CATEGQREQAL PERSEPTEON m A ’
cmmm PALSIEQ mumm ; .

Thesis feg‘the Degi'eg a? M. A.
Ee‘éECHEGfiE SEAR E3 i‘fEfiSiTY
iANET B. Ei‘kaE‘f

£925 -

In.“ .,‘» ”,1 us,“ - H L,” , in," ‘ ‘ V v,"
-»:9.-.’I?~;s:‘3z33‘«-§¥u—f;eifi‘i’azza‘h-Saws" “ - _ ‘~ = - -

"~' ,".-'e‘ ‘ :‘ n
1". :{cqfimp 11.3ug‘l»|

 

 

 

THESIS

‘3.

5r

,. : Ar 5-?

9 .. .‘f’ILL \ _:

j’Ms‘m“ ,. _

- WWW;

g3:'£lr-C§..; A '.'

  

magma iv
HUM: & SDNS'
800K BINDERY 1ND.

LIBRARY an N DEF‘S
sputum-t. 219mm

  

 

 
  

 

 

MSU

LIBRARIES
W

\-

 

 

RETURNING MATERIALS:
Place in book drop to
remove this checkout from
your record. FINES wil]
be charged if book is
returned after the date
stamped be1ow.

 

 

 

 

 

 

 

 

ABSTRACT

CATEGORICAL PERCEPTION IN A CEREBRAL
PALSIED POPULATION

BY
Janet B. Baldrey

A review of the literature concerning categorical
perception and the Motor Theory of Speech Perception has
revealed that these concepts have ramifications with
respect to the cerebral palsied. The literature is lacké
ing in elaboration on this area. The presence of cate-
gorical perceptual patterns in the cerebral palsied popu-
lation would suggest the applicability of the Motor Theory
of Speech Perception to a pathological population.

The purpose of this study was to ascertain the
presence of categorical perception in a cerebral palsied
population, through the use of identification and discrimi-
nation tasks similar to those used in previous studies.
The stimuli were rapid and £3§£§_temporally varied at
the silent intervocalic segment. The tasks were admin-
istered to three groups: a control group and two
experimental groups. The two experimental groups were

cerebral palsied with good speech and cerebral palsied

Janet B. Baldrey

with dysarthric speech patterns. The assignment to
groups was based on performance on a standard articu-
lation task and the rating of spontaneous speech. There
were fifteen subjects in total; five in each group.

The results were computed totals of responses for
all stimuli in each group; these were graphed. The dis-
plays indicated that all groups perceived categorically.
The results were then divided in each group according
to what the original word had been. These displays
indicated that the groups did perceive categorically but
that they perceived differently. The phoneme boundaries
were placed at different points along the temporal con-
tinuum. The results of the discrimination task indi-
cated that while the subjects perceived categorically
in an identification situation, they did not in a dis-
crimination situation.

The Motor Theory of Speech Perception provided
the best explanation for the results. Conclusions drawn
were that there is not a learning component to speech
perception ability and that there was not necessarily a
production/perception link for accurate and rapid per-
ception of Speech sounds, but it would seem that pro-
duction abilities did make a difference in the pattern
exhibited. The results also indicated that individuals
with cerebral palsy do exhibit different perceptual

patterns from normal individuals and that cerebral

Janet B. Baldrey

palsied individuals with good speech and those with
dysarthric speech exhibit different perceptual patterns
from each other.

Directions for further research lie in the areas
of delineating these differences more carefully and dis-
covering what factors, i.e., type of speech therapy,

could influence this difference.

CATEGORICAL PERCEPTION IN A CEREBRAL

PALSIED POPULATION

BY

Janet B. Baldrey

A THESIS

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

MASTER OF ARTS

Department of Audiology and Speech Sciences

1975

Accepted by the faculty of the Department of
Audiology and Speech Sciences, College of Communication
Arts, Michigan State University, in partial fulfillment

of the requirements for the Master of Arts Degree.

 

 

 
    

Thesis ///7

Guidance Committee: ’ ““.‘_l _ ’/ Chairman

 

 

ACKNOWLEDGMENTS

The writer wishes to express sincere thanks to
the many people without whose help this thesis would not
have been possible, but especial thanks to the following:
Dr. Daniel S. Beasley for all his time and effort in try-
ing to make a researcher out of the writer; Dr. Leo V.
Deal and Dr. John M. Hutchinson for their time and con-
tributions as committee members; Mrs. Patricia Bainbridge
for help and ideas from her wealth of knowledge about
cerebral palsy; and to Ms. Ingeborg Vedder without whose
help I never would have gotten any subjects.

The writer wishes to express sincere thanks to

her parents for their support and understanding.

ii

TABLE OF CONTENTS

Chapter
I 0 INTRODUCTION 0 O O O O O O O .
Categorical Perception . . . . .

Identification Tasks . . . . .
Discrimination Tasks . . . . .

Motor Theory of Speech Perception .
Criticisms of the Motor Theory

Categorical Perception and the Cerebral

Palsied . . . . . . . . .
Statement of the Problem . . . .

II. EXPERIMENTAL PROCEDURES . . . . .

Subjects. . . . . . . . . .
Rating Scales . . . . . . . .
Experimental Stimuli. . . . . .

Stimuli Preparation . . . . .
Experimental Tape Preparation . .

Testing Procedures . . . . . .
Calibration. . . . . . . . .

III. RESULTS AND DISCUSSION. . . . o .

Identification Task . . . . . .
Discrimination Task . . . . . .

Predicted Discrimination Peaks. .
Obtained Discrimination Functions.

Ratings of the Speech of the Cerebral
Palsied . . . . . . . . .

Conclusions. . . . . . . . .

Implications for Further Research .

iii

Page

15

18
21

23
23
25
26

26
27

29
31

33

33
50

50
51

54
56
56

Chapter Page

IV. SUMMARY . . . . . . . . . . . . . 58
APPENDICES
Appendix
A. TASK TRIAL COMPOSITION . . . . . . . . 60
B. IDENTIFICATION TASK DATA FOR INDIVIDUAL
SUBJECTS . . . . . . . . . . . . 63
C. DISCRIMINATION TASK DATA FOR INDIVIDUAL
SUBJECTS . . . . . . . . . . . . 65
LIST OF REFERENCES. . . . . . . . . . . . 76

'iv

LIST OF TABLES

Table Page
1. Scales for rating experimental subjects . . . 25
2. Order of test presentations . . . . . . . 29

3. Cross-over points for data from the Identifi-
cation Task for temporally collapsed data,
in msec . . . . . . . . . . . . . 38

4. Predicted discrimination peaks from data
obtained by collapsing the judgments into
groups of two and three: these were also
separated by original word. . . . . . . 51

5. Ratings of the Overall Intelligibility of the
two Experimental Groups; ratings of their
functional capacity and of their need for
therapeutic appliances: and mean number of
errors on articulation test . . . . . . 55

Figure

1.

2a.

2b.

2c.

3a.

3b.

3c.

4a.

4b.

LIST OF FIGURES

Schematic representations of assumed stages in

speech production . . . . . . . . .

Identification task function for Group I:
Control for Intervocalic /p/ and /b/. . .

Identification task function for Group II:
Cerebral Palsied/Good Speech for Inter-
vocalic /p/ and /b/ . . . . . . . .

Identification task function for Group III:
Cerebral Palsied/Dysarthric Speech for
Intervocalic /p/ and /b/. . . . . . .

Identification task function for Group I:
Control for Intervocalic /p/ and /b/,
for combined data in groups of 2 and 3 for
frequency of /p/ and /b/ identifications .

IdentifiCation task function for Group II:
Cerebral Palsied/Good Speech for Inter-
vocalic /p/ and /b/, for combined data in
groups of 2 and 3 for frequency of /p/ and
/b/ identifications . . . . . . . .

Identification task function for Group III:
Cerebral Palsied/Dysarthric Speech, for
Intervocalic /p/ and /b/, for combined
data in groups of 2 and 3 for frequency
of /p/ and /b/ identifications. . . . .

Identification task function for Group I:
Control. Separated functions by: original
word for stimulus production and by group-
ings 2 and 3 stimuli . . . . . . . .

Identification task function for Group II:
Cerebral Palsied/Good Speech. Separated
by: original words for stimulus production
and by groupings of 2 and 3 stimuli . . .

vi

Page

10

34

35

36

39

40

41

43

44

Figure

4c. Identification task function for Group III:
Cerebral Palsied/Dysarthric Speech.
Separated by: original words for stimulus
production and by groupings of 2 and 3
stimuli . . . . . . . . . . . .

5a. Discrimination task function displayed by a
one-step comparison using results from
stimuli produced from original word
”rapid" . . . . . . . . . . . .

5b. Discrimination task function displayed by a
one-step comparison using results from
stimuli produced from original word
"rabid" . . . . . . . . . . . .

vii

Page

. 45
. 53
. 53

CHAPTER I

INTRODUCTION

In dealing with an individual with a speech
and/or language problem, the adequacy of the input
channel for speech and language processing must be
determined. Although the individual will be audiologi-
cally evaluated, normal hearing does not insure an_
intact language input, perceptual processing channel.

The individual must be able to perceive what he hears.

To use an analogy, an English speaker may hear French but
may not perceive what is being said. From this, speech
perception, then, would seem to be synonymous with under-
standing. In fact, speech perception has been defined in
a broad sense as recognition and interpretation of
incoming information or as the attaching of meaning to
stimuli to which senses respond (McDonald, 1964; Darley,
1964: Travis, 1971). In a more narrow sense, speech per-
ception has been defined as the implied transduction of
acoustic energy to neural patterns (Williams, 1972).
Because speech production is a motor function, it has

been suggested that the stored perceptual patterns

referred to may be, in part, at least motor patterns
(Liberman, et al., 1967). This approach to the study

of speech perception has been most thoroughly developed
within the framework of the Motor Theory of Speech Per-
ception. In turn, the Motor Theory has relied heavily
on the perceptual pattern phenomenon called categorical
perception. The ramifications of this definition and
theoretical framework to explain Speech perception for a
cerebral palsied population are particularly important

and far reaching.

Categorical Perception
. Identification Tasks

There are single aspects of the acoustic continuum,
eg. transitions of the second formant and silent duration,
that can signal the difference between phonemes and,
therefore, between words. For example, the duration of
the silent interval present in stop consonants is suf-
ficient for distinguishing two phonemes (Lisker, 1957;
Liberman, et al., 1961); and the second formant transition
is a sufficient signal for distinguishing the sounds /b/,
/d/, and /g/ (Delattre, et al., 1955: Liberman, et al.,
1957). One method of studying this phenomenon is through
the use of identification tasks. An identification task
is performed in order to determine whether an individual
perceives categorically or whether a phoneme is categori-

cally perceived. The listener will be presented with a

word and asked to identify it. The listener is given the
alternative identification names; for example, he is
asked to identify the word as either "rapid" or "rabid."
Thus, this is a forced-choice task.

If a phoneme stimulus is systematically varied
along one continuum, the sound will be acoustically simi-
lar to one phoneme at one end of the continuum and to
another phoneme at the other end. For example, using the
silent interval duration continuum, the duration can be
altered for the intervocalic /p/, shifting the perception
from /p/ to /b/ and vice versa. Lisker (1957), using
recorded speech words £2232 and ruby, found that the
perception shifted abruptly from £2223 to ruby during
an identification task as a result of the length of the
silent interval duration. At what point along the tem-
poral continuum the shift occurred was dependent upon
what the original word had been. If the stimuli had been
produced from £2229, the boundary was between 70 and 80
msec; for those stimuli produced from £221, the boundary
was at about 105 msec. Liberman et a1. (1961) used the
synthetically produced words rapid and rabid that varied
only in the silent interval duration of the intervocalic
segment. The phoneme boundaries lay at approximately
70 msec and were "reasonably sharp" (Liberman et al.,
1961, p. 183). When the alteration is of the second

formant transition, similar boundaries occur. Using an

identification task, Liberman et a1. (1957) determined
that sharp boundaries existed in the perception of /b/,
/d/, and /g/ along a continuum where the second formant
transition range was 1320 cps to 2980 cps. Although
increments in the silent interval duration or the tran-
sition remained constant, there was a place along the
temporal continuum at which point the alternative stimuli
identifications diverged. The listener appears to sort
the stimuli into categories. Although, for example,
there was no more difference between a word with a 60 msec
interval from one with a 70 msec interval and between one
with 70 msec and 80 msec, those of 60 msec and 70 msec

were perceived as rabid and those of 80 msec as rapid.

Discrimination Tasks

 

Another aspect of categorical perception is the
listener's discriminative abilities. When asked to dis-
criminate those sounds that have categorical identifi-
cation boundaries, the listener can discriminate no
better than he can label the stimuli. That is, discrimi-
nation within the phoneme categories has been found to
be much poorer than across phoneme boundaries (Liberman
et al., 1957). The listener may judge stimuli that are
closer on the temporal continuum as different, while
stimuli that are farther apart on that same continuum
may be judged as the same. Using an example of stimuli

with varied silent interval durations, the stimuli with

a 50 msec duration will be judged the same as one with

70 msec silent interval duration more often that the

70 msec stimulus will be judged the same as the stimulus
with an 80 msec silent interval duration (Liberman et al.,
1957: 1961; 1967).

A typical discrimination task would require that
the stimuli be presented in ABX traids, with X being
identical to either A or B. A and B would be stimuli
that were a certain number of steps apart on the con-
tinuum. The listener would then have to make a forced
choice of whether x was identical to either A or B.

To account for this type of discrimination

pattern, the psychological concepts of acquired simi-

 

larity and acquired distinctiveness have been proposed.

 

Dollard and Miller (1950) have described acquired simi-
larity in this manner: "Attaching the same cue-producing
response to two distinctive stimulus objects gives them

a certain learned similarity increasing the extent to
which instrumental and emotional responses will generalize
from one to the other [p. 101].” They have described
acquired distinctiveness as follows: "Attaching dis-
tinctiveness cue producing responses to similar stimulus
objects tends to increase their distinctiveness [p. 101]."
The listener, through his long association with the
language, has learned to name some of the variations

that he hears as /p/ and others as /b/. The place on

the temporal continuum of the identification boundaries
are determined through a learning process. That is,

the discrimination peaks appear at the phoneme boundaries
and support an acquired similarity and distinctiveness
explanation; these peaks also support that the placement
of the boundaries is through a learned process (Liberman
et al., 1961; 1967).

The types of discrimination and identification
patterns described here are unique because they are found
only when phonemes are the stimuli. Also, they are unique
because only certain phonemes can be used as stimuli.

The aspects of the speech continuum that are varied to
produce the categorical perception of the phonemes, i.e.,
the second formant transitions or silent interval dur-
ations, when heard outside of the speech context do not
produce the same type of identification and discrimination
gradients (Liberman, et al., 1957, 1961). However, cate-
gorical perception has only appeared as a perceptual
pattern associated with the perception of consonants,

particularly the stop consonants.

Motor Theory of Speech Perception
Research into the phenomenon of categorical per-
ception has led to the development of the Motor Theory
of Speech Perception that attempts to account for: (1)
the rapidity with which a listener can perceive speech;

(2) control of the variation of speech sounds that are

PM .ka ymu

t:
d

J u n u . r.\ . k.‘
a... .nu — u l s C... s w... .1. a a1 Tn cl. .hu 5 a

produced; and (3) the fact that the identification and
discrimination patterns are associated with a limited.
number of speech sounds. The Motor Theory proposes that
"speech is special" and, therefore, is perceived in a
special way. .

A purely "auditory" processor for speech is inade-
quate to account for the rapidity and accuracy with which
speech sounds are perceived. A listener can process as
many as thirty phonemes per second. The temporal resolv-
ing power of the ear cannot process thirty discrete
acoustic events per second and obtain anything more than
an unanalyzable buzz (Liberman, 1970). Consequently, a
model of speech perception would have to account for this.

The model would also have to account for the con-
trol of the variation of the speech sounds that are pro-
duced. The identification ability of listeners demon-
strates that this control does occur. The variation
associated with speech sound production are of at least
two types: allophonic variation and coarticulation.

Each distinct speech sound that we hear and can
discriminate from other speech sounds is called a phoneme.
Each phoneme is actually a group of sounds, called 21127
phones. The allophones are each different from the other
but more similar to members of their phoneme class than
to other phonemes. However, categorical perception tasks

have demonstrated that this is not always true; some

allophones are similar and yet assigned to different
phoneme classes. (This seeming contradiction will be
explained in the section dealing with the encoder/decoder
relationship.)

Allophonic variation is accounted for, at least
in part, by reference to the mechanical constraints of
the speech mechanism; that is, the movements are never
exactly the same for any two productions of the same
sound. Allophones are also the result of coarticulation,
which is variation in the production of a sound as a
result of contextual variation. The articulatory
positions of the surrounding phones in a given context
affect the production of the phone in question. There-
fore, it is not only the characteristics of the phone,
per se, but the phonetic context that aides in phonetic
discrimination. Liberman et al. (1967) stated: ”The
same phoneme is most commonly represented in different
phonemic environments by sounds that are vastly dif-
ferent. . . . This is a central fact of speech per-
ception [p. 432]." Thus, Ali, Gallagher, Goldstein, and
Daniloff (1971) found that listeners could predict the
presence of nasal consonants at the end of a CVC syllable
and a CVVC syllable, for example /keI/, with better than
chance results when the consonants and vowel-consonant
transitions had been spliced away from the end. This

finding suggested that not only were phones influenced

by adjacent sounds but that some information about a
phoneme is present in those adjacent phonemes.

The Motor Theory of Speech Perception postulates
the concept of the encoding of speech to allow for this
parallel transmission of information about more than one
phoneme at a time. This parallel transmission in turn
could overcome the temporal resolving power of the ear
and explain the ability to perceive phonemes more rapidly
than other classes of sounds. The model for production
of Liberman et a1. (1967) may serve as a useful reference
for this discussion (Figure 1). Each phoneme is a com-
bination of subphonemic features; each of these features
is associated with a specific set of "signals" that are
programmed throughout the speech production and per-
ception mechanism. True encoding occurs as a consequence
of the interaction of these features' signals. In the
model, as the signals for a speech movement move from
level to level, the conversions are monotonic. However,
at the level of conversion from muscle contraction to
vocal tract shape, a very considerable amount of encoding
must occur. This would be related to the mechanical and
contextual constraints described earlier. Thus, encoding
is the merging of past and present instructions to the
vocal tract, and the subsequent loss of segmentability
of the acoustic signal into discrete phonemes (Liberman

et al., 1967). Encoding would not seem to occur below

10

SCHEMA FOR PRODUCTION

LEVELS CONVERSIONS
Soacho
Syntactic
Rules
PI P293----------Pn

 

 

Pumas (or Soto of t.
Distinctive rectum) '2 -—

 

 

t3 _. __

t4 __

Monotor Ruloa
Motor Commands c. - c3

 
  

Myomotor Rules

 

 

ChaocNddmteuuuna I 293
Articulatory Rulos
Place and Manner i oz 3

Acoustic Rules
WWW»

(Spooch)

 
  
     

DESCRIPTIONS OF
SIGNALS

Roprountod by pottorna of
carrot activity in tho Con-
"mlunmNIQnNm.

(as do»)

Naval Signal: to Honcho

Mosaic Contraction.

Woo! Tract m, oto.

Sounds of Spooch

Fig. 1. Schematic representations of assumed
stages in speech production (Liberman et al., 1967,

p. 445).

11

the leVel of conversion from muscle contraction to vocal
tract shape because of the monotonic relationship of the
shape of the vocal tract to the sound signal. This mono-
tonic relationship is measurable, but above the pr0posed
level of encoding such a relationship is inferred and is
more difficult to measure. It is possible, therefore,

that some encoding occurs at higher levels in this model.

The encoding of phonemes by the speaker necessi-
tates that the listener be able to extract the correct
information from this coded signal; that is, he must be
able to decode what he hears.

In perceiving or decoding speech the listener
must have some reference that is consistent or invariant
in the sound to allow for accurate perception and
identification. It has already been explained that the
acoustic signal has no such invariance as a result of the
encoding process. The invariance necessary for accurate
decoding must be present prior to the encoding, above the
level of vocal tract shape. It is unlikely that such
invariance occurs at the muscle contractions level for:
"If muscles contract in accordance with the signals sent
to them, then this conversion is essentially trivial
(Liberman et al., 1967)."

Perhaps the neuromotor signals provide the neces-
sary invariance for the decoder. Through electromyographic

recordings of the muscles of the speech mechanism it has

12

been demonstrated that enough invariance is present to
fulfill the expectation that the motor commands are suf—
ficient to be used by the decoder in the perceptual pro-
cess (Liberman, et al., 1967). This concept, however,
necessitates a close association between speech production
and speech perception. Shankweiler and Studdert-Kennedy
(1967) were of the opinion that the most practical system
for decoding would be as part of the encoding system, both
with appropriate linkages to the sensory and motor com-
ponents of the speech production centers. They stated

" . . . it would be unparsimonious to assume that the
speaker/listener employs two entirely separate processes
of equal status, one for encoding language and the other
for de-coding [p. 452]." Consequently, one hears the

same /d/ because perception is mediated by the neuromotor
correlates of gestures that, in essence, are the same.

The /d/ gesture, or some important characteristic of that
gesture, may be the same in any two cases (Liberman, et
al., 1967).

The idea of mediating neural correlates of articu-
latory gestures is the basis of the explanation of the
differences in the perception of speech sounds. Earlier
it was noted that only certain phonemes were categorically
perceived. Only the phonemes that are encoded are cate-
gorically perceived. The listener after long experience

associates the phonemes with the appropriate articulations;

13

over time these articulations, or the neuromotor patterns
of these articulations, become part of the perceiving
process. When significant acoustic cues that occupy
different positions along a single continuum are produced
by essentially discontinuous articulations, i.e., /p/ and
/b/, the perception becomes discontinuous, or categorical.
When acoustic cues are produced by movements that vary
continuously from one articulatory position to another,
e.g., the vowels, perception tends to change continuously
with no apparent perceptual categories (Liberman, et al.,
1967).

The association of the production and perception
of speech leads to the hypothesis that the placement of
the phoneme boundaries and the discrimination of encoded
speech sounds was a learned ability mediated by the
neuromotor correlates of articulatory production (Liber-
man et al., 1961; 1967). This is supported by the con-
cepts of acquired similarity and acquired distinctiveness.
Through experience with the production of speech, the
listener begins to categorize the speech sounds. Some of
the allophones are very similar acoustically, and the
listener must learn where to place the phoneme boundary
to perceive them accurately. Those within the category
acquire similarity; those across the boundary acquire
distinctiveness. This explains why two similar allophones

are perceived as different phonemes. The concept of

l4

learning playing a part in speech perception is further
supported by evidence from discrimination tasks. The
discrimination of encoded speech sounds is largely con-
trolled by the phoneme labels available (Liberman, 1961).
Further, the position that perception is mediated via
stored neuromotor patterns suggests that Speech, in some
way, is a Special phenomenon associated with specific
characteristics of man's neurological system.

The "speech is Special" tenet is supported by dif-
ferences between the perception of speech and nonspeech
stimuli. The nonencoded speech sounds, vowels, and non-
linguistic stimuli, tones and so on, are perceived in a
continuous manner. There is also another difference
between encoded sounds and nonencoded sounds with respect
to place of perception. Through dichotic listening tasks
it has been shown that speech stimuli are more readily
perceived by the right ear, which dominates the auditory
tract to the left hemisphere (Kimura, 1961). The left
hemisphere has been found to be pre-potent for speech
at a very early age (Kimura, 1967). Studies have been
conducted that not only demonstrate a stronger right ear
advantage for speech sounds but a stronger right ear
advantage for phonemes that are highly encoded (Day and
Vigorito, 1973; Cutter, 1973). Nonlinguistic stimuli,
e.g., melodies, have been Shown to be more readily per-

ceived in the right hemisphere (Broadbent and Gregory,

15

1964; Bryden, 1963). These data would support the pro-
posed close association of the language and speech pro-
duction centers with the Speech perception centers and,
concommitantly, a close association of the encoding and
decoding of speech.

Consequently, it has been suggested that the per-
ception of Speech is Special. It is carried out by a

special processor, and it has a strong learning component.

Criticisms of the Motor Theory

The presence of stored motor patterns for cate-
gorical perception for certain speech sounds has been
challenged by Lane (1965). In researching the categorical
perception of nonspeech stimuli, he has obtained identifi-
cation boundaries and discrimination peaks at these
boundaries. He used stimuli that were produced by
inverting the Spectrograms of speech stimuli and than
playing these through a Pattern Playback device to con-
vert the picture to sound. In this way, Lane obtained
stimuli that varied as the speech sounds varied but were
nonlinguistic in nature. Similar stimuli were used by
Liberman et a1. (1961), but no phoneme boundaries or
discrimination peaks were obtained in that experiment.
Beare (1963) and Ekman (1963; in Lane, 1965) obtained
labelling gradients comparable to those of the stop
consonants with stimuli of Spectral colors. The dis-

crimination of the colors was the most acute at the

16

boundaries between color classes, just as discrimination
is most acute at the boundaries between phoneme classes.
These data would appear to cast doubt on the premise '
that categorical perception is peculiar to phonemes and,
therefore, that speech has a special processor.

The premise that there is a learning component to
speech perception is challenged by data from experiments
of infant perception. The left hemisphere is pre-potent
for speech and language at an early age (Kimura, 1967).
Some would go so far as to say that this specialization
is present at birth (Molfere, 1973). If this is the
case, a certain undefinable amount of speech and language
ability is innately determined. For example, it has been
found that infants of one and four months of age can
categorically perceive StOp consonants (Eimas, et al.,
1971). The placement of the phoneme boundaries would
appear to be one of the innate characteristics of speech,
at least to some extent. These results would not support
a special process for Speech perception that depended on
the use of learned articulatory patterns.

This information could cast serious doubt on the
Motor Theory of Speech Perception. The criticism that
nonlinguistic stimuli are categorically perceived has
been responded to by Studdert-Kennedy et a1. (1970).
Obtaining the same results for nonencoded stimuli as for

encoded stimuli for labelling functions is possible. The

17

important factor in these experiments is the degree to
which the stimuli are categorically perceived. Any set
of stimuli, if spaced widely enough along a sound con-
tinuum, can produce labelling functions that are categori-
cal. For the encoded speech sounds, the Space necessary
for this type of function is very small. In examining
the labelling gradients, the sharp decline in alternate
identifications produces a valley on the function; a
valley also appears to either Side of the discrimination
peaks. These valleys are as important as the peaks;

they depict the lack of alternate identifications or the
chance discriminations. These valleys are missing in the
Lane (1965) study. The absence of the valleys indicate
inefficient sorting of within category stimuli.

There is no rebuttal for the fact that there is
categorical perception in the infant. Presence of pho-
neme boundaries at birth does not preclude a special
processor; it does not preclude a close association
of production to perception. It does indicate that
the learning component of speech perception may not be
critical. However, a learning component could still be
active; learning could serve as a means to refine the

Perceptual speech and accuracy of speech perception.

l8
Categorical Perception and the
Cerebral Palsied
If transduction of acoustic energy to neural
patterns, learned or innately given, is a component of
speech perception, an individual with brain damage could
have difficulty perceiving speech. Cerebral palsy is
defined according to Westlake and Rutherford (1961) as
" . . . a group of disturbances of motor function which
occurs as a result of involvement of the cortical and
subcortical motor control areas [p. 1]." The proposed
close association of the production patterns of Speech
sounds with the perception of speech would suggest certain
important ramifications for the cerebral palsied population.
Speech is a motor function, resulting from neuromotor pro-
gramming activity; and the cerebral palsied individual, by
definition of the disorder, suffers from motor dysfunction.
In fact, one of the problems frequently associated with
cerebral palsy is disordered speech. These problems can
be caused by a wide range of lesions at various levels of
integration of the neurological tracts. They can be
symbolization difficulties, muscular involvement, or motor
programming problems. Referring to Figure 1, it is obvious
that cerebral palsy can disrupt production of Speech at
any level. If the Motor Theory is valid and if the dis-
ruption occurs at the level of the conversion of muscle
contraction to vocal tract Shape, then the disruption in

production should in fact result in a disruption in

19

perception. Those speech problems due to muscular
involvement of dysarthrias would cause the individual to
have production problems, with absent or distorted pro?
duction patterns for the speech sounds; this would be
especially true if there is a learning component to
Speech perception. Ability of the cerebral palsied to
perceive categorically would bear directly on a Motor
Theory of Speech Perception. This in turn could bear
directly on the approach of the intervention program used
with the cerebral palsied in speech therapy.

McNeilage, Rootes, and Chase (1967) conducted a
study of a seventeen-year-old female with severe impair-
ment of somesthetic perception and motor control. On
articulation, or production tasks, she performed poorly.
In an attempt to assess her categorical perception ability
several different types of stimuli were used in identifi-
cation and discrimination tasks. She was asked to identify
and discriminate voiced stOp consonants in syllables and
front vowels. The identification task only was adminis-
tered for voiced and voiceless stop consonant initial
position, voiceless stOp consonant in initial cluster,
and voiced/voiceless stop consonant in intervocalic
position. She achieved the same degree of unanimity of
identification for only the EEEEE/EEEX.93 did the sub-
jects, who were normal college students, in a previous

experiment using the same stimuli. However, McNeilage

20

et al. concluded ”no tendency for the overall trend of
the patient's perceptual judgments to be influenced by
her productive disabilities [p. 464]."

The presence of a categorical perception pattern
in the identification and discrimination of an individual
with a neurological disorder lends support to the critics
of the Motor Theory of Speech Perception. The data from
the McNeilage et a1. (1967) study support a nonlearning
based theory of speech perception. The data also could
support an approach to speech perception where speech is
not considered to be special. The presence of categorical
perception in a brain damaged population could suggest
that the Motor Theory of Speech Perception is wrong; that
the neuromotbr correlates are unnecessary for speech per-
ception. Or, this perceptual pattern in a braindamaged
population could suggest a modified Motor Theory of Speech
Perception; one where the system is innately tuned to
develop speech and language, both for perception and pro-
duction, and the production_of the phonemes does not aid
in perception, thereby eliminating the learning component.
That is, the female subject used in the McNeilage et al.
(1967) study may or may not have the propensity to
develop the neural patterns for production innately
given, and she may or may not use them for perception
of phonemes. The presence of the categorical perceptual

pattern in a neurologically impaired individual does

21

indicate that she is perceiving as a normal individual,
whatever the process may be.

Categorical perception tasks could help determine
whether the cerebral palsied population are perceiving
as the nonbrain damaged population does. This information
could raise the question of the effect of different per-
ception on speech and language and the effect of Speech
and language on perception. The ability of a cerebral
palsied person to perceive normally would benefit speech
and language pathologists working with these individuals.
Also, this information could provide valuable information
to investigators of speech perception. That is, if the
Motor Theory of Speech Perception is valid, than persons
who exhibit neuromotor problems, such as cerebral palsy,
should also exhibit difficulty in performing certain

speech-Specific perceptual tasks.

Statement of the Problem

From a review of the literature, there would
appear to be a need for a study that would explore the
speech perception of the cerebral palsied population
with respect to the phenomenon called categorical per-
ception. This is because (1) this population has a
high probability of speech perceptual difficulties due
to brain damage, and (2) information about categorical
perception in pathological populations may help clarify the

issues concerning the Motor Theory of Speech Perception.

22

Specifically this study seeks to answer these

questions:

1.

Does the cerebral palsied individual exhibit a
categorical perception pattern, as measured by
identification and discrimination tasks using
the words rapid and Egbig_and varying the inter-

vocalic silent interval duration?

If so, how does this pattern compare with non-

brain damaged individuals?

Are there any differences between the perceptual
patterns of cerebral palsied individuals with

good Speech and those with dysarthric speech?

How do judges rate the intelligibility of cerebral
palsied speakers, and does this correlate in any

way with severity of involvement?

CHAPTER II

EXPERIMENTAL PROCEDURES

This study used fifteen subjects divided into
three groups. Each subject listened to a set of tests
designed to demonstrate his ability to perceive categori-

cally.

Subjects

The subjects of this study were fifteen adult
volunteers, all of whom exhibited normal hearing (+20dB)
in at least one ear at the frequencies of 250, 500, 1000,
2000, 4000, and 8000 Hz. All had normal intelligence as
measured by completion of at least a high school edu-
cation. They were divided into three groups of five
each: a control group and two experimental groups.
The control group was comprised of normal young adults.
The experimental groups were comprised of subjects who
were diagnosed as cerebral palsied, having sustained
brain damage pre- or peri-natally. Each subject was
asked to read the Sentence Articulation Test Section of
The Fisher-Logemann Test of Articulation Competence and

to speak extemporaneously for three minutes. Three

23

24

trained judges rated the subject on over-all intelligi-
bility, articulatory errors, and the presence of dysar-
thric speech patterns.

More specifically, the composition of the groups

was as follows:

Group I: Control. This group consisted of three
females and two males. The age range was from 21 years
9 months to 28 years 3 months with a mean age of 24 years
2 months. Each subject was free from a history of neuro-

logical impairment and speech and language impairment.

Group II: Cerebral Palsied/Good Speech. This
group was comprised of one male and four females. The
age range was 20 years 4 months to 47 years 5 months with
a mean age of 29 years 5 months. Each subject was rated
by all three judges as being free of articulation errors
and dysarthric behavior, and all received a high positive
overall intelligibility rating of 1 (see Table 1). All
subjects except one were able to mark their own score
sheets. The one subject who required assistance was
visually impaired; the examiner marked as the subject
responded verbally. This group consisted of four sub-

jects diagnosed as spastic and one as athetoid.

Group III: Cerebral Palsied/Dysarthric Speech.

 

This group was comprised of one male and four females.

The age range was 22 years to 45 years 4 months with a

25

mean age of 41 years. Each subject was rated by all
three judges as having at least three articulatory errors
and exhibiting some behavior descriptive of dysarthric
speech patterns. All sounds on the articulation task
could be counted as errors. The range of number of
errors for these subjects was 11 to 130; the mean number
of errors was 43.6. The subjects received varied over-
all intelligibility ratings. (See Table 1.) All sub-

jects were able to mark their own score sheets.

TABLE 1. Scales for rating experimental subjects

 

 

Functional Therapeutic InteIIigggility
l-no practical l-no bracing or 1-always intelligible
limitation of apparatus 2-almost always
activity Z-minimal bracing intelligible
-2-slight limi- or apparatus 3-usually intel-
tation 3-bracing & ligible
3-moderate limi- apparatus 4-usually unin-
tation 4-institution- telligible
4-great limi- alized S-almost always
tation unintelligible
5-no useful 6-always unintel-
activity ligible

 

Rating Scales
Three listeners, all Master's level students at
Michigan State University's Department of Audiology and
Speech Sciences, listened to recorded samples of the
experimental subjects' speech. They scored each subject
on the presence of articulatory errors, dysarthric speech

patterns, and overall intelligibility. The experimenter

26

rated the experimental subjects on two separate scales
similar to those of Denhoff and Robinault (1960). One
of these dealt with functional ability, as defined by
what the individual could do without assistance (mechani-
cal or human). The other scale can be described as thera-
peutic, defined here as the amount of bracing or apparatus
used by the individual. These three scales are described

in Table 1.

Experimental Stimuli

Stimuli Preparation

In order to determine whether categorical per-
ception existed, identification and discrimination tasks
were employed. These tasks consisted of words whose
silent intervocalic segments were temporally modified.
The words £2212 and Espid_were spoken by a trained adult
male speaker in a sound-proofed recording suite using an
Electro-Voice 635A Dynamic Omni-directional microphone.
The words were recorded on a full-track tape deck
(Ampex 601, Model 652) at 7 1/2 ips. These words were
then dubbed from the original recording to a second tape,
using a full-track Ampex Model AG 600 and Ampex 601 tape
recorders. These second generation recordings were used
for generating the experimental stimuli.

The stimulus items for the experimental tasks

were prepared by a mechanical cut and splice technique as

27

described by Lisker (1957). Using a tape editor, the
silent intervocalic interval was aurally determined for
each word. The silent interval was cut from each word
and a segment of blank leader tape was Spliced between the
segments preceding and following the silent interval.

To insure that only the silent interval had been cut away
and that the Spliced segment was of the desired duration,
a spectrogram (Kay-Sonograph) was made of each stimulus
item. These were compared with Spectrograms of the orig-
inal words. The silent intervals were measured and found
to be accurate to within :4 msec. Twenty-four stimuli
with silent intervals ranging from 40 msec to 150 msec

in 10 msec steps were made from the words 52212.336 52219“
There were twelve stimuli per word.

Experimental Tape
Preparation

 

Two tasks were administered to each subject.
One task required an identification or labelling
response; the other was a discrimination task where
the subject was asked to make a same/different judgment.
Both tests were prepared by dubbing the prepared stimuli
from an Ampex 601 Model 652 recorder to an Ampex Model
AG 500 tape deck, half-track, using both tracks simul-
taneously.

Test I was the Identification Task. It consisted

of three presentations of each of the 24 stimuli, for a

28

total of 72 trials. The subject heard the number of the
trial, a 2-second pause, and then the test stimulus.

The subject was given 10 seconds in which to respond.
All trials were on one tape. The length of time to
administer this task was approximately 15 to 20 minutes.
Randomization was achieved by selecting a number from

a set of 72 numbers where each number from 1 to 24 was
represented three times.

Test II was the Discrimination Task, designed
according to a paired comparisons Ax paradigm. It con-
sisted of pairing all stimuli prepared from one word with
all other stimuli prepared from that word, that is, all
stimuli prepared from the original word £3pi§_were paired
with all other stimuli prepared from the word Egpig.

The same was true for all stimuli prepared from the
original word £221§° Stimuli prepared from different
words were not paired, for example, one from Egpig_to

to one from £2219! in order to eliminate potential per-
ceptual cuing by a difference other than silent interval
duration.

Test II consisted of 78 pairs for each set of
stimuli, resulting in a total of 156 trials. Each trial
consisted of the trial number, a 2-second pause, test
stimulus A, a 3-second pause, test stimulus B, and a
lO-second pause for responding. The pair presentations
and the ordering of the stimuli within each pair were

randomized.

29

This experimental task was presented in two
parts: Part A consisted of trials 1-78 and Part B con-
sisted of trials 79-156. Each part required approximately
20 to 25 minutes to administer. (See Appendix A.)

Thus three experimental tapes were constructed.
One consisted of the stimuli to be used for the identifi-
cation task. Two tapes were constructed for the discrimi-
nation task, parts A and B. With three tapes, there were
six possible presentation orders. The presentation
order for each subject was determined on a random basis.
The order of presentation was kept constant between

experimental and control groups; see Table 2.

TABLE 2. Order of test presentations

 

Subject Number Test Order

 

test I; test IIA; test IIB
test I; test IIB; test IIA
test IIA; test IIB; test I
test IIB; test IIA; test I
test IIA; test I; test IIB

U'IDUJNI"

 

Testing Procedures
The examiner screened each subject's hearing
prior to administering the articulation or listening
tasks. A portable Beltone audiometer was used in a
sound-treated room using standard audiometric procedures.

The examiner then asked each subject to read the sentences

30

on The Fisher-Logemann Test of Articulation Competence.l

After this, the subject was asked to talk about his job,

school, or something of interest to him for approximately
three minutes. This was all recorded on a Sony portable

tape recorder, Model TC-106A.

The examiner then asked for pertinent information
from all subjects such as birthdate and for information
from the experimental subjects such as type of diagnosis
and time of onset of cerebral palsy.

The examiner then went into the control booth.
The instructions were printed on a test booklet in front
of each subject. Inside each test booklet were response
sheets associated with each of the experimental tasks.
The instructions were read by the examiner through a loud
speaker located in the sound-treated listening room. The
directions for the Test I: Identification Task were as
follows:

Test I: You will be listening to single words. You are
being asked to decide if this word is "rapid" or “rabid."
For each word you will first hear the number of the trial,
then the word. Then mark on the answer sheet which word
you think it is. Please listen carefully. If you want
more time on any word, or want to stop the tape, let the
examiner know. Thank-you.

The directions for the Test II: Discrimination Task were

as follows:

 

1It was necessary to use the picture portion of
this measure with the visually impaired subject.

31

Test II: You will be listening to pairs of words. Some
will be the same word and some will be different words.
You are being asked to decide if the pair is the same or
different. If you decide they are the same, mark SAME on
your answer sheet; if you decide they are different, mark
DIFFERENT. Please listen carefully. If you want more
time on any pair, or want to stop the tape, let the
examiner know. Thank-you.
After the directions were read, the subject was advised
which answer sheets to turn to.

Both tests were presented at 72 dB SPL through
a single loud speaker in a sound-treated room. The sub-
ject sat facing the Speaker at a distance of five feet.
The test tapes were played on a tape recorder of a Maico
Audiometer (Model MA-24). The experimenter monitored the
subjects both visually and auditorily from the control

room.

Calibration

Prior to testing, all equipment was calibrated.
After eight of the 15 subjects had been tested, the
equipment was again calibrated. At the end of the
testing procedure, the equipment was checked for con-
sistency of calibration.

The calibration procedure was as follows:
a Brfiel-Kjaer sound level meter, type 1613, with a 4145
condenser microphone was fixed to a tripod at the same
height as the head of a person sitting in a chair and
placed five feet from the speaker. White noise was

generated at 70 dB HTL and presented through the

32

speaker. Using the A scale on the sound level meter,
the reading was 72 dB SPL.

The tape recorder was calibrated by playing a
tape of a 1000 Hz tone at 70 dB HTL through the loud
speaker and attenuating the Speech gain to peak at

0 dB VU.

CHAPTER III

RESULTS AND DISCUSSION

Identification Task

The total judgments for each silent interval
duration were computed (see Appendix B). The results
of the identification task are shown in Figures 2a to
2c. The results of the identification task indicate that
the cerebral palsied population does perceive categori-
cally. Phoneme boundaries are evident for each of the
three separate groups. Because the task was a forced
choice between two alternatives, the graph lines are
mirror images of one another. From these graphs it is
obvious that the boundaries are not as clear cut as in
previous experiments. Previous experiments (Liberman
et al., 1957, 1961) were concerned with the concept of
categorical perception and not with the formulation of
normative data. Therefore, in previous studies subjects
were eliminated because they failed to label consistently
and/or failed to exhibit perceptual boundaries. None
of the subjects in the present study were selected on
the basis of ability to label consistently or were given

any training in this type of task. This could be the

33

34

.mcowumowmwucoow \b\ one \m\ we hocosvoum o>wumaou msonm ucflom zoom .\b\ one \Q\
owamoo>uoucH How Houusou «H macho How sowuocsm xmmu SOADMOAmaucooH .cw .mwm

dun! z. Asap—.2. .qunzm mo zogdmao
co. 0! On. ON. 0: 00. 00 on O» 00 0o 0?

 

2

(3
GI

I)
N

QIILG 32.2.3. 36$...

OIIIO 32399. ..o.a<¢..
.2230 u. 9.20

'3 8
(02-1) smuasonr 1v101 :IO assume

 

35

.mcoHuMOAMHucoow
\b\ one \m\ we mocoavouw msonm ucHom comm .\n\ one \Q\ oflamoo>uoucH How nooomm
©000\oowmamm Hmubouou «HH macaw How sowuocsm xmmu SOHHMOAmeucmoH .nm .mflm

dun: z. 456mb:- kzunzm “.0 20:53.0
0! 0». ON. 0: 00. OO 00 Oh 00 On 0?

2

 

<> (3
0 N
(02-1) Slususoonr ‘IVIOJ. :IO UBBNnN

GIG 32.6%! gems...
OIIO 32.2.8. 33$...
58% 38\......o._ .2830 s 92....

36

.msowumofimwucmofi \n\ one
\m\ we mucosvmum msonm unwed zoom .\n\ can \m\ owamoo>umucH How nummmm OAHSHHMmha
\oowmamm Hangouou "HHH msouo How SOHHUSSM xmmu cowumowmwucooH .oN .mam

.omm: z. #355. .536 ...o 222%..

on. O! on. ON. 0: 00. cm 00 ON 00 on 0.... o

9

AV AU. AV \\\\MV nU\\\\kU\\\\nyllllhv
/O\ /4/<\<

I)

N

8 8
(02 cu Sluauaoonr mm so assume

GIIIQ 82892. .905...
OnIIO 32.83. .052:
5.26 2.5326 \ .350... .9523 “.= 39.0

9

 

37

reason for the more variable boundaries. However, there
are also differences between the groups in phoneme boun-
dary placement. For Group I (Fig. 2a) the boundary is
most distinct, occurring at approximately 130 msec, for
the durations used. As the variability in the groups
increased, the type of configuration became more con-
fusing. Although there still appeared to be phoneme
boundaries, rather than one or two cross-overs, there

is a wide range along the continuum of what appears to
be "random" identification (Figs. 2b and 2c).

Group II data showed cross-overs in the same
region on the continuum as Group I. The cross-over points
for this group were associated with shorter silent inter-
val durations, the first appearing between 100 msec and
110 msec. For Group III the cross-over points on the
continuum were at 90 msec and continue through the same
silent interval durations as for Groups I and II.

Liberman et a1. (1961) found, for synthetic
speech, that 70 msec was the phoneme boundary between
the perception of /p/ and /b/. Lisker (1957), who used
recorded live speech, also determined that between 70-80
msec was the boundary for some stimuli, but that approxi-
mately 105 msec served as the boundary for other stimuli.
These cross-over points in the present study were all
higher than those previously found, especially for

Group I, which was the group most comparable to these

38

other studies. This difference could be related to the
stimuli. It could also, and this seems more likely,
be due to the nonpreselection of the subjects.

To determine whether the boundaries for the three
groups converged, the data were collapsed over temporal
intervals. .That is, the total judgments for two steps
along the continuum (i.e., 40 msec and 50 msec) were
computed and then graphed together; and subsequently
results for three steps along the continuum were handled
similarly. These data are shown in Figures 3a to 30.

For each graph, the section on the right is the plotting
of two stimuli simultaneously; on the left is the plotting
of three stimuli simultaneously. A clearer picture was
gained of the subjects phoneme boundary placement.

Groups I and II placed the phoneme boundary similarly.
Group III varied from the other two groups in placement.
The phoneme boundaries appeared to be closer together

than when the stimuli were plotted as individual points.
The points at which the data crossed over for each
grouping of stimuli are listed in Table 3.

TABLE 3. Cross-over points for data from the Identifi—
cation Task for temporally collapsed data, in

 

 

msec.

Group 2 Stimuli 3 Stimuli
Group I 130-140 120-130
Group II 110-120 120-130

Group III 90-100 90-100

 

39

.mcoauMUAMAucocw
\n\ one \Q\ mo mucoswoum How m can N we mmsoum cw sumo pmcwnaoo now .\n\ one \Q\
usamoo>uoucH How aouucoo "H macho mom cowuocdm Mme» coaumowmwusocH .Mm .mHm

dun! i .3255 baa-o no 8:55:
8:0». ONTOO. catch 86* 0910.! 09:0“. 0.700. 00.00 2.100 006*

(D
V’

SINWI' W101 d0 UNION

GIG 32.282 :07 GIG .889... .93..

OIO 3.892 .99). G OIO 328a... .b.a<¢..

:22 £8.58... on. :28 .558... on.
use... .2... ea 8.85 . :3...» at. Go 3.3.6

38238

c
2

 

.3230 .- .35

40

.mSOHumowmwucoow \n\ mam \m\ we mososonm
How m osm «.mo mmsoum cw mumo cocwnfioo now .\n\ cam \m\ owamoo>umucH mom nommmm
UOOU\anmHmm Hmunouou “HH.msouo How sowuossm xnmu sowumowmeusooH .nm .mwm

.OUOS :- Izzy—.2. FszE “—0 20ng
8:0». 02:00. 00:2. 0010' 09.0.! 0070“. 0.700. 00.00 2.00 090'

o
o.
:m
s m
2m
0
a...
8 m
GIG 8.582 .93.. GIG .8882 .205... com
OIO 82282 .05.... OIO 8.892 sees...
:28 2.858... oo. :22 8.35.2.3 03 co
=2...- oox: .6 .996- =33. 2: 3 83.5 co.

 

so...» .33 328.233 .3.

41

 

.msowumowmwucoofl \n\ com \m\ 00 mocmsvoum How m cam N
no mmsoum cw mumc omswnEoo How .\n\ cum \0\ owamoo>uoucH How .cooomm oeunuummho
\oowmamm Hmunouoo ”HHH msouw How sowuosSm xmmu sowumoamwusooH .om .mflm

dual z. 455...! badwflo 80.5.5.8
00700. 0N100. 00:05 80¢ 0010! 03.02 0:00. 00.00 0W8 00:00

GIG 82.8.9. 285....
0 IO 82.8.2 .053...

GIG 38.2.3. .903...
OIO 82699. .35...

:22 255.8... 08
.35.. as. .o .929

:22 .2258... as
.35... .2... 3 836

 

5....» staggers... .528 ”a 92o

42

There was still a decelerating trend, toward
shorter durations, from Group I to Group II to Group III.
However, when the data were plotted in groups of three,
Groups I and II converge. That is, the phoneme boundaries
of these two groups were closer than for Group II and
Group III.

Lisker (1957), using the words 52222 and pppy,
found the range of closure durations for /p/ 90-140 msec
with an average of 120 msec, and that for /b/, the range
was 65-90 msec with an average of 75 msec. If the stimuli
used had been made from the word £2222! judgments were
divided between 70 and 80 msec. If the stimuli had been
produced from 5221! the boundary line was at 105 msec.
Therefore, a 30 msec closure duration difference can be
expected for identification gradients made from different
words. In the present study the words were pgpig and
53239. To determine whether a boundary shift as described
by Lisker (1957) was present, the data were separated on
the basis of what the original word had been. These data
were then computed in groups of two and three points
along the temporal continuum. Graphs were made for the
data from each word; there were two graphs per word, one
for the groups of two stimuli and one for the groups of
three stimuli. These are shown in Figures 4a to 4c. The
boundaries for those stimuli made from 52219 are not dif-

ferent from Group I to Group II to Group III. So,

43

ooumnmmom .Houusou «H macho How newness“ xmmu sowumoamwusooH .me .mwh
. due... 2. Sent... Eu...» .3 22.5.8

mm:mm:wm::ww:m.
uu««m«.. -muuu.»mm«..

    

drIAV.ie§xr9.n=m<¢
nYIAu.xzsxto.n:m<¢

  
   

 

 

 

$32.. .835 532.2395 .22.... 3525 $3222.25
:22 £553.... :22 £2.32... on.
.3...» n 3 a .2...» u .o 2.35

4095.00 ._ .593

.HHSEHum m cam N mmcwmaoum an cam sowuosooum msassflum new ouo3 Hmcwmfluo «an mcowuosSm

O D
Q N
SLNBNSOOOI' W101 :IO 8385.0”

0
G'

44

.HHSEHum m mam N
no mmcfimsoum an cam cofiuospoum mSHSEHum Mom mouo3 Hmcwmfiuo “an Umpmummmm .noommm
ooou\coflmamm Hmubmuoo “HH msouw How cowuocsw xmmu cowumoHMHucooH .Qv .mwm

dual 2. J<>¢uh== him...” no 20.5.5.5

H m t H m H 3 m N M m
.m... ma...mmm...mmm
I Z 6 I I I I H I I H
m o. o m m M M m m M o m m m m M o m m Mm

.\ o...\./. ..
0/0 I /.\.V

 

s 3

Ci
G’

I:
N
smauasonr ""101 JO aasmu

GIG .388... 90$.
OIO 82.88.. 9&5.

ID
‘.

 

 

22.2.. .325 .23... .325 :28... 352.6 .539... .853
:22 £23333... :2... £553.... on.
.35..» n .o 2.35 .2...» u .o 2.25

.33...» 32593.... immumuo S .323

45

can N no mmcflmsoum an can cowuosvoum msaafiwum How mvuo3 Hmcflowuo

oaunuummmo\umfimamm Hmunonmo

OZI‘OOI
06" Ol.
09"0?
OGI'OEI
OZI'OOI

m
m
w

 

 

.=zo._=s:5::sao.°
:25.» n 3 2.3.6

 

"HHH muonw How GOHuUGSM xwmg :owuonMHuchH

dual 2. J<>¢mh5= mam-cw 8 202.530

.'

06"01.

m
9
O

 

O
o m m m

.
9
O

m w
m m

 

 

4'4 82.092 9m<¢
OIO 8.80... 93¢

.532. 352.6

 

OEI'OZI

   

.flaseflum n

"ma cmumnmmmm
00v omnfih

w
w

06‘09

.28....osoto

:22 £358.... on.

:35.» N 3 3:20

tomato o_¢§>o\om_04(a 4<¢ow¢uo "=- maoca

   

.nommmm

 

2

46

consequently these boundaries can be said to be the same.
For those stimuli made from rapid, the downward trend

was still apparent. Group I showed no cross-over points
for these stimuli. Group II had cross-overs at 90 msec
to 100 msec, 110 msec to 120 msec for groups of two
stimuli; for groups of three stimuli, the cross-over was
between 120 msec and 130 msec. Group III data for groups
of three stimuli placed the phoneme boundary between 60
msec and 70 msec; for groups of two, the phoneme boundary
was between 70 msec and 80 msec. Only the data for

Group III showed the 30 msec difference found by Lisker
(1957), a difference which he attributed to other cues.
He did not elaborate on what these other cues might

have been.

The perceptual patterns do show a difference
among the groups. Whereas each group appeared to per-
ceive categorically, the categories are quite different,
indicating some difference in perception. Because the
two cerebral palsied groups differed from the control
group, it would seem that a part of the difference could
be due to the presence of cerebral palsy, and more spe-
cifically to the motor dysfunction associated with
cerebral palsy. This contention is supported by the
fact that the two cerebral palsied groups differed from
each other whereby the more severe group was less apt

to show categorical perception. However, the more

47

severely involved subjects, with respect to speech
involvement, in Group III, were not less apt to demon-
strate categorical perception patterns than other sub-
jects in Group III.

Although there is categorical perception in all
three groups, a Motor Theory of Speech Perception can
help to explain the differences. These results could
also weaken the Motor Theory of Speech Perception.
Because Group III did categorically perceive, neuromotor
patterns are not necessary; and this nullifies the con-
nection of production and perception. However, if the
Motor Theory of Speech Perception is used as a base, it
provides an explanation for the difference among the
groups. Cerebral palsy is the result of brain damage and
is a motor dysfunction. Even if the speech is not dis-
ordered, as in Group II, it is possible for there to
have been some rearranging of the neuromotor patterns
necessary for speech perception. This does not mean
that the individual would have a perceptual disability,
but it could mean a perceptual difference. Group II's
perceptual pattern was different from Group I.

By assuming that the brain damage could disorder
the stored neuromotor correlates for perception, it must
also be assumed that these correlates are innately
given, as suggested by previous research (Eimas, 1971).

The present study does support this contention, because

48

of the fact that Group III did categorically perceive.
These subjects have not had long experience with the
speech sounds which Liberman et al. (1957, 1961) stated
were necessary for categorical perception. Further,
they were able to place phoneme boundaries, while
experiencing minimal motor speech history. This lack
of experience could weaken the link between production/
perception.

However, Group III did perceive more differently
from Group I than did Group II. Consequently, there
could be some advantage to be gained in speech perception
when the listener is able to produce the speech sounds.
However, this is not necessary for accurate and rapid
perception. It could make a difference in degree of
accuracy or rapidity of perception, however.

Motor Theory does provide the basis for an
explanation of these data. These data also can tell
us something about Motor Theory. First, as stated
above, the data support the contention that the system
is innately tuned to speech + language production and
perception; secondly, there is possibly some advantage
to be gained by being able to produce the speech sounds;
and third, the data provide implications about the level
at which the encoding occurs. Because of the difference
between Group I and Group II, it seems likely that some

encoding occurs above the level of the conversion of

49

muscle contractions to vocal tract shape. This would be
because Group II, which has no difficulty at that level,
does have a different perceptual pattern from Group I.
The likely place for this additional encoding to occur
would be at the level of the neural correlates, the
presumed invariants necessary for the decoder to operate.
The data from electromyographic studies strongly support
the invariance of the motor commands (Liberman et al.,
1967). Liberman et al. (1967), however, state that only
further investigation can determine if any significant
restructuring or encoding occurs at the higher levels of
production. The differences between Groups I and II
indicate the possibility of some degree of restructuring
may occur at this higher level. In this case, the cor-
relates could be different, not enough to affect the
speech of the individual, but only enough to affect the
encoding that takes place at this level. The difference
in encoding between the Group I and Group II could be
accounted for on the basis of this contention. For
Group III there would be a difference of the encoding
at the level proposed by Liberman et al. (1967) and at
the higher level proposed here. This would explain the
greater difference between Group I and Group III than
between Group I and Group II.

The groups do not differ in even steps along the

temporal continuum. The difference between Group III

SO

and Group II is greater than the difference between
Group II and Group I. This would indicate that the
greatest amount of encoding occurred at the level of
conversion from muscle contraction to vocal tract shape.
This is because the greatest difference in perceptual
pattern occurred in the group that had difficulties at
that level; the disturbance of the encoding at that
level caused more difference; therefore, more encoding

must occur there.

Discrimination Task

Predicted Discrimination
Peaks

 

 

Liberman et al. (1967) stated that the points of
maximal and minimal differential sensitivity should be
displaced along the continuum to correspond with the
phoneme boundaries. It should then be possible to
predict these points given the phoneme boundaries.
Liberman et al. (1957) devised a test to predict the
entire discrimination function for phonemes. The dis-
crimination peaks were located at the phoneme boundaries.
In the present study, the ABX triad was not used, so the
predictive test of discrimination function was not
applied. However, since the previous discrimination
peaks have fallen at the phoneme boundaries, it can be
predicted that the discrimination peaks of the present

study should fall at the phoneme boundaries. The data

51

from the separate sets of stimuli grouped into two's and

three's, as presented in Figs. 4a to 4c, were the data

used to predict the discrimination peaks. The predicted

discrimination peaks are listed in Table 4.

TABLE 4. Predicted discrimination peaks from data
obtained by collapsing the judgments into

groups of two and three; these were also
separated by original word.

 

 

 

 

Rapid Rabid
Group
Sets of 2 Sets of 3 Sets of 2 Sets of 3
Group I none none 110-120 120-130
Group II 90-110 120-130 110-120 120-130
110-120
Group III 70-80 60-70 110-120 120-130

120-130

 

thained Discrimination
Functions

 

 

The actual discrimination functions should provide
information about the learning component proposed by
Motor Theory. From the Identification Task there
appeared to be an advantage to being able to actually
produce the phonemes and to benefit from the learning
that took place in the development of speech production.

In order to further delineate the assumption
of the Motor Theory, the discrimination functions were
plotted by comparing each stimulus in a set with the
stimulus next to it (Appendix C) in what is referred to

as a “one—step" comparison. These data appear in

52

Figure 5. The discrimination is below chance within the
categories, many fewer than 50% of the judgments were
"different." There are peaks at some of the phoneme
boundaries as predicted from the identification task.
The peak for Group II, taken as the highest point, is
between 120 msec and 130 msec. However, the peak for
Group III occurred much further along the continuum than
predicted, being at 90-100 msec point rather than at the
predicted 60-70 msec point. For each set of data, one
for each original word, there is no correlation between
the predicted and obtained peaks. There also are none
of the valleys that are important within the categories
to indicate the efficient sorting of within-the-category
stimuli. The discriminations were predominately of the
”same” judgment. The stimuli would, therefore, seem to
exhibit acquired similarity, because of the fact that
such diverse cues all produced the same response in so
many_listeners.

These data indicate that the labelling of speech
sounds and the discrimination of the speech sounds are
not related, as had been pr0posed (Liberman et al., 1957;
1961; 1967). Further, the learning component of speech
perception is not supported by these findings. These
data would support one of Lane's (1965) objections to
categorical perception tasks. That is, if the subject

is allowed to label a stimulus as anything he wishes, the

53

0mm 81m mo; 'RAPID"

enuolk C>——C>
Camp w E] ------ C]
Group III: A---A

A

° ' ’/“°\ /"°’Ko

e駧§_

M110” OF SILENT MN ”SEC.

C)

 

mean or'omuf‘manms (TI-B)

Fig. 5a. Discrimination task function displayed
by a one-step comparison using results from stimuli pro-
duced from original word "rapid."

onto-w. muuws wono. "new"

OnwpI:ID-—<D
a“, u: D ...... D
GNNWIWHA“"flA

    

 
      

[00‘

   

   

n-«A curd—o—pm- \
o—Dwu‘uz—d/ "wu cr

   

C)

  

mum omrreaeuf‘anmts (1'85)

0
x .t
O
\

DURKTKHIOfiflLEflTINWERHM.fllIflflfl3

Fig. Sb. Discrimination task function displayed
by a one-step comparison using results from stimuli pro-
duced from original word "rabid."

54

perception is not going to be categorical. In the
present study, the subject had to label either rapid
or rabid. Although the discrimination task was also
forced choice, the subject was making a judgment on
similarity or difference to only one word. Without
another reference (i.e., the two words of an ABX para-
digm) the subject did not easily sort the stimuli into
two categories. The listener had difficulty on the
basis of silent interval duration alone in discriminating
these stimuli. It would seem from this study that more
than one subphonemic feature must differ for the lis-
tener to easily sort out the phonemes.

The three groups performed comparably on this
task. Although the three groups identified in a cate-
gorical mode, they did not discriminate in a categorical
mode. Thus, the results of the present investigation
would tend to weaken the arguments for the learning com-
ponent of speech perception, the production/perception
link, and the idea of labelling and discrimination being
closely associated.

Ratings of the Speech of the
Cerebral Palsied

The mean ratings of the judges for overall

speaker intelligibility and number of articulation

errors, and the ratings of the examiner of functional

55

capacity and therapeutic appliances, can be found in

Table 5 .

TABLE 5. Ratings of the Overall Intelligibility of the
two Experimental Groups; ratings of their
functional capacity and of their need for
therapeutic appliances; and mean number of
errors on articulation test.

 

Speech

 

Group Functional Therapeutic Intelligi' ggéogg
bility
Group II
1 4 3 1 °
2 2 2 1 0
3 4 3 1 °
4 1 1 1 0
5 -2 ‘2 '1 o
M 2,5 M 2.2 M l 0
Group III
3 3 3 1 11
5 -1 —1 -4 54
M 3.2 M 2.6 M 2.8

 

The mean ratings for the Group II functional and
therapeutic ratings are somewhat higher than for Group III.
However, inspection of the individual ratings for each
group's subjects show low variability between the groups.
The difference between the two groups in the mean ratings
for overall speech intelligibility was significant. This
would support the conclusions that the differences
between groups are related to something that is related

to the difference in speech ability. However, it must

56

be kept in mind that the experimenter was the one to rate

the subjects on functional ability and the therapeutic

scale, whereas three judges rated the overall speech

intelligibility. This could be a biasing factor.

of the

Conclusions

Conclusions that have been drawn from the results

study can be summarized as follows:

There is not a learning component to speech per-

ception ability.

There is not necessarily a production/perception
link for accurate and rapid perception of speech

sounds.

Individuals with cerebral palsy do exhibit dif-
ferent perceptual patterns from normal indi-

viduals.

Cerebral palsied individuals with good speech
and those with dysarthric speech exhibit dif-

ferent perceptual patterns from each other.

It would seem that production abilities did make

a difference in the pattern exhibited.

Implications for Further Research

From the results of this study some directions

for further research can be recommended. A weak point

in this study was the preparation of the stimuli; the

57

manual cut and splice technique is not as accurate as
synthetic speech. Replicating this study using synthetic
speech would give a more precise picture of the perceptual
differences among groups.

In an attempt to define the benefit gained from
production ability, a study involving children and adults
with cerebral palsy would give information on this topic.
Also, the type of previous speech therapy the subjects
had received could play a part in their ability to per-

ceive categorically.

CHAPTER IV

SUMMARY

The phenomenon of categorical perception was
studied with a cerebral palsied population. The impli-
cation that the results could have for a Motor Theory of
Speech Perception and the explanations this theory could
provide for the results were also considered.

Fifteen adult subjects were evenly divided into
three groups: control, cerebral palsied/good speech, and
cerebral palsied/dysarthric speech. All subjects were
of normal intelligence and had hearing. The cerebral
palsied groups were comparable except for their speech
production abilities. Each subject responded to identifi-
cation and discrimination tasks.

The results indicated categorical identification
for all groups. Discrimination results were similar for
all of these groups, but they were not related to the
predictions associated with categorical perception
studies and Motor Theory. The results indicated that
production abilities were not necessary for accurate

perception but that the ability to produce the speech

58

59

sounds made a difference in the perceptual pattern of
these listeners. Those listeners who could not produce
the phonemes exhibited greater variability. There was
a difference between the two cerebral palsied groups
and the control group. This was indicative of a com-
ponent of perception that was altered by the cerebral
palsied involvement. It was hypothesized that this
indicated encoding takes place at a higher level than

that proposed byva Motor Theory of Speech Perception.

APPENDICES

 

APPENDIX A

TASK TRIAL COMPOSITION

Trial #

COGDVO‘U'I-bMNH

Task Trial Composition

Stimulus

rabid-110
rapid-40
rapid-60
rapid-100
rapid-50
rabid-4O
rapid-110
rabid-140
rapid-140
rabid-120
rapid-90
rabid-130
rapid-130
rabid-7O
rapid-80
rapid-150
rabid-90
rabid-80
rabid-SO
rapid-70
rabid-60
rabid-150
rapid-120
rabid-100
rapid-130
rapid-4O
rapid-140
rabid-90
rabid-140
rapid-50
rabid-SO
rabid-110
rapid-100
rapid-150
rabid-100
rapid-60

APPENDIX A

60

Identification Task-Trial Composition

Trial #

Stimulus

rabid-4O
rabid-50
rapid-110
rapid-90
rabid-130
rabid-150
rabid-7O
rabid-80
rapid-80
rapid-120
rabid-120
rapid-70
rapid-100
rabid-110
ra id-SO
ra id-SO
rapid-70
rapido130
rapid-40
rabid-140
rapid-120
rabid-90
rapid-90
rabid-130
rabid-40
rabid-70
rapid-140
rabid-150
rabid-100
rapid-110
rapid-150
rapid-80
rabid-120
ra id-60
ra id-80
rabid-60

Trial #

61

Task Trial Composition

Discrimination Task-Trial Composition

Word

rapid
rabid
rabid
rapid
rapid
rapid
rabid
rabid
rabid
rapid
rapid
rabid
rapid
rabid
rapid
rabid
rapid
rapid
rapid
rabid
rabid
rapid
rabid
rabid
rapid
rapid
rapid
rapid
rabid
rabid
rapid
rabid
rapid
rabid
rabid
rapid
rapid
rabid
rabid
rabid
rapid
rapid
rapid
rabid
rapid

Interval

 

20-130
90-90
150-110
40-70
40-60
90-110
80-80
140-50
60-70
150-80
100-40
100-50
40-90
140-60
80-70
110-70
110-110
110-40
150-140
150-40
120-90
120-60
130-40
120-60
90-100
100-50
140-110
40-40
150-150
70-130
120-50
120-70
40-80
90-70
140-120
140-50
130-80
120-100
130-90
80-40
90-140
150-50
100-70
110-110
130-140

Trial #

Word

 

rabid
rabid
rapid
rapid
rabid
rabid
rapid
rapid
rapid
rapid
rabid
rabid
rabid
rapid
rapid
rabid
rapid
rabid
rabid
rabid
rabid
rapid
rapid
rapid
rabid
rabid
rapid
rapid
rapid
rapid
rabid
rabid
rapid
rapid
rabid
rabid
rapid
rabid
rabid
rabid
rapid
rapid
rapid
rapid
rapid

Interval

 

150-130
110-40
60-70
40-50
90-150
70-140
120-100
130-150
110-150
110-70
100-40
40-90
100-140
150-100
90-90
140-140
70-50
50-110
40-120
150-50
60-40
150-60
110-100
100-140
130-130
70-70
40-150
80-120
130-50
50-50
60-50
50-130
150-90
130-100
140-130
50-120
70-50
110-100
120-120
130-120
100-100
130-60
60-60
150-120
140-70

Discrimination Task-Trial Composition (cont'd.)

Trial #

Word

rapid
rabid
rabid
rabid
rabid
rapid
rapid
rapid
rapid
rabid
rabid
rabid
rapid
rapid
rapid
rapid
rabid
rabid
rabid
rabid
rapid
rapid
rapid
rapid
rabid
rabid
rabid
rabid
rapid
rapid
rapid
rapid
rabid
rapid
rapid
rapid
rapid
rabid
rabid
rabid
rabid
rabid
rabid
rabid
rabid
rapid
rabid
rabid
rabid

Interval

70-130
120-110
150-80
60-100
100-130
140-120
140-140
80-100
90-70
90-110
80-90
140-110
120-130
120-120
140-80
130-90
130-80
60-150
80-50
60-80
110-120
90-50
80-50
40-140
50-40
70-40
150-120
100-150
40-120
60-110
80-60
70-70
110-130
80-80
100-60
60-90
110-50
60-60
50-70
140-40
150-150
50-90
100-120
100-90
120-120
70-120
70-80
60-110
90-60

Trial #

140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156

Word

 

rabid
rapid
rabid
rabid
rabid
rabid
rabid
rapid
rapid
rabid
rabid
rapid
rapid
rapid
rapid
rabid
rabid

Interval

140-90
80-110
60-130
70-100
100-80
40-40
50-50
130-40
60-140
80-110
140-150
90-120
130-130
80-90
60-50
70-150
140-80

APPENDIX B

IDENTIFICATION TASK DATA FOR INDIVIDUAL SUBJECTS

APPENDIX B

Stimuli Produced From "Rapid"
Silent Interval Duration msec

Judgment

Identification Task Data for Individual Subjects

Control

Group

1212033021

1221033003

0321033003

0312032103

0330033003

1221033012

0312033003

0321030303

0321031212

0303030303

0303030312

0303030312

PBPBPBPBPB

34S

2112033030
0321030321
2112302121
2103030321
0312211230

2103122121
0303121212
0303122112

0303032121

h
C0303030321

e

e

0.0303120321
S

d
0
3
1
2
0
3
1
Z
0
3

PBPBPBPBPB

Cerebral Palsied/Goo

345

2103032121
2121120330
1221211221
3021121230

3012303030

3012121230

h .
C2112122130

e
e
p2103120330

c S
2
1
O
3
0
3
1
2
3
0

2112211230

0312031221

PBPBPBPBPB

Cerebral Palsied/Dysarthri

1
2

345

63

64

Identification Task Data for Individual Subjects
Stimuli Produced From "Rabid"

Silent Interval Duration msec

Judgment

Grou

Control

2130302112

2121123003

2130212121

2121211221

1230122103

1203120321

3021030321

0303030303

2103030321

0303030312

0303030321

0303030312

PBPBPBPBPB

9345

2112123030

3021031230

3030302130

3030032112

2121032112

0321031203

1212032130

0312032103

0303032103

h
C0312121203

e
e
P0321032112
S
@0312211203
0
G
/
d
e
.1
BPBPBPBPBPB
a
P
l
a
r
b
e
r.
8
C1 2 3 4 5

2121212112
2130211203
2121301203
2121211203
1221121203

0321211212

1230210303
h

C1212211203
e

e
p2112210312

S

C1203122112

hr1
0
3
0
3
3
O
0
3
0
3

0303120303

PBPBPBPBPB

Cerebral Palsied/Dysart

34S

1
2

APPENDIX C

DISCRIMINATION TASK DATA FOR INDIVIDUAL SUBJECTS

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

H H H a
H H H» H H H H H H H H H m oo-om
H H H H H a
H H «H, H H H H), H H H m Hm-om
H H H H H a
)H H H, ,H H H, H, ,H H ,H m omH-ov
H H H H H H H a
«H H H H H H) [H H m OHH-DH
H H H H H H H H H Q
Hi, H H, H H H m pHHuDH
H H H H H H a
H H H H H H H ,H H m ONH-oe
H H H H H .H H H n
H H H H H H H m oHH-oH
H H H H H H n
H H H H H H, H H H m ooH-oH
H n
H H rH, H H H H H H ,H Ht H H H, m ,bm-bH
H H H n
H H H» H) H ,iH, H, H, bH, H HF, up m, ,pmhph
H H a
H H H H H H. H H H H, H H ,HI m OHJDH
H H H H a
H H H H H[ H H, , H H H H .m bo-oe
H H H H H a
H H H H gr H ,H H H H m om-ow
a
H H H H\ H H) H H \H H H H \bH H H m ovuow
m H m N H m I. m H H m H m N H :39?! 523
uHuspHmmxa\.m.u nuoomw voou\.m.o Houucou IHwoma :wq

Honazz uuonnsm and macho unnamvsm

U NHDmem<

gHmm msHseHum

mpuonnam HmzuH>HucH mom mama HmmH :oHpmaHEHuumHa

65

66

 

Hn4

1.0.4.4.4H

U
.H

.4H4 .4 Pa .4 .4 r4 .4 .4

pH

HWH H H

:3
u
H

raH. .4 H. L4 H. .4

afflrd .4 H4 .4 L4 .4 Hn4

5
\NH H H H H H

Hn4 H4 .4 +4 .4 L4

HH

HH

HH

.4 Ha .4rd .4 H4 .4 H1 .4 .4+4
HuH .4.4 .4 H. .4 .4 H4 .4
.4.4 H. .4 H. .4H4 .4 .4.4 Hqu .4 H4

Hl-I

H

m
numomw voou

VHHHH

MMH

HHHHHH

H.4.4.-4

H
HH
HH
H
H
NH
\.@.U
0

.H

Hn4
.HFH L. .4 .4 r‘ .4

rd .4
LnL4 L4 .4 H4 .4 Pa .4 r‘ .4 H. .4 H4 .4 H1 .4

H

rap. .4 r4

Hu4 F4 .4 H.

H
H.
H

v

HH

H

.4H

H

.4 H! .4 r4 .4 H4 .4 r4 .4 H4 .4 .4 H. .4 F1

#1 .4

H

n N .H
Honuaou

nesz uuonpam wad usage

.4.4.-4.4.4.-4.4.4.4

r4

mamnmnmamnbnmobnmamnmamnmnpnbe

ucosmvam

OHHwbw
.DOH-oc
om-oo

oH-oo

po-oo
,bmH-om
.bHH-om
an-om
.bNH-om
DHHMom
OOH-om

om-om
omhbm
H:memu:m%mm_

Adena :Hg
uHmm maHseHuw

 

H.v.u:ouv muumnnsm HanuH>chH How mama meH :oHpmcHEHHumHQ

67

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

H H H a
H H H H H H H H ,H H H H« m omhbm
H . a
,H. H H H H H H H H H H H H« .H m, pmhbm
H H H H H a
H, H H H H H H H«, H ,H m« omHuoH
H H H H H H H H H a
H H H H H «H .m oanoH
H , H H H a
H H, .H H, H« H H H H H .H m bmH-OH
H H H H H H H a
H H H« H H H HI. H Hw omHnoH
H H H H H H a
H H .H H H. H H H .«H m, oHH-OH
H H H H H H H H a
H H H H H H .H m, ooH.oH
H H H H H H o
H H H, H H ,H H H« H« m« conch
H H H a
H H H H, H H H H ,H H H ,H. m om-pn
H n
H, H «H ,H H H H H ,H. H H, H H «H m. DH-OH
H H H H H H H H H a
H «H H H H .H« HP. omH-om
H H H H H H a
H H H H H H H H« H .m« ovH-om
H H H H H H H H H H H H a
H , H H« ml omeom
H H H H H a
H H H, H, H H H H H H ,m DNH-o
V H m N H m H b N. H« m P m N H fifiqfifaoum.
uHunuuumHa\.m.u :uoomw vooo\.m.u Houucoo ,Huoma cH.
Honasz puonnsm van macho unoEmvan HHmm msHsaHum

H.w.ucouv muuonnsm HmsvH>chH How mama HmMH :oHuwcHEHhumHa

 

 

 

 

 

 

 

 

 

 

68

 

 

 

 

 

 

 

H H a
H H H H H H H H H H .H H H m oHH-ooH
H a
H H H H H H «H H H H« H H H «H ,m ooH-ooH
H H H H H H a
H, H H H H ,H H H, H m omH-om
H H H a
H H« H H H H H H H H H« H m« .oHH-om
H H H H H H H a
H H H H H H H .H m oMH-pm
H H H H H a
H H. H H H. H H, H H H, m omanm
H H H H n
H H H H H. H H H H H H m ,bHHwom
H H H H H a
H H H H H H H H H. .H .m ooH-om
H a
H, H«, H H H H H H H H H, «H H ,H m om-om
H H H H H H H a
H, H H« H H, H H H m oMH-pm
H H H H H H H H H a
H H H H H H m oHH-pm
H H H a
H H, H H H« H« H«, H H , H H H« m oanpm
H H H H H a
H H H H« H H H« H H H m bNH-om
H H . H H H H a
H, H H H H H H «H H« m oHHuom
H H H H H H H H H a
H H H H, «H .H ,m ooH-m%
m H m N H m H m N H m H F N H HERE? a8
uHHnuHmmxa\.m.o nuoomw vooo\.m.u Houucou momma :Hg
Honasz uuonnsm cam macho ucosmwsn HHam msHssHpm

H.u.u:ouv muuonnam Hmst>chH How mama xmmH :oHumcHEHHUmHQ

 

 

 

 

 

 

 

 

 

 

69

 

 

 

 

 

 

 

 

H H a

H H«‘ H H H H! H H .H H H. H m oHH-pHH
H a

H H H H. H H H H H H H H«. H H, m omH-oMH
H a

H H H H H H H H H H H H H .H, m omH-oNH
H H a

H, H H H H H H H H H H« H H m DHH-omH
H H H H H H H a

H H H« H H H H H m oHH-pNH
n

H H H H H H H H H H H H H H H , m oNH-oHH
H H o

H H H H H H, H H H H H H« m pmH-oHH
H H H . H H H a

H H H H H H H H m DHH-oHH
. a

H, H H H. H H H H H H H H H H m, oHH-pHH
H H H H a

H H, H H H H H H H H m «HNH-oHH
H H a

H« H H H ,H ,H H H H H H H m ,HHH-oHH
H H H a

H H H H H H ,«H H H« H H .m omH-ooH
H H H . H . H H a

H H H H H «H, H m oHH-ooH
H H a

H H H ,H H H .H H« H H «m OMH-o0H
H H H a

H H .H H H H .H .H H m o~H-oo

m H m H. H m H .F H« H Emma? 5on .

uHusu m :uoomw voou\.m.u Hoppcou awome :HH

Honsaz pumnnsw cam msouu “cosmvsw HHmm msHseHum

H.v.u:ouV muumhnsm HmsuH>HucH How mums HmHH :oHumcHeHHuHHo

70

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

H H H H H H H H a
H.. H H H H H H m HHH-oH
H H H H H H H H a
H. H H, H H, H H m. .HNHHHH
H H H H H a
H H H H H H, H H H H m bHH-oH
H H H H a
H H H .H H H H H H H H., m .boH-oH
H H H H a
H H H H H H. H H H H H m Hm-pH
H H a
H H H H H H .H H H. H. H. H.. H m. HH-HH
H a
H H H .H H H .H H H H. H H H .H m 0H-oH
H a
H H H H H H H, H H H. H H. H H m. HH-OH
H H a
H ,H. H. H H H H H H, H. H H H. m om-oH
H a
.H H H H H H. H H H H H. H H H m HHuH
HHHHHHH: aoHH.
. H H a
H. H H H H H H H H H H H H m pmH.bmH
H H . H H a
H H H H H H H H H, H. H m omHnoHH
H a
H .H H H ,H, .H H H H H. H H H H. m oHH-oHH
H H . a
H H H H H H H H H H H H H m omH-om
m H m H H m H m H H H. H m N H FEES: 52H
UHHHHHHHHQ\.H.U Huoomw Hooo\.m.u Houpaou mems nHH

H.v.u:ouv muuonnnm HmsvH>chH How mama xmmH :oHHmnHEHHumHQ

Hmnesz Huonnsm Ham 9:0H0

Hamsmush

 

HHmm mszsHHm

71

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

a

H H H H H H H H. H H H .H H H H m HHuom
H a

H H H H H H. H H H. H H H. .H H m Hb-oo
H H H H H H H H H H H n

H H H. H m. omH-bm
H H H H H H H H H H a

H H H. H H .m .bHH-om
H H H H H H H H a

H. H H, H .H H, H. m oHHhom
H H H H H H H a

H H, H, H H .H H H m oNHJHm
H H H H a

H H H H. H .H H H H H ,H. m oHHnHm
H H n

H H H .H H H H H H H H H H m OOH-om
H H H a

H H H, H H .H, H H H H. H ,H .m omnbm
H a

H H H H H H H. H H H H H H H m obhbM.
H H H a

H .H H H H H, H H H H H H .m, oH-om
H a

H ,H H .H .H H, .H H. .H H H H H H m om-om
H H H a

H H H H H H H H .H H H ,H m om-om
H H H H H H H H a

H H H H H H .H ,m oMHnoH
H H H H H H H a

H H H H H H H H m oeH-o

m H, m N .H m H n H H .m H m N H. HHHHHMH: Scum.

UHHHHHmmHa\.H.u :uommw vooo\.m.u HOHpcou momma :HH

Honesz Huomnsm Ham monu ucosmvan HHmm msHaaHum

H.v.u:ouv muuonnsm Hm=HH>chH How «Ham HmmH :oHumcHEHHumHa

 

 

 

 

 

 

 

 

 

 

72

 

 

 

 

 

 

 

 

H H H H H a
H H H .H H H H oHH-oH
H H H a
H H H H H, H H H H m o~H-oH
H H H H H H a
H H H H, .H H H ,HHH-HH
H H a
H H H H H H H H H H H m oHH-oH
H H H H H a
H H. H. H H H H H H .bm-HH
H a
H. H H H H ,H H H H .H H H .H m HthH
H
H H H .H H. H H H H. H H H H. H H oH-oH
H H H H H H H H H H a
H H H. H. H m .HHH-oH
H H H H H H H a
H.. H H H H H H H H HHH-oH
H H H H H H a
H. H H H. H H H H H.. .m OHH-oH
H H H H H n
H H H H H H. H H H H H ONH-HH
H H H H H a
H H H H H H. H H H H H HHH-OH
H H H H a
H .H. H H .H H H H H H. H m pHanH
H H H H H H H H H a
H H H H, H. H m om-pH
H H a
H H H H H .H H H H H H. m HH-H
H H H H H H N H H H H H H H:HHHHH= 50»
u.HnuHm nuoumw uooo\.m.o Houucou momma :H.
Honesz Hummnam cc“ Q=OHU acoEmwah HHam msHssHum

H.H.H:ouv muuoszm HH3HH>HH=H HOH Hume HHHH coHHHaHeHHumHa

 

HHHHHH

HHHH

HHHH

LOP!

bhhhbb
omHuom

....
HH
F" H
HH

H H
HH

DHH-om

Ha rdri LH

.boHuob

HH H H'
HH H H H
H H H H

pmnbm

HH H H
H.”

r* r4 r4 r4
HHHHHHH

omHbe

H
HH H H H

H HH H H

H HL—I H H H H
H

DHH-ow

HHHHHHHH
H

HH

HH
r4 L4 H1 :4 rd .4 +4 .H

H

DMHuow

H H HH
HH
H H H HF“ H

H
HHHHr—Q

F0

DmHuDW

HF!

DHHupm

HHH'HHHHHHHH

H

H
HHHH

HH

OOHubm

HH
f...
H

omuom

r!

'UMHHHHHH
HHHH

om-bm

HHHLHHH

HH
HHHHH

omHJDH
H

anomnmnmnbnpomnmnmnpanMQann

H H oeHn

H H H FHHHHH: 995
oou\.m.u HOHpcou momma :HH
Honasz puonnam cam msouo pcosmcsw HHmm msHserm

~1- HH H H r.

MH H H H H H H

m HH H

H- Hr—a H

H H HH

hH H H H H H

:1- HH H H H H

N H HH L—C H H H H

 

0

H

H

H

H
F

H H

\.H.u ,mw

uHHsuHmmHn gum

H.v.ucouv muuonnzm HmsvH>chH How «use HmMH coHumcHEHHumHn

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

H H a

H Hi» H H H H H H »H H H» H H m onH-pNH
a

H H H H, H H H» H H H H H» H H» Ht m» »bNH-oNH
H H H H H n

H H H H» H, H» H H H H m oManHH
H n

H» H H H H H H H H H H H H HI m oeH-oHH
H a

H H H H» H H H H H H H H» H »H m »anhoHH
H H H H H n

H H H H H H, H H H» H »»m DNH-DHH
H H H H a

H H H H H H H H H H fl m »bHH-pHH
H H H H H H H a

.H H H H H H H H m omH-OOH
H H H H H H a

4.H H H H H H H H H m ovH-OOH
7‘ H H H H H a

H H H, H» H H, H H H H »m OMHHPPH
H H H H H n

H H H H H H H+» H» H H m oNH-90H
H H H H n

H H H H H H» »H H H H H m oHH-ooH
H a

H H H H H, H H H H »H H H H H m» boHuooH
H H a

H H H H» H H H H »H H H »H »»H RP omH-om
H H H H H H H a

H H H H H H H H m oeH-o

m H m H H m H n H H »m H »n N H HzmemH:,aon.

UHunuHmmNa\.m.o gumomw wooo\.m.u Houunou momma :HH

H.w.ucouv muuonnsm HmsvH>HucH How mama xmmh coHumcHeHHumHm

Hmpesz pumnpsm dam macho

pamswvah

HHmm msHseHum

 

J

 

 

 

 

 

 

 

 

 

 

H H H a

H Hy H H «\H H H H H H, H H m pmeoMH
H H a

"MH H H H H H H H H H H H H! m OmH-bHH
H H n

H H H H H H H H» H H H H H mx panoeH
H a

H H» H \H H H H H H H H H HI H m omH-omH
H H H a

H H H H H H H H H H H H m OHH-bmH
H a

H H H H» H H H H H \H H H H H ml omH-omH
H H H a

H \H H H H H H H H H H» H m \bMH-oNH
H H H H H n

H H H H H! H H H, H» H m4 owH-oN

m u m H H m a m N H m H m P H Htfirpfi: 58m .

annuummHn\.m.u nuoomw v000\.m.o Houucoo momma cHg

penssz uuonnsw wad macho uaovaah “Hum msHssHpm

H.w.p:ouv mpuonnsm HmsvH>chH pom mama meH :oHumcHEHHUmHQ

LIST OF REFERENCES

LIST OF REFERENCES

Ali, L., Gallagher, T., Goldstein, J., and Daniloff, R.,
Perception of Coarticulated Nasality. J. Acoust.
SOC. Amer., 49' 538-540, (1971).

 

Aungst, L. F., and Prick, J. V., Auditory Discrimination
Ability and Consistency of Articulation of /r/.
J. Speech Hearing Res., 29, 76-85, (1964).

Beare, A., Color Name as a Function of Wavelength.
Amer. J. of P§1., 76, 248-256, (1963).

Broadbent, D., and Gregory, M., Accuracy of Recognition
for Speech Presented to the Right and Left Ears.
Quar. J. Exp. Psx., 16, 359-360, (1964).

Bryden, M., Ear Preference in Auditory Perception. g,
EXE. P820, 65' 103-105, (1963).

Cutter, J., A Parallel Between Degree of Encodedness and
the Ear Advantage: Evidence from an Ear-Monitor-
ing Task. J. Acoust. Soc. Amer., 53, 368, (1973)
(Abstract).

Day, R. and Vigorito, J., A Parallel Between Degree of
Encodedness and the Ear Advantage: Evidence from
a Temporal-Order Judgment Task. J. Acoust. Soc.
Amer., 53, 368, (1973) (Abstract).

 

Delattre, P., Liberman, A., and Cooper, P., Acoustic Loci
and Transitional Cues for Consonants. J. Acoust.
SOC. Amara] 27' 769-773, (1955).

 

Denhoff, E. and Robinault, I., Cerebral Palsy and Related
Disorders: A Develo mental Approach to Dys-
function,Hc§raw-HIIE BooE Co., Tnc., New York,

 

 

Dollard, J. and Miller, N., Personality and Psycho-
therapy, McGraw-Hill 500E Co., Inc., New York,

76

77

Eimas, P., The Relation Between Identification and Dis-
crimination Along Speech and Non-Speech Continua.
Lang. & 82o, 6' 206-217, (1963).

Eimas, P., Siqueland, Jusczyk, and Vigorito, J., Infants
Can Categorically Perceive Speech Sounds.
Science, 171, 303-306, (1971).

Kimura, D., Some Effects of Temporal Lobe Damage on Audi-
tory Perception. Cana. J. Psy., 15, 156-165,
(1961).

Kimura, D., Functional Assymmetry of the Brain in Dichotic
Listening. Cortex, III, 163-177, (1967).

Lane, H. The Motor Theory of Speech Perception: A Criti-
cal Review. Psy. Rev., 72 (4), 275-309, (1965).

Liberman, A., Harris, R., Hoffman, H., and Griffith, B.,
The Discrimination of Speech Sounds Within and
Across Phoneme Boundaries. J. Exp. Psy., 54,
358-368, (1957).

Liberman, A., Cooper, F., Shankweiler, P., and Studdert-
Kennedy, M., Perception of the Speech Code.
Psy. Rev., 74, 431-461, (1967).

Liberman, A., Some Results of Research on Speech Per-
ception. J. Acoust. Soc. Amer., 29, 117-123,
(1957).

Liberman, A., Harris, K., Eimas, P., Lisker, L., and
Bastian, J., An Effect of Learning on Speech
Perception: The Discrimination of Durations of
Silence with and without Phonemic Significance.
Lang. and Sp., 4, 175-195, (1961).

Liberman, A., The Grammars of Speech and Language.
C02. P820, 1 (4), 301-323, (1970).

Locke, J., Discriminative Learning in Children's Acqui-
sition of Phonology. Speech Hearing Res., 11,
428-434, (1968).

MacNeilage, P., Rootes, T., and Chase, R., Speech Pro-
duction and Perception in a Patient with Severe
Impairment of Somesthetic Perception and Motor
Control. J. Speech Hearing Res., 10, 449-467,
(1967).

78

Mattingly, I., Liberman, A., Syrdal, A., and Halwes, T.,
Discrimination in Speech and Nonspeech Modes.
C09. P810, 2' 131-157, (1971).

Molfere, D., Cerebral Asymmetry in Infants, Children and
Adults: Auditory Evoked Response to Speech and
Music Stimuli. J. Acoust. Soc. Amer., 53, 363,
(1973) (Abstract).

Monnin, L. and Huntington, D., Relationship of Articula-
tory Defects to Speech-Sound Identification.
J. Speech Hearing Res., 17, 352-366, (1974).

Schatz, C., The Role of Context in the Perception of
St0ps. Lang., 30, 47-56, (1952).

Shankweiler, P. and Harris, K., An Experimental Approach
to the Problem of Articulation in Aphasia.

Aphasia: Selected Readin 3, Mary Taylor-Sarno,
Appleton Century Erofts, N.Y., (1972).
Stevens, K. and House, A., Studies of Formant Transitions

Using Vocal Tract Analog. J. Acoust. Soc. Amer.,
28, 578-585, (1956).

 

Studdert-Kennedy, M., Liberman, A., Harris, R., and
Cooper, F., Motor Theory of Speech Perception:
A Reply to Lane's Critical Review. Pay. Rev.,

 

Studdert-Kennedy, M., Liberman, A., and Stevens, R.,
Reaction Time During the Discrimination of Syn-
thetic Stop Consonants. J. Acoust. Soc. Amer.,
36, (1963).

Van Riper, C., S eech Correction: Princi les and
Methods. refitice-Hall, N.3., (1563).

weiner, P., Auditory Discrimination and Articulation.
J. Speech Hearingypis,, 32, 19-28, (1967).

Westlake, H. and Rutherford, D., S eech Thera for the
Cerebral Palsied. NationaE Easter Seaf Society,
Chicago, (1961).

 

MITIHIIIWINHHSIWIWIIUIINIIWIWIIWIUIW IWIITtWIWES
3 1293 030615888