The Psychological Record, 1964, 14, 327-334.

PHONEMIC ANALYSIS-SYNTHESIS SKILLS OF NORMAL

AND SPEECH DEFECTIVE CHILDREN

LOUIS LEREA
Northern Illinois University

A group of 18 normal speaking Ss and 18 children with

developmental articulatory speech defects were matched and
exposed to three series of tape recorded utterances which were
associated with 47 minimal triplet words, In condition I, each
child heard one word from each of the 47 triplets; progressively
fragmented words were given in conditions II and III. The
child repeated the words or completed the mutilated members of
the triplets and pointed to one of three pictures depicting each of
the. triplets. The groups did not differ in condition I. The speech
defective children, however, were inferior to the normal Ss in
perceptual integration of fragmented words. These findings were
discussed in terms of phonemic 'cloze' and suggestions for further
research were offered.

Accurate speech perception-scanning, discriminating and inte-

grating complex 'patterns of sounds-is a fundamental requisite for
acquiring acceptable oral language. The auditory signals which the
child must learn to decode and eventually encode are seldom complete.
Speech bears the distortions imposed by the speaker, the listener, the
interaction of phonetic, linguistic and physical environments, etc. The
normal child develops the ability to fill in the appropriate sounds and
answer meaningfully to a visitor's question, "-ello -s mo-er -ome?" Several
researchers and clinicians have suggested that one important factor
underlying developmental articulatory speech defects may be a de-
ficiency in identifying and integrating the phoneme elements within
words.
Van Riper and Irwin (1958) speculated that ". . . there may be
a strong possibility that the articulation case has not been able to
master correct pronunciation because he has not learned the basic
phonetic skill of combining and analyzing sound sequences." Speech
sound discrimination and perception tests were administered to 100
college students by Summers (1958). The discrimination test required
that his 55 listen to a nonsense syllable and select this key syllable
from among three syllable options. In the 35 item perception test, the
examiner presented a key sound and his Ss were instructed to locate
the sound within a set of four typewritten words. Summers reported a
correlation of .264 between speech sound discrimination and percep-
tion and intimated that articulation problems may be a function of a
latent deficiency in analyzing and synthesizing phonemes.
328 LEREA

Van Riper (1963), Bryngelson and Mikalson (1959), Nemoy and

Davis (1937), and West, Ansberry and Carr (1957) have recommended
procedures which are intended to give children with articulatory de-
fects a greater awareness of the distinctive sounds within words. In
essence, the speech therapist says a nonsense syllable or a word which
includes a sound that the child is attempting to correct; the child
signals each time he hears the sound and then indicates whether the
sound occurred at the "beginning, middle or the end" of the utterance.
Variations on this analysis-synthesis theme are many. Despite the wide-
spread endorsement by speech clinicians of this therapeutic regimen,
only speech sound discrimination has been subjected to extensive experi-
mental investigation.
The intent of the present study was to assess the phonemic ability
of children with developmental articulation problems and children with
normal speech. In this paper, phonemic ability refers to the skills
required to differentiate among the distinctive sound elements of an
utterance and to integrate these sequential linguistic cues so that a
contiguous whole configuration is perceived.

METHOD
Subjects
Thirty-six elementary school children, 18 speech defective and an
equal number of normal Ss, participated in this investigation. The two
groups were matched in sex, chronological age, grade level and in-
telligence. The 15 boys and 3 girls comprising the experimental group
misarticulated three or more consonant sounds, two of which were
( s ), (r) or (I). These children were receiving speech therapy in the
public schools. The mean age of the experimental group and the normal
Ss, serving as their controls, were 93.8 and 91.9 months respectively.
Each member of the control group was free of speech problems and
was a classmate of a speech defective S; the children ranged in grade
level from one through three. The intelligence quotients, as determined
by the WIse (verbal), the California Test of Mental Maturity or the
Ammons Full Range Picture Vocabulary Test, were within normal limits
for both groups.
An S was disqualified if: ( 1) his medical history suggested an
etiologically significant organic abnormality or (2) he was unable to
pass a pure tone sweep check test, in either ear, at frequencies from
250 to 8000 cps present at 15 dB. The children with developmental
articulation defects 'possessed speech problems which were of un-
determined origin.
Phonemic Analysis-Synthes1.3 Series
A preliminary test consisting of 64 sets of three 'minimal' triplet
words was prepared. Each word within a particular set was distinguished,
ostensibly, from its two accompanying words by a change in a phoneme,
such as: mail, nail, jail; trick, track, truck; race, rake, rain. Other criteria
 PHONEMIC ANALYSIS-SYNTHESIS SKILLS 329

governing the selection of a minimal triplet were: (1) the words should
not exceed the vocabulary level of most children who were eligible to
enter first grade and (2) the three words comprising a triplet could
be pictorially represented.

A Shure microphone, Model 556S and an Ampex 601 tape recorder

with its companion amplifier, Model 620, were used to record and
administer the trial test at a tape speed of 7~~ inches per second. An
adult male, judged to have distinct speech articulation, received eight
hours of training before recording three variations of a 192-word list.
In the initial series, the speaker was directed to pronounce one key
word from each of the 64 triplet sets. After completing this series, he
returned to the first triplet and recorded again. In this second recording,
the speaker omitted one phoneme of a word from each minimal set:
mai-; -ick; -ace. He mouthed the missing elements of a word and vocal-
ized the remaining sounds of the words, The speaker, in the final series,
produced only the essential phoneme which altered the meaning of the
word, i.e. n--; -a-; --no The correct responses were randomized within
each triplet with the exception of the third series which was keyed,
whenever possible, to vowels, diphthongs and those consonants classified
as continuants. Consonants which could not be articulated in isolation
were terminated with a schwa vowel. The speaker was discouraged from
prolonging a sound beyond the duration in which the phoneme could
be identified clearly and monitored his loudness by means of a V-U
meter. An eight second interval was permitted between each vocalization.
To alert S, the speaker said the word "listen" before giving the whole
or fragmented utterance.

A copy of the master tape was presented to a class of 24 university

students who were enrolled in a course which included a unit in de-
scriptive phonetics. Each student was requested to listen to the three
recordings and to write the words and fragmented stimuli in the IPA.
A minimal triplet was discarded if two of the students failed to tran-
scribe accurately anyone of the three utterances .associated with a
given triplet. The final phonemic analysis-synthesis series consisted of
47 sets (141 words). The distinctive phoneme sets occurred as follows:
24 in the initial, 8 in the medial and 15 in the final positions. This
categorization, particularly between sets representing beginning and
middle positions, was arbitrary since the significant phonemes of these
minimal triplet words alw-ays initiated the utterance in the second re-
cording. The 47 minimal triplets constituting the phonemic analysis-
synthesis series are given in Table 1.

These triplet sets were sketched in color on 4.5 X 15 inch cards;

each rectangular card depicted one minimal triplet. Five public school
speech therapists reviewed the pictures and listened to the three taped
series. They were unanimous in their agreement regarding the clarity
of the sketches which provided the context for the recorded utterances.
330 LEREA

TABLE 1
PHONETIC ANALYSIS-SYNTHESIS SE".{IES

Position Item Minimal Triplets

1 hat cat rat
2 boat coat goat
3 man fan pan
4 car jar star
5 nail mail jail
6 wing ring king
7 lake rake bake
8 pail sail tail
9 light night write
10 dream cream stream
11 vase face lace
Initial 12 gum thumb plum
13 ship lip whip
14 gun run sun
15 door floor roar
16 wire fire tire
17 keys cheese peas
18 wall ball call
19 nose hose toes
20 bell well shell
21 chair hair bear
22 chop top mop
23 bone phone cone
24 school stool spool
25 soup soap sip
26 boot beet boat
~7 sleep slip slap
Medial 28 trick truck track
29 book back bike
30 pearl pill pool
31 cape cap cup
32 kite cat coat
33 pin pig pick
34 badge bath bag
35 coal coat comb
36 can cab calf
37 rose rope road
38 match mat map
39 cage cane cake
Final 40 cup come cut
41 beet bead beef
42 race rake rain
43 neck nest net
44 bug bus buzz
45 bell bed belt
46 path patch pack
47 wash watch wasp

The instrumentation and methodological precision in the preparation

of the master tape are admittedly questionable. Nevertheless, the task
bears some resemblance to the clinical situation in which the speech
 PHONEMIC ANALYSIS-SYNTHESIS SKILLS 331

therapist says one or more words, syllables or sounds and directs the
child to analyze and synthesize the phonemic elements of the utterance.
Procedure
This study on phonemic ability may be considered as a dual in-
vestigation of speech sound discrimination among minimal triplets
(condition I) and phonemic integration (conditions II and III). Condition
I always was followed by condition II and then III. These three experi-
mental trials were administered individually in a quiet room of the
public school which the child attended. The signal was set at a con-
versational level. E presented all 47 rectangular cards in each of the
three conditions and the child was required to point to the picture
which was associated with each recorded utterance. Before designating
his choice, the child was instructed to repeat the word or complete the
mutilated word. TIlls step was interjected in an effort to secure responses
which perhaps would reflect the child's subjective comparison of his
own auditory feedback with his internalized "intra-personal" speech
model (Van Riper & Irwin, 1958). Antecedent to conditions II and III,
the child was cautioned that, "The pictures you point to may, or may
not, be the same as the pictures you pointed to before."
To summarize, in condition I, the child heard the word hat; he
repeated the word and pointed to one of three pictures (hat-cat-rat).
Following the entire 'presentation of this word discrimination series,
the rectangular card associated with item 1 was shown again for Con-
dition II and the child heard the syllable ea-. He attempted to say the
whole word and designated the picture which he associated with that
recorded signal. The third series was essentially the same as the previous
47 item trials with the exception that, in condition III, the child was
given only the phoneme r--. Two practice items were included to intro-
duce the three conditions. A standardized form was prepared for the
purpose of tabulating the child's performances. If the child pointed to
the correct picture but had said a word which was unrelated to the
key utterance, the response was considered to be a guess and scored as
an error. The three test series usually were completed in approximately
45 minutes.

RESULTS

The mean and standard deviations of the scores obtained in the

three experimental conditions for the normal and speech defective
children are given in Table 2. Since the first condition constituted a
speech sound discrimination task of minimal triplet words and therefore
similar to other studies which have been concerned with this skill, the
statistics computed for condition I were treated separately. The t-test
for related measures was employed to determine whether the difference
between the two groups in differentiating among whole words could be
attributed to random sampling from a population of differences with
332 LEREA

TABLE 2
MEANS AND STANDARD DEVIATIONS FOR THE 1WO GROUPS IN
SPEECH SOUND DISCRIMINATION (CONDITION I)
AND INTEGRATION (CO'NDITIONS II AND III)

Groups Conditions
I II III
Normals Mean 46.33 40.11 35.22
S.D. 6.58 4.44 8.54
Speech Mean 43.39 35.61 28.39
Defectives S.D. 6.15 4.13 9.53

a mean of zero. The magnitude of the twas 1.60 (df==17) and failed
to achieve statistical significance.
A 2 X 2 repeated-measures mixed design, classified by Lindquist
( 1953) as Type I, was used in evaluating the data for the two groups
in conditions II and III. As shown in Table 3, the F values for groups
and conditions exceeded the .05 and .01 per cent levels of confidence,
respectively. These results suggest that the normal Ss participating in
this study were superior to the speech defective children. Furthermore,
the combined mean of both groups in oondition III was markedly
smaller than the mean in condition II. The interaction of groups and
conditions was not significant.
The Pearson Product Moment correlations between speech dis-
crimination and phonemic integration are given in Table 4. All of the

TABLE 3
SUMMARY OF THE ANALYSIS OF VARIANCE OF THE PERFORMANCES
OBTAINED FROM THE 18 NORMAL AND 18 SPEECH DEFECTIVE
CHILDREN IN CONDITIONS II AND III

Source df ms F
Between Subjects 35 87.70
B (Groups) 1 480.50 6.30t
Error (b) 34 76.32
Within Subjects 36
A ( Conditions) 1 388.32 11.01 0
AB 1 50.00 1.42
Error (w) 34 35.28
Total 71

t Significant beyond the .05 level of confidence

o Significant beyond the .01 level of confidence
 PHONEMIC ANALYSIS-SYNTHESIS SKILLS 333

TABLE 4
CORRELATION COEFFICIENTS BETWEEN SPEECH SOUND
DISCRIMINATION AND THE TWO INTEGRATION SERIES
FOR THE NORMAL AND SPEECH DEFECTVE GROUPS (N==18)

Groups Conditions
II III

Normals .357 .515

Speech Defectives .527 .442

The tables value of r at the .05 level of significance is .468 (df==16)

correlations were low and, despite the statistical significance of two,

speech sound discrimination and phonemic integration probably should
not be construed as measuring identical skills.

DISCUSSION
Although the performance of the normal group in differentiating
among minimal triplet words was somewhat better than the speech
defective Ss, the mean difference did not warrant rejecting the null
hypothesis. Mindful of the procedural limitations and the range of
the perceptual demands of this investigation, the performance of this
small sample of speech defective children when compared with the
normal Ss reflected a deficiency in perceptual integration of phonemes.
These findings may be viewed from another vantage point. The
Ss in this study were required to analyze complete words and synthesize
fragments of a linguistic unit. The groups did not differ in their ability
to differentiate whole 'percepts; the performances of the speech defective
children, however, deteriorated markedly in comparison to the normal
children when the tasks required greater effort to achieve integration
or closure. The Gestalt concept of closure refers to the individual's
capacity to organize disjointed stimuli into unified and contiguous wholes.
The ease with which an individual synthesizes an incomplete pattern
is a function of the number of elements remaining in the perceptual
field. It was to be expected, therefore, that both groups would experience
increasing difficulty in the three experimental conditions since the
'perceptual cues were progressively deleted in each condition.
This principle of closure was adopted by Taylor (1956; 1957). He
developed the 'cloze procedure' as a method of assessing the readability
and listenability of a communication. Taylor omitted words from a
passage and discovered that the difficulty level of the material could
be determined by counting the number of words or 'eloze units' which
the reader could insert based on the context of the passage. Conditions
II and III of the present study may be interpreted as a phonemic
334 LEREA

clozure task. Ss with articulatory defects did not experience appreciable

difficulty in responding appropriately to complete words; the focus of
their perceptual disturbance became evident, however, in situations
which demanded a synthesis of fragmented utterances.
Whether this problem of phonemic clozure is a relevant deficiency
among children with developmental articulatory speech defects, or
merely a fortuitous correlate, must await further investigation. It also
should be mentioned that perceptual integration is associated with the
degree of familiarity of the perceptual field. It would seem fruitful,
therefore, to consider in future research the relationship between
phonemic clozure of words and oral language facility of children with
developmental articulatory speech problems.

REFERENCES

BRYNGELSON, B. & MIKALSON, E. Speech correction through listening. Chicago:

Scott Foresman, 1959.
LINDQUIST, E. F., Design and analysis of experiments in psychology and educa-
tion. Boston: Houghton-Mifflin, 1953.
NEMOY, H. & DAVIS, H. The cor-rection of defective consonant sounds. Boston: Ex-
pression, 1937.
SUMMERS, R., Perceptive vs productive skills in analyzing speech sounds from
words. J. speech hear. Disord., 1953, 18, 140-148.
TAYLOR, W. L., Recent developments in use of "Cloze Procedure." Journalism
Quart. 1956, 33, No. 1.
TAYLOR, W. L., "Cloze" readability scores as indices of individual differences
in comprehension and aptitude. J .appl. Psychol., 1957, 41, No.1.
VAN RIPER, C., & IRWIN, J. V., Voice and articulation. Englewood Cliffs:
Prentice-Hall, 1958. P. 29.
VAN RIPER, C., Speech correction, principles and methods. (4th ed.). Englewood
Cliffs: Prentice-Hall, 1963.
WEST, R., ANSBERRY, M., & CARR, A., The rehabilitation of speech, (3rd ed.).
New York: Harper, 1957.