the same position was always used, a plumb bob was dropped
fron the microphone to a spot marked on the floor. During the calibra-
tion the subject's chair (the only furniture in the test room) was re-
moved and the door was closed. The condenser microphone was calibrated
at the beginning and end of each day with a Bruel and Kjaer 4230
microphone calibrator and the calibration of both the speech and noise
systems was checked with a Bruel and Kjaer Type 2606 measuring amplifier.
Calibration checks revealed that speech and noise levels were almost
always within _+0.5 dB of the desired levels. Data were eliminated when
calibration of either system was more than 1 dB off (which happened in
two cases) .
Audiometric calibration was performed daily in accordance with the
requirements of the ANSI S3.6 (1969) specifications for audiometers.
Hearing Protective Devices
In order to achieve a monaural condition, or at least to minimize
any binaural contribution, a V-51R earplug was inserted by the investi-
gator in the subject's poorer ear. Four sizes of the plug were avail-
-------
31
able, and the plugs were fitted according to the procedures recommended
by Guild (1966). In order to achieve maximum attenuation a set of David
Clark 117 earmuffs was rendered monaural and the remaining muff was
placed over the plugged ear. Table IV lists attenuation values for the
V-51R earplug, the David Clark 117 earmuffs, and the combination of
the two devices. These data represent mean values minus one standard
deviation, and were supplied by the U.S. Air Force (1973).
Table IV. Attenuation values in dB for hearing protectors
as reported by U.S. Air Force (1973). Data represent mean values
minus one standard deviation.
Frequency
Hearing
Protector
David Clark 117
V-51R
Combined
125
13
22
29
250
21
20
36
500
36
22
35
Ik
37
24
36
in Hz
2k
33
33
38
3k
34
34
48
4k
35
29
52
6k
33
33
46
8k
39
31
38
Method of Presentation
All subjects were instructed about the general purpose of the ex-
periment, and the length and nature of the test session. They were then
asked to read and sign a standard consent form used by the U.S. Air Force,
Aerospace Medical Research Laboratory (See Appendix B.).
Air conduction and bone conduction audiometry were performed ac-
cording to the modified Hughson-Westlake technique described by Newby
(1964). Subjects with differences of more than 10 dB between air con-
duction and bone conduction thresholds were not used in the study.
The poorer ear was then occluded with an earplug and earmuff, and
the subject was seated in the test room. The floor had been marked with
masking tape to ensure a constant position for the subject's chair. In
order to prevent the subject from turning the unoccluded ear toward the
loudspeakers a plumb bob was adjusted to be about two inches in front of
the subject's nose. The subject was asked not to touch it, and to glance
at it periodically to make sure that it was in the center of his or her
line of vision. Standard instructions were then read aloud by the exper-
imenter while the subject read them silently (see Appendix C.) Two prac-
tice lists were administered, one for each type of material at speech-to-
rat ios of -1 and -3 dB. After the practice session the subjects were given
a short rest, and then instructed about the remainder of the test. In
the middle of the experiment they were given a ten-minute rest between
administration of the two speech materials. The total test lasted
approximately two hours.
-------
32
During the main experiment the speech-to-noise ratios were changed
by adjusting the attenuator of the noise system. This was done between
each list (or pair of UM Test #1 lists) according to a predetermined
program for each experimental subject. The order of difficulty (speech-
to-noise ratio of -1 or -3) of the practice lists was counterbalanced
across subjects, as was the order of presentation of the experimental
speech materials. The order of lists within each type of material was
also counterbalanced, and the order of speech-to-noise ratio was pseudo-
randomized so as to minimize the chance occurrence among subjects of the
same speech-to-noise ratio with the same list.
Subjects responded to the MET by selecting the correct response from
one of six alternatives for each test item. They responded to the UM
Test #1 by repeating each sentence immediately after it was presented.
During the instructions all subjects were encouraged to guess. They
were told to complete every item on the MET, even though they were not
sure of the stimulus word. For the UM Test #1 they were told to repeat
those words they did hear, even though they might not have heard the
entire sentence, and always to guess when unsure.
Responses to the UM Test #1 were monitored by the investigator
through the talkback system of the Grason-Stadler audiometer, and were
recorded by a Sony TC-850 tape recorder. Later these responses were
scored by two other listeners, so as to minimize the possibility of error
on the part of a single listener. Since product-moment correlations of
the scorings of the investigator with each of the two other listeners were
0.995 and 0.990, the investigator's scorings were used.
Analysis of the Data
Speech discrimination scores in percentage correct were tabulated
for all of the experimental data. The data were then subjected to a
three-factor analysis of variance using a repeated measures model (Winer,
1971). Simple main effects were examined to clarify certain interactions,
and Newman-Kuels tests were performed on pairs of means. The parameters
studied were hearing loss (according to groups), speech-to-noise ratio,
and type of speech material.
Pearson product-moment correlations were performed in order to test
the predictive capabilities of different combinations of frequencies.
A correlational matrix was prepared which related hearing loss in the
following combinations of frequencies to speech discrimination scores:
500, 1000, and 2000 Hz; 500, 1000, 2000, and 3000 Hz; 1000, 2000 and
3000 Hz; 1000, 2000, and 4000 Hz; 2000, 3000, and 4000 Hz; and 3000,
4000, and 6000 Hz. The results of these analyses are discussed in the
next chapter.
-------
33
Corrections for Guessing
In order to minimize the occurrence of correct responses due to
chance the MRT was corrected for guessing as recommended by Kryter
(1972) and in the proposed ISO standard for measuring the intelligibility
of speech (1975):
I in % = 100 R - W
T N-l
where I = intelligibility
T = number of items in the test
N = number of alternatives (6 for MRT)
R = number right
W = number wrong
Such a correction was not considered appropriate for the UM Test #1,
since the number of alternatives was so large, and virtually unknown.
When this correction had been performed the mean discrimination scores
for the two speech materials became somewhat more similar to each other for
all conditions except quiet. Table V shows the mean discrimination scores
for the MRT before and after correcting for guessing. Mean scores for the
UM Test #1 are shown for comparison.
Three-Factor Design
Slightly different speech-to-noise ratios were used for the two
speech materials because of the differences between the shape of their
performance-intensity functions. Because of these differences it was
originally planned that the data for each speech material would be
analyzed separately in two two-factor analyses. However, on inspection
of the mean data for each speech-to-noise ratio in each group, it was
decided that the speech-to-noise ratios could be combined into four
levels, thus facilitating the study of speech materials in a three-
factor analysis of variance. They were combined as follows:
Speech-to-Noise Ratio (dB)
UM Test #1 Quiet -1 -3 -5
MRT Quiet 0 -3 -6
Combined Quiet 0-1 -3 -5-6
-------
Table V. Mean discrimination scores in percent correct
for MRT before and after corrections for guessing, and for DM Test # 1.
MRT before MRT after
Correction Correction
S/N in dB Q 0-3-6 Q 0 -3 -6 Q -1 -3 -5
Group I 94 86.8 80.1 69.1 92.8 84.1 76.1 62.9 99.6 87.7 78 57.9
Group II 89.1 69 59.4 45 87 62.8 51.3 34 97.8 67.7 50.8 28.6
Group III 82.8 61.1 43.1 32.4 79.3 53.4 31.8 19.7 93.9 53.2 36.8 15.3
S/N = Speech-to-noise ratio
Q = Quiet
OJ
-------
CHAPTER IV
RESULTS
This chapter will present the results of the analysis of the data.
The outcome of the analysis of variance will be shown, as well as the
tests of simple main effects and the Newman-Keuls procedures. Two
additional procedures will be presented—a series of correlations
between combinations of audiometric frequencies and discrimination
scores, and an ad hoc partitioning of the groups according to high-
frequency, rather than mid-frequency hearing acuity. Mean scores and
standard deviations in terms of percent correct responses are shown in
Table VI. Data are given for the three groups of subjects on each of
the two speech materials in four speech-to-noise ratios.
Analysis of Variance
The data were subjected to a three-factor analysis of variance
using a repeated measures model (Winer, 1971). Tne three main effects
studied were hearing loss, speech-to-noise ratio, and speech materials.
There were three levels of hearing loss, represented by groups I, II
and III; four levels of speech-to-noise ratio: quiet, 0 and -1 dB
combined, -3 dB, and -5 and -6 dB combined; and two speech materials,
the LM Test #1 and the MRT. The dependent variable was discrimination
score in percent correct responses. The level of significance that was
determined for the study was the .05 level of confidence. However, the
.01 level was reported when it occurred. The results of the analysis
displayed in Table VII show that differences in discrimination scores
for the main effects of hearing loss and speech-to-noise ratio were
significant at the .01 level of confidence (hearing loss F = 37.9, df
2,45; and speech-to-noise ratio F = 630, df 3,135). The difference in
discrimination scores due to the main effect of speech materials was
not significant at the .01 level but was significant at the .05
level of confidence (F = 5.47, df 1,45). Tne interaction of hearing
loss and speech-to-noise ratio was significant at the .01 level (F =
25.7, df 6,135), and the interaction between speech materials and
speech-to-noise ratio was significant at the .01 level (F = 16.6, df
3,135). The interaction between speech materials and hearing loss, and
the three-way interaction between speech materials, hearing loss, and
speech-to-noise ratio were not significant.
Since two of the four interactions were statistically significant,
tests on the simple main effects of the hearing loss and speech materials
factors were performed (according to Winer, 1971) in order to examine
their interactions with speech-to-noise ratio. The results of these
tests and tests using the Newnan-Keuls procedure for assessing the
difference between ordered means (Winer, 1971) are displayed in Table
VIII. Examination of the hearing loss factor as it interacted with
speech-to-noise ratio revealed significant differences among the three
35
-------
36
Table VI.
responses.
Mean scores and standard deviations in percent correct
Group I
Group II
Group III
S/N
Mean
SD
S/N
Mean
SD
S/N
Mean
SD
S/N
Mean
SD
S/N
Mean
SD
S/N
Mean
SD
Q
92.8
5.0
Q
99.6
0.5
Q
87.0
6.3
Q
97.8
2.6
Q
79.3
9.1
Q
93.9
7.0
MET*
0
84.1
4.1
UM#1
-1
88.2
6.4
MRT*
0
62.8
12.4
UM#1
-1
67.7
15.4
MRT*
0
53.4
16.3
UM#1
-1
53.2
22.2
-3
76.2
7.0
-3
78.1
8.2
-3
51.3
14.3
-3
50.8
18.6
-3
31.8
18.3
-3
36.8
20.7
-6
63.0
9.4
-5
58.4
14.3
-6
34.0
14.0
-5
28.6
16.4
-6
19.7
21.1
-5
15.3
13.4
S/N = Speech-to-noise ratio in dB
*MRT scores have been corrected for guessing,
-------
Table VII. Summary of analysis of variance to determine effects of hearing
loss, speech-to-noise ratio and speech materials on speech discrimination scores.
Source of Variation
Between Subjects
Factor A Hearing loss
Error A (Subjects within groups)
Within Subjects
Factor B S/N
AB Hearing Loss x S/N
Error B (B x Subjects within groups)
Factor C Speech Materials
AC Hearing Loss x Speech Materials
Error C (C x Subjects within groups)
BC SAl x Speech Materials
ABC Hearing Loss x S/N x Speech Materials
Error BC (BC x Subjects within groups)
Sum of
Squares
67,473
40,078
156,971
12,793
11,214
738
55
6,054
2,913
435
7,897
df
2
45
3
6
135
1
2
45
3
6
135
Mean
Squares
33,737
891
52,324
2,132
83
738
27.5
135
971
72.5
58.5
F
37.9**
630**
25.7**
5.47*
.20
16.6**
1.24
S/N = Speech-to-noise ratio
* Significant at .05 level of confidence.
** Significant at .01 level of confidence.
-------
38
Table VIII. Tests on significant interactions of hearing loss with
speech-to-noise ratio, and speech materials with speech-to-noise ratio.
Simple Main Effects
Hearing loss
Quiet
S/N 0-1 dB
S/N -3 dB
S/N -5-6 dB
df 2,180
df 2,180
df 2,180
df 2,180
Speech materials
Quiet
S/N 0-1 dB
S/N -3 dB
S/N -5-6 dB
df 1,180
df 1,180
df 1,180
df 1,180
F = 7.93 **
F = 32.10 **
F = 53.14 **
F = 55.78 **
F = 35.9 **
F = 2.63
F = 1.51
F = 7.03 **
Newman - Keuls
Quiet
Difference in % correct
between means of groups
S/N 0-1 dB
Difference in % correct
between means of groups
S/N -3 dB
Difference in % correct
between means of groups
S/N -5-6 dB
Difference in % correct
between means of groups
I and II
I and III
II and III
I and II
I and III
II and III
I and II
I and III
II and III
I and II
I and III
II and III
3.8
9.6 **
5.8 *
20.6 **
32.6 **
12 **
26 **
42.7 **
16.7 **
29.1 **
42.9 **
13.8 **
S/N = Speech-to-noise ratio
* Significant at the .05 level of confidence.
** Significant at the .01 level of confidence.
-------
39
groups, even in the quiet condition. These differences were significant
at the .01 level of confidence for all four speech-to-noise ratios and
they became more pronounced as speech-to-noise ratios decreased.
Further probing using the Newman-Keuls technique showed that the means
of all three groups were significantly different from each other at the
.01 level of confidence in the three noise conditions. In the quiet
condition groups I and III were significantly different at the .01
level, also groups II and III at the .05 level, but not groups I and
II. The interaction between hearing loss and speech-to-noise ratio is
graphically displayed in Figure 3. Since the increment in noise level
between quiet and the first noise condition is much greater than those
between the three noise conditions the curves have been broken between
quiet and the speech-to-noise ratio of 0-1 dB.
Examination of the speech materials factor as it interacted with
speech-to-noise ratio revealed differences significant at the .01
level of confidence between the two materials in the quiet condition
with higher scores on the UM Test #1, and in the most difficult noise
condition with higher scores on the MRT. Since only two means were
involved the Newman-Keuls procedure was not performed. The interaction
between the speech materials and speech-to-noise ratio factors is
graphically displayed in Figure 4. Because neither the interaction
between hearing loss and speech materials, nor the three-way interaction
between hearing loss, speech-to-noise ratio, and speech materials was
significant, additional tests were not performed.
Correlational Tests
In order to determine which frequencies, or groups of frequencies,
best predicted speech discrimination scores, groups II and III were
combined. Each subject's audiogram was divided into various frequency
combinations (500, 1000, and 2000 Hz; 500, 1000, 2000, and 3000 Hz;
1000, 2000, and 3000 Hz; 2000, 3000, and 4000 Hz; 3000, 4000, and 6000
Hz). For each subject each of these combinations was correlated
with the subject's speech discrimination scores (averaged across the
two speech materials) for the four experimental conditions, and with a
composite discrimination score obtained by averaging across the two
materials and the four conditions. Pearson product-moment correlations
were calculated, which are displayed in a correlational matrix in Table
IX. Most of the different audiometric combinations show high positive
correlations with each other, which is to be expected. All of the
combinations show negative correlations that are significantly different
from zero with the discrimination scores resulting from all of the con-
ditions tested, indicating that the greater the hearing loss, the lower
the discrimination score. Those combinations that include frequencies
above 2000 Hz show particularly high correlations with discrimination
scores, especially in the higher levels of background noise.
Since all of the frequency combinations showed correlations with
discrimination scores that were significantly different from zero, it
was necessary to explore the relationships of the correlations to each
-------
40
1
o
a
ifi
100-H
90 -\
80 H
70 H
60 H
50 H
40 H
30-^
20 H
Group I
Quiet
0-1
-3
-5-6
Speech-to-Noise Ratio (dB)
Figure .3. Mean percent correct responses of the three groups as a function
of speech-to-noise ratio. Scores are averaged across the two speech
materials.
-------
CA
0)
CC.
o
o
*^
100-
90-
80-
70-
60-
50-
40-
30-
20-
10-
41
MRT
Quiet
0-1
-3
-5-6
Speech-to-Noise Ratio (dB)
Figure 4. Mean percent correct response in the two speech materials as a
function of speech-to-noise ratio. Scores are averaged across the
three groups.
-------
Table IX. Correlational matrix among six combinations of audiometric frequencies and
discrimination scores in various speech-to-noise ratios. Data are for subjects in groups II and III.
Scores Averaged Across
.5,1,2 .5,1,2,3 T7I73 T7Y7I 17174" 17176 S/N dB Q 0-1 -3 -5-6 All Conditions
.5J72 .86 .84 .73 .57 .32+ -.53 -.48 -.55 -.51 -.54
.5,1,2,3 .99 .89 .89 .69 -.62 -.74 -.77 -.75 .77
1,2,3 -89 .90 .67 -.64 -.77 -.79 -.76 -.80
j^j .94 .82 -.76 -.81 -.85 -.82 -.87
2^74 .90 -.65 -.83 -.85 -.84 -.86
^5-j-g- -.56 -.79 -.81 -.82 -.82
3,4,6
+ All correlations are significant at the .05 level of confidence except the correlation of
.5,1,2 with 3,4,6.
.5,1,2 = Average of audiometric frequencies 500, 1000 and 2000 Hz.
S/N = Speech-to-noise ratio
Q = Quiet
-------
43
other. The Hotelling formula (Guilford, 1965) was used in order
to test the significance of differences between the correlations.
The outcome is in the form of a t-ratio that takes into account the
correlation of frequency combinations with each other, as well as
the correlation of frequency combinations with discrimination scores.
The matrix in Table X displays the significance of differences among
correlations of frequency combinations with the composite discrimi-
nation scores (the last correlation displayed in Table IX). According
to the Hotelling procedure significant differences exist between the
500, 1000, and 2000 Hz combination and all of the other combinations
employing higher frequencies. The other frequency combinations are not
significantly different from each other with the exception of 500,
1000, 2000, and 3000 Hz in comparison with 1000, 2000, and 4000 Hz.
It can be concluded that for the present experimental conditions, the
average of 500, 1000, and 2000 Hz is the poorest predictor of speech
discrimination, while the other combinations are about equally efficient,
with 1000, 2000, and 4000 Hz appearing to be slightly superior in
quiet, and 2000, 3000, and 4000 Hz a better predictor in noise.
Further Ad Hoc Comparisons
During the experiment it had become evident that there were
some members of Group II whose discrimination scores were poorer than
certain members of Group III. High-frequency hearing acuity appeared
to be the critical factor. Therefore groups II and III were divided on
the basis of high-frequency, rather than mid-frequency hearing levels
to see if the differences in discrimination scores increased.
Groups II and III were combined and then partitioned according to
whether subjects' thresholds were better or worse than the median
hearing level (47 dB) for the averaged frequencies 2000, 3000, and 4000
Hz. Those subjects whose average hearing levels fell on the median
were eliminated, leaving a total number of 29 subjects, 14 of whom had
hearing levels better than median (Group Y) and 15 worse (Group Z) .
Mean discrimination scores of these two groups as a function of speech-
to-noise ratio are shown as solid lines in Figure 5. Mean scores for
groups I, II, and III are shown by dotted lines for comparison. The
differences between the newly partitioned groups are greater than when
the groups are divided by average hearing level at 500, 1000, and 2000
Hz.
The causes and implications of the results presented above will
be discussed in Chapter V.
-------
44
Table X. Matrix displaying significance of differences among
correlations of various frequency combinations and speech discrimination
scores averaged over the four experimental conditions.
"27374
.57T72 .5,1,2,3 T7273 1,2,4 2,3,4 3,4,6
.•57172 3.96** 4.51** 5.11** 3.67** 2.63"
.5,1,2,3 1.97 2.33* 2.03 .69
1/273- 1.64 1.43 .29
1,2,4 .33 .98
* Significant at the .05 level of confidence.
** Significant at the .01 level of confidence.
.5,1,2 = Average of audiometric frequencies 500, 1000, and 2000 Hz (etc.)
-------
100-
90-
80-
70-
(/> 60-
I
O
a
QC
t»
o
O
4->
§
£.
50-
40-
30-
20-
10-
45
Group I
\
Quiet
\
0-1 -3
Speech-to-Noise Ratio (dB)
-5-6
Figure 5. Mean speech discrimination scores in percent correct as a func-
tion of speech-to-noise ratio. Groups II and III have been combined
and partitioned according to whether subjects have more or less hearing
loss than the median (47 dB) at the average of 2000, 3000, and 4000 Hz.
Group Y = <47 dB, Group Z = >47 dB.
-------
CHAPTER V
DISCUSSION
The results that were presented in the previous chapter will be
discussed in the following pages. The discussion will include infer-
ences about the causes and implications of the results. The statistical
procedures, the analysis of variance and correlational tests will be
discussed, along with the implications of partitioning groups II and
III according to their high-frequency, rather than mid-frequency
hearing levels. Tne results of the present study will then be conpared
to the those of other investigations. Comparisons will be made of
differences between groups, the effect of decreasing speech-to-noise
ratio, the similarity of scores for the two speech materials and the
results of correlational tests. The final sections will deal with the
results of the present study with respect to the monaural condition,
the term "speech frequencies" as it has been used traditionally, and
the concept of an appropriate "low fence" or point of beginning handicap.
Statistical Analysis
Analysis of Variance
Tne present study examined speech discrimination scores in noise
backgrounds of forty-eight individuals whose mean average hearing
levels at 500, 1000, and 2000 Hz ranged from -3dB to 28 dB. They were
divided into three groups with mean average hearing levels as follows:
Group I 2.5 dB
Group II 13.4 dB
Group III 24.7 dB
Tne results showed large differences in speech discrimination ability
among subjects whose hearing was within the range of "no impairment" as
defined by the AAGO. Differences in mean discrimination scores between
groups I and II ranged from 20.9% to 29.4% and differences between groups
I and III ranged from 32.9% to 43.2%, in the noise conditions. Mean
speech discrimination scores of each group were statistically different
from each other, making it impossible to conclude that the subjects in
groups II and III could understand speech as "normally" as those in
Group I, especially in a background of noise. The only condition in
which all of the group means were not significantly different from each
other was quiet, where the difference between mean scores of groups I
and II was not statistically significant. However, even in quiet
groups I and II, and II and III were significantly different. These
findings indicate that differences between individuals traditionally
46
-------
47
categorized as unimpaired may appear even under listening conditions
that are favorable. The differences between groups increased markedly
as listening conditions become less favorable, indicating that the
introduction of noise exacerbated the effects of hearing loss.
Differences between groups and between speech materials may have
been underestimated due to truncation effects in the quiet and -5-6
dB conditions. Eighteen subjects scored 100% correct on the sentences
in the quiet condition, and five subjects (all from Group III) scored
0% on the UM Test #1, four of whom scored 0% on the MET, in the -5-6
dB condition.
Differences between discrimination scores for the two speech
materials were fairly small, especially after the scores on the MET
had been corrected for guessing. However, the differences were
significant at the .01 level of confidence in quiet and in the most
difficult noise condition, with subjects scoring higher on the UM Test
#1 in the former, and lower on the UM Test #1 in the latter condition.
Since the interaction between hearing loss and speech materials was
not significant, it is assumed that the differences between groups
were not dependent upon speech materials, and the differences between
speech materials were not a function of hearing loss. Also, since the
three-way interaction between hearing loss, speech materials and
speech-to-noise ratio was not significant, it is assumed that neither
the hearing loss nor the speech materials factors were dependent upon
an interaction of the other variable with speech-to-noise ratio.
The fact that the two speech materials were presented at slightly
different speech-to-noise ratios does not explain the differences in
mean scores of the two materials in the -5-6 dB condition. This
procedure probably decreased rather than increased the difference in
the -5-6 dB condition, and may have masked a difference in the 0-1 dB
condition. Since the MET was presented at a speech-to-noise ratio of 0
dB, it can be assumed that the scores would have been slightly poorer
at -1 dB, where the UM Test #1 was presented. Likewise, since the UM
Test #1 was presented at a speech-to-noise ratio of -5 dB, it can be
assumed that the scores would have been slightly poorer if it had been
presented at -6 dB as was the MPT.
These results imply that the two materials could be considered
equivalent only in the -3 dB condition. However, the lack of interaction
between the hearing loss and speech materials factors indicates that
the two materials did not differ in their ability to elicit differences
between the three groups.
Other Tests on the Data
Efearson product-moment correlations were calculated to assess
the abilities of various frequency-averaging methods to predict speech
discrimination scores. The combinations of 1000, 2000, and 4000 Hz and
2000, 3000, and 4000 Hz showed the highest correlation with discrimination
ability (or more rightly disability, since the correlations were negative)
-------
48
It appeared that the average of 1000, 2000, and 4000 Hz was a slightly
better predictor in the quiet condition, while the average of 2000,
3000, and 4000 Hz was slightly better in the noise conditions. All of
the combinations that included frequencies above 2000 Hz showed high
correlations with speech discrimination that were not significantly
different from each other, with the exception of 1000, 2000, and 4000
Hz, which was a significantly better predictor than 500, 1000, 2000,
and 3000 Hz. However, all of the combinations tested were significantly
better than 500, 1000, and 2000 Hz. At least for the present experimental
population and conditions the combinations of 1000, 2000, and 4000 Hz and
2000, 3000, and 4000 Hz are slightly superior to the others, especially
for predicting speech discrimination scores in a background of noise.
It is reasonable to conclude that these frequency combinations are
superior to the traditionally-used 500, 1000, and 2000 Hz for describing
a population with high-frequency hearing loss.
In order to assess further the contribution of high-frequency
hearing acuity groups II and III were combined, and then partitioned
according to hearing levels that were greater or less than the median
hearing level (47 dB) at the averaged frequencies 2000, 3000, and
4000 Hz. This procedure revealed differences of approximately 30
percentage points between the two groups in all but the quiet condition.
These differences are much greater than those between the original
groups II and III, divided according to average hearing level at 500,
1000, and 2000 Hz. The magnitude of these differences suggests that a
more appropriate method of grouping the hearing-impaired subjects would
have been according to hearing level in the higher frequencies, rather
than the method followed, which was based on average hearing levels at
500, 1000, and 2000 Hz.
Comparison of Results with Other Studies
Differences Among Groups
The division of hearing-impaired subjects into groups resembles
the studies of Acton (1970) and Kuzniarz (1973). The designs of the
three studies are roughly analogous, with some exceptions. Acton and
Kuzniarz each used three hearing-impaired groups instead of the present
study's two. The two more severely-impaired groups had hearing levels
that were slightly greater than the present study's groups II and III
respectively. Other differences are that both Acton and Kuzniarz used
PB monosyllables instead of the MRT, they did not employ sentences
(although Kuzniarz did use sentences in another part of his experiment),
and Kuzniarz' speech materials were in Polish. In spite of these
differences, the overall results of the present study are quite con-
sistent with those of Acton and Kuzniarz in most respects. Clear
differences were evidenced between hearing-impaired and normal listeners,
and among the various experimental groups. The only exception was
Acton's Group A, whose hearing levels were between those of Group I
and Group II of the present study. Acton's Group A performed nearly
as well as his control group in speech-to-noise ratios of +10 dB and
-------
49
above, and they surpassed the control group in less favorable speech-to-
noise ratios. The author attributes the phenomenon to "conditioning"
that may have occurred in Group A's experience in noisy industrial
environments.
The scores of all subjects in the present experiment are slightly
higher than those of the other two investigators, which is somewhat
surprising since the monaural condition was employed in the present
study but not in the others. The most likely explanation is the fact
that the other investigators used PB monosyllables instead of the MET
and sentences. According to Kryter and Whitman (1965), the difference
between PB words with a large number of alternatives (1000) and the
MRT is about 20 percent at a speech-to-noise rai:io of 0 dB. The ANSI
S3.5 (1969) standard for calculation of the Articulation Index (AI)
shows a difference of 25 to 30 percent between 1000 Harvard PB words
and rhyme words and sentences, respectively at an AI of 0.4, which is
a speech-to-noise ratio of approximately 0 dB in the present experiment.
Similarity between Acton's PBs (Fry's monosyllabic word lists), Kuzniarz'
PBs (in Polish), and the Harvard word lists, can only be assumed since
the author knows of no comparative data. Other explanations for the
differences in scores between the present study and those of Acton and
Kuzniarz could be in the slight differences in hearing levels between
analogous groups, possible differences in the methods of measuring the
speech signals, and in the masking capabilities of the different
noises, (although the spectra of the three noises appear to be fairly
similar). Aside from the differences in scores, certain common
conclusions can be drawn for the three studies, namely that there are
clear differences between the discrimination scores of individuals
that traditionally have been considered either mildly impaired or
unimpaired and those with truly normal hearing (as defined by these
investigators) , and that high frequency hearing acuity plays an
integral part in defining these differences.
Effect of Speech-to-Noise Ratio
Other investigations as well as the present one have shown increas-
ing differences between normal-hearing and hearing-impaired groups as
speech-to-noise ratio decreases. The scores of Acton's (1970) hearing-
impaired groups moved further apart as speech-to-noise ratio decreased
from +20 dB toward 0 dB, but the control group's scores remained
between experimental groups A and B. Kuzniarz (1973) found increasing
differences between groups as the speech-to-noise ratio grew more
difficult, and also Dickman (1974) found that differences between the
normal and hearing-impaired groups increased as they progressed from
"easy" to "difficult" speech-to-noise ratios for different types of
noises.
Both Ross et al (1965) and Elkins (1971) found that hearing-
impaired individuals performed more poorly than normal-hearing subjects
in noise backgrounds, but did not find greater differences as noise
levels increased. In fact, for certain conditions the relative shift
-------
50
for hearing-impaired subjects was smaller than it was for normals. The
most likely explanation for this phenomenon is that these investigators
presented the stimuli at 40 dB above each subject's speech reception
threshold, whereas the other investigators (Acton, Kuzniarz, Dickman,
and the present author) presented the stimuli at the same level for all
subjects. (See discussion of the effects of this procedure in Chapter
II).
Another explanation may involve the introduction of reverberation.
According to Sabine (1950), the effect of reverberation is to cause
an overlapping or blurring of speech segments by their predecessors in
time. This effect is achieved by the reflection of sound by hard
surfaces. The faster the sound decays (or the shorter the reverberation
time) the less likely is this blurring effect to occur. Farther
complicating the situation is the fact that the signal consists of both
direct and reflected components. Bullock (1967), and Nabelek and
Pickett (1974a) have shown that hearing-impaired listeners are more
sensitive to increased reverberation than are normal listeners.
Nabelek and Pickett (1974a) hypothesized on the basis of their data
that hearing-impaired listeners are not as efficient as their normal-
hearing counterparts at integrating delayed speech reflections with the
direct sound. It is possible that increasing the level of the babble
in relation to the speech signal produced a greater degree of reflected
speech-like sound, which in turn increased the difficulty of hearing-
impaired subjects in selecting the desired signal from the unwanted
background.
Similarity of Speech Materials
The fact that there were only small differences between the two
types of speech materials is not surprising. The ANSI S3.5 (1969)
standard for calculating the Articulation Index contains curves
for the two materials that are very similar. In the ANSI figure
sentence scores are about 5% higher than scores on rhyme tests for
more favorable AI values. The curves overlap at an AI of about 0.3 and
below this point scores on the rhyme test are slightly higher for
comparable AI values. The relationship between the two materials
(shown in Figure 4) is comparable to the relationship between the
curves shown in the ANSI standard.
Comparison with Other Correlational Studies
VJien a variety of frequency averages were correlated with speech
discrimination scores, the results showed the importance of high-
frequency hearing. The averages of 1000, 2000, and 4000 Hz and 2000,
3000, and 4000 Hz appeared to be the most efficient predictors of
speech discrimination, and all of the combinations that included
frequencies above 2000 Hz showed significantly higher correlations
than the average of 500, 1000, and 2000 Hz.
-------
These findings are not consistent with some of the earlier cor-
relational studies, such as those of Carhart (1946), Harris et al
(1956) and Quiggle et al (1957). However, as mentioned in Chapter II,
these studies were conducted in quiet conditions. Also, "hearing for
speech" was defined in terms of speech reception thresholds rather
than discrimination scores. Carhart (1946) and Quiggle et al (1957)
compared pure-tone hearing acuity to speech reception thresholds for
spondee words. It is not surprising that the lower frequencies (500,
1000, and 1500 Hz) were the ones that correlated most highly with
speech reception thresholds, since spondee words are primarly dependent
upon vowels and minimally dependent upon consonants for intelligibility.
Also of interest is the fact that Quiggle et al (1957) based their
conclusions on random samples drawn from the Wisconsin State Fair
of 1954, rather than from hearing-impaired populations. Harris et
al (1956) developed a multiple regression formula that was based
on speech reception thresholds for PB words, (the level at which
50% were identified correctly), and therefore consonant energy was a
more important factor than it was with spondee tests. Interestingly,
the investigators found that the most important frequencies were 1000,
2000, 4000, 500, and 6000 Hz in that order. The resulting regression
formula was found to be the best of a variety of formulae for predicting
hearing loss for speech in patients with sloping losses (20 dB or more
difference between 500 and 2000 Hz). But the best predictor for all
types of hearing loss was determined to be an adjusted version of the
500, 1000, and 2000 Hz average.
The present study's findings are more consistent with the results
of other investigators, when speech discrimination score rather than
speech reception threshold was the dependent variable. Mullins and
Bangs (1957), although they used quiet instead of a noise background,
found that 2000 and 3000 Hz were the audiometric frequencies that
best predicted speech discrimination scores. Harris (1965) found that
the frequency region of 2000-4000 Hz was the most important for
understanding distorted speech (without noise in the background).
However, since everyday speech is not always distorted, the authors
concluded that the average of 1500, 2000, and 3000 Hz would be the
best predictor of discrimination scores when the speech was distorted
about 50% of the time. Harris1 study was criticized by Webster
et al (1965) for distorting the material so heavily, and then for
presenting it at 40 dB above the subjects' speech reception thresholds.
These procedures might respectively increase or decrease the importance
of high-frequency hearing in correlational analyses.
Results of correlations performed on data that had been gathered
in backgrounds of noise have been inconclusive. Ross et al (1965)
found that 500 Hz correlated significantly with "relative discrimination
shift" in noise, but poorly with speech discrimination in quiet, while
the opposite was true of 4000 Hz. The authors suggested that one
reason for this unexpected finding may have been the fact that the
hear ing-impaired subjects already had poor discrimination scores
in quiet, and therefore the introduction of masking did not produce a
very large relative discrimination shift. Another explanation may be
the fact that the experimenters presented the stimuli at a level of
-------
52
40 dB above speech reception threshold, the implications of which
practice have been discussed earlier*
These explanations may also apply to the studies by Elkins
(1971), Murry and Lacroix (1972), and Myers and Angermeier (1972), all
of which used the Modified Rhyme Hearing Test (MRHT) and presented the
stimuli at 40 dB above speech reception threshold. In Elkins1 study
the relative discrimination shift for hearing-impaired listeners was
smaller than expected. Correlations of all audiometric frequencies
with speech discrimination in noise were fairly low, but the individual
frequencies 2000, 3000, and 4000 Hz did correlate significantly with
speech discrimination in the least noisy conditions. The results of
Murry and Lacroix were similar to those of Elkins. They found that
correlations between the average of 1000, 2000, and 3000 Hz and
discrimination scores "approached reliability" for the easy lists of
the MRHT, but that correlations for the noisier conditions were
uniformly low, Myers and Angermeier found large amounts of scatter
when discrimination scores were plotted as a function of audiometric
frequency. (Correlations were not computed.) The authors concluded
that it was impossible to explain the variance in scores by any
audiometric index.
The results of the present study are in agreement with two other
correlational studies of speech discrimination in noise. Lindeman
(1971) tested discrimination of monosyllables (Dutch) in "cocktail
party" noise. The author found that the best predictors of speech
discrimination in noise were the audiometric frequencies 2000 and 6300
Hz, respectively. Kryter et al (1962) found that the frequencies
2000, 3000, and 4000 Hz were consistently good predictors of speech
discrimination scores both in high noise (-3 dB S/N) and low noise
(+10 dB S/N) conditions. Although correlations were higher for PB
words than for sentences, these frequencies were also superior to the
others) (500, 1000, and 6000 Hz) for predicting sentence discrimination.
Multiple correlations for various combinations of frequencies showed
that the combination of 1000, 2000, and 3000 Hz was only slightly
less efficient than the combination of 2000, 3000, and 4000 Hz. The
authors concluded that, in light of earlier studies that advocated
greater importance of the lower frequencies, the combined frequencies
1000, 2000, and 3000 Hz would be a good compromise.
Webster (1964) criticized the study of Kryter et al (1962) for
the application of correlational techniques to a population with
predominantly high-frequency hearing loss. Webster pointed out that
correlation coefficients would be influenced by the range and number
of cases distributed throughout the range of measurement. Since
the population in question showed greater ranges and numbers of cases
with losses above 2000 Hz than below 2000 Hz, Webster maintained that
the correlations of discrimination scores with higher frequencies
were artificially high. This criticism would also apply to the present
study since there was considerably more variability and a wider range
of thresholds in the higher audiometric frequencies than in the lower
ones. However, it would be inappropriate to perform these tests on
a population with similar ranges and variability in low-frequency as
-------
53
well as high-frequency thresholds, if the results were to apply to
individuals with sensori-neural hearing losses. In spite of Webster's
implication to the contrary in 1964, the vast majority of today's
compensation cases involve sensori-neural hearing losses, nearly all
of which are noise-induced. To include subjects with significant
losses at 500 Hz would necessitate changing the character of the
population by including a substantial number of conductive losses.
Thus, the population would no longer be suitable for studying the
formula which is supposed to describe it. Webster does admit that if
Kryter's recommendation (of a 1000, 2000, and 3000 Hz method) is
confined to noise-induced-hearing loss subjects, then it is "optimal
for the sample to be studied."
Monaurality
The monaural condition was used in this experiment, even though
it was not representative of common, everyday conditions. However,
monaurality was needed to accurately assess the relationship of audio-
metric threshold to speech discrimination scores. The only way to have
avoided monaurality would have been to select subjects with completely
symmetrical hearing losses binaurally (not a very common condition),
which was not possible within the practical constraints of this experiment.
Pilot Study #2 showed a binaural advantage of 13% for normal listeners.
Therefore, hearing-impaired listeners could conceivably have scored
from 0 to 13% more poorly than the normal-hearing subjects if the
binaural condition were used, simply because of the difference in
thresholds between the two ears. Since the exact amount of influence
that this disparity would cause would be unknown, the naturalness
of the binaural condition was sacrificed in favor of accuracy. The use
of the monaural condition does not diminish the significance of the
differences between experimental groups. In fact, such differences
would most likely be eve'n larger in the binaural condition because of
the relatively smaller contribution of the hearing-impaired subjects'
poorer ears.
However, when making comparisons of these data with binaural
data collected in a reverberant environment it should be kept in mind
that differences between the monaural and binaural data of up to about
13 percentage points could occur, and also that the discrimination
scores of all subjects in this experiment could be expected to be
somewhat higher in real-life, binaural listening conditions.
The Speech Frequencies
Results of the statistical analyses presented in Chapter IV indi-
cate that the traditional label of "speech frequencies," applied to
500, 1000, and 2000 Hz, is inappropriate. Actually, the term has been
applied in a variety of ways in past years. Sabine (1942) referred to
all the frequencies between 128 and 4096 Hz as the "important speech
range of frequencies" or just the "speech range." The 1947 AMA
-------
54
standard (Carter, 1947) used the frequencies 512, 1024, 2048, and 4096
Hz, and the weighted percentages of hearing loss were "based on the
existing data bearing on the relative importance of the auditory
frequency and intensity range in the hearing of speech." Even the
Committee on Conservation of Hearing of the AAOO (DeForest and Lierle,
1955) referred to these four frequencies as the "speech frequencies."
Fletcher (1950) probably had considerable influence on the use
of the term. He found that the frequencies 500, 1000, and 2000 Hz bore
the closest relationship to "hearing loss for speech," which was
defined as 50% correct responses. As mentioned above, Carhart (1946),
Harris et al (1956), and Quiggle et al (1957). defined "hearing for
speech" similarly. Davis (1970) supported the practice by referring to
the average of 500, 1000, and 2000 Hz as the "central speech range".
However, certain investigators have continued to include frequencies
above 2000 Hz in the term, tor example, Kryter et al (1962) stated
that "information in the speech range above 2000 cps contributes
significantly to the understandability of sentences in the presence of
noise " Myers and Angermeier (1972) referred to the "speech range"
as 500 to 3000 Hz. On the basis of his research, Kuzniarz (1973)
objected to "the present concept of so-called 'most important speech
frequencies': 500-2000 Hz," and recommended the average of 1000,
2000, and 4000 Hz, which had been accepted by the Ministry of Health
in Poland.
The term "speech frequencies" is equated with the average of 500,
1000, and 2000 Hz in most audiology clinics, since this average hearing
level corresponds so well with the speech reception threshold level for
spondee words (Newby, 1964). It must be remembered, however, that
higher frequencies should be included in the definition when predicting
speech discrimination ability under everyday conditions. Perhaps in
order to avoid confusion the term "mid-frequencies" could be applied to
the average of 500, 1000, and 2000 Hz.
Fences
The results of the present investigation have not resolved the
question of the location of the point of beginning handicap, or the
"low fence". It is evident that the subjects in Group III, whose mean
average hearing level at 500, 1000, and 2000 Hz was 25 dB, had consider-
able difficulty understanding speech in the backgrounds of noise used
in this experiment, up to 43 percentage points more difficulty than
Group I. This difficulty was encountered for sentences and monosyllables
alike. But Group II, whose mean average hearing level at 500, 1000, and
2000 Hz was only about 13 dB also experienced considerable difficulty
understanding speech in noise, in this case up to 30 percentage points
more difficulty than Group I. Figure 5 showed that a more effective way
of dividing the hearing-impaired groups involves higher frequency
hearing acuity. This fact, along with the discussion of correlations
above, would indicate that whatever fence is selected should include
frequencies above 2000 Hz.
-------
55
In a discussion of the fence issue, Kryter (1970) presented
speech intelligibility functions for individuals with various average
hearing levels at 500, 1000, and 2000 Hz and at 1000, 2000, and 3000
Hz. The functions were drawn from calculations based on the Articulation
Index, where the amount of audible sound pressure in each speech band
was determined according to the typical configuration of a noise-induced
hearing loss.
Using these AI calculations and speech intelligibility predictions,
Kryter devised a method of calculating the percentage of "hearing
impairment for speech" for persons with sensori-neural or conductive
losses. To account for the added discrimination difficulties that
accompany noise-induced hearing loss, even when speech is at an optimal
loudness level, Kryter proposed a correction which assigned twice as
much handicap to these individuals as compared to those with conductive
losses. He also presented different percentages of handicap based on
different levels of unamplified speech ("everyday" speech at 65 dB,
"conversational" speech at 55 dB, and "weak conversational" speech at
50 dB measured at the listener's ear). For these calculations a fence
of 0 dB at the average of 500, 1000, and 2000 Hz or 10 dB at the
average of 1000, 2000, and 3000 Hz is implied. In this respect Kryter's
proposed method resembles Fletcher's (1929) original "Point-Eight
Rule," except that the percentage loss per dB is quite different, the
slope being much more steep in Kryter's method.
Assuming that a long-term rms level of 65 dB reflected the level
of "everyday" speech, Kryter determined that the hearing level at
which individuals could hear 100% of sentences was 15 dB for the
average of 500, 1000, and 2000 Hz or 25 dB for the average of 1000,
2000, 3000 Hz. These levels were based on the assumptions that
everyday speech environments are quiet (even though the level of 65 dB
was chosen to reflect a slightly-raised voice due to ambient noise),
that everyday speech is undistorted, and that individuals with noise-
induced hearing losses have normally shaped performance-intensity
functions (as do individuals with conductive losses). In order to
account for lower voice levels, Kryter (1973) proposed as a fence
average hearing levels of 6 dB at 500, 1000, and 2000 Hz or 16 dB at
1000, 2000, and 3000 Hz. These levels were consistent with 100%
intelligibility of sentences at a level of 55 dB in quiet, but Kryter
stated that even these individuals would be disadvantaged in comparison
with normal-hearing persons when the speech was distorted, or the
level was weaker than usual.
The above proposals have not been accepted by the medical com-
munity or by governmental bodies, and the extent to which they are
being considered seriously is not known. Most of the fences that
Kryter proposed are lower than those presently in use by the various
States or Federal agencies.
Cne way to approach the problem of an appropriate fence would be
to find the hearing level at which hearing-impaired subjects begin to
perform differently from their normal-hearing controls. A signif-
icant finding in Acton's (1970) study was the fact that the most
-------
56
mildly-impaired group (Group A) performed more efficiently than the
control group in most of the noise backgrounds, and virtually as well
in quiet. By way of explanation the author suggested that in situations
where the redundancy of speech is reduced by noise, individuals are
able to "get-by" so long as their hearing levels do not reach those of
a "critical hearing loss." Once this critical level is passed individuals
show increasing difficulty in understanding speech, especially as
redundancy is further reduced. Acton hypothesized that the critical
hearing loss was somewhere between Group A and Group B. However, both
Acton's Group B and this study's Group II performed significantly more
poorly than their normal-hearing counterparts in all of the noise
conditions, and judging by the magnitude of the differences could be
considered past the point of "critical hearing loss."
Mean hearing levels of Acton's groups A and B, and the present
study's Group II are as follows:
.5, 1, 2k Hz 1, 2, 3k Hz 1, 2, 4k Hz
Acton's Group A 4 dB 11.5 dB 14.8 dB
Acton's Group B 13.2 dB 27.6 dB 30 dB
Present Study Group II 13.4 dB 25.9 dB 29.6 dB
Acton's hearing levels have been converted from British Standard 2497
(1954) to ANSI (1969) reference values.*
It could be hypothesized that the appropriate fence lies some-
where between Acton's Group A and the present study's Group II (whose
mean thresholds are almost identical with Acton's Group B). If the
midpoints between these two groups are selected, then the estimated
fences would be approximately 9 dB at 500, 1000, and 2000 Hz, 19 dB at
1000, 2000, and 3000 Hz, and 22 dB at 1000, 2000, and 4000 Hz. These
values are very close to the mean minus one standard deviation of Group
II hearing levels, which are 10.5 dB, 18.7 dB and 24.6 dB, respectively.
The selection of a fence is ultimately dependent upon the defini-
tion of hearing handicap and the conditions under which handicap is
assessed. Davis (1965) defined handicap as "the disadvantage imposed
by an impairment sufficient to affect one'"s personal efficiency in
the activities of daily living." Since speech communication is the
activity most likely to be impaired by hearing loss, the AAOO defined
the absence of hearing handicap (or "impairment" in 1959) as, "The
ability to hear sentences and repeat them correctly in a quiet environ-
ment....," (Lierle, 1959). As mentioned previously, many investigators
have pointed out that undistorted speech in quiet is not typical of
Audiometric thresholds given in personal communication from
Dr. W. I. Acton.
-------
57
everyday conditions. Therefore, the determination of hearing handicap
should be based on speech discrimination in noise, even if tests in
quiet are included.
Basically, the selection of a fence is a social issue. It rests
on the question of how much speech communication ability is needed in
order to conduct the activities of daily living in a satisfactory
manner. The answer will undoubtedly be influenced by such variables as
an individual's age, occupation, lifestyle and personal preference.
Field, rather than laboratory research will probably be needed in order
to solve the problem, but research in this area has been inconclusive
to date. Until more information is forthcoming, the decision on an
appropriate fence will necessarily be somewhat arbitrary.
-------
CHAPTER VI
SUMMARY AND CONCLUSIONS
Summary
The purpose of the investigation was to explore the relationship
between hearing level at various audiometric frequencies and speech
discrimination in different noise backgrounds. Although these relation-
ships had been investigated numerous times before, the studies had not
been designed specifically to test the adequacy of the American Academy
of Ophthalmology and Otolaryngology's (AAOO) selection of 26 dB as the
"low fence". This hearing level, averaged over the audiometric fre-
quencies of 500, 1000, and 2000 Hz, is said to be the point above which
hearing handicap occurs. It has been incorporated into many state
compensation statutes, and has also been widely used in the U.S. and
abroad for purposes of damage-risk criteria and the setting of occupa-
tional noise standards.
The AAOO method for computing hearing handicap has been brought
into question during the past few years, both by researchers and by
policy-makers, for two primary reasons: a) that the 26-dB fence is
too high, and b) for the exclusion of frequencies above 2000 Hz.
The present study, therefore, has investigated the relationship between
hearing level and speech discrimination to see if there are differences
among individuals whose hearing is at or better than the low fence, or
whether they are all indeed "not handicapped."
The AAOO low fence is based on the assumption that "hearing
impairment should be evaluated in terms of ability to hear everyday
speech under everyday conditions." Consequently, the present experiment
has employed "everyday" sentences as well as a closed-set test of
monosyllables, presented in a quiet background and in various levels of
noise, in a mildly reverberant sound field. In designing the study the
following experimental questions were posed:
1. What is the relationship between average hearing level at
500, 1000, and 2000 Hz and speech discrimination scores in
noise for individuals whose average hearing levels are at or
better than the AAOO low fence?
2. Is the relationship dependent upon speech-to-noise ratio?
3. Is the relationship between average hearing level and speech
discrimination scores differently described by different
speech materials?
4. Which combination of audiometric frequencies best predicts
speech discrimination scores?
58
-------
59
Forty-eight subjects between the ages of 21 and 56 were tested
with two types of speech materials: the University of Maryland lest
#1 (DM Test #1), which employs the CID "everyday" sentences, and
the Modified Rhyme Test (MRT), a closed-set test of rhyming monosyl-
lables. Subjects were divided into three equal groups according to
their better-ear average hearing levels at 500, 1000, and 2000 Hz.
Group I had mean average hearing levels of 2.5 dB and hearing at all
frequencies (250 - 6000 Hz) of 20 dB or better. Group II had mean
average hearing levels of 13.4 dB and Group III had mean average
hearing levels of 24.7 dB at 500, 1000, and 2000 Hz. Both groups II
and III were unrestricted for hearing loss in the higher frequencies.
Since most of these subjects had been exposed to noise their losses were
considerably greater in the frequencies above 2000 Hz than in the
mid-frequencies.
Each subject listened to the LM Test #1 in a quiet condition and
in speech-to-noise ratios of -1, -3, and -5 dB, and to the MRT in quiet
and in speech-to-noise ratios of 0, -3, and -6 dB. The reverberation
time of the room was 0.625 second and remained unchanged.
Results of the tests and the statistical analyses provided answers
to the above questions as follows:
1. Significant differences were found in mean speech discrimination
scores among all of the three groups, showing that within the "normal"
area under the AAOO fence there was considerable individual variability in
the ability to discriminate speech in noise. Groups with mean average
hearing levels of 24.7 dB and even 13.4 dB performed significantly more
poorly than the control group, whose mean average hearing level was 2.5 dB.
2. The relationship between average hearing level and speech discrimi-
nation scores proved to be dependent upon speech-to-noise ratio. The
discrimination scores of all three groups were depressed as the speech-
to-noise ratio became lower, and the differences between groups increased.
These differences were apparent even in the quiet condition, although
they became much larger as noise was introduced.
3. Examination of the mean scores of the UM Test tl and the MRT
(after correcting for guessing) showed that subjects scored very
similarly on the two kinds of materials in the two intermediate noise
conditions, but not in the quiet condition, or in the most difficult
noise condition. Mean scores in quiet were generally lower on the MRT
than on the LM Test #1, and scores in the most difficult noise condi-
tion were lower on the LM Test #1 then on the MRT. The two materials
appeared to be equally effective at delineating differences among
the three groups in the various noise conditions.
4. Correlational tests revealed that frequency combinations that
included frequencies above 2000 Hz were significantly better predictors
of speech discrimination scores than the combination of 500, 1000,
and 2000 Hz.
-------
60
Conclusions
On the basis of this experiment it is possible to draw the
following conclusions:
1. Individuals whose hearing levels are at or better than the
AAOO low fence may have considerably more difficulty in
understanding speech than those whose hearing is normal, as
defined in this experiment. This is true even for those
whose average hearing levels in the mid-frequencies (500,
1000, and 2000 Hz) are approximately 14 dB.
2. Increased levels of noise, in relation to the speech signal,
tend to exacerbate the adverse effects of hearing loss.
3. Simple sentences, such as the CID "everyday" sentences,
and a closed-set test of monosyllables, such as the MRT, can
be considered roughly equivalent for measuring speech discrimina-
tion in certain conditions of noise, but not in quiet.
4. Combinations that include frequencies above 2000 Hz are
significantly better predictors of speech discrimina-
tion score than the combination of 500, 1000, and 2000 Hz
for persons with noise-induced hearing loss.
Recommendations
Four principle recommendations can be made as a result of this
investigation. First, frequencies above 2000 Hz should be included in
any technique for assessing the ability of hearing-impaired individuals
to understand speech in "everyday" listening conditions. For the assess-
ment of hearing handicap in a noise-exposed population similar to that
of this experiment, the average of 1000, 2000 and 4000 Hz appears to be
the most appropriate simple average, since this average has been shown
to correlate highly with speech discrimination in quiet as well as in
noise. Unequal weighting of the different frequencies was not considered
in this report, although it would be reasonable to explore this method
in the future using these or other data. It is also recommended that
the term "speech frequencies" should not be applied to 500, 1000 and
2000 Hz alone, but should be used broadly to include all of the audible
frequencies of the speech spectrum (through 8000 or 10000 Hz), and that
frequently-used combinations of audiometric frequencies be specified,
such as ".5, 1 and 2k Hz" or "1, 2 and 4k Hz". The combination of 500,
1000, and 2000 Hz could be termed the "mid-frequencies".
The second recommendation pertains to the height of the fence, or
the point of beginning handicap. The present 26-dB fence, averaged over
500, 1000 and 2000 Hz has been shown in this investigation to be above
the point of beginning handicap. Even 26 dB averaged over 1000, 2000
and 3000 Hz appears to be too high since that level corresponds to the
mean hearing level of this study's Group II (at those frequencies), who
-------
61
showed significantly more difficulty in understanding speech than the
normal-hearing group. Data gathered in another study (Acton, 1970) showed
that individuals with average hearing levels of 12 dB at 1000, 2000 and
3000 Hz, or 15 dB at 1000, 2000 and 4000 Hz performed as well as their
normal-hearing controls, even in noise conditions. Therefore, until
further research defines this point more precisely, it is suggested that
the midpoint between the "handicapped" and "not handicapped" groups is
selected, namely, 19 dB at 1000, 2000 and 3000 Hz or 22 dB at 1000, 2000
and 4000 Hz.
The third recommendation is for further research into the concept
of the fence as a social issue. Techniques should be developed to deter-
mine the amount of speech communication ability that is needed in order
to conduct the activities of daily living in a satisfactory manner.
Various lifestyles and various activities should be studies. This kind
of research would benefit the development of speech communication criteria
for normal-hearing as well as hearing-impaired individuals.
A final recommendation pertains to clinical as well as laboratory
tests of speech discrimination in noise. In order to assess accurately
an individual's ability to understand speech in various "everyday" con-
ditions, speech materials should be presented at a level that reflects
life-like listening conditions. The speech level should be the same
for all clinical patients or experimental subjects rather than being
adjusted to an optimal level for each listener. It appears that the
adjustment of presentation level for speech discrimination material to
each patient's optimal listening level (PB max) is primarily intended
to assess the patient's ability to understand speech with a hearing aid.
While this procedure is definitely useful for the intended purpose, it
is not appropriate for assessing the speech discrimination abilities
of persons who are not suited for or do not intend to wear hearing
aids. Instead, individuals should be tested at speech levels of about
38 dB above speech audiometric zero (or 58 dB long-term rms , the
speech levels that they typically must listen to in everyday life.
-------
APPENDIX A
PI IDT EXPERIMENTS
-------
62
Instrumentation
Pilot Studies #1 and #2 were conducted about four months earlier
than Pilot Study #3, and the placement of equipment was slightly
different. For the earlier tests, two of each type of loudspeaker were
used, located in corners of the room, and equidistant from the subject
at 45 degree angles to the front of the subjects' midline. The noise
signal (babble) was routed through KLH Model Six speakers, which were
placed on the floor, and the speech signal was routed through Grason-
Stadler 162-4 speakers, located directly above the others. Later, for
the purpose of creating a more evenly diffuse sound field, one of each
type of loudspeaker was removed. For Pilot Study #3 and for the final
experiment the subject was seated facing the two remaining loudspeakers,
which were placed in one corner of the room. It is believed that this
difference in instrumentation does not affect the applicability of
Pilot Studies #1 and #2 to the final experiment.
Pilot Study #1
The purpose of this experiment was to examine the equivalency
of the UM Test #1 lists in a noise background, and to see whether or
not a practice effect occurred, (an increase in scores as a function of
familiarity with the task). The ten lists were presented to ten '
normalhearing subjects in a background of 12-speaker babble. The lists
were presented in the sound field and both ears were unoccluded. The
level of the speech signal was 55 dB (long-term rms measured at a point
corresponding to the listener's ear) and the babble was 62 dB, resulting
in a speech-to-noise ratio of -7 dB. Order of presentation of the lists
was informally counterbalanced. Subjects' responses were monitored by
the investigator through the talkback system of the Grason-Stadler 1701
aud iometer.
The resulting speech discrimination scores varied considerably among
subjects and among lists for the same subject, but there appeared to be
no practice effect. The range of mean scores for the various lists was
approximately 20% (not including List A, the recording of which had
become defective and subsequently was eliminated). Because of the
intra-subject variability and the differences between means of lists,
it was decided to pair them on the basis of mean scores for use in the
final experiment. This would provide 20 sentences, including 100 key
words, for each listening condition. Pairing of the lists brought the
mean scores to within 1-1/2% of each other for the speech-to-noise
ratio tested (see Table A-l). The selected pairs were lists B + F, D +
I, G + H and J + E.
Pilot Study #2
The purposes of this study were to assess potential differences
between monaural and binaural discrimination of MRT words by subjects
-------
63
Table A-l. Pilot Study #1. Mean scores and standard
deviations in percent correct of ten normal-hearing subjects
for various single and combined lists of the UM Test #1.
Speech-to-noise ratio = -7 dB.
List Mean SD
B 75.5 11.7
C 67.4 14.1
D 66.7 14.6
E 59.6 7.1
F 55.6 14.8
G 75.6 11.2
H 55.4 11.8
I 67.3 12.7
J 72.8 10.5
Selected Pairs
B + F 65.6
D + I 67.0
G + H 65.5
J + E 66.2
-------
64
with normal hearing, and to determine the equivalency of the MRT word
lists as recorded for this experiment.
Eight MRT word lists were recorded, six from MRHT Form 3, and two
from Form 2, using the vocabulary and techniques described by Kreul et
al (1968). The eight lists were presented to thirteen normal-hearing
listeners at a speech-to-noise ratio of -7 dB (speech level 55 dB
long-term rms and noise level 62 dB measured at the listener's ear as
before). Four lists were presented monaurally and four binaurally and
the order of presentation of lists was counterbalanced across subjects
and listening mode. Monaurality was achieved by the use of a V-51R
earplug and a David Clark type 117 earmuff over one ear whose combined
attenuation capabilities were described in the section on measurement.
Although the subjects were asked not to move their heads, they were not
physically restrained. In order to minimize head movements a plumb
line was hung from the ceiling with the bob centered about two inches
in front of the head at eye level. Subjects were asked not to touch
the plumb bob and to keep it, to the extent possible, centered in the
line of vision.
The results showed a mean increase of 13% for binaural over
monaural scores (see Table A-2). Consequently, the speech was presented
monaurally in the final experiment. Those lists with the smallest
standard deviations and the most similar means were selected for use in
the main study. Lists B, C, D and F from Form 3 were the best candidates.
List A of Form 2 was used as a practice list.
Pilot Study #3
The purpose of this phase of the pilot work was to determine
appropriate speech-to-noise ratios so that the scores of the hearing-
impaired listeners, and to a lesser extent those of the normal-hearing
listeners, would be likely to fall along the linear portion of the
performance-intensity function. The subjects consisted of four normal-
hearing and four hearing-impaired individuals, (see Table VI) . Two of
the hearing-impaired subjects would be categorized as Group II (subjects
#5 and 6) and two as Group III members (subjects |7 and 8), according
to the criteria described earlier. With speech stimuli presented at a
fixed level of 60 dBA (measured at the listener's ear), the following
speech-to-noise ratios were explored:
__ Speech-to-noise ratio (dB)
Quiet +2 0 -1 -2 -3 -4 -5 -6 -8 -9
UM Test #1 Quiet +2 +1 -1 -3 -4 -5 -6 -7 -9
The four selected MRT lists and the four pairs of UM Test #1
lists were presented monaurally, preceded by a practice list of each
type delivered at a mildly difficult speech-to-noise ratio. The data
are given in Table A-3. In this table the MRT scores have been corrected
-------
Table A-2. Pilot Study #2. Mean scores and standard deviations in percent
correct on the Modified Rhyme Test of thirteen normal-hearing subjects in monaural
and binaural listening modes. Speech-to-noise ratio = -7dB. Scores are not cor-
rected for guessing.
MRT
List
A
* B
* C
* D
E
* F
A
B
Form
3
3
3
3
3
3
2
2
N
7
7
7
7
6
6
6
6
Monaural
Mean
56.6
55.7
51.7
53.1
49.3
57.7
64.7
61.3
X = 56.3
SD
12.5
8.8
8.4
6.0
13.0
8.0
10.2
7.9
N
6
6
6
6
7
7
7
7
Binaural
Mean
69.7
70.0
69.3
67.0
62.9
64.9
76.0
74.3
X = 69.3
SD
13.6
4.7
10.6
6.8
9.6
5.6
9.9
7.8
*Lists selected for final experiment
-------
Table A-3. Pilot Study #3. Individual scores on MET and UM Test #1 in a variety
of speech-to-noise ratios. MRT scores have been corrected for guessing.
Speech-
to- noise-
Ratio (dB)
Quiet
+2
+1
0
-1
-2
-3
MRT -4
-5
-6
-7
-8
-9
Quiet
+2
+1
0
-1
UM -2
Test #1 -3
-4
-5
-6
-7
-8
-9
123
Average hearing
-5 dB -2 dB 2 dB 3
80.8
92.8
80.8 83.2
92.8
80.8
66.4 80.8
68.8
54.4
66.4 30.4
98
90 94
82 84 83
70 71
53 54
25 30
Subject #
45 6 7 ° 8
level at 500, 1000 and 2000 Hz
dB 12 dB 15 dB 20 dB 28 dB
95.2 95.2 90.4
68.8 61.6
80.8 83.2
64 40 64
78.4 61.6
37.6 40
52 35.2 42.4
44.8
16
6.4
99 99 96
85
93 63 70 71 57
84 41 48
19 50
54 35 29
4 47
30
-------
67
for guessing. It appeared that the linear portion of the hearing-
impaired subjects' performance-intensity functions was between speech-
to-noise ratios of approximately +2 and -7 dB for the MRT and +1 and
-5 dB for the UM Test #1. It also appeared that the normal-hearing
subjects would have relatively little difficulty with these speech-to-
noise ratios. In order to make sure that some degradation would occur
in the performance of the normal-hearing subjects, and to highlight
differences between groups should such differences occur, the positive
speech-to-noise ratios were eliminated and speech-to-noise ratios of
0, -3, and -6 dB were chosen for the MRT, and -1, -3, and -5 dB for
the UM Test #1.
-------
APPENDIX B
CONSENT OF VOLUNTEER
-------
REPLY TO
ATT N OF
68
DEPARTMENT OF THE AIR FORCE
657OTH AEROSPACE MEDICAL RESEARCH LABORATORY 'AFSC1
WRIGHT-PATTERSON AIR FORCE BASE. OHIO 45433
SUBJECT
Consent of Volunteer
1. I hereby volunteer to participate as a test/experimental subject
in the following investigation/test which has as its purpose
2. .has
discussed with me to my satisfaction the reasons for this investigation/
test and its possible adverse and beneficial consequences.
3. This consent is voluntary and has been given under circumstances
in which I can exercise free power of choice. I have been informed
that I may at any time revoke my consent and withdraw from the experi-
ment without prejudice and that the investigator or physician may
terminate the experiment at any time regardless of my wishes.
4. I understand that before my use as a test subject, I must inform
the principal investigator and/or project physician of any change to
my medical status. This information will include any medications I
have taken and any medical or dental care/treatment received since my
last use as a test subject.
(Signature of Volunteer)
TFroject Officer)
(Witness)
-------
APPENDIX C
INSTRUCTIONS TO SUBJECTS
(MPT)
-------
SPEECH INTELLIBILITY TEST
INSTRUCTIONS
EXaring the test it is important that you do not move your head
around. You may move it up and down a little so that you are free to
write, but please do not move it sideways. The plumb bob has been
arranged so that it would be in the center of your head. Try not to
touch it with your head, and as much as possible, keep it in the
center of your line of vision.
You will hear some words and sentences in a background of noise.
There are 3 parts to the test. The first part is fairly short. First
an announcer will read a list of words. Each time the announcer will
say "Number one ) (or so); you will mark the (word) please." You are
to look at the square corresponding to that number, decide which word
he said, and circle the word. Try to ignore the noise, even if it
sounds rather loud, and concentrate on the voice reading the list of
words. If you are not sure which word was said, take a guess.
After that you will hear a list of short sentences. This time
you will just repeat each sentence. Please be sure to talk clearly
so that I will be able to hear exactly what you have said. If you
are not able to hear all the words in the sentence, repeat the ones you
do hear. If you are in doubt about some of the words, take a guess.
Now you will hear a series of word lists, and you will circle
the test words. Turn the page each time the announcer completes the
list of 50 words. Try to ignore the noise and concentrate on the voice
reading the words.
After that you will have a few minutes to relax and then you will
hear a series of sentences.
During the tests there may be a short pause while I rewind the
tapes.
Remember, in all cases if you're not sure of the words you hear,
please guess whenever you can.
69
-------
APPENDIX D
SPEECH lyiATERIALS
-------
MODIFIED RHYME TEST
-------
NAME
EAR
DATE
MODIFIED RHYME HEARING TEST 3
LIST
fang
rang
gang
1
bang
hang
sang
mark
park
lark
2
bark
hark
dark
peel
feel
reel
3
keel
eel
heel
tang
tarn
tack
1
tab
tap
tan
sick
sing
sill
5
sit
sin
sip
mass
mad
mat
6
map
man
math
pup
putt
pun
7
pug
puff
pub
hop
top
shop
3
pop
cop
mop
best
nest
test
9
west
rest
vest
10
cuff
cud
cub
cup
cut
cuss
11
sale
safe
sane
sake
save
same
12
dust rust
just gust
bust must
13
heave heal
heath heap
hear heat
dim
did
dip
14
din
dig
dill
15
took
cook
book
look
hook
shook
16
sap
sag
sack
sat
sass
sad
gun
bun
sun
17
run
nun
fun
18
page pale
pane pay
pave pace
19
got hot
tot pot
lot not
20
tick wick
pick sick
kick lick
wit
sit
bit
21
fit
hit
kit
22
kith kit
kiss kid
king kill
23
foil
coil
soil
oil
toil
boil
24
fig rig
pig wig
big jig
25
peach peas
peal peak
peat peace
26
pill
pig
pit
Pip
pin
pick
sup
sun
sud
27
sung
sum
sub
28
fizz
fill
fig
fit
fib
fin
29
bent
went
sent
tent
dent
rent
30
pat
pass
pad
pang
pan
path
31
teach
teak
teal
tear
team
tease
32
dud dun
dub dull
dug duck
33
beak
beat
beach
beam
-bead
bean
34
way say
may day
gay pay
35
then
pen
ten
hen
men
den
36
paw
thaw
jaw
say
law
raw
37
lane
lake
lame
lace
lay
late
38
pale tale
bale gale
male sale
39
till
fill
hill
bill
kill
will
bed
fed
red
40
wed
led
shed
41
hold
fold
sold
gold
cold
told
bun
bug
but
42
buff
buck
bus
43
seed seem
seep seen
seethe seek
44
sin tin
win din
fin pin
45
neat heat
beat meat
seat feat
fame
came
game
46
name
same
tame
sip
hip
lip
47
rip
tip
dip
48
bath
ban
bass
back
bad
bat
49
cake cape
case cane
cave came
50
race
rake
raze
rate
ray
rave
70
-------
UNIVERISTY OF MARYLAND TEST #1
LISTS B THROUGH J
-------
71
LIST IB
1. The water's too cold for
swimming.
2. Why should _! get up jso early.
3. j>hine your own shoes this time.
4. It's raining right here in the room.
5. Where are you going this morning?
6. You should come here when I call.
7. Don't try to get out of it.
8. We let little children go to the movies.
9. There isn't enough paint to finish.
10. Do you want eggs for breakfast?
-------
72
LIST 1C
1. Everybody should brush teeth before meals.
2. Once a year everything's all right.
3. Don 't use _up all the letter paper.
4. Anything like that's all right with me.
5. Those people outside ought to see a doctor.
6. The windows are _so dirty this month _I can 't see.
7. Please pass the bread and butter first.
8. Don't forget to write and pay your bill.
9. Don't let the dog out of the house.
10. There 's a good ballgame this afternoon.
-------
73
LIST ID
1. If_ you want to go it's all right.
2. Throw these old Time magazines out.
3. Do you want to wash _up in the stream?
4. It's a real dark night so watch your driving.
5. I'11 carry your package for you.
6. Don't you forget to shut off the water.
7. Mountain fishing is my idea of a good time.
8. Fathers used to spend more time with their children.
9. E5e_ careful not to break the glasses.
10. I'm sorrier than you for the mistake.
-------
LIST IE
1. You can catch the bus across the street.
2. Tell her the news on the phone.
3. I '11 catch _up with you later.
4. I '11 think it over and call her.
5. I_ don 't want to ^o to the movies.
6. See a dentist _if your tooth hurts.
7. Put that cookie back in the box.
8. You ought to stop fooling around so much.
9. Tonight that extra time's up.
10. How do you spell your name?
-------
75-
LIST IF
1. Music always makes me cheer up.
2. My brother's in town for a short while.
3. We live a few miles off the main road.
4. This suit needs to cjo to the cleaners.
5. They ate enough green apples.
6. Have you been sick all this week?
7. Where have you been working lately?
8. There 9s not enough table room in the kitchen.
9. It's hard to see where he is.
10. Look out for new business.
-------
LIST 1G
1. I '11 see you right after lunch.
2. I '11 see you later this afternoon.
3. White shoes are awful to keep clean.
4. You stand over there until I move.
5. There's a piece of cake left for dinner tonight.
6. Don't wait for me at the front corner.
7. It's _no trouble at all to tell.
8. Hurry up_ with the morning paper.
9. It didn't say anything about a big rain.
10. That drugstore phone call's for you.
-------
LIST 1H
1. Believe me it's too late.
2. Let's get that cup of coffee.
3. Let's get out of here before long.
4. I hate driving if it's at night.
5. There was water in the cellar yesterday.
6. She'11 only be gone a few minutes.
7. How do you-know we'll have it soon?
8. Children like candy after heavy meals.
9. No grass grows when we don't get rain.
10. They 're not listed in the new phone book.
-------
78
LIST II
1. Where can I find a place to park?
2. I like those big red apples.
3. You '11 get fat by eating candy.
4. The color show's over in the Fall.
5. Why don 't they paint their other walls?
6. How come you always get to cp first?
1. What are you hiding under your coat?
8. I_ should always buy new cars.
9. What's wrong with sugar and cream in my coffee?
10. I'll wait just one minute.
-------
79
LIST 1J
1. But we won't be ready to start.
2. I don 't know what's wrong with the car.
3. It sure takes a sharp knife to cut meat.
4. I haven't read a newspaper since we got television.
5. The weeds are spoiling this yard.
6. Call me a little later for breakfast.
7. Do you have change for a five-dollar bill?
8. How are the things _we bought?
9. I "d like some ice cream with my pie.
10. I don't think I'll have dessert.
-------
APPENDIX E
RAW DATA
SUBJECTS' HEARING LEVELS AND
DISCRIMINATION SCORES
-------
GP.OUP I
Sul^ect Ago
No. (Years)
Sex
Educational
Level (Yrs.)
250
Hearing Levels (dB)
500 Ik 2k 3k 4k
*MRT scores have been corrected for guessing
6k
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
46
43
21
21
33
36
55
38
44
41
38
31
52
34
30
36
M
F
M
F
F
M
M
F
M
F
M
F
F
M
F
M
12
16
12
15
12
12
12
15
16
12
12
18
12
12
13
17
5
0
10
-5
5
0
5
5
-5
10
0
-5
5
5
-5
0
5
0
5
0
5
5
10
10
0
5
5
-5
5
5
-5
5
0
0
5
0
0
0
5
0
0
0
0
-5
0
5
-S
5
0
0
5
0
0
15
10
0
5
5
-5
0
5
15
0
0
20
0
5
-5
5
10
15
-5
5
15
0
— S
10
20
0
10
15
10
5
5
0
10
10
0
10
20
10
5
15
20
5
10
15
0
10
-5
15
15
20
0
0
20
15
20
15
15
10
10
Hz
Discrimination Scores in % Correct
U.M. Test #1 MKT*
Q -1 -3 -5 Q 0 -3
-6 (dU)
100
99
99
100
100
99
100
99
100
100
100
100
100
99
99
100
88
93
85
96
93
90
91
93
92
76
84
95
78
94
82
81
79
83
64
82
83
80
81
84
82
90
76
85
64
81
62
74
53
73
50
59
44
71
77
54
59
45
70
77
26
45
69
63
97.6
90.4
95.2
100
88
97.6
97.6
95.2
90.4
80.8
95.2
92.8
95.2
88
88
92.8
88
80.8
78.4
92.8
85.6
88
80.8
80.8
85.6
85.6
78.4
88
80.8
85.6
85.6
80.8
66.4
66.4
83.2
83.2
76
80.8
73.6
78.4
80.8
80.8
78.4
88
68.8
71.2
64
78.4
64
61.6
68.8
73.6
49.6
71.2
61.6
52
61.6
52
73.6
83.2
56.8
61.6
52
64
CD
O
-------
GROUP II
Subject Age
No. (Years)
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
52
48
54
52
47
47
47
46
37
47
37
54
38
27
54
38
Sex
M
M
M
M
M
M
M
M
M
M
M
M
M
Educational
Level (Yrs.)
12
9
16
12
10
12
10
12
10
10
20
12
18
14
12
12
250
Hearing Levels (dB)
500 Ik 2k 3k 4k
6k Hz S/N
0
20
10
5
5
5
5
15
15
10
5
5
10
5
10
0
0
15
10
10
10
5
10
15
10
15
10
10
10
5
15
5
15
15
10"
10
10
5
15
10
10
5
10
10
15
10
15
15
15
20
35
15
10
30
25
20
10
10
10
30
10
30
10
25
25
70
55
40
25
60
40
30
25
85
30
60
55
55
45
55
35
70
60
50
65
50
50
30
70
95
60
70
65
45
65
60
20
75
80
45
65
60
65
35
45
90
55
60
55
25
60
55
Discrimination Scores in % Correct
U.M. Test #1 MRT*
Q -1 -3 -5 Q 0 -3
-6 (dB)
100
98
98
93
98
91
100
98
98
100
97
100
99
99
100
96
81
50
68
57
82
47
75
89
92
53
83
60
53
70
76
47
88
50
47
32
42
41
46
81
86
30
51
37
37
48
60
37
63
18
16
23
39
18
41
55
44
14
16
32
18
29
30
1
92.8
71.2
90.4
92.8
88
85.6
85.6
92.8
78.4
88
90.4
85.6
88
95.2
88
78.4
66.4
47.2
54.4
83.2
64
56.8
52
76
68.8
56.8
76
47.2
66.4
83.2
61.6
44.8
68.8
40
49.6
61.6
56.8
32.8
47.2
66.4
59.2
44.8
76
28
42.4
61.6
54.4
30.4
54.4
30.4
25.6
30.4
37.6
23.2
23.2
54.4
47.2
23.2
56.8
11.2
13.6
42.4
35.2
35.2
*MRT scores have been corrected for guessing
oo
-------
GROUP III
Subject
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Pqe
(Years)
53
55
44
50
55
56
55
46
53
52
56
52
54
50
37
42
Sex
M
M
M
M
M
M
M
F
M
M
M
M
M
M
M
M
Educational
Level (Yrs.
12
10
12
8
12
12
12
17
12
12
13
12
12
9
12
8
. ) 250
20
15
0
15
5
15
15
10
5
25
15
5
15
10
5
5
Hearing Levels (dB)
500
15
20
5
15
5
15
15
10
10
20
20
20
10
15
5
15
Ik
15
20
25
20
10
20
20
20
20
15
25
25
20
25
10
20
2k
30
45
55
30
70
50
40
45
30
30
25
35
30
35
65
45
3k
80
55
70
60
70
75
50
65
35
50
55
50
25
50
90
45
4k
100
55
75
80
65
90
60
70
45
50
60
40
35
60
85
35
6k
100+
70
75
85
50
95
80
80
60
40
75
55
50
45
90
35
Hz
S/N
Discrimination Scores in % Correct
U.M. Test #1 MRT*
Q -1 -3 -5 Q 0 -3 -6 (dB)
94
96
73
98
85
87
94
94
100
100
95
100
99
96
97
95
29
49
13
46
25
43
42
42
82
75
62
73
75
72
37
86
18
30
1
45
26
10
38
21
81
56
24
50
54
46
27
72
8
0
0
4
0
0
2
0
41
31
7
36
37
29
5
44
68.8
76
61.6
68.8
80.8
68.8
83.2
76
85
90.4
88
80.8
90.4
73.6
85.6
90.4
32.8
59.2
23.2
56.8
61.6
42.4
56.8
40
83.2
73.6
37.6
61.6
54.4
49.6
44.8
76
11.2
44.8
1.6
18.4
30.4
11.2
23.2
25.6
47.2
64
23.2
54.4
49.6
35.2
16
52
1.6
13.6
0
8.8
6.4
0
20.8
0
61.6
56.8
11.2
35.2
25.6
20.8
0
52
*MRT scores have been corrected for guessing
-------
REFERENCES
Acton, W. I. Speech intelligibility in a background noise and noise-
induced hearing loss. Ergonomics, 13, 546-554 (1970).
MA American Medical Association. Committee on Rating of Physical
Impairment. Guides to the evaluation of permanent impairment:
ear, nose and throat and related structures. J. Amer. Med. Assoc.,
177, 99-108 (1961).
AMA American Medical Association. Committee on Rating of Physical
Impairment. Ear, nose and throat and related structures. Guides
to the Evaluation of Permanent Impairment, Chicago: AMA (1971).
ANSI American National Standard for Acoustical Terminology. ANSI Sl.l
(1960)(R1976).
ANSI American National Standard Criteria for Background Noise in
Audiometer Rooms.' ANSI S3.1 (1960) (R1971).
ANSI American National Standard Methods for the Calculation of the
Articulation Index. ANSI S3.5 (1969).
ANSI American National Standard Specifications for Audiometers. ANSI
S3.6 (1969).
Aniansson, G. Binaural discrimination of "everyday" speech. Acta
Otolaryngol., 75_, 334-336 (1973).
Atherley, G. R. C. and Nobel, W. G. Clinical picture of occupational
hearing loss obtained with the Hearing Measurement Scale. In
D. W. Robinson (Ed.), Occupational Hearing Loss. New York:
Academic (1971) .
BBN Bolt, Beranek and Newman, Inc. Impact of Noise Control at the
Workplace. Report No. 2671 submitted to U.S. Dept. Labor (1974).
Beranek, L. L. (Ed.), Noise and Vibration Control. New York: McGraw-
Hill (1971). ~~
Bergman, M. Hearing and Aging. Audiology, 10, 164-171 (1971).
Blumenfeld, V. G., Bergman, M., and Millner, E. Speech discrimination
in an aging population. J. Speech Hearing Res., 12_, 210-217 (1969)
BSI British Standards Institute. Method of Test for Estimating the
Risk of Hearing Handicap Due to Noise Exposure. BS 5330 (1976).
83
-------
84
Bullock, M. D. The effects of different reverberation times upon
the intelligibility of PB words as perceived by subjects with
normal hearing and subjects with sensorineural impairments and
concomitant discrimination losses. Masters thesis, Chio State
Univ. (1967).
Burns, W. Noise and Man. Philadelphia: Lippincott (1973).
Carhart, R. Speech reception in relation to pattern of pure tone
loss. J. Speech Pis., !!_, 97-108 (1946).
Carter, M. A. (for AMA Council On Physical Therapy). Tentative
standard procedure for evaluating the percentage of useful
hearing loss in medicolegal cases. J. Amer. Med. Assoc., 133,
396-397 (1947).
CHABA, Committee on Hearing, Bioacoustics and Biomechanics. Speech
Understanding and Aging, Report of Working Group 75. National
Academy of Sciences, Washington, D. C. (1977).
Cooper, J. C. and Cutts, B. P. Speech discrimination in noise.
J. Speech Hearing Res., 14, 322-337 (1971).
Davis, H. The articulation area and the Social Adequacy Index for
hearing. Laryngoscope, 58, 761-778 (1948).
Davis, H. Guide for the classification and evaluation of hearing
handicap. Trans. Amer. Acad. Ophth. and Otol., 69, 740-751
(1965).
Davis, H. Hearing handicap, standards for hearing, and medicolegal
rules. In H. Davis and S. R. Silverman (Eds), Hearing and
Deafness. New York: Holt, Rinehart and Winston (1970).
Davis, H. A historical introduction. In D. W. Robinson (Ed.),
Occupational Hearing Loss. New York: Academic (1971).
Davis, H. Some comments on "Impairment to hearing from exposure to
noise" by K. D. Kryter. J. Acoust. Soc. Amer., 53, 1237-1239
(1973).
De Forest, R. E. and Lierle, D. M. (for AMA Council on Physical
Medicine and Rehabilitation). Principles for evaluating hear-
ing loss. J. Amer. Med. Assoc., 59, 550-552 (1955).
Dickman, D. M. The effects of three noise backgrounds on the sen-
tence intelligibility scores of hearing-impaired listeners.
Doctoral dissertation, Univ. Maryland (1974).
Elkins,, E. Evaluation of Modified Rhyme Test results from impaired-
and normal-hearing listeners. J. Speech Hearing Res., 14, 589-
595 (1971).
-------
85
Elkins, E. Causey, G. D., Beck, L. B., Brewer, D., and de Moll, J.B.
Normal and impaired listener performance on the University of
Maryland Revised CID Sentence Lists. (Unpub. report) Univ. of
Maryland (1975).
EPA, U.S. Environmental Protection Agency, Occupational noise exposure
regulation: Request for review and report. Fed. Register, 39,
43802-43809 (1974).
EPA, U.S. Environmental Protection Agency, Comments on Occupational
Safety and Health Administration's Draft Environmental Impact
Statement, "Proposed Regulation: Noise" (1975).
Fairbanks, G. Test of phonemic differentiation: The Rhyme Test,
J. Acoust. Soc. Amer., 30, 596-600 (1958).
Feldman, R. and Reger, S. , Relations among hearing, reaction time, and
age. J. Speech Hearing Res., 10, 479-495 (1967).
Fletcher, H. Speech and Hearing. New York: Van Nostrand (1929).
Fletcher, H. A method of calculating hearing loss for speech from an
audiogram. J. Acoust. Soc. Amer., 22, 1-5 (1950).
Fletcher, H. The perception of speech sounds by deafened persons.
J. Acoust. Soc. Amer., 7A_, 490-497 (1952).
Fletcher, H. and Gait, R. H. The perception of speech and its rela-
tion to telephony. J. Acoust. Soc. Amer., 22, 89-151 (1950).
Fowler, E. P. A Simple method of measuring percentage of capacity
for hearing speech. Arch. Otolaryngol., ^6_, 874-890 (1942).
Fowler, E. P. Symposium on noise: (a) the percentage of capacity to
hear speech and related disabilities. Laryngoscope, 57, 103-113
(1947).
French, N. R. and Steinberg, J. C. Factors governing the intelligi-
bility of speech. J. Acoust. Soc. Amer., 1.9, 90-119 (1947).
Gardner, M. B. Effect of noise, system gain, and assigned task on
talking levels in loud speaker communication. J. Acoust. Soc.
Amer., 40, 955-965 (1966).
von Gierke, H. E. Testimony at the public hearings on the proposed
standard for occupational exposure to noise, U. S. Department of
Labor, OSHA (1975).
Ginnold, R. Workmen's compensation for hearing loss in Wisconsin.
Labor Law J., 682-697 (Nov. 1974).
Giolas, T. G. Comparative intelligibility scores on sentence lists
and continuous discourse. J. Aud. Res., 6y 31-38 (1966).
-------
86
Giolas, T. G. and Epstein. Comparative intelligibility of word lists
and continuous discourse. J. Speech Hearing Res., 6, 349-358
(1963).
Glorig, A. and Baughn, W. L. Basis for percent risk table. In
Proceedings of the International Congress on Noise as a Public
Health Problem, U.S. EPA Eeport 550/9-73-008 (1973)."
Goetzinger, C., Proud, G., Dirks, D., and Embrey, J. A study of hearing
in advanced age. Arch. Otolaryngol., 73, 662-673 (1961).
Guild, E. Personal protection. In Industrial Noise Manual, (2nd Ed.).
Detroit: AIHA (1966).
Guilford, J. P. Fundamental Statistics in Psychology and Education,
(4th Ed.). New York: McGraw-Hill (1965).
Harris, C. M. (Ed.), Handbook of Noise Control. New York: McGraw-
Hill (1957).
Harris, J. D. Pure-tone acuity and the intelligibility of everyday
speech. J. Acoust. Soc. Amer., .37, 824-830 (1965).
Harris, J. D., Haines, H. L., and Myers, C. K. A new formula for
using the audiogram to predict speech hearing loss. Arch.
Otolaryngol., 63_, 158-176 (1956), also Erratum. 64, 477 (1956).
Harris, J. D., Haines, H. L., and Myers, C. K. The importance of
hearing at 3kc for understanding speeded speech. Laryngoscope,
70_, 131-146 (1960).
Harris, J. D., Haines, H. L., Kelsey.- P. A., and Clack, T. D. The
relation between speech intelligibility and the electroacoustic
characteristics of low fidelity circuitry- J. Aud. Res., _!,
357-381 (1961). ~
High, W. S., Fairbanks, G., and Glorig, A. Scale for self-assessment
of hearing handicap. J. Speech Hearing Pis., 29, 215-230 (1964).
Hirsh, I. J., Reynolds, E. G., and Joseph, M. Intelligibility of
different speech materials. J. Acoust. Soc. Amer., 26, 530-
538 (1954).
House, A. S., Williams, C. E., Hecker, M. H. L., and Kryter, K. D.
Articulation-testing methods: Consonantal differentiation
with a closed-response set. J. Acoust. Soc. Amer., 37, 158-
166 (1965).
ISO International Organization for Standardization, ISO Standard
1999 - Acoustics - Assessment of Occupational Noise Exposure for
Hearing Conservation Purposes (1971).
-------
87
130 International Organization for Standardization, Recommended
Methods for Measuring the Intelligibility of Speech. First
draft proposal of ISO/TC 43/SC 1/WG 12 (Nov. 1975).
Jerger, J. Audiological findings in aging. Advances in Oto-rhi-
nolaryngology, _20, 115-124, Basel (1973).
Klockhoff, I., Drettner, B., Hagelin, K. W., and Lindholm, L. A
method for computerized classification of pure tone screening
audiometry results in noise-exposed groups. Acta Otolaryngol.,
75_, 339-340 (1973).
Knudsen, V. 0. The sensibility of the ear to small differences of
intensity and frequency. Phys. Rev., _n, 84-102 (1923).
Korn, T. S. Effect of psychological feedback on conversational
noise reduction in rooms. J. Acoust. Soc. Amer., 26, 793-794
(1954).
Kreul, E. J., Nixon, J. C., Kryter, K. D., Bell, D. W., Lang, J. S.,
and Schubert, E. D. A proposed clinical test of speech discrim-
ination. J. Speech Hearing Res., 11, 536-552 (1968).
Kryter, K. D. Effects of ear protective devices on the intelligibility
of speech in noise. J. Acoust. Soc. Amer., 18, 413-417 (1946).
Kryter, K. D. Hearing Impairment for speech. Arch. Otolaryngol.,
77, 598-602 (1963).
Kryter, K. D. The Effects of Noise on Man. New York: Academic (1970).
Kryter, K. D. Speech communication. In H. P. Van Cott, and R. G.
Kinkade (Eds.), Human Engineering Guide to Equipment Design.
Washington, DC: Amer. Inst. for Res. (1972).
Kryter, K. D. Impairment to hearing from exposure to noise.
J. Acoust. Soc. Amer., ^3, 1211-1234 (1973).
Kryter, K. D. Hearing impairment and damage risk from noise exposure.
Testimony at the public hearings on the proposed standard for
occupational exposure to noise. U.S. Department of Labor, OSHA
(1975).
Kryter, K. D. and Whitman, E. C. Some comparisons between rhyme and
PB-word intelligibility tests. J. Acoust. Soc. Amer., 37, 1146
(1965).
Kryter, K. D., Williams, C., and Green, D. M. Auditory acuity and
the perception of speech. J. Acoust. Soc. Amer., 34, 1217-
1223 (1962).
Kuzniarz, J. J. Hearing loss and speech intelligibility in noise.
In Proceedings of the International Congress on Noise as a
Public Health Problem. U.S. EPA Report 550/9-73-008 (1973) .
-------
88
Liden, G. Medicolegal problems in the assessment of deafness.
Acta Otolaryngol., Suppl. 158 (1960).
Lierle, D. M. (for the Subcommittee on Noise of the AAOO Committee
on Conservation of Hearing). Guide for the evaluation-of
hearing impairment. Trans. Amer. Acad. Ophthal. Otolaryng.,
63, 236-238 (1959).
Lindeman, H. E. Relation between audiological findings and complaints
by persons suffering noise-induced hearing loss. J. Amer. Indus.
Hyg. Ass., _32, 447-452 (1971).
Littler, T. S. Physics of the Ear. London: Pergamon (1965).
MacKeith, N. W. and Coles, R. R. A. Binaural advantage in hearing of
speech. J. Laryng. Otol., 85_, 213-232 (1971).
Mayer, C. The perception of speech in noise as a function of age.
Doctoral dissertation, City University, N. Y. (1975).
Merluzzi, F. and Hinchcliffe, R. Threshold of subjective auditory
handicap. Audiol., 12_, 65-69 (1973).
Meyer, A. F. Testimony at the public hearings on the proposed standard
for occupational exposure to noise. U.S. Department of Labor,
OSHA (1975).
Miller, G. A. The masking of speech. Psychol. Bull., 44, 105-129
(1947).
Millin, J. P. The effect of small room reverberation on discrimination
tests. Doctoral dissertation, Case Western Reserve Univ. (1968).
Miner, R. and Danhauer, J. L. Articulation functions for normal and
hearing impaired subjects on the MRT and SSIT in multitalker noise.
Paper presented at the 91st meeting of the Acoustical Society of
America. Washington, DC (1976).
Moncur, J. P. and Dirks, D. D. Binaural and monaural speech intelli-
gibility in reverberation. J. Speech Hearing Res., 10, 186-195
(1967).
Mullins, C. J. and Bangs, J. L. Relationships between speech discrim-
ination and other audionetric data. Acta Otolaryngol., 47,
149-157 (1957). —
Murry, T. and Lacroix, P. G. Speech discrimination in noise and hear-
ing loss at 3000 Hz. Nav. Sub. Med. Res. Lab. Report No. 719,
Groton, Conn. (1972).
Myers, C. K. and Angermeier, C. Hearing loss at 3kHz and the Qiaba
"Proposed Clinical Test of Speech Discrimination in Noise."
Nav. Sub. Med. Res. Lab. Report No. 720 Groton, Conn. (1972).
-------
89
Nabelek, A. K. and Pickett, J. M. Fteception of consonants in the
classroom as affected by monaural and binaural listening, noise,
reverberation, and hearing aids. J. Acoust. Soc. Amer., 56, 628-
639 (1974a) . —
Nabelek, A. K. and Pickett, J. M. Monaural and binaural speech per-
ception through hearing aids under noise and reverberation with
normal and hearing-impaired listeners. J. Speech Hearing Res.,
11, 724-739 (1974b).
Nett, E. M., Doerfler, L. G., and Matthews, J. The relationships
between audiological measures and handicap. Unpub. report,
Univ- of Pittsburg Sch. Med. (1959).
Newby, H. A. Audiology, (2nd ed.). New York: Appleton-Century-
Crofts (1964).
Niemeyer, W. Speech discrimination in noise-induced deafness. Int.
Aud., 6., 42-47 (1967).
NIOSH, U.S. Department of Health, Education, and Welfare, National
Institute for Occupational Safety and Health, Criteria for a
recommended standard: Occupational exposure to noise. HSM 73-
1001 (1972).
Northern, J. L. and Hattler, K. W. Evaluation of four speech discrim-
ination test procedures on hearing impaired patients. J. Aud.
Res., suppl. 1 (1974).
OSHA, U.S. Department of Labor, Occupational Safety and Health Adminis-
tration, Draft Environmental Impact Statement: Proposed Regula-
tion, Noise. (1975).
Pearsons, K. S., Bennett, R. L., and Fidell, S. Speech levels in vari-
ous environments. Bolt, Beranek and Newman, Inc., BBN Report No.
3281 (1976).
Pickett, J. M. Effects of vocal force on the intelligibility of
speech sounds. J. Acoust. Soc. Amer., 28^, 902-905 (1956).
Pickett, J. M. Limits of direct speech communication in noise.
J. Acoust. Soc. Amer., 30_, 278-281 (1958).
Quiggle, R. R., Glorig, A., Delk, J. H. and Summerfield, A. B.
Predicting hearing loss for speech from pure-tone audiograms.
Laryngoscope, 67, 1-15 (1957).
Quist-Hanssen, S. and Steen, E. Observed and calculated hearing
loss for speech in noise-induced deafness. Acta Otolaryng.,
suppl. 158, 277-281 (1960).
-------
90
Risberg, A. and Martony, J. A method for classification of audio-
grams. In G. Fant (Ed.), Proceedings of the Symposium on
Speech Communication Ability and Profound Deafness, Washington,
DC: A. G. Bell Assoc. for the Deaf (1972).
Robinson, D. W. (Ed.) Occupational Hearing Loss. New York: Academic
(1971).
Ross, M. Huntington, D. A., Newby, H. A., and Dixon, R. F. Speech
discrimination of the hearing-impaired individual in noise.
J. Aud. Res., _6, 47-72 (1965).
Rudmose, W. Hearing loss resulting from noise exposure. In C. M.
Harris (Ed.), Handbook of Noise Control. New York: McGraw-Hill
(1957).
Sabine, H. J. Less Noise Better Hearing. Chicago: Celotex Corp.
(1950).
Sabine, P- E. On estimating the percentage loss of useful hearing.
Trans. Amer. Acad. Ophthal. Otolaryng., ^6_, 179-196 (1942).
Scott, W. A. and Wertheimer, M. Introduction to Psychological
Research. New York: John Wiley & Sons (1964).
Shower, E. G. and Biddulph, R. Differential pitch sensitivity of
the ear. J. Acoust. Soc. Amer., _3/ 275-287 (1931).
Silverman, S.R. and Hirsh, I.j. Problems related to the use of
speech in clinical audiometry. Ann. Otol. Rhinol. Laryng.,
64_, 1234-1244 (1955).
Silverman, B., Bilger, R. C., Hanley, T. D., and Steer, M. D. The
relative intelligibility of male and female talkers. J. Educ.
Psychol., 44_, 418-428 (1953).
Simonton, K. M. and Hedgecock, L. D. A laboratory assessment of
hearing acuity for voice signals against a background of noise.
Ann. Otol., 62_, 735-747 (1953).
Steinberg, J. C. and Gardner, W. B. On the auditory significance of
the term hearing loss. J. Acoust. Soc. Amer., 11, 270-277 (1940).
Suter, A. H. and von Gierke, H. E. Evaluation and compensation of
occupational hearing loss in the United States. In Proceedings
of the Congress, Man and Noise. Turin, Italy: Edizioni Minerva
Medica (1975).
Thomas, W. G., Testimony at the public hearings on the proposed
standard for occupational exposure to noise. U.S. Department of
labor, OSHA (1975).
-------
91
Til Man, T. W. and Carhart, R. An expanded test for speech discrimi-
nation utilizing CNC monosyllabic words. USAF School of Aero-
space Med., Brooks AFB, SAM-TR-66-55 (1966).
UK-Member Body of ISO/TC 43/SC1, Proposal for a revision of ISO 1999 -
Acoustics - Assessment of Occupational Noise Exposure for Hearing
Conservation Purposes (1975).
U.S. Air Force, Aerospace Medicine, Hazardous Noise Exposure. APR
161-35, Attachment 9 (27 July 1973).
Utley, J. The relationship between speech sound discrimination and
percentage hearing loss. J. Speech Hearing Dis., 9, 103-113
(1944). ~
VA, U.S. Veterans Administration, Rating procedure relative to specific
issues: 50.07 rating of hearing impairment. Dept. Med. and
Surgery Manual, M21-1 (1976).
Ward, W. D., Fleer, R. E., and Glorig, A. Characteristics of hearing
losses produced by gunfire and by steady noise. J. Aud. Res.,
.2JU 325-356 (1962).
Webster, J. C. Important frequencies in noise-masked speech.
Arch. Otolaryngol., 8C^, 494-504 (1964).
Webster, J. C. Effects of noise on speech intelligibility in W. D.
Ward and J. E. Fricke (Eds.), Noise as a Public Health Hazard,
ASHA Report No. 4. Washington, DC: American Speech and Hear-
ing Association (1969)-
Webster, J. C. Compendium of speech testing materials and typical
noise spectra for use in evaluating communications equipment.
Naval Electronics Laboratory Center, Human Factors Technology
Division, Tech. Doc. 191 (13 Sept 1972).
Webster, J. C. The effects of noise on the hearing of speech.
In Proceedings of the International Congress on Noise as a
Public Health Problem, U.S. EPA Report 550/9-73-008 (1973).
Webster, J. C. and Klumpp, R. G. Effects of ambient noise and nearby
talkers on a face-to-face communication task. J. Acoust. Soc.
Amer., 34_, 926-941 (1962).
Webster, J, C., Davis, H., and Ward, W. D. Everyday speech intelligi-
bility. J. Acoust. Soc. Amer., 38, 668 (1965).
Winer, B. J. Statistical Principles in Experimental Design (2nd ed.).
New York: McGraw-Hill (1971).
-------
I. REPORT NO.
TECHNICAL REPORT DATA
fi-cse read Instructions on the reverse before completing)
EPA-55Q/9-7R-inn
2.
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
The Ability of Mildly Hearing-Impaired Individuals to
Discriminate Speech in Noise
5. REPORT DATE
Jan. 1978
6. PERFORMING ORGANIZATION CODE
Project #7321, Task #03
7. AUTHOR(S)
Alice H. Suter
8. PERFORMING ORGANIZATION REPORT NO.
AMRL-TR-78-4
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Aerospace Medical Research Laboratory Wright-Patterson
Air Force Base, Ohio
10. PROGRAM ELEMENT NO.
62202F
11. CONTRACT/GRANT NO.
12. SPONSORING AGSNCY NAME AND ADDRESS
EPA-ONAC
Washington D.C. 20460
13. TYPE OF REPORT AND PERfOO COVERED
Final
14. SPONSORING AG_EJSICY COQE,
EPA/ONAC
15. SUPPLEMENTARY NOTES
16. ABSTRACT
The investigation explores the relationship between hearing level at various audio-
metric frequencies and speech discrimination in different noise backgrounds." The
study was designed specifically to test the American Academy of Ophthalmology and
Otolaryngology's (AAOO) selection of a 26-decibel average of 500, 1000 and 2000 Hz
as the'point above which hearing handicap occurs. The AA00 method for computing
hearing handicap has lately been brought into question for-two primary-peft&Qfts; that
the 26-dB fence is too high, and for the exclusion of frequencies above 2000 Hz.
The present study, therefore, attempted to see if there were differences among
individuals whose hearing was at or better than the low fence, arrd if so,what factors
caused or affected the differences. Forty-eight subjects were tested with two types
of speech materials: the University of Maryland Test #1 which employs simple, "every-
day" sentences, and the Modified Rhyme Test, a closed-set test of rhyming monosyll-
ables. Speech stimuli were presented at 60 dBA measured at the listener s ear. The
noise stimulus, a babble of twelve voices, was presented at levels of 60 to 66 .dBA.
Subjects were divided into three groups according to their hearing levels at 500,
1000, and 2000 Hz. One group had normal hearing at all tested audiometric frequencies
and the other two had mild hearing losses in the mid-frequencies and considerable
amounts of loss in the high frequencies, which is typical of noise-hearing loss.
Subjects listened to both speech materials in a quiet condition and in (cont. on back)
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b.lDENTIFIERS/OPEN ENDED TERMS C. COSATI field/Group
Hearing Impairment,
Speech discrimination
Noise, hearing handicap fence
13. DISTRIBUTION STATEMENT .... rnft/minr
Limited Supply Available at EPA/ONAC
and Research Traingle Park, S.C.
19. SECURITY CLASS (ThisReport)
Unclassified
20. SECURITY CLASS (Thispagt)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
-------
three levels of background noise. Group scores were compared in a three-
factor analysis of variance and correlations between audiometric freq-
uencis and individual discrimination scores were performed. Conclusions
are drawn about the differences between the three groups, the effect of
increasing noise on the discrimination of speech by hearing-impaired
individuals, the equivalence of two kinds of speech materials, and the
importance of certain audiometric frequencies to the discrimination of
speech. Recommendations are made for computing hearing handicap in
terms of the height of the fence and the inclusion of frequencies
above 2000 Hz.
U S. GOVERNMENT PRINTING OFFICE: 1978— 260-830/5
------- |