[*«EQNaSFEECH mam DESCQEMMAFEGN
EN SUSfiECW WE‘EH EMPAEREQ
E€EARENG

“tasks éor i'iua Degree (925 pk. D.
MECHEGAN HATE WIVE‘ES‘LE“.

P'aui S‘ Niswwéer
1968

“LIBRARY ”“
mags ‘ Michigan State
’ ” University (

 

This is to certify that the

thesis entitled

NON-SPEECH SOUND DISCRIMINATION
IN SUBJECTS WITH IMPAIRED
HEARING

presented by

Paul S. Niswander

has been accepted towards fulfillment
of the requirements for

(lflXQM/L/

’ mg; ”Gamay

Date /4//;/C/D 9
/

0-169

 

 

 

 

 

fx ‘,'- ”3'1 may .
’ ‘ "fi' ‘ —.. ' ' A
l 1‘ ‘ .- ~_ 7“ 1’ .
'u 4.1. 0.1
‘n
In
t/
‘ l
.
1% ' ’ 331.4,. ‘
.
“2

‘ ‘. “‘3’. J‘v‘_- ,y _‘ ‘ ; ‘ ‘ .:
‘Wft g, (71.“ ‘ ‘ L 1
i ‘ “Hts . .~ , ‘ :

Stmdtthq.‘ ~ . , _ ‘ ‘ ‘ I.

' l
.. fit
., . ,\
L“. as»; ,L :. ‘ .

35' “atrium“ ;".

 
   
   
   
    

saw-tun) : .
.li. ‘, :20.)

’ mm}. thing “2' ._ - ”.-
9 M, (f-
.~ ”Lawytcm L‘-‘ :1 . 3 ‘

‘I

NWCQIA tn 3.171..

'.
, A teat: What“.

 

ml Con :7 t r neg»- , w, 9

_-.UI1 1‘83: 35-? T. 4-. _ 5; - .737

 

.-

, .

"la-'KM‘ - ,fi.
)‘ - I”. _ ‘_‘.'»_.'- .‘ ‘

mi

ABSTRACT
NON-SPEECH SOUND DISCRIMINATION

IN SUBJECTS WITH IMPAIRED
HEARING

BY

Paul S. Niswander

Although speech stimuli provide the basis for much
human interaction, a large proportion of an individual's
sound environment is composed of meaningful, non-speech
sounds. Some non-speech sounds serve warning or alerting L
functions and thereby are important to the physical well- '
being of the individual. Other non—speech sounds provide

additional information and assist man to adjust more suc-

cessfully to his environment.

The characteristics important in discriminations

-r ...,_ s,

involving non-speech sounds are not known. This research
was conceived in an attempt to evaluate some of these
parameters in subjects with impaired hearing. h

A test consisting of fifty familiar, non-speech It
sounds was constructed. The stimuli included animal cries,
mechanical noises, musical instruments, and human non-speech
sounds.

The test was administered to sixteen subjects with

impaired hearing. The subjects were divided into two

 

T—_———

Paul S. Niswander

.groups of equal size on the basis of the extent of the
hearing loss. Class I subjects had average thresholds for
the speech frequencies (500 Hz, 1000 Hz and 2000 Hz)

.between 20 dB and 40 dB ISO in the better ear. Subjects
assigned to Class II had average thresholds between 40 dB
and 60 dB ISO in the better ear at these frequencies.

Initial pure-tone thresholds were determined for
each ear at 250 Hz, 500 Hz, 1000 Hz, 2000 Hz, 4000 Hz,
6000 Hz, and 8000 Hz. The speech reception threshold was
determined for binaural earphone presentation. Lists of
CID W—22 words were administered binaurally through ear—
phones at -8 dB, 0 dB, +8 dB, +16 dB, and +24 dB relative
to the SRT. A non-speech sound list was presented at each
of the same sensation levels. Each subject was tested
twice with at least two weeks between the test sessions.
Spectral analysis of the non-speech sounds provided the
frequency characteristics of each sound.

The results of the study indicated that the dis-

’ crimination functions (sensation level vs. discrimination

score) for the non-speech sounds very closely resembled

those obtained from the speech test. The slope of the
linear portion of the non-speech function was slightly
less than the slope of the corresponding portion of the
speech function. In addition, the correlation between the

speech and non—speech scores at the higher sensation levels

 

   

‘W———_——'

is

Paul S. Niswander

was quite high. The results suggest that similar processes
mediate the discriminations of these stimuli.

High reliability coefficients were obtained for the
speech and non-Speech tests of auditory discrimination.
There was no significant difference between the correla—
tions obtained from the two tests.

No significant correlations were found between the
audiometric configurations of the subjects and the spectra
of the sounds which were identified correctly at each sen—
sation level. Moreover, the various frequency bands of
the non-speech sounds were discovered to have no differen-

tial effects in their contributions to intelligibility.

 

 

NON-SPEECH SOUND DISCRIMINATION
IN SUBJECTS WITH IMPAIRED

HEARING

BY

\0
6
Paul SieNiswander

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of,

DOCTOR OF PHILOSOPHY
Department of Audiology and Speech Sciences

1968

 

#—
- 455/0 67
343" 67

TABLE OF CONTENTS
Page

LIST OF TABLES I I o I I I I o I I I I I I I I I I I iV

LIST OF FIGURES C O O O I O O O O O I Q C O O I O 0 v
Chapter

I. STATEMENT OF THE PROBLEM . . . . . . . . . . 1

Introduction . . . . . . . . . . . . . . 1

Purpose of the Study . . . . . . . . . . . 6

Definition of Terms . . . . . . . . . . . 7

II. REVIEW OF SENSORY DISCRIMINATION

I
I
I
I
O
o
00

Introduction . . . . . . . . . . . . . . . 8
Speech Discrimination . . . . . . . . . 8
Non- -Speech Sound Discrimination . . . . . 27

III. SUBJECTS, MATERIALS, EQUIPMENT AND
PROCEDURES . . . . . . . . . . . . . . . . 30

Introduction . . . . . . . . . . . . . . . 30
Subjects . . . . . . . . . . . . . . . 30
Stimulus Materials . . . . . . . . . . . . 31
Equipment and Procedures . . . . . . . . . 34
Response Scoring . . . . . . . . . . . 37

Additional Equipment and Procedures for
Analyzing the Sound Tapes . . . . . . . 33

IV. ANALYSIS OF THE DATA AND DISCUSSION . . . . 42

Introduction . . . . . . . . . . . . . . 42
Summary of the Data . . . . . . . . . . . 42
Discussion . . . . . . . . . . . . . . . . 51

V. SUMMARY, CONCLUSIONS AND RECOMMENDATIONS . . 65

Introduction . . . . . . . . . . . . . . . 65

Summary . . . . . . . . . . . . . . . . . 65

Conclusions . . . . . . . . . . . . . . . 68

Recommendations . . . . . . . . . . . . . 69
ii

 

     

fiIIOGRAPHYAI I I I I I I I I'I I I I I I I I‘I I
9&2: dix A Summary of the Non-Speech Sound Lists,

Sva;ndix B Sounds Ranked from Least Difficult
to Most Difficult I I ‘I I I I I I I

111 ‘

 

 

Page
71
‘73

74

 

LIST OF TABLES

Table Page
1. Non-Speech Test and Retest Discrimination
Scores for Class I Subjects . . . . . . . . 43
2. Speech Test and Retest Discrimination
Scores for Class I Subjects . . . . . . . . 44
3. Non-Speech Test and Retest Discrimination

Scores for Class II Subjects . . . . . . . 45

4. Speech Test and Retest Discrimination

Scores for Class II Subjects . . . . . . . 46
5. Slopes in the Linear Region of the Dis-

crimination Function for Class I and

Class II Subjects . . . . . . . . . . . . . 52
6. Differences in Discrimination Scores

. Between Class I and Class II Subjects
for the Non-Speech and Speech Tests . . . . 54

7. Test-Retest Reliability Coefficients of the
Speech and Non-Speech Sounds at the
Various Sensation Levels . . . . . . . . . 57

8. Product—Moment Correlations Between the
Results of the Speech and Non-Speech
Discrimination Tests . . . . . . . . . . . 59

9. Summary of Statistic for Testing Differences
Between the Correlations Obtained from
Class I and Class II Subjects . . . . . . . 59

10. Product-Moment Correlations Between the
Relative Intensity Levels of the Ident-
ified Sounds at a Given Frequency and
the Threshold of the Subjects at the
Same Frequency . . . . . . . . . . . . . . 62

. ll. Ninety-Five Percent Confidence Intervals

. Around Zero Correlation for the Various
1 Sensation Levels . . . . . . . . . . . . . 62

iv

-‘amw. ‘-

Figure

1.

2.

10.

11.

LIST OF FIGURES

Page

Average Percent Articulation as a Function

of Number of Sounds per Syllable . . . . . 11
Relative Intelligibility of Three Types of

Speech Stimulus Material as a Function

of Relative Intensity . . . . . . . . . . l4
Articulation Scores for Three Different

Types of Test Material as a Function

of Signal-to—Noise Ratio . . . . . . . . . 15
Average Relative Intelligibility of Words

and Sentences Under a Wide Variety of

Listening Conditions . . . . . . . . . . . 15
Frequency-Power Distribution of Speech . . . 17

Articulation Score for Monosyllabic Words
as a Function of the SPL of the Speech . . 18

Syllable Articulation for Speech Passed
Through High Pass and Low Pass Filters
in the Quiet O I I O I O O D C O O O O I O 20

Block Diagram of Equipment Used for
Spectral Analysis of the Non—Speech
Sounds 0 I I O I I I O O O I I O O O I I C 4 o

Non-Speech Discrimination Score as a Function
of Intensity for Class I Subjects . . . . 47

Non—Speech Discrimination Score as a Function
of Intensity for Class II Subjects . . . . 48

Non-Speech Discrimination Scores as a Function
of Intensity for Class I and Class II
SllbjeCts O O I O O I O I O I O I I I I C O 48

"v.

Figure Page

12. Speech Discrimination Score as a Function
of Intensity for Class I Subjects . . . . . 49

13. Speech Discrimination Score as a Function
of Intensity for Class II Subjects . . . . 50

14. Speech Discrimination Scores as a Function
of Intensity for Class I and Class II
Subjects . . . . . . . . . . . . . . . . . 50

15. Mean Speech and Non-Speech Discrimination

Scores as a Function of Intensity for
all Subjects . . . . . . . . . . . . . . . 51

vi

 

CHAPTER I

STATEMENT OF THE PROBLEM

Introduction

Pure—tone and speech audiometry comprise the basic
armamentarium of the clinical audiologist. More recently,
special tests have been developed for the differential
diagnosis of auditory pathologies and now are administered
routinely in many clinics as part of the audiological
assessment battery.

The pure-tone tests determine certain performance
characteristics of the auditory mechanism. Depending on
the site of the lesion, the impaired hearing patient gives
certain characteristic responses to various types of audi-
tory stimuli. The special hearing tests were developed to
investigate these pecularities and thereby provide site of
lesion information. None of these tests provides directly,
however, information regarding an individual's performance
in complex and meaningful auditory environments. This can
only be inferred from the results of such tests.

Tests utilizing speech stimuli were developed
originally as a means of investigating the characteristics
of various communication systems. Reasoning that the

human auditory mechanism could be regarded as a special

1

7.

type of (or vital link in any) communication network,
scientists applied speech tests to the assessment of human
audiological performance. In conjunction with the pure-
tone tests, the speech tests provided vital site of lesion
information. Perhaps even more important, however, was the
fact that these tests assessed human auditory performance
with meaningful auditory stimuli.

DeSpite the fact that speech stimuli provide the
basis for much human interaction, a large proportion of an
individual's sound environment is composed of meaningful
non—speech sounds. Some non-speech sounds serve warning
or alerting functions and thereby are important to the
physical well-being of the individual. The automobile and
train horns, the air raid siren and the railroad crossing
bell all are examples of formal alerting devices. That
is, they exist mainly to provide alerting information.
Other sounds may serve an informal alerting function, pro-
viding alerting information only within a specific context.
The gun shot at the firing range or in the hunting field
may have very little alerting value. But gun shots
shattering a quiet summer evening in a residential neigh-
borhood may convey highly alerting information.

Other non-speech sounds provide additional informa-
tion and thereby assist man to adjust more successfully to
his environment. The sound of rain falling on the roof;

of the neighbor's power mower; of the clock ticking in the

hall; all provide information regarding the condition of

the environment. Man's knowledge of the true state of his
environment is based on successful reception and recognition
of such sounds.

Some human non-speech sounds may serve as an impor-
tant vehicle of communication between individuals. A sigh
or audible yawn may, within a given situation, communicate
as much information as a verbal remark. Likewise, a hearty
laugh may convey much more information than a whole series
of words.

Because of the importance of non—speech sounds, it
is indeed unfortunate that so little research effort has
been expended in specifying the relevant parameters of such
stimuli. Such investigations could provide knowledge
having both theoretical and clinical implications.

Consider, for a moment, studies of speech intelli-
gibility. Investigations using speech stimuli have revealed
some of the physical and psychological factors important in
speech discrimination.

The physical parameters of the stimuli can usually
be specified exactly. Intensity, frequency bandwidth, and
frequency spectrum have been accurately measured for the
various speech sounds. Manipulation of these parameters,
singly and in combination, has revealed the relative con-

tribution of each to discrimination.

The psychological factors associated with speech
material are well known but their effect on discrimination
is more difficult to quantify. Type of stimulus material,
familiarity of items, and the effect of practice all con-
tribute their share to speech intelligibility. The physical
and psychological factors of the stimulus material interact
to provide the cues for discrimination.

In addition, the state of the auditory mechanism
of the listener is a relevant factor in the measured dis-
crimination performance. Some types of impairments affect
speech discrimination more than other types. Physical
manipulation of the physical parameters of the speech
stimuli has more effect in some types of impairments.

In contrast, none of the above information is
available for situations in which meaningful non-speech
sounds were utilized. One would expect many of the same
parameters to be relevant for both speech and non-speech
material.

The intensity of any auditory stimulus reaching
the listener's ears is of paramount importance in detection
and discrimination. Certain relationships exist between
the intensity and intelligibility of speech materials. Do
the same relationships hold for meaningful non-speech
stimuli? If the relationships are similar, with what type
of speech stimulus material do the non-speech functions

most directly compare?

Frequency filtering affects speech intelligibility.
Does this type of filtering also affect the discrimin-
ability of non-speech stimuli? How do different types of
filtering change the intelligibility? Are some frequency
bands more important than others?

It has been noted that psychological factors may
be associated with meaningful non-speech sound stimuli.
How do these factors affect intelligibility? Are the same
factors important for speech and non-speech stimuli? Are
non—speech sounds which have a formal alerting connotation
more intelligible under a given set of listening conditions
than those which do not? How do changes in listening
situations differentially affect sounds of various
psychological connotation?

Is the discriminability of meaningful non-speech
sound stimuli related to the type of hearing loss in a
subject? Is there a relationship between the spectra of
identified sounds and the measured pure-tone thresholds in
a listener? Can the results of non-speech discrimination
tests be used to predict an individual's success in adjust—
ing to his auditory environment?

The quantity and variety of unanswered questions
such as posed above regarding the discrimination of mean-
ingful non-speech sounds reveal the need for research in

this area. The present study was conceived in such a light.

 

,-~..r-_

It was not meant to be an investigation of all relevant

factors in non—speech discrimination. Rather, the research

was intended to be an initial step in evaluating some of

these parameters.

Purpose of the Study

The purpose of the present study was to evaluate

some of the characteristics of non-speech sound discrimina—

tion in hearing impaired subjects. A tape recorded test

consisting of fifty meaningful non-speech sounds was

developed and administered to the subjects. The following

questions were posed as a means of specifying some of the

relevant parameters of non-speech sound discrimination in

these subjects:

1.

What is the relationship between the measured
discrimination score and the intensity of
presentation?

What is the relationship between the measured
discrimination score and the extent of the
individual's hearing loss?

What is the relationship between the individual's
pure—tone threshold configuration and the fre-
quency spectra of the sounds correctly identi-
fied?

Is there a difference in the relative contri-
butions to intelligibility of the various
frequency bands of the sounds?

What is the relationship between speech and
non-speech discrimination scores?

What is the test-retest reliability of the test?

Definition of Terms

Meaningful.—-The discrimination data described in
the previous section were obtained by means of a test in-
corporating meaningful non-speech sound stimuli. In this
context, the term "meaningful" can be used interchangeably
with "environmental." That is, the sounds were not syn—
thesized or generated in the laboratory specifically for
the purposes of this study.

Many of the sounds were those which the subjects
might have encountered in their everyday activities. How-
ever, some of the sounds might have been encountered only
in a highly specific situation. The sounds of some musical
instruments, while familiar to most people, might be heard
only in the concert hall. Likewise, the cries of particular
animals might be encountered in a visit to a zoo.

Non-Speech.--The term "non—speech" was sufficiently
unambiguous for most of the sounds, since they were gener-
ated mechanically. Several human sounds of the non-speech
variety were included, however. It has been mentioned that
such sounds can, in certain situations, convey information
between individuals. These sounds are nevertheless non-
speech in that they do not contain formal, linguistic con-

tent. Thus, they are meaningful but non—speech.

CHAPTER II

REVIEW OF SPEECH AND NON-SPEECH DISCRIMINATION

Introduction

This chapter briefly reviews some of the
characteristics of speech and non-speech sound discrimina-
tion. The literature concerned with speech discrimination
is quite vast and no attempt has been made to cite all of
the relevant studies. Rather, the review covers some of
the more salient features of speech discrimination.

Very little information is available concerning
the important parameters of non-speech sound discrimination.
This review includes the few studies that have concentrated

on this type of discrimination.

Speech Discrimination

Introduction

Discriminations involving speech stimuli are quite
complex and involve higher-order neural mechanisms. The
results of intensity and frequency discrimination studies
can be analyzed neatly in terms of the physical parameters
of the stimuli involved. Performance is predictably re—

lated to changes in these parameters. Speech discrimination

performance is not so intimately related to the physical
parameters of the stimuli, however. Psychological factors
such as meaning, context, and familiarity enter the picture.
Physical and psychological factors interact to produce a
very complicated process. This section covers, very

briefly, some of these factors.

Stimulus Materials

 

The measured discrimination score is a function of
the type of speech material presented to the listener.
Isolated speech sounds, nonsense syllables, words, phrases,
and sentences all have been used. Each involves separate
perceptual factors and has particular advantages and dis-
advantages as a test tool. As is discussed in succeeding
sections, the scores obtained with the various types of
materials under a given set of testing conditions are in-
terrelated.

The early researchers called the score obtained
with any type of speech material the "articulation score."l
More recently, speech articulation has been reserved for

describing the perception of speech sounds.2 Intelligibility,

 

1James P. Egan, "Articulation Testing Methods," The
Laryngoscope, LVIII (1948), pp. 955-991.

2Karl D. Kryter, "On Predicting the Intelligibility
of Speech from Acoustical Measures," The Journal of Speech
and Hearing Disorders, XXI (1956), pp. 208-217.

 

 

10

then, is used to refer to the discrimination of words,
phrases, or sentences.

Nonsense syllables were originally the most common
speech stimulus material. The first extensive lists were
constructed by Fletcher and Steinberg at Bell Telephone
Laboratories.1 These lists comprised sixty-six syllables
of the consonant-vowel-consonant variety. The method of
testing was to read an introductory sentence followed by
three of the syllables. Subjects wrote their responses in
phonetic symbols. Sound articulation and syllable articu-
lation (intelligibility) scores were obtained from a care-
fully selected and trained listening crew under a variety
of testing conditions. It was found that the sound articu-
lation was not dependent on the order of the sounds even
though some sound pattern combinations were more common in
the language than others. Sound articulation was discovered
to depend on the number of sounds in the syllable, espe—
cially as the number increased beyond three sounds per
syllable. Figure 1 illustrates this relationship.

An optimal transmission system was constructed and
various types of distorting systems were compared against

this system. The efficiency of an unknown system could

 

lH. Fletcher and J. C. Steinberg, "Articulation
Testing Methods," The Bell System Technical Journal, VIII
(1929), pp. 806-854.

 

p... “L— ..
n

 

11

then be quickly evaluated with respect to the optimal
reference system by means of the obtained syllable articu—

lation scores.

   

 

100
E
o c 90
OO
H-H
H 80
H
(D5
m3
544-) 70
$2
“1 60
50

 

3 4 5 6
Average Number of Sounds Per Syllable

Figure l.--Average Percent Articulation as a Function of
Number of Sounds per Syllable.

Monosyllabic words currently are the most common
test material for measuring speech discrimination ability.
Perhaps the best known series of word lists was developed
at the Psycho-Acoustic Laboratory at Harvard University
for wartime research on communication equipment.1 These
are referred to as the PB-50 lists. A rather similar set

of fifty-word lists, using a somewhat smaller and more

 

lHallowell Davis and S. Richard Silverman, Hearin
and Deafness (New York: Holt, Rinehart and Winston, 1965;,
pl

12

familiar vocabulary, was prepared at the Central Institute
for the Deaf.1 Recorded versions of both lists are avail-
able. The recorded CID W-22 lists yield somewhat higher
scores than the PB-SO lists due to greater word familiarity
and better recorded quality.

The development of the Harvard word lists proceeded
on several premises.2 First, all or nearly all of the
fundamental speech sounds should be represented by a test.
Furthermore, these speech sounds should occur within the
lists with a frequency equal to their occurrence in every—
day speech. Second, there should be a satisfactory distri-
bution of difficulty in the lists. The material should be
made reasonably sensitive to small differences in intellig-
ibility by excluding items which might be too easy or too
difficult. In addition, however, there should be a wide
distribution of difficulty. Third, the test items could
be single nonsense syllables, meaningful words, or meaning-
ful phrases and sentences in which there would be contextual
relations among the words. Each type of material has
relative advantages and disadvantages. Nonsense syllables
measure phoneme intelligibility better but require trained

speakers. For words, intelligibility is dependent on the

 

lIbid.

2Egan, The Laryngoscope, LVIII, pp. 955-991.

 

13

number of sounds. Sentence material provides the only
adequate method of satisfactorily measuring inflection,
rhythm and intonation. Responses to sentences may be
difficult to score, however, and the lists too often are
highly dependent on the intelligence or education of the
subject.

The development of the Harvard monosyllabic lists
proceeded through several steps. Directly preceeding the
PB—50 lists was a pool of 1200 words (revised monosyllabic
or RM) divided into 24 lists of 50 words each. From these
words, the PB-SO lists were constructed to satisfy the
following criteria: (1) monosyllabic structure; (2) equal
average difficulty; (3) equal range of difficulty; (4)
equal phonetic composition; (5) composition representative
of English Speech; and (6) common usage. The result of
this project was the lists of fifty words comprising the
PB-50 series.

Spondaic words are useful for establishing the
level at which speech can just be heard. These words are
of equal difficulty. Auditory tests W—l and W-2, consist-
ing of Spondaic words, were developed at the Central
Institute for the Deaf as a modification of Auditory Test

No. 9 of the Harvard Psycho-Acoustic Laboratory.1 The

 

1Davis and Silverman, pp. 182-183.

l4

latter test was developed at Harvard for wartime use at
the Army and Navy Aural Rehabilitation Centers.

Several sentence intelligibility tests have been
developed, the best known of which is the set constructed
by Fletcher and Steinberg.l Sentence tests present parti-
cular cultural and intelligence problems and are not often
used in discriminatién testing.

The relationships between the relative intellig-
ibilities of several types of speech stimuli as a function

of intensity are illustrated in Figures 2 and 3.

 

 

 

 

8 101 Selected
g Spondees
H
8 8Q Unselected Spondees
m
gee
H “—-Monosyllables
“540
4.1
8
o 20
H
o
m
0 10 20 30 40 50 60
Relative Intensity in dB
Figure 2.-- Relative Intelligibility of Three Types of

Speech Stimulus Material as a Function of Rela-
tive Intensity.

 

1Fletcher and Steinberg, The Bell System Technical
Journal, VIII, pp. 806-854.

 

 

15

   

 

“ 100
o I l
o Digits
H
3 80 Sentences
U
m
5 60 Nonsense Syllables
f.
g 40
4.)
5 20
o
H
0
D4
0

 

 

-18 -12 -6 O 6 12 18
Signal-to-Noise Ratio in dB

 

Figure 3.--Articulation Scores for Three Different Types
of Test Material as a Function of Signal-to—
Noise Ratio.
As might be expected, sentence intelligibility is
higher than word articulation in all but the most optimal

listening conditions. Figure 4 shows this relationship.

 

 

 

100
w
o
s
m.
.1.)
g.
m
U).
Q.
s
m
o
H
o
m

0

0 20 40 6U 80 100

Percent Word Articulation

Figure 4.--Average Relative Intelligibility of Words and
Sentences Under a Wide Variety of Listening
Conditions.

 

16

The contextural clues of sentences make it possible to miss
sounds or even whole words while still receiving the general
idea of the sentence.

Acoustical Correlates of
Speech Discrimination

 

Speech stimuli comprise a wide range of intensities,
not only between different levels of speaking but also be—
tween the different speech sounds. A speaker talking as
loudly as possible can radiate an average speech power of
1000 microwatts. The power in the faintest whisper may
decrease to as little as 0.001 microwatt. Thus there is a
range of about sixty decibles detween the loudest shout and
the faintest whisper.

Fletcher has calculated the relative powers of the
different speech sounds.1 The loudest vowel, /9/, was
assigned a relative phonetic power of 680. The faintest
consonant, /6/, has a relative phonetic power of 1. Thus
the power ratio between the strongest sound and weakest
sound is 680 to 1. This is a range of about 28 dB.

The range of frequencies in speech sounds also is
great.. Figure 5 illustrates the spectral distribution of

speech for a group of talkers. Although the frequency

 

lHarvey Fletcher, Speech and Hearing in Communica-
tion (Princeton, New Jersey: D. Van Nostrand Company,
Inc., 1953), p. 86.

 

l7

spectrum is quite broad, it is to be shown that not all
frequencies are necessary for high intelligibility.

-2o‘
,-30
-40
-50
-60
-70

-80

Relative Intensity-dB

 

 

-90

 

50 100 200 500 1000 2000 5000 10000
Frequency in Hz
Figure 5.—-Frequency—Power Distribution of Speech.

The distribution of energy is not constant through-
out the frequency range for most sounds. Rather, the
principal energy components tend to group into five or six
prominent frequency regions. These "formant" regions are
the result of acoustical resonances within the vocal tract.
Even though there may be a wide range of vocal qualities
between speakers, the formant regions for a given sound will
remain relatively constant.

The intensity of the speech material reaching the
listener's ear is of great importance in intelligibility.

If some type of noise masking is mixed with the speech
signal, the intelligibility becomes a function of the

signal-to-noise ratio. Figures 2 and 3 illustrated the

18

effect of intensity on the intelligibility of several types
of speech material. Figure 6 shows the relationship between
sound pressure level and the intelligibility of monosyllabic
words. The intelligibility of the material increases with
increases in intensity until the threshold of pain is
reached. At this level, further increases in intensity

may result in slightly reduced discrimination scores.

100
’8
o 80
u
H
8
6C
U)
o
3
3 4C
4.)
8
0 2(
H
o
m

 

 

 

0 20 40 60 80 100
Average Sound Pressure Level of Speech (dB)

 

Figure 6.--Articulation Score for Monosyllabic Words as a
Function of the SPL of the Speech.

The sound pressure level necessary for identifica-
tion of half the test items depends on the speech material
presented. Figure 6 shows that this value is about 33 dB
for monosyllables. This value is for monaural, earphone
presentation. If the words are presented binaurally, twice
the sound pressure is available for processing at the input

of the auditory system. Consequently, the threshold is

19

lowered (improved) by about 3 dB. Spondees are reported
correctly fifty percent of the time at 17 dB monaurally

and 14 dB binaurally. Connected discourse presented through
earphones can be followed by the listener when the speech
level is about 24 dB.

The relationship between word intelligibility and
intensity is somewhat complicated by the differential in-
tensities of the various speech sounds noted earlier. In
less than optimal listening conditions, some sounds will
be missed more consistently than others. Fletcher has
determined the articulation functions for the individual
speech sounds.1 He estimates that the sounds /v/, /f/, and
/6/ account for more than half of the phonetic mistakes
made in listening to ordinary conversations.

The intelligibility of Speech material depends also
on the range of frequencies presented to the listener. The
results of filtered speech experiments have contributed
substantially to the knowledge of frequencies important for
speech discrimination. As noted earlier, the entire speech
spectrum ranges from below 50 Hz to over 10,000 Hz. Many
of these frequencies can be deleted with but little decrease

in intelligibility.

 

lIbid., pp. 415-418.

20

Figure 7 illustrates the effect of frequency
filtering on syllable intelligibility. Although they con-
tain most of the speech power, the lower frequencies
apparently contribute little to intelligibility. When all
of the components of a speech signal below 1000 Hz are
removed, the speech power is reduced approximately eighty
percent, but the articulation score drops only ten percent.
The following figure illustrates that frequencies above
1900 Hz contribute about the same to syllable articulation

as do frequencies below 1900 Hz.

   

 

 

100
High Pass Low Pass

80

a) ' .
H
«‘3

H 60
H
m

p 40
s
o
S

0 20
04

0

100 300 1000 3000 10,000

Cutoff Frequency (Hz)

 

Figure 7.--Syllable Articulation for Speech Passed Through
High Pass and Low Pass Filters in the Quiet.

It has been shown that removal of the high fre-

quency components affects consonant articulation more than

 

21

vowel articulation.l Conversely, removal of the low
frequencies has more effect on vowel articulation scores.
Another way of viewing the frequencies important
to speech discrimination is to inspect the formants
necessary for high intelligibility. It has been noted
that most speech sounds contain five or six well defined
formant regions.2 Research has indicated that elimination
of all but the three lower formants has but little effect
on intelligibility. The higher formants appear to be more
important in transmitting particular characteristics of a
speaker's voice, i.e. in speaker identification tasks.
Investigations dealing with the effects on intelli-
gibility of duration, amplitude limiting (peak clipping),
modulation, interruption and masking are too numerous to
cite here. They have contributed to the body of knowledge
concerning relevant acoustical parameters in speech intelli-
gibility. They have shown that speech, because of redund—
ancy due to frequency, intensity, and psychological
characteristics, is remarkably immune to intelligibility

reducing distortion.

 

1Ibid., p. 86.

2Peter B. Denes and Elliot N. Pinson, The S eech
Chain (Baltimore: Waverly Press, Inc., 1964), pp. -
122.

 

22

Efforts to predict speech intelligibility on the
basis of the physical characteristics of the speech waves
and the basic discriminative properties of the ear have
not been particularly successful. One such study made the
basic assumption that the interpretation of speech sounds
was largely dependent upon differential pitch and intensity
sensitivity and this could be used as a basis for predict-
ing results of intelligibility tests.l By the use of
empirical methods, a functional relation was obtained
between the energy distribution of speech, the differential
sensitivity of the ear, the masking properties of the
sounds, and the number of correctly perceived syllables in
an articulation test. A comparison was made between these
empirical functions and the corresponding fundamental speech
and hearing characteristics. In some cases, the computed
intelligibility compared favorably with experimental re-
sults. In many cases, large discrepancies were noted.
Because of psychological factors, it seems doubtful whether
speech intelligibility can ever be accurately predicted

from acoustical characteristics.

 

1W. A. Munson, "Relation Between the Theory of
Hearing and the Interpretation of Speech Sounds," The
Journal of the Acoustical Society of America, XVII (1945),
p. 103.

 

23

Psychological Correlates of
Speech Discrimination

 

The psychological correlates of speech discrimina-
tion are somewhat more difficult to evaluate quantitatively
than are the acoustical correlates. Yet they are, in most
situations, of significant proportion.

It has been mentioned that the intelligibility of
syllables increases as a function of the number of sounds
per syllable. In addition, the intelligibility of words
increases as the number of syllables increases. Part of
the increased intelligibility may be attributed to a
fundamental physical increase in the duration of the
syllables and words. As the length of the word is increased,
however, the listener has more contextural cues at his dis-
posal on the basis of which he is able to make a more
accurate identification.

The importance of such contextural clues is shown
by the higher intelligibility of continuous discourse over
isolated words.l Whole sounds or even complete words can
be missed in continuous discourse with no significant

reduction in intelligibility.

 

1John O'Neill, "Recognition of Intelligibility
Test Materials in Context and Isolation," The Journal of
gpeech and Hearing Disorders, XXII (March, , pp. -
0

 

 

24

The familiarity of the test material affects the
measured discrimination score.l Owens composed lists where
the phonetic content was held constant but the familiarity
was varied. He found that lists characterized by greater
familiarity, even to a slight degree, were significantly
more intelligible.

Related to the factor of familiarity is the effect
vocabulary size. If the total vocabulary size from which
the test items are chosen is reduced, the intelligibility
score will show a corresponding increase. The listener's
chances of matching a poorly discriminated word with the
known test item it resembles are increased.

The intelligibility score will increase as a func-
tion of practice.2 This again is related to the familiarity
effect. As the familiarity of the lists is increased by
repeated presentations, the discrimination scores will
increase.

Another psychological correlate of speech discrimina-

tion is the effect of group pressure.3 As the amount of

 

1Elmer Owens, "Intelligibility of Words Varying in
Familiarity," The Journal of Speech and Hearing Research,
IV (June, 1961), pp. 113-129.

 

2Fletcher and Steinberg, The Bell System Technical
Journal, VIII, p. 854.

3G. R. Miller and W. R. Tiffany, "The Effects of
Group Pressure on Judgements of Speech Sounds," The Journal
of Speech and Hearing Rgsearch, VI (1963), pp. 149-156.

 

 

25

group pressure is increased, the number of incorrect
responses offered by subjects will increase. The effect
of group pressure is correspondingly increased as the
difficulty of the speech discrimination situation is in-
creased.

Further review of the psychological correlates of
speech discrimination is not within the scope of this
chapter. It can be seen that their effects are many and
varied. Their existence contributes heavily to the diffi-
culty encountered in quantitatively predicting speech dis-
crimination performance in specified situations.

Neural Correlates of
Speech Discrimination

The discrimination of speech is based not only on
the physical parameters of the stimuli but also on psycho-
logical factors inherent in the speech material and
listening situation. Hence, in speech discrimination, the
fundamental discriminative capabilities of the auditory
system are supplemented by higher-order cortical function-
ing.

The auditory pathway proceeds from the medial geni-
culate body by way of the auditory radiations (geniculo-
temporal fibers) to terminate on the auditory cortex
(Brodmann's areas 41 and 42). This region lies on the
dorsal surface of the superior temporal convolution, buried

in the floor of the lateral sulcus.

26

Area 41 comprises the principal auditory reception
area. Area 42 is largely an auditory association area.
Another region of interest, area 22, is located on the
superior temporal convolution posteriorly.

The neural correlates of the more basic discrimina-
tive processes are equally represented bilaterally. That
is, ablation of portions of the auditory areas in either
hemisphere produces a deficit in performance. Because of
unequal cortical representation, however, the deficit
normally is greatest in the contralateral ear. In contrast,
discriminations involving meaningful speech associations
appear to be mediated largely in the dominant hemisphere.
Ablations of the non-dominant hemisphere normally produce
little deficit in such performance.

Area 22 appears to be responsible for mediating
the associations involved in recognizing auditory sounds.
This may include speech, musical sounds, or any such
familiar noises. Lesions of area of 22 in the dominant
hemisphere produce the loss of ability to distinguish these
sounds known as auditory agnosia.

Wernicke's area, area 42 in the dominant hemisphere,
appears to be involved in the more complicated mechanisms
underlying comprehension of language. Lesions of this
area result in an inability to perceive spoken language,

termed auditory or receptive aphasia.

27

Non-Speech Sound Discrimination

Previous Studies Utilizing
Non-Speech Sounds

 

 

Very little effort has been expended in the
quantitative study of non-speech sound discrimination.
One of the few tests which has been developed utilized
non-speech sounds as a means of screening the hearing of
pre-school children.1

The sounds included in the test were: a dog's
bark, an auto horn, a gun shot, a bird song, a cat meow,
and a telephone bell. These sounds were filtered so that
two of the sounds contained only frequencies between 250
Hz and 750 Hz; two of the sounds contained frequencies
between 1000 Hz and 2000 Hz; and the remaining two sounds
included only the band of frequencies between 3000 Hz and
5000 Hz.

The children were shown pictures representing the
six sounds. The sounds were presented at 50 dB ASA, then
at 15 dB ASA. If the child was not able to identify all
the sounds at the lower level, he was given a follow-up

pure-tone test.

 

1Marion P. Downs, "The Familiar Sounds Test and
Other Techniques for Screening Hearing," The Journal of
School Health, XXVI (1956). pp. 77-87.

 

28

It was found that the familiar sounds screening
test was nearly as efficient as the complete pure-tone
threshold test in uncovering hearing impaired children.

In addition, it took much less time to administer.

Sound lists have been used in psychological assess-
ment.l’ 2 Projective techniques use largely visual
association materials. The sounds were used to assess
auditory association. The Stone study required subjects
to compose stories based on three heard sounds. The
listeners were to include: (1) what caused the sounds;
(2) what was happening; and (3) the outcome of the action.
The responses in story form were thus comparable to those
obtained with the Thematic Apperception Test.

Neural Correlates of Non-Speech
Sound Discrimination

 

 

Very little information is available regarding the
neural mechanisms responsible for discriminations involving
non-speech material. The section which discussed the neural
correlates of speech discrimination indicated that the
dominant hemisphere (left hemisphere in normal persons)

mediated speech discriminations. Areas 22 and 42, in

 

1D. R. Stone, "A Recorded Auditory Apperception
Test as a New Projective Technique," Journal of Psychology,
XXIX (1950), pp- 349—353.

2Harry Wilmer, "An Auditory Sound Association
Technique," Science, CXIV (1951), pp. 621-622.

29

particular, were responsible for this function. It was
also assumed that area 22 in the dominant hemisphere was
responsible for mediating the associations involved in
recognizing non-speech stimuli. Such views were well
founded, for clinical studies had shown that lesions in
the dominant hemisphere could disturb recognition of both
speech and non-speech sounds.

There is a growing body of evidence to suggest
that the right or non-dominant hemisphere is responsible
for non-speech discriminations. For example, the percep-
tion of melodies presented dichotically is better in the
left ear.1 Since the crossed auditory pathways are stronger
than the uncrossed, it was hypothesized that the right
hemisphere was most important in such discriminations.
Other writers also suggest that the right hemisphere is

predominant in the processing of non-verbal sounds.2

Summary

The lack of studies to review concerning the dis-
crimination of non-speech sounds reveals the need for re-
search in this area. Both theoretical and clinical infor—

mation could be contributed by such studies.

 

lDoreen Kimura, "Left—Right Differences in the Per-
ception of Melodies," The Quarterly Journal of Experimental
Psychology, XVI (1964), pp. 355-358.

2V. Mountcastle, ed., Interhemispheric Relations
and Cerebral Dominance (Baltimore: Johns Hopkins Press,
1962).

 

CHAPTER III

SUBJECTS, MATERIALS, EQUIPMENT
AND PROCEDURES

Introduction

This chapter summarizes the test procedures.
Subject selection and classification are reviewed. Speech
and non-speech test materials have been discussed. A list
of the equipment employed has been included. Finally, the
procedural details of the subject testing sessions and

subsequent sound analysis have been included.
Subjects

Subject Selection

 

Sixteen hearing impaired individuals were selected
as subjects for the present investigation. They were
chosen from among patients having received a relatively
recent hearing evaluation at the Speech and Hearing Clinic

of Michigan State University.

Subject Classification

 

The subjects were subdivided into two equal size
groups on the basis of their pure-tone threshold averages

at the three speech frequencies. Class I subjects had

30

VWWW

‘-

 

31

average pure—tone thresholds between 20 dB and 40 dB ISO
in the better ear at 500 Hz, 1000 Hz, and 2000 Hz. Class
II subjects displayed average thresholds between 40 dB and
60 dB in the better ear at these three frequencies.

No attempt was made to classify the subjects on
the basis of type of hearing impairment. Site of lesion
testing was not conducted for the purpose of differential
diagnosis. It was likely that both groups included con-
ductive and sensorineural type pathologies.

Although subjects were selected without regard to
sex classification, approximately equal numbers of males
and females participated. Class I included four males and
four females. Class II included three males and five
females.

All subjects were adults at least eighteen years
of age. No maximum age limit was set. The only restric-
tion was that subjects be sufficiently mentally alert to
provide the necessary discrimination responses. Informal
observation of the subjects before and during the test

sessions suggested that all satisfied this criterion.

Stimulus Materials

Speech Materials

Tape recorded lists of CID W-l words were used to
establish the speech reception threshold. The speaker of
the words was a member of the Audiology and Speech Sciences

faculty at Michigan State University.

32

Tape recorded lists of CID W—22 words were used to
obtain the speech discrimination scores at the various in-
tensity (sensation) levels. Lists 1A, 2A, 3A, and 4A were
used. The Speaker for these lists was the same as for the
W-l lists. Both the W-l and W-22 lists were c0pied
directly from the master tapes prepared for use in the

Speech and Hearing Clinic at Michigan State University.

Non-Speech Materials

 

The non-speech sounds were recorded on magnetic
tape. The final test comprised a master and four random-
izations of fifty familiar non-speech environmental sounds.
These included animal cries, musical instruments, mechanical
sounds, and non-speech human sounds. The sounds are
summarized in Appendix A.

The sounds were drawn from several sources. Most
were recorded from sound effects records and auditory
training tapes. Three of the sounds were recorded directly
from the source especially for the purposes of this tape.

An attempt was made to minimize cultural and ex—
perience differences between the subjects by including only
quite familiar sounds. An initial tape of seventy sounds
was prepared. This was presented to a group of eight
graduate students and faculty members in a very informal
and less than Optimal listening Situation. The fifty most
frequently identified sounds were retained for inclusion

in the final test tapes.-

33

The sounds were recorded at 7-1/2 inches per second
on one track of a dual-track tape recorder. An identifica-
tion number for each sound ("number one, number two," etc.)
was recorded on the second channel. A male talker was used
for the identification numbers. This procedure allowed the
identification to be maintained at a comfortable listening
level (SRT + 40 dB) independent of the presentation level
of the sounds.

For recording the master tape, the intensity of
each sound was adjusted so that the average of the peaks
for all sounds was the same. This procedure did result in
some apparent loudness differences between the sounds.

This was due to the fact that the intensity adjustment
procedure did not take into account the total energy of
the sounds. In addition, the damping of the measurement
apparatus did not allow the true peaks to be registered
for impulse sounds with extremely short rise times.
Nevertheless, no gross loudness differences were noted
between the sounds.

The durations of the different sounds varied be-
tween approximately eight seconds and thirty seconds.
Although no attempt was made to match the sounds in dura-
tion, a segment that seemed to be reasonably long for
.correct identification was allowed for each sound. Some
sounds, such as animal cries, varied in the length required
for a complete "cycle." A five second silent interval was

allowed between sounds.

34

The order of presentation of the sounds in the
random lists was determined from a table of random numbers.

These lists were dubbed directly from the master tape.

Equipment and Procedures

quipment

 

The following equipment was used for testing the
subjects:
Suttle pre-fabricated sound treated room
Allison Model 22 Research Audiometer
Telephonic Model TDH-39 headphones
In addition, the following equipment was used for the
preparation and analysis of the non-speech sound tapes:
Garrard Lab-80 automatic turntable
Ampex Model 601 dual-track tape recorder
Ampex Model 350-AG dual-track tape recorder
Magnecord Model 1022 dual-track tape recorder
Bruel and Kjaer Model 2305 power level recorder
Bruel and Kjaer Model 2112 audio frequency spectro—

meter

Procedures

 

The subject was seated in the sound treated room
and given a brief introduction to the nature of.the study
in an informal nature. He was told that the investigation

was concerned with finding new ways to evaluate adequately

35

patients with hearing loss like himself, and that the first
part of the session would involve tests such as he had
taken before, and that pure-tone thresholds would be
determined. Following this, Speech reception thresholds
and speech discrimination scores at various loudness levels
would be obtained. The subject was then told that this
would conclude the first portion of the session. Following
a short break, he would be given instructions for the
second part of the study.

Pure-tone air conduction thresholds were determined
at 250 Hz, 500 Hz, 1000 Hz, 2000 Hz, 4000 Hz, 6000 Hz, and
8000 Hz for each ear. The descending level technique of
pure-tone threshold evaluation was utilized.

Following this, the speech reception threshold for
binaural earphone presentation was determined. The tape
recorded lists of CID W-l words were used to establish the
SRT. Oral responses were obtained from the subjects.

Tape recorded lists of CID W-22 words were pre-
sented binaurally through earphones at -8 dB, 0 dB, +8 dB,
+16 dB, and +24 dB re SRT in ascending order of loudness.
Lists 1A, 2A, 3A, and 4A were utilized and one list was
presented at each sensation level. The subjects provided
oral reSponses to the stimulus words.

The order of presentation of the lists was rotated
so that each list was given at each sensation level. For

example, Subject 1 received list 1A at -8 dB, 2A at 0 dB,

36

3A at +8 dB, 4A at +16 dB, and 1A at +24 dB. Subject 2
received list 2A at -8 dB, 3A at 0 dB, 4A at +8 dB, 1A at
+16 dB, and 2A at +24 dB. Such rotation of lists was em-
ployed to compensate for any inherent differences in
intelligibility among the lists.

Because there were five sensation levels and only
four lists, it was necessary to present one of the lists
at two sensation levels. As noted in the rotation descrip-
tion given above, the same list was presented at the lowest
and highest levels. The scores at the lowest level
generally were less than ten percent, so it was assumed
that repeating this list at the highest level would not
affect significantly (by means of word familiarity) the
score at the highest level.

This concluded the first portion of the session.
Following a short rest period, the subject again was
seated in the test room and given instructions pertaining
to the remainder of the session. He was told that he would
hear a number of familiar sounds. Animal cries, musical
instruments, human sounds and other environmental sounds
would be included. A complete tape of the sounds would be
played for him at a very soft level. Then a second tape
would be played at a somewhat higher level, and so forth,
until a total of five lists had been presented. It was
pointed out to the subject that all the lists contained

the same sounds, but they were ordered in somewhat different

37

manners on the various tapes. The subject was instructed
that his task would be to identify, as precisely as poss-
ible, each sound. If the sound was a musical instrument,
he was to give the name of the instrument; if an animal
cry, the name of the animal was to be furnished, etc. If
he could not think of the specific name, a more general
description should be given. The subject was requested to
write his responses for each test item on an answer sheet
which was furnished.

The non-speech sound tapes were then presented
binaurally through headphones at -8 dB, 0 dB, +8 dB, +16
dB, and +24 dB re SRT, again in ascending order of loudness.
The presentation time for each tape was approximately fif-
teen minutes.

Each subject participated in two test sessions.

At least two full weeks were allowed between the test ses-
sions. The procedures in the second session, including
the instructions, were identical to those of the first.

The calibration of the pure-tone and speech audio-
meters was checked prior to and periodically during the

course of the data collection.
Response Scoring

The procedure of allowing the subject to furnish
his own responses to the non-speech stimuli did not, in
most cases, result in any scoring problem. Some responses

did present difficulties, however.

38

Dogs, cats, pigs, sneezes, and squeaking doors are
familiar to most peOple and their responses to these are
quite precise. What the investigator conceived of as a
church bell may have been identified, certainly correctly,
as a clock hell or carillon chime. The sound of a tele-
phone ringing, which seemed unambiguous enough, may have
been identified simply as a ringing bell. While the sound
was certainly that of a ringing bell, it was something in
addition, and it seemed reasonable to assume that the sub-
ject should have recognized those cues which distinguished
the telephone from just another bell ringing. Since the
study was concerned with discrimination, some preciseness
in that discrimination was necessary.

The eXperimenter decided that some latitude in
responses scored correct would be allowed, provided the
response made it reasonably clear that the subject knew
the correct source of the sound to which he was listening.
Thus "bell ringing“ would not be an acceptable answer to
a telephone ringing. On the other hand, the distinctions
between a trumpet, cornet, and horn are sometimes not

really clear and all must be allowed as correct responses.

Additional Equipment and Procedures
Analyzing tfigrSound Tapes
Preparation of the Tapes
The sounds which were found originally on records

were transferred to magnetic tape by means of the Garrard

39

turntable feeding an Ampex 601 recorder. Onto this same
tape were transferred additional sounds found on various
auditory training tapes. This tape-to-tape transfer was
accomplished by means of the Magnecord recorder feeding an
Ampex 350-AG recorder.

From the tape thus recorded, the master sound tape
was prepared. This involved not only transferrence of the
sounds from the original tape to the master tape but also
adjustment of the intensity of each sound so that, as
mentioned previously, the average of the peaks for all the
sounds was the same.

The Magnecord and Ampex 350-AG recorders were used
for this part of the recording procedure. The original
tape was played on the Magnecord recorder. This recorder
was connected to the Ampex machine, on which the master
tape was prepared. The intensities of the sounds on the
original tape were monitored by feeding an output from the
Magnecord recorder to the power level recorder. Each sound
on the original tape was played several times and the out-
put level of the Magnecord recorder adjusted so that the
average of the peaks, as monitored by the power level
recorder, reached a certain Specified level. The sound
was then transferred to the master tape on the Ampex
recorder. This resulted in equal average peak intensities
on the master tape. A 1000 Hz calibration tone was recorded
onto the master tape at a level corre3ponding to the average

peak intensity of the various sounds.

40

The randomized lists were prepared from the master
tape using the same pair of recorders. Although the inten-
sities of the sounds on the master tape were monitored as
the randomized tapes were being prepared, no further change

in any of the sound levels was required.

Analysis of the Sounds

 

The block diagram of the analysis equipment appears
in Figure 8. Tape 100ps were prepared for use on the
recorder. Each 100p allowed about three seconds recording
time. Since all of the sounds were of greater duration
than this, the sounds were analyzed by section from several
tape loOpS. Hence, three to five 100ps were required to
obtain the complete spectral characteristics for most of
the sounds. The overall Spectra of these sounds were ob-
tained by averaging the spectra obtained from each section

(100p).

 

Tape Power Level
Recorder Spectrometer Recorder

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 8.--Block Diagram of Equipment Used for Spectral
Analysis of the Non-Speech Sounds.

41

The spectra of the sounds were obtained from one-
third octave band analyses. For continuous sounds, the
average power within any given one-third octave band was
taken as the relative sound pressure within that band.
The average of the peaks within a given one-third octave

band was recorded as the pressure level within that band

for impulse sounds.

CHAPTER IV

ANALYSIS OF THE DATA AND DISCUSSION

Introduction

This chapter summarizes and discusses the statistical
procedures and results of the investigation. Conclusions are
drawn regarding non-speech sound discrimination in hearing
impaired subjects and the implications for such discrimina-

tion testing in clinical situations.

Summary of the Data

Table 1 summarizes the test and retest non-speech
sound discrimination scores from the individual subjects
in Class I (20 dB - 40 dB loss) at the various sensation
levels. The first score listed is that obtained in the
first session; the second that obtained in the second
session. For the purpose of comparison, Table 2 summarizes
in the same format, speech discrimination scores from the
same subjects. Tables 3 and 4 contain the non-speech and
speech discrimination scores, reSpectively, from the indi-

vidual subjects in Class II, (40 dB - 60 dB loss).

42

43

 

 

 

 

N.m H.HH m.oH m.vH v.m $.5H N.mH m.vH m.oa o.NH om

N.vw m.wm m.mh N.mh o.Hh m.vm N.H¢ o.mm m.m N.m M
No mm mm Nm Nh Nm om ow o v m
mm mm mm om Nb mm ow 0v m OH 5
m5 mh on mm on vm mm om mm mm m
¢m om vm mm Nm vb Nv mN OH m m
om mm mm mm mu ou Nm mN N N v
m5 mm ov om mv «N NH NN N o m
om om mm Nm on vb Nv 0% v m N
mm vm mm 00 Nb Nm mm vm vH m H
mm B mm H mm B 0% B mm B

vm+ oa+ m+ mu pomflbsm

ABMm mu mvv Hw>ma

 

 

.muownnsm H mmmHU How mwuoom coaumcflfifluomfln umwumm can ummB nommmmlcozll.H magma

44

 

 

 

 

m.h m.h m.b N.h m.v m.m m.0H h.m m.m h.N Om

m.om N.Hm o.mm m.¢m o.mo m.mm m.ov m.Nv m.N N.v M
«m mm om mm mm on mm mm o g m
cm mm mm mm mm vb mN Nv o m h
vm om vm mm we mm No ow OH w m
mm mm gm mm Nb on mm vw m w m
mm mm vm om em on om om N N v
ow mm Nb Nb Nm om wv Nm v N m
mm Nm Nm mm Nb on mm vv o m N
on om on Nm mm on Nm Nm 0 o H
mm 9 mm 9 mm 9 mm 9 mm a

¢m+ ma+ m+ o mu pommnsm

Afimm mu mvv Hm>wq

 

 

.muoanSm H mmmau How mwuoom coaumcwfifluomfla ummumm cam umma

sommmmuu.m magma

 

m2: m.m.~ m.N._.. N.:.. m.NH v.3” H.m._n 0.3” m.m ¢.m Om

 

45

 

 

m.mn o.os m.mn o.mm m.mm o.nm m.m¢ m.m~ m.aa m.¢ m
mm om we we we we mm mm m ma m
em mm mm mm om me am ma m m 5
mm om mm we mm mm as ea em o m
mm mm mm om on am am om m o m
we we on an mm mm as we we a w
om mm mm om we we om we NH m m
mm mm mm mm we om om ma m o m
an «a am we mm mm om om ma 0 H
mm 9 mm 9 mm 9 mm 9 mm a

¢~+ wa+ m+ o m: pomflnsm

Aemm mu moo Hm>mq

 

 

.muomhasm HH mmmau Mom mmuoom Goaumcweflnomwa ummpmm can umma nommmmlc02I|.m magma

46

 

 

 

 

N.m m.m o.m m.oa m.m H.va N.m m.> H.N m.N om

m.om m.bm m.mm N.Hm m.mm m.mw m.nm 0.0m N.N N.H M
mm mm om mu om mm Nv mm o o m
mm mm mm mm om mm Nm mm N o n
em mm om mm vs mo Nm ma o o m
mm vm om om Nb om Nv mm v o m
om vm wm om mm om mm mm N w v
mm mm vm om Nw em om vN o o m
vm Nm mo om mv om vN NN m N N
vb Nu mm mm on em Nv mN v o H
mm B mm 9 mm 8 mm 8 mm B

vN+ ma+ m+ 0 ml nomnnsm

ABMW mH mflv Hm>mq

 

 

.muomnnsm HH mmmao How mmuoom COHuMGHEHHOmHQ umwumm paw umma sommmmll.v mHQMB

47

Figure 9 illustrates the relation between non-speech
discrimination score and intensity of presentation (sensa-
tion level) for Class I subjects. The brackets on the
curve delimit i one standard deviation unit. Figure 10
illustrates the same features for Class II subjects.

Figure 11 summarizes, for the purposes of comparison, the

functions shown in Figures 9 and 10.

Sensation Level (dB re SRT)

 

100
8 80 ‘1
o z/z”””
U)
8:: 60 I
--I C:
+3 0)
2:: 40
. . I
En:
.HV
‘6 20
.1”.
o o I

 

-8 0 +8 +16 +24

Figure 9.--Non-Speech Discrimination Score as a Function of
Intensity for Class I Subjects.

48

Sensation Level (dB re SRT)

 

100
(D
8
U 80 .—
u: x 1
‘HC: 60 I/
«90)
MO
.533"?
.I—{V
H
a I
,H 20
Q

o I

 

 

-8 0 +8 +16 +24

Figure lO.--Non-Speech Discrimination Score as a Function of
Intensity for Class II Subjects.

Sensation Level (dB re SRT)

 

 

 

 

100
(D
H
8
m 80
8::
«4:: 6O
4J0)
€60
C334
«am
an. 40
.Hv
u
U
«'3 20
D

)k
-8 0 +8 +16 +24

Figure ll.--Non-Speech Discrimination Sc0res as a Function of
Intensity for Class I and Class II Subjects.

49

Figures 12 and 13 illustrate the speech discrimina-
tion functions from Class I and Class II subjects respect-
ively. Figure 14 shows, on the same graph for ease of
comparison, the speech discrimination functions from Class
I and Class II subjects.

Figure 15 summarizes the mean speech and non-Speech
discrimination scores for all subjects as a function of

intensity.

Sensation Level (dB re SRT)

 

100

80 ’,,»”"%I

40 :I

20

Discrimination Score
(Percent)

 

0 I

-8 0 +8 +16 +24

Figure 12.--Speech Discrimination Score as a Function of
Intensity for Class I Subjects.

50

Sensation Level (dB re SRT)

100
3
0
CA
.22 /
+,m 60 1
rd 0
C: H
'28
w, 40
H
O
.2 I
Q 20

0 «I

 

 

-8 0 +8 +16 +24

Figure l3.--Speech Discrimination Score as a Function of
Intensity for Class II Subjects.

Sensation Level (dB re SRT)

o/.
80 Class V$/$
60 as II

40

 

Discrimination Score
(Percent)

O

at

 

-8 0 +8 +16 +24

Figure l4.--Speech Discrimination Scores as a Function of
Intensity for Class I and Class II Subjects.

51

Sensation Level (dB re SRT)

 

100

3 0 Speech /).(
o/

1‘;
60 - Non-Speech

Discrimination Score
(Percent)

 

 

4o
1 \L
20
0 3:.
-8 0 +8 +16 +24

Figure 15.--Mean Speech and Non-Speech Discrimination Scores
as a Function of Intensity for all Subjects.

Discussion

Non-Speech Data

The pooled non-speech discrimination function
relating sensation level with discrimination score, shown
in Figure 15 increases monotonically and tends to reach
an asymptote at higher intensity levels. The function is
approximately linear between -8 dB and +8 dB sensationw
levels. The slope of the curve in this linear region is
about 3.4%/dB.

The standard deviation remained relatively constant,

the smallest occuring at the lowest intensity level of

52

pre-sensation. The fact that zero percent discrimination
score served as a lower boundary at this sensation level
undoubtedly resulted in the smaller standard deviation
here.

Figure 11 reveals that the functions for the two
groups of subjects were quite similar in shape. Both were
approximately linear between -8 dB and +8 dB sensation
level. Table 5 summarizes the slopes in this linear
region for both groups of subjects, The curve from Class
II subjects showed a slightly smaller slope. This caused
the functions to diverge somewhat at the higher sensation
levels. In general, however, the same function appeared
to describe the data from both groups, the main difference
being that the Class II function was shifted to the right
along the intensity axis. The average discrimination
score difference between Class I and Class II subjects

was 8.0 percent.

Table 5.--Slopes in the Linear Region of the Discrimination
Function for Class I and Class II Subjects.

 

 

GROUP SLOPE (%/dB)

 

Class I 3.5

Class II 3.4

 

53

Comparison of Speech and
Non-Speech Data

 

 

Figure 15 illustrates the differences between the
pooled speech and non-speech discrimination scores. The
speech function started at a lower level but, because of
a larger lepe, intersected the non-speech function.
Hence, at the higher sensation levels, the Speech discrim-
ination scores were higher than the non-speech discrimina-
tion scores. Comparison of Figures 12 and 13 with Figures
9 and 10 reveal that the standard deviations for the speech
scores were smaller at all intensity levels than the
corresponding non-speech score standard deviations.

Figure 14 shows that the speech discrimination
functions from the two groups of subjects, like the non-
speech functions, were similar in shape but shifted rela-
tive to each other. The speech materials did not appear
to be as discriminating with respect to extent of loss,
however. As opposed to an average non-speech discrimination
score difference of 8.0 percent, the average speech score
difference was 5.1 percent between the two groups of sub-
jects. Table 6 summarizes the discrimination score
differences between Class I and Class II subjects for the
two tests at the various sensation levels.

It is interesting to Speculate on the reasons why
the non-speech items were more discriminating than the

Speech material with reSpect to extent of loss. Part of

54

this might be attributed to differences in the normal

listening eXperiences of the subjects.

Table 6.--Differences in Discrimination Scores Between
Class I and Class II Subjects for the Non-Speech
and Speech Tests.

 

 

Sensation Level (dB re SRT)

-8 0 ‘+8 +16 +24

 

Speech 3.0 12.8 3.0 3.3 3.4

 

The majority of Class I subjects exhibited a mild
hearing loss. Such a loss would be handicapping in many
listening situations. Most of the subjects apparently
did not regard the loss to be of sufficient magnitude to
warrant use of a hearing aid, however, for only two indi-
viduals in this group used such an instrument. Therefore,
the majority of these individuals were not accustomed to
listening under aided conditions.. Although some distortions
in acoustic stimuli would have been present for these in-
dividuals due to the existing hearing pathology, they were
unfamiliar with the additional distortions created by a
hearing aid.

Six of the eight Class II subjects wore hearing

aids regularly. They undoubtedly had learned to discriminate

55

both Speech and non—speech sounds which were colored by
the distortion characteristics of the hearing aid.

The sound stimuli delivered to the subjects in the
test sessions were of a moderately "normal" quality.. That
is, they contained little distortion and hence their
characteristics very closely resembled the corresponding
sound heard in the natural environment. Class II subjects
were at a disadvantage in trying to discriminate such
sounds, for they had learned to identify the distorted
sounds delivered by the hearing aid. For them, the natural
or normal sound had become the sound they heard through the
_hearing aid. Thus the stimuli sounded unnatural to these
subjects. It is reaSonable to believe that the effect of
unnatural quality would be more pronounced for non-Speech
than for Speech stimuli.

The results of the studies cited in Chapter II
revealed that quite distorted and unnatural sounding speech
can be identified accurately. It is not necessary for the
Speech to sound normal to be discriminated. Even though
Class II subjects did not hear Speech which sounded normal
to them on the basis of their listening experiences, this
dimension of the speech stimuli probably had little effect
on their discrimination scores.

On the other hand, this aspect appeared to be more
important for non-speech stimuli. Sounds which should

have been familiar did not sound normal to Class II subjects.

56

They were not distorted by the hearing aid as customarily
was the case. The effects of familiarity were reduced
correSpondingly. Hence, the Class II subjects had more
difficulty than Class I subjects on non-Speech than on
Speech stimuli.

On the basis of this discussion, two tentative
conclusions can be advanced. First, subjects accustomed
to identifying non-Speech sounds distorted by a hearing
aid will have greater difficulty than subjects not so
accustomed in identifying the undistorted sounds, even
though both groups of subjects have comparable hearing
losses. Second, these results seem to indicate that non-
speech sounds may be more susceptible than speech stimuli
to the effects of distortion.

One other factor which might have contributed to
the larger discrimination score differences from non-Speech
items between Class I and Class II subjects should be men-
tioned here. The fifty non-Speech items were selected from
a larger group of seventy sounds which was presented pre-
liminary to the investigation to a small number of graduate
students and faculty members. The fifty most frequently
identified sounds were retained for inclusion on the final
test tape. Most of these panel members were relatively
young. Likewise, most subjects in Class I were relatively
young. Class II, because of the extent of loss required,

comprised generally older individuals. The sounds which

57

were familiar to the initial panel and Class I subjects

may have been not quite so familiar to the Class II sub-
jects. Hence, for the non-Speech sounds, there may have
been a familiarity factor operating which did not influence
the speech discrimination scores.

Pearson product-moment correlation coefficients
were computed between the non-speech discrimination scores
obtained in the two test sessions at the various sensation
levels to estimate test-retest reliability. For the pur—
poses of comparison, Similar correlations were calculated
for the Speech scores. Table 7 presents the results of

these computations.

Table 7.--Test-Retest Reliability Coefficients of the
Speech Non-Speech Sounds at the Various Sensation

 

 

 

 

Levels.
Sensation Level (dB re SRT)
-8 0 +8 +16 +24
Non-Speech
Class I .99 .46 .91 .94 .97
Class II -.08 .25 .76 .95 .99
Speech
Class I .09 .10 .60 .76 .90

Class II .12 .78 .81 .88 .81

 

58

It can be seen that, with the exception of the
lower sensation levels, the reliability coefficients of
both lists were quite high. Moreover, there was little
difference apparent between the values obtained either from
the two subject classes or from the Speech and non—speech
tests.

The rather deviant results noted at -8 dB sensation
level, especially for the non-Speech test, probably should
be interpreted in terms of data limitations imposed by

this level rather than in terms of unusual test or auditory

 

characteristics. At this level, zero percent served as a
lower boundary, and Since several subjects obtained this
score, the values did not achieve a wide distribution.
This reduced the standard deviation, producing a rather
unstable and perhaps misleading value of correlation co-
efficient. The existence of one extremely high test and
retest score was largely responsible for the inflated co-
efficient noted for Class I subjects at this level.

Pearson product-moment correlations were computed
between the speech and non-Speech discrimination scores at
the various intenSity levels. Table 8 summarizes the
results of these calculations.

With the exception of the lower sensation levels,
the correlations were quite high. Again, data limitations
at the lower level Should be considered when interpreting

the results from this level.

59

$
Table 8.--Product-Moment Correlations BetWeen the Results
of the Speech and Non-Speech Discrimination Tests.

 

 

Sensation Level (dB re SRT)

 

-8 0 +8 +16 +24
Class I .08 .56 ' .61 .55 .72
Class II -.09 .35 .54 .83 .92

 

The correlation coefficients obtained from Class
II subjects at +16 dB and +24 dB were higher than the
corresponding values obtained from Class I subjects. To
test for significant differences, a technique suggested
‘ by Hays was employed. A r to Z transformation was performed
for each correlation figure. The resultant 2 values were

used to compute the following test statistic:

 

Z - Z

Statistic = r(2) r(2)
N-3

The results of these calculations are summarized in Table 9.

Table 9.--Summary of Statistic for Testing Differences

Between the Correlations Obtained from Class I
and Class II Subjects.

Level (dB) Statistic

 

+16 1.26

 

+24 1.51

 

 

60

In a normal sampling distribution, the value of both Z and
the test statistic required for rejection of an hypothesis
at the .05 level of Significance is 1.96. Since the com-
puted values did not exceed this figure, the observed
differences in the correlations were not significant.

The Similiarity of the non-speech and Speech dis-
crimination functions, together with the high correlations
between the tests, tend to suggest that similar neural
function is involved in both types of discrimination. This
is an attractive hypothesis, but it is, at best, only par-
tially supported by results such as those being presented.
Further evidence, based on various approaches, Should be
at hand before a definitive statement can be made.

It is known that the neural mechanisms involved in
various types of discriminations have only a degree of
Specificity. Some of the neural correlates seem to be
especially important for judgments involving a particular
type of discrimination. Many of the correlates, however,
are involved in many types of discriminations.

One would eXpect that the same would hold for non-
speech and speech discrimination. While both would involve
many common mechanisms, neural correlates especially import-
ant for only one of the types of discriminations undoubtedly
exist.

Correlation techniques also were employed to in-

vestigate the possible relationship between a subject's

61

pure-tone thresholds and the spectra of the sounds which
he identified correctly at each sensation level.

The thresholds for each subject were determined at
250 Hz, 500 Hz, 1000 Hz, 2000 Hz, 4000 Hz, 6000 Hz, and
8000 Hz. The threshold was established for each ear and
the threshold of the better ear used in the analysis. By
means of Spectral analysis, the relative intensity level
of the one-third octave bands containing these frequencies
was obtained for each sound.

If, at a given sensation level, a particular sound
was identified by a subject, the relative intensity of that
sound at each frequency cited above was matched with the
threshold of the subject at the same frequency. This
matching between thresholds and sound Spectra was made at
each sensation level for all sounds and all subjects. A
product-moment correlation coefficient was calculated
between the resultant pairs of values. Table 10 summarizes
the results of this part of the study.

As a means of interpreting the results in Table
10, ninety-five percent confidence intervals around zero
correlation were calculated for each sensation level.

These intervals are Shown in Table 11.

Comparison of Tables 10 and 11 reveals that none
of the obtained correlation values for a given level ex-
ceeded the confidence interval limits for that level. Hence,

no correlation was Significantly different from zero.

62

Table 10.--Product—Moment Correlations Between the Relative
Intensity Levels of the Identified Sounds at a
Given Frequency and the Threshold of the Sub-

jects at the Same Frequency.

 

 

Frequency (Hz)

 

Level (dB) 250 500 1000 2000 4000 6000 8000
-8 .114 .029 .086 -.130 .058 -.043 -.115

0 .051 -.045 -.052 -.094 -.012 .015 .062

+8 -.008 -.025 -.032 -.027 -.031 -.047 -.034
+16 .008 -.007 -.003 -.018 -.009 -.018 -.015
+24 .000 .001 -.007 -.016 -.030 .003 -.007

 

Table ll.--Ninety-Five Percent Confidence Intervals Around
Zero Correlation for the Various Sensation

 

 

 

Levels.
Level (dB) 95% Confidence Interval
-8 -.2685r5+.268
o -.1185r5+.118
+8 —.0805r5+.080
+16 -.0805r5+.080
+24 -.o785r5+.078

 

These results indicate that there was no relation-
ship between the threshold configuration of a subject and

the Spectra of the sounds which he identified correctly.

63

The perceptual effects of the sounds apparently were
dominant in determining correct identification.

In speech discrimination tests, the words identified
correctly depend somewhat on the threshold configuration of
the subject. Words which contain high frequency sounds such
as /s/, /v/, and /f/ are missed more frequently by patients
having predominantly higher frequency losses. Given a
list of words of Similar familiarity, the physical charac-
teristics of the sounds comprising the words will be of

great importance in determining correct identification.

 

In contrast, no such effects were found in the
present study for non-Speech sounds. This might, in part,
be due to familiarity differences between the sounds. The
psychological differences might have masked the physical
variables.

There might be some quite basic differences in the
way speech and non-Speech stimuli are perceived. Words
are comprised of separate Speech sound entities. It is
necessary to identify all, or nearly all, of these sounds
correctly before the word is discriminated. Each Speech
sound might be identified somewhat independently of the
others, with restrictions imposed by the physical influence
of its sound environment in the word and limitations im-
posed by the probability of certain letter combinations

in the language.

64

For non-Speech sounds, however, the identification
is more of an all-or-none nature. The separate cues are
not available and the discrimination must be based on the
long-term spectrum of the sound. Such differences could
account for the apparent reversal of physical and psycho-
logical effects noted in the data.

An item analysis provided listings of the percent-
age of subjects correctly identifying a sound at a given
sensation level. Comparison of these data with the spectra
of the sounds identified did not reveal any obvious rela-
tionships. There did not appear to be any differences
between the Spectral configurations of the sounds ident-
ified and of the sounds missed. Again, the results lead
to the conclusion that psychological variables were more
important than spectral configuration in determining
identification.

For a given sound, some frequencies undoubtedly
are more important than others for discrimination. The
results of the present study did not reveal such effects,
but further investigations might. Frequency filtering
techniques which presented various bandwidths to the sub-
jects would be a more definitive way of approaching this

aspect of the problem.

CHAPTER V

SUMMARY, CONCLUSIONS AND RECOMMENDATIONS

Introduction

The purpose of the present study was to evaluate
non—speech sound discrimination in subjects with impaired
hearing. The results of Speech and non-Speech discrimina—
tion tests were compared. This chapter summarizes the
results of the investigation and discusses recommendations

for further research.

Summary

A test consisting of fifty non-speech sounds was
constructed and administered to Sixteen hearing impaired
subjects. The subjects were divided into two groups on
the basis of the extent of the hearing loss. Class I
subjects had average thresholds for the speech frequencies
(500 Hz, 1000 Hz and 2000 Hz) between 20 dB and 40 dB ISO
in the better ear. Subjects assigned to Class II had
average thresholds at these frequencies between 40 dB and
60 dB ISO in the better ear.

Initial pure-tone thresholds were determined for

each ear at 250 Hz, 500 Hz, 1000 Hz, 2000 Hz, 4000 Hz,

65

66

6000 Hz, and 8000 Hz. The speech reception threshold was
determined for binaural earphone presentation. Lists of
CID W-22 words were administered binaurally through ear-
phones at -8 dB, 0 dB, +8 dB, +16 dB, and +24 dB relative
to the speech reception threshold. A non-speech sound
list was presented at each of the same sensation levels.
Each subject was tested twice with at least two weeks
between the test sessions. Spectral analysis of the non-
Speech sounds provided the frequency characteristics of
each sound.

Statistical analysis of the data was performed to
determine the reliability of the non-Speech test. Correla-
tion coefficients were obtained between the results of the
Speech and non-Speech tests. Correlation coefficients also
were computed between the threshold configurations of the
subjects and the spectra of the non-Speech sounds identified
correctly at each sensation level.

An item analysis provided a listing of the percent-
age of subjects correctly identifying each non-Speech sound
at the different sensation levels. This was compared to
the Spectral characteristics of the various sounds to deter-
mine which, if any, frequency bands contributed most heavily
to correct sound discrimination.

The results of the study indicated that the discri-
mination functions (sensation level vs. discrimination

score) for the non-speech sounds very closely resembled

67

those obtained from the Speech test. The lepe of the
linear portion of the nonespeech function was Slightly less
than the SlOpe of the corresponding portion of the Speech
function, however.

A high correlation was found between the results
of the speech and non-Speech tests, particularly at the
higher sensation levels. No significant differences were
discovered between Class I (20 dB to 40 dB loss) and Class
II (40 dB to 60 dB loss) subjects with regard to correla-
tions between the Speech and non-Speech tests. The non-
Speech test did appear to be slightly more discriminating
than the speech test (CID W—22) with respect to extent of
loss.

High test-retest reliability was obtained for both
tests at the higher sensation levels. No Significant
differences existed between the reliability coefficients
of the two tests.

No Significant correlations were found between the
threshold configurations of the subjects and the Spectra
of the sounds which were identified correctly at each
sensation level. Moreover, no frequency band of the sounds
appeared to be particularly important in its contributation

to correct identification.

68

Conclusions

Within the limits of the present study, the
following conclusions appear warranted:

l. The Similiarity of the discrimination functions
(sensation level vs. discrimination score) and the high
correlations between the speech and the non-Speech sound
tests suggests that highly similar processes mediate the
discriminations involving these types of stimuli.

2. The psychological effect of familiarity appears
to be more important for non-speech than for Speech sound
discrimination. Less redundancy in the physical character-
istics of non-Speech sounds may account for this result.

In addition, individual differences in familiarity may be
greater for non-Speech than for Speech stimuli.

3. Individual subjects, on the basis of their
listening eXperiences, learn to associate certain physical
characteristics with each familiar sound. These criteria
of physical normality are more important for non-Speech
than for Speech sounds. Moreover, the non-Speech stimuli
appear to be more susceptible to the effects of distortion.

4. There seems to be no relationship between the
pure-tone audiometric configuration of a subject and the
frequency Spectra of the non-Speech sounds which he dis-
criminates correctly.

5. The discriminability of non-Speech stimuli

does not appear to be related to the frequency composition

69

of the sounds. In addition, differential contributions of
the various frequency bands of the non-Speech sounds to
intelligibility were not discovered.

6. Scores obtained on the non-speech test of
auditory discrimination seem to be more highly affected by
extent of hearing loss than are the scores obtained from
the speech test.

7. The reliability of the non-Speech test of
auditory discrimination appears to be sufficiently high to

warrant its use in clinical Situations.

Recommendations

Much work remains to be done regarding non-Speech
discrimination in hearing impaired subjects. This investi-
gation perhaps has raised more questions than it has
answered.

Since no attempts were made to classify type of loss,
such an investigation should be conducted. Non-Speech dis-
crimination, like Speech discrimination, may be quite
sensitive to type of impairment.

The effects of noise masking on non-Speech discrim-
ination should be investigated. Differential effects of
various types of hearing pathologies in identifying noise
masked non-speech sounds should be researched. Such pre-
sentation of stimuli might be even more discriminating with

respect to extent and type of loss.

70

Studies should be designed to investigate the
effects of various types of frequency filtering on non—
speech discrimination. Such research also would provide
more definitive information regarding the contributions of
the various frequency bands to intelligibility.

Finally, much work remains to be done relative to
the importance of psychological variables such as familiar-
ity on non-Speech discrimination. Such data could provide
further information regarding the interrelationships of

Speech and non-Speech discrimination mechanisms.

BIBLIOGRAPHY

Books

Davis, Hallowell and Silverman, S. Richard. Hearing and
Deafness. New York: Holt, Rinehart and Winston,
1965.

 

Denes, Peter B. and Pinson, Elliot N. The Speech Chain.
Baltimore: Waverly Press, Inc., 1964.

 

Fletcher, Harvey. Speech and Hearing in Communication.
Princeton, New Jersey: D. Van Nostrand Company,
Inc., 1953.

 

Hays, William L. Statistics for Psychologists. New York:
Holt, Rinehart and Winston, 1963.

 

Mountcastle, V., ed. Interhemispheric Relations and
Cerebral Dominance. Baltimore: John Hopkins Press,
1962.

 

 

Stevens, S. 8., ed. Handbook of ExPerimental Psychology.
New York: John Wiley and Sons, Inc., 1966.

Truex, Raymond C. and Carpenter, Malcolm B.‘ Human Neuro-
anatomy. Baltimore: The Williams and Wilkins
Company, 1964.

 

Wever, Ernest G. and Lawrence, Merle. Physiglogical
Acoustics. Princeton, New Jersey: Princeton
University Press, 1954.

 

 

Articles and Periodicals

Downs, Marion P. "The Familiar Sounds Test and Other
Techniques for Screening Hearing," The Journal of
School Health, XXVI (1956), pp. 77-87.

 

 

71

72

Egan, James P. "Articulation Testing Methods," The
Laryngoscgpe, LVII (1948), pp. 955-991.

 

Fletcher, H. and Steinberg, J. C. "Articulation Testing
Methods," The Bell System Technical Journal,
VIII (1929), pp. 806-854.

 

Kimura, Doreen. "Left-Right Differences in the Perception
of Melodies," TheQuarterlyJournal of Experimental
Psychology, XVI (1964): PP. 355-358.

 

 

Kryter, Karl D. "On Predicting the Intelligibility of
Speech from Acoustical Measures," The Journal of
Speech and Hearing Disorders, XXI (1956), pp. 208-
217.

 

 

Miller, G. R. and Tiffany, W. R. "The Effects of Group
Pressure on Judgements of Speech Sounds," The
Journal of Speech and Hearinngesearch, VI (1963),
pp. 149-156.

 

Munson, W. A. "Relation Between the Theory of Hearing and
the Interpretation of Speech Sounds," The Journal
of the Acoustical Society of America, XVII (I945),
p. 103.

 

O'Neill, John. "Recognition of Intelligibility Test
Materials in Context and Isolation," The Journal
of Speech and Hearing Disorders, XXII (1957), pp.
87-900

 

Owens, Elmer. "Intelligibility of Words Varying in
Familiarity," The Journal of Speech and Hearing
Research, IV (1961), pp. 113-129.

 

Stone, D. R. "A Recorded Auditory Apperception Test as a
New Projective Technique," The Journal of Psycho-
1092, XXIX (1950), pp. 349-3530

 

Wilmer, Harry. "An Auditory Sound Association Technique,"
Science, CXIV (1951), pp. 621-622.

APPENDICES

\DQQONUI-waH
O

N no N no N «are H'P‘ H rd H F4 H +4 H
m w»:» n::a c>xo a>~q oscn a on N r4 o
O O O O O O O O O O O O O O O O

APPENDIX A

SUMMARY OF THE NON-SPEECH SOUND LISTS

Sawing

Clock ticking
Hammering

Pig

Doorbell chime
Goat

Horse

Cat

Duck

Cow

Coyote

Dog

Boat whistle

Jet airplane

Car horn

Train

Dialing telephone
Toy whistle
Telephone busy Signal
Typewriter
Telephone ringing
Ping-pong game
Trumpet

Banjo

Snare drum

Master List

26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.

73

Accordion

Flute

Church bell

Gun Shots

Siren

Snore

Cough

Whistling (human)
Footsteps
Rooster

Crickets

Bird song

Music box

Sneeze

Crow

Telegraph key
Alarm clock ringing
Laugh

Glass breaking
Crowd cheering
Baby crying
Squeaking door
Knocking on door
Auto starting
Xylophone

\DCDQQUIQUJNH
O

rd H‘lfl #4 H
n on N +4 o
O O O O O

15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.

APPENDIX B

SOUNDS RANKED FROM LEAST DIFFICULT TO MOST DIFFICULT

Siren

Laugh

Dog

Trumpet

Knocking on door
Bird song

Church bell

Cough

Snare drum
Telephone busy signal
Crow

Dialing telephone
Auto starting
Squeaking door
Sawing

Cat~

Rooster

Train

Cow

Whistling (human)
Horse

Car horn
Footsteps
Telephone ringing
Doorbell chimes

26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.

74

Baby crying
Snore

Clock ticking
Banjo
Typewriter
Coyote

Pig

Gun shots
Music box
Accordion

Duck

Hammering
Xlephone
Alarm clock ringing
Ping-pong game
Toy whistle
Goat

Boat whistle
Crowd cheering
Crickets

Jet airplane
Telegraph key
Sneeze

Flute

Glass breaking

 

 

,‘1 .. ?.—_—.rr— 7—-