- Email: [email protected]
The effects of phonetic context on speech-sound discrimination
JOURNAL OF COMUUNICATION
DISORDERS
153
13 (1980), 153-151
THE EFFECTS OF PHONETIC CONTEXT ON SPEECH-SOUND DISCRIMINATION RONALD GOLDMAN Centerfor
Developmental
and Learning Disorders,
University of Alabama in Birmingham
LUCY LONG ARMOUR Philadelphia,
Pennsylvania
ARTHUR H. SCHWARTZ Department of Biocommunication,
University ofAlabama
in Birmingham
This study was designed to examine the influence of contextual elements on speech-sound discrimination performance. The test employed was comprised of ten stimulus words located within three sentence structures providing initial, medial, and final placement of stimuli within sentences. All sentence items were presented in two forms, naturally articulated and mechanically assembled, to 25 normal preschool children. Disruption of the suprasegmental feature of the assembled test was achieved by splicing isolated words into sentences previously recorded with silence in the stimulus word position. Statistical analysis indicated that error scores on the assembled sentence items were significantly greater. Disruption of suprasegmental features appeared to significantly influence the discrimination of target sounds within stimulus words when the words were in the medial and final positions of the sentences.
Introduction In assessing speech-sound discrimination, there are dimensions other than the acoustic properties of phonemes which may affect performance (Hirsh, 1954; Miller and Isard, 1963; Schwartz and Goldman, 1974; Goldman and Harris, 1977). Additional features such as stress, intonation, juncture, duration, and coarticulatory variables are dimensions contributing to accurate processing of speech and language. This study was designed to determine if contextual variations in speech influence speech-sound discrimination performance.
Address correspondence to: Ronald Goldman, Ph.D., Center for Disorders, University of Alabama in Birmingham, University Station Alabama 35294. This work was supported in part by Project 910, U.S. Maternal and ment of Health, Education and Welfare as awarded to the Center for Disorders, University of Alabama in Birmingham. 0 Elsevier North Holland,
Inc., 1980
Developmental and Learning P. 0. Box 313, Birmingham, Child Health Service, DepartDevelopmental and Learning
002
l-9924/80/02 153-05$0 1.75
154
R. GOLDMAN,
L. L. ARMOUR,
and A. H. SCHWARTZ
Procedures Ten monosyllabic nouns were employed as stimulus words appearing in three different sentence structures. The words used as stimulus items were:,/hce, ring, bee, nuil, hear.,fun, .sock, whale, ruke, and coat. Each stimulus word appeared in the initial, medial, and final position of the sentences. The sentences used were: Sentence
1: [Stimulus word] is on the wall.
Sentence 2: She showed the [stimulus word] to the boy. Sentence 3: She showed the boy the [stimulus word]. Six subtests, each consisting of ten sentences containing the ten stimulus words. were administered. In three of the subtests sentences were assembled by splicing identical copies of stimulus words into the designated position. These stimulus words had been recorded independently of the sentences by a trained speaker. Splicing of stimulus words into position was done to most naturally approximate the temporal pattern of speech. The other three subtests, consisting of the same sentences, were recorded, but naturally articulated. To increase the task difficulty and approximate a realistic communication task, a background of cafeteria noise was included. The competing background noise was presented at a level approximately 12 dB less intense than the stimulus iterns. A total of 25 preschool children, ranging in age from 3 yr, 4 mo to 4 yr, 6 mo, served as subjects. All children participating in this study were required to pass a hearing screening test, a vocabulary comprehension measure, and an assessment of articulatory proficiency. The subjects selected were then administered a picture-word association training procedure to confirm that they were able to recognize the pictures of all stimulus items incorporated in the test. Any items missed were trained prior to test administration. Upon completion of the training procedure, all six subtests were administered in a random order. Children were shown a page containing pictures of the stimulus words (i.e., ring) and two similar sounding foils (i.e., wing and king). The child was instructed to identify the stimulus words in the sentence by pointing to the appropriate picture. The sets of stimulus/foil words were as follows: (1) face-vase, base; (2) ring-wing. king; (3) bee-pea, tea; (4) bear-pear, tear: (5) nail-rail, mail; (6) fanvan, pan; (7) sack-shack, tack; (8) whale-rail, sail: (9) rake-lake, shake; and (IO) coat--boat, goat. Results Mean error scores for the 25 subjects were computed for each subtest. Error scores for stimulus words in the initial, medial, or final positions in both assembled and natural sentence formats are shown in Fig. I.
PHONETIC
Natural
O-0
Assembled
t----i
Position
Fig. 1.
CONTEXT
AND SPEECH-SOUND
I
I
I
Initial
Medial
FInal
of Stimulus
F,ord
within
the
DISCRIMINATION
155
Sentence
Mean number of errors for natural and assembled presentations medial, and final positions within sentences.
of stimulus words in initial,
A r-ratio for related measures (Winer, 1962) was calculated to compare performance scores between natural and assembled sentences for each position in which a stimulus word appeared. The results presented in Table 1 indicated no significant difference in mean error scores between the sentences in the assembled and natural formats when the stimulus words appeared in the initial position. Comparison of natural and assembled subtests when stimulus words were presented in the middle or final positions yielded statistically significant t-ratios of 4.21 for the medial position and 6.24 for the final position. This suggests that significantly more errors occurred in the assembled format in which coarticulatory cues are minimized. Comparisons among the three subtests indicated that for the natural format, initial placement of the stimulus word resulted in a mean error score significantly greater than error scores for the medial or final placements (t-ratios = 3.56 and 4.53). There were no significant differences, however, between the performance on the assembled format subtests where the stimulus words were located in the initial, medial, and final positions of the sentences.
156
Comparison
R. GOLDMAN,
L. L. ARMOUR,
and A. H. SCHWARTZ
TABLE 1 of Mean Number of Errors Between Natural and Assembled Presentation Words in Initial, Medial, and Final Positions Within Sentences
Word Position
of Stimulus
Mean Number of Errors
Standard Deviation
Initial position subtests: Natural Assembled
2.20 2.28
1.41 1.81
0.30
Medial position subtests: Natural Assembled
1.36 2.52
1.03 1.58
4.21’
Final position subtests: Natural Assembled
1.04 3.00
1.24 1.68
6.24*
*Significant
r-ratio
at the 0.01 level of confidence.
Discussion The results of this investigation suggest the contribution of factors other than the acoustic properties of phonemes to speech-sound discrimination performance. Specifically, error scores for sentences in the assembled format exceeded error scores for sentences presented with natural contextual features. This was a consistent trend exhibited by the subjects. Furthermore, the subjects averaged 3.23 more errors for all sentences in the assembled format regardless of stimulus word position. These data provide potentially useful information as to the influence of contextual variables on speech-sound discrimination test performance. If these features do influence processing in a consistent manner, they must be taken into consideration in assessment. In the present study, the stimulus words were inserted in three different sentence structures to determine if initial, medial, or final placement affected discrimination. Comparison of performance scores for the three natural sentence contexts indicated that position does, indeed, influence response accuracy. Initial placement of stimulus words resulted in error scores significantly higher (_? = 2.20) than either the medial (2 = 1.36) or final (2 = 1.04) placements. While it is possible that this difference was a function of the sentence structures serving as an alerting device for the words in the medial or final position, it is unlikely, since the opposite effect was noted in the assembled format sentences. That is, for assembled sentences, error scores increased when the stimulus words occurred in the medial (x = 2.52) and final positions (x = 3.00). Perhaps contextual information is simply not as important to initial words
PHONETIC
CONTEXT
AND SPEECH-SOUND
DISCRIMINATION
157
as to later positioned elements. This interpretation is based on the lack of significant differences in performance for the assembled and natural subtests when stimulus words were located in the initial position. Subjects were frequently observed to respond as soon as the stimulus word appeared, possibly making the discrimination independent of subsequent parts of the sentence. The findings of this study seem to suggest that certain suprasegmental features and word positions may influence a child’s performance for sound discrimination on items where the stimulus word is embedded in the normal language context. Language is considerably more than appropriate syntax, semantics, and phonology (Glucksberg and Danks, 1975; Bates, 1976, Rees, 1978); to some extent, the same consensus has been reached relative to phonology. Increasing attention is being given to the relevance and complexity of dimensions other than the acoustics of the target phonemes in speech-sound discrimination testing. Features such as stress, intonation, duration, and coarticulation are being recognized as fundamental requisites to efficient processing of speech and language. The belief that speech is neither produced nor perceived as a linear sequence of discrete sound units should be given consideration when clinical techniques for evaluation or remediation of speech-sound discrimination are developed. References Bates, E. (1976). Pragmatics and psycholinguistics in child language. In Morehead, D.M., and Morehead, A.E. (eds.), Normal and Deviant Child Language. Baltimore: University Park Press. Glucksberg, S., and Danks, J. (1975). Experimental Psychoknguisfics. New York: John Wiley and Sons. Goldman, R., and Harris, G. Influence of vocabulary and picture word association training on speech-sound discrimination test results. Paper presented at the Annual Convention of the .4merican Speech and Hearing Association, Chicago, Illinois. Hirsh, I.J., Reynolds, E.G., and Josephs, M. (1954). Intelligibility of different speech materials. J Acoustical Sm. Amer. 26:530-538. Miller, G.A., and Isard, S. (1963). Some perceptual consequences of linguistic rules. J. Verb. Learn. and Verb Behav. 3:217-228. Rees, N. (1978). Pragmatics of language: Applications to normal and disordered language development. In Schiefelbusch, R. (ed.), Bases ofLanguages Intervention. Baltimore: University Park Press. Schwartz, A.H., and Goldman, R. (1974). Variables influencing performance on speech-sound discrimination tests. J. Speech HearingRes. 17:25-32. Winer, B .J. (1962). Statisrical Principles in Experimental Design. New York: McGraw-Hill.
DISORDERS
153
13 (1980), 153-151
THE EFFECTS OF PHONETIC CONTEXT ON SPEECH-SOUND DISCRIMINATION RONALD GOLDMAN Centerfor
Developmental
and Learning Disorders,
University of Alabama in Birmingham
LUCY LONG ARMOUR Philadelphia,
Pennsylvania
ARTHUR H. SCHWARTZ Department of Biocommunication,
University ofAlabama
in Birmingham
This study was designed to examine the influence of contextual elements on speech-sound discrimination performance. The test employed was comprised of ten stimulus words located within three sentence structures providing initial, medial, and final placement of stimuli within sentences. All sentence items were presented in two forms, naturally articulated and mechanically assembled, to 25 normal preschool children. Disruption of the suprasegmental feature of the assembled test was achieved by splicing isolated words into sentences previously recorded with silence in the stimulus word position. Statistical analysis indicated that error scores on the assembled sentence items were significantly greater. Disruption of suprasegmental features appeared to significantly influence the discrimination of target sounds within stimulus words when the words were in the medial and final positions of the sentences.
Introduction In assessing speech-sound discrimination, there are dimensions other than the acoustic properties of phonemes which may affect performance (Hirsh, 1954; Miller and Isard, 1963; Schwartz and Goldman, 1974; Goldman and Harris, 1977). Additional features such as stress, intonation, juncture, duration, and coarticulatory variables are dimensions contributing to accurate processing of speech and language. This study was designed to determine if contextual variations in speech influence speech-sound discrimination performance.
Address correspondence to: Ronald Goldman, Ph.D., Center for Disorders, University of Alabama in Birmingham, University Station Alabama 35294. This work was supported in part by Project 910, U.S. Maternal and ment of Health, Education and Welfare as awarded to the Center for Disorders, University of Alabama in Birmingham. 0 Elsevier North Holland,
Inc., 1980
Developmental and Learning P. 0. Box 313, Birmingham, Child Health Service, DepartDevelopmental and Learning
002
l-9924/80/02 153-05$0 1.75
154
R. GOLDMAN,
L. L. ARMOUR,
and A. H. SCHWARTZ
Procedures Ten monosyllabic nouns were employed as stimulus words appearing in three different sentence structures. The words used as stimulus items were:,/hce, ring, bee, nuil, hear.,fun, .sock, whale, ruke, and coat. Each stimulus word appeared in the initial, medial, and final position of the sentences. The sentences used were: Sentence
1: [Stimulus word] is on the wall.
Sentence 2: She showed the [stimulus word] to the boy. Sentence 3: She showed the boy the [stimulus word]. Six subtests, each consisting of ten sentences containing the ten stimulus words. were administered. In three of the subtests sentences were assembled by splicing identical copies of stimulus words into the designated position. These stimulus words had been recorded independently of the sentences by a trained speaker. Splicing of stimulus words into position was done to most naturally approximate the temporal pattern of speech. The other three subtests, consisting of the same sentences, were recorded, but naturally articulated. To increase the task difficulty and approximate a realistic communication task, a background of cafeteria noise was included. The competing background noise was presented at a level approximately 12 dB less intense than the stimulus iterns. A total of 25 preschool children, ranging in age from 3 yr, 4 mo to 4 yr, 6 mo, served as subjects. All children participating in this study were required to pass a hearing screening test, a vocabulary comprehension measure, and an assessment of articulatory proficiency. The subjects selected were then administered a picture-word association training procedure to confirm that they were able to recognize the pictures of all stimulus items incorporated in the test. Any items missed were trained prior to test administration. Upon completion of the training procedure, all six subtests were administered in a random order. Children were shown a page containing pictures of the stimulus words (i.e., ring) and two similar sounding foils (i.e., wing and king). The child was instructed to identify the stimulus words in the sentence by pointing to the appropriate picture. The sets of stimulus/foil words were as follows: (1) face-vase, base; (2) ring-wing. king; (3) bee-pea, tea; (4) bear-pear, tear: (5) nail-rail, mail; (6) fanvan, pan; (7) sack-shack, tack; (8) whale-rail, sail: (9) rake-lake, shake; and (IO) coat--boat, goat. Results Mean error scores for the 25 subjects were computed for each subtest. Error scores for stimulus words in the initial, medial, or final positions in both assembled and natural sentence formats are shown in Fig. I.
PHONETIC
Natural
O-0
Assembled
t----i
Position
Fig. 1.
CONTEXT
AND SPEECH-SOUND
I
I
I
Initial
Medial
FInal
of Stimulus
F,ord
within
the
DISCRIMINATION
155
Sentence
Mean number of errors for natural and assembled presentations medial, and final positions within sentences.
of stimulus words in initial,
A r-ratio for related measures (Winer, 1962) was calculated to compare performance scores between natural and assembled sentences for each position in which a stimulus word appeared. The results presented in Table 1 indicated no significant difference in mean error scores between the sentences in the assembled and natural formats when the stimulus words appeared in the initial position. Comparison of natural and assembled subtests when stimulus words were presented in the middle or final positions yielded statistically significant t-ratios of 4.21 for the medial position and 6.24 for the final position. This suggests that significantly more errors occurred in the assembled format in which coarticulatory cues are minimized. Comparisons among the three subtests indicated that for the natural format, initial placement of the stimulus word resulted in a mean error score significantly greater than error scores for the medial or final placements (t-ratios = 3.56 and 4.53). There were no significant differences, however, between the performance on the assembled format subtests where the stimulus words were located in the initial, medial, and final positions of the sentences.
156
Comparison
R. GOLDMAN,
L. L. ARMOUR,
and A. H. SCHWARTZ
TABLE 1 of Mean Number of Errors Between Natural and Assembled Presentation Words in Initial, Medial, and Final Positions Within Sentences
Word Position
of Stimulus
Mean Number of Errors
Standard Deviation
Initial position subtests: Natural Assembled
2.20 2.28
1.41 1.81
0.30
Medial position subtests: Natural Assembled
1.36 2.52
1.03 1.58
4.21’
Final position subtests: Natural Assembled
1.04 3.00
1.24 1.68
6.24*
*Significant
r-ratio
at the 0.01 level of confidence.
Discussion The results of this investigation suggest the contribution of factors other than the acoustic properties of phonemes to speech-sound discrimination performance. Specifically, error scores for sentences in the assembled format exceeded error scores for sentences presented with natural contextual features. This was a consistent trend exhibited by the subjects. Furthermore, the subjects averaged 3.23 more errors for all sentences in the assembled format regardless of stimulus word position. These data provide potentially useful information as to the influence of contextual variables on speech-sound discrimination test performance. If these features do influence processing in a consistent manner, they must be taken into consideration in assessment. In the present study, the stimulus words were inserted in three different sentence structures to determine if initial, medial, or final placement affected discrimination. Comparison of performance scores for the three natural sentence contexts indicated that position does, indeed, influence response accuracy. Initial placement of stimulus words resulted in error scores significantly higher (_? = 2.20) than either the medial (2 = 1.36) or final (2 = 1.04) placements. While it is possible that this difference was a function of the sentence structures serving as an alerting device for the words in the medial or final position, it is unlikely, since the opposite effect was noted in the assembled format sentences. That is, for assembled sentences, error scores increased when the stimulus words occurred in the medial (x = 2.52) and final positions (x = 3.00). Perhaps contextual information is simply not as important to initial words
PHONETIC
CONTEXT
AND SPEECH-SOUND
DISCRIMINATION
157
as to later positioned elements. This interpretation is based on the lack of significant differences in performance for the assembled and natural subtests when stimulus words were located in the initial position. Subjects were frequently observed to respond as soon as the stimulus word appeared, possibly making the discrimination independent of subsequent parts of the sentence. The findings of this study seem to suggest that certain suprasegmental features and word positions may influence a child’s performance for sound discrimination on items where the stimulus word is embedded in the normal language context. Language is considerably more than appropriate syntax, semantics, and phonology (Glucksberg and Danks, 1975; Bates, 1976, Rees, 1978); to some extent, the same consensus has been reached relative to phonology. Increasing attention is being given to the relevance and complexity of dimensions other than the acoustics of the target phonemes in speech-sound discrimination testing. Features such as stress, intonation, duration, and coarticulation are being recognized as fundamental requisites to efficient processing of speech and language. The belief that speech is neither produced nor perceived as a linear sequence of discrete sound units should be given consideration when clinical techniques for evaluation or remediation of speech-sound discrimination are developed. References Bates, E. (1976). Pragmatics and psycholinguistics in child language. In Morehead, D.M., and Morehead, A.E. (eds.), Normal and Deviant Child Language. Baltimore: University Park Press. Glucksberg, S., and Danks, J. (1975). Experimental Psychoknguisfics. New York: John Wiley and Sons. Goldman, R., and Harris, G. Influence of vocabulary and picture word association training on speech-sound discrimination test results. Paper presented at the Annual Convention of the .4merican Speech and Hearing Association, Chicago, Illinois. Hirsh, I.J., Reynolds, E.G., and Josephs, M. (1954). Intelligibility of different speech materials. J Acoustical Sm. Amer. 26:530-538. Miller, G.A., and Isard, S. (1963). Some perceptual consequences of linguistic rules. J. Verb. Learn. and Verb Behav. 3:217-228. Rees, N. (1978). Pragmatics of language: Applications to normal and disordered language development. In Schiefelbusch, R. (ed.), Bases ofLanguages Intervention. Baltimore: University Park Press. Schwartz, A.H., and Goldman, R. (1974). Variables influencing performance on speech-sound discrimination tests. J. Speech HearingRes. 17:25-32. Winer, B .J. (1962). Statisrical Principles in Experimental Design. New York: McGraw-Hill.