- Email: [email protected]
The relationship between auditory temporal analysis and receptive language development: Evidence from studies of developmental language disorder
Nruropsychologuz, Vol. 23. No. 4, pp. 527 534, 1985. Pnnted tn Great Britain.
i
002X-3932/85 $3M)+0.00 1985 Pergamon Press Ltd.
THE RELATIONSHIP BETWEEN AUDITORY TEMPORAL ANALYSIS AND RECEPTIVE LANGUAGE DEVELOPMENT: EVIDENCE FROM STUDIES OF DEVELOPMENTAL LANGUAGE DISORDER PAULA Department
of Psychiatry,
University
RACHEL Department
of Neurology,
Johns Hopkins
TALLAL
of California,
San Diego, La Jolla, CA 92093, U.S.A.
E. STARK
School of Medicine 21205, U.S.A.
and J. F. Kennedy
Institute,
Baltimore,
MD
and DAVID Department
of Biostatistics
and Pediatrics,
MELLITS
Johns Hopkins
School of Medicine,
(Accepted 9 November
Baltimore,
MD 21205, U.S.A.
1984)
Abstract-The relationship between sensory, perceptual and motor abilities and receptive language abilities was studied in developmental dysphasic children. The tests administered included experimental auditory, visual and cross-modal perceptual tests. In addition, a battery of neurodevelopmental “soft sign” sensory, perceptual and motor tests were also given. Demographic and case history data were collected. Receptive language was derived based upon a battery of standardized language tests. Multivariate analyses were employed to examine the relationship between receptive language and sensory, perceptual and motor abilities. Results demonstrated that auditory perceptual variables, specifically those requiring rapid temporal analysis, were most highly correlated with the degree of receptive language deficit of the dysphasic children.
INTRODUCTION the diagnosis of specific developmental language delay (developmental dysphasia) is made by a process of elimination. That is, having determined that a child is a year or more delayed in receptive and/or expressive language development, further assessment is based on ruling out deficits of hearing, nonverbal intelligence, motor development, oral structural development, middle ear functioning, language environment and emotional development. A child who demonstrates a significant delay in language development in the absence of all of these factors is diagnosed as having a specific developmental language delay or disorder known as developmental dysphasia [ 11. In an attempt to understand better the nature of developmental language disabilities, research studies have focused on either a psycholinguistic approach [ 11, 12, 201 or on investigating the basic perceptual, motor and cognitive mechanisms that are presumed to subserve language [19]. In general, these two research approaches have progressed in isolation from each other. That is, studies evaluating the linguistic abilities of developmentally dysphasic children have not simultaneously investigated the more primary perceptual, motor and cognitive abilities of the same children. Similarly, although studies of the development of perceptual, motor and cognitive abilities of dysphasic children have
CLINICALLY,
521
528
PAULA TALLAL, RACHEL E. STARK and DAVID MELLITS
hypothesized a theoretical relationship between these abilities and the language disorders of these children, direct experimental assessment of this theorized relationship has not been attempted. The purpose of the present study was to determine the extent to which basic perceptual and motor deficits correlate with the degree of receptive language deficit in developmentally dysphasic children. More specifically, in our research we have previously demonstrated that dysphasic children are primarily impaired in their ability to discriminate and sequence both rapidly presented nonverbal signals and also speech sounds that incorporate rapidly changing acoustic spectra, such as stop consonant-vowel syllables [21, 22, 241. These subjects were unimpaired in their ability to respond correctly to the same nonverbal signals when they were presented more slowly and were also unimpaired in responding to speech sounds that were either steady state in nature or had been computer synthesized in such a way as to slow down the rate of change of formant transitions within stop consonant--vowel syllables [23, 241. These studies demonstrated that children with developmental language deficits are not equally impaired in processing all speech sounds. Rather, they have specific difficulty processing those speech sounds that are characterized by rapidly changing spectra which are critical for their discrimination. Further studies [16,25] demonstrated that these subjects’ errors in speech production were similar to their errors in speech perception. In these studies, it was found that those speech sounds that rely on brief temporal cues for their discrimination were not only most often misperceived by dysphasic children, but were also most frequently misproduced or omitted in their speech output. It was these results that led us to hypothesize that deficits in temporal acoustic analysis may preclude or delay the development of normal speech perception and production, and subsequently result in delayed or disordered language development, specifically receptive language. The present study was designed to investigate the hypothesis that deficits in temporal analysis are specifically related to the degree of receptive language disorder in dysphasic children.
METHODS Subjects
Twenty-six developmental dysphasic children between the ages of 5 and 9 yr participated as subjects in this study. Potential subjects had been previously diagnosed clinically as language impaired and were re-examined for inclusion in this study using standardized clinical tests. All subjects had a performance IQ as measured by the Wechsler Intelligence Scale for children [28] of at least 85 or above. The mean performance IQ for the dysphasics was 100 (range 87 123). The mean socioeconomic status was 3.4 based on the Hollingshead Scale [6]. All subjects had normal absolute thresholds for hearing, normal oral peripheral structure and oral motor abilities and none had frank neurological “hard signs”. Each subject was given a large battery of receptive and expressive language tests assessing phonology, semantic and syntactic linguistic functions, an articulation test, and for subjects 7 yr and older, a reading test assessing vocabulary, comprehension and phonetic decoding skills. Details of this test battery have been reported elsewhere [17] and receptive language assessment is described below. In order for a child to be included as a developmentally dysphasic subject, his or her receptive language age had to be at least six months behind performance mental age (MAP) and chronological age (CA). Expressive language had to be at least 12 months behind MAP and CA and composite language age (expressive and receptive language ages combined) had to be at least 1 yr behind MAP md CA. Articulation age and reading age for all subjects was within six months of expressive language abilities. The characteristics of the dysphasic children are shown in Table 1.
PROCEDURES The main procedures used in this study to assess central processing abilities are essentially the same as those that have been used in our previous studies with language impaired and normally developing subjects. These procedures, designated the Repetition Method, have been described in detail in previous publications [ 18, 211. The Repetition
AUDITORY
TEMPORAL
Table
ANALYSIS
1. Description
Number of subjects Mean CA (months) Verbal IQ Performance IQ Full Scale IQ SES (Hollingshead Scale) Language “Age” (months) Receptive Expressive Speech articulation Raw score (Templin Darley)
AND
RECEPTIVE
LANGUAGE
of demographic
DEVELOPMENT
529
variables 26 78 82 100 90 3.4
(S.D. (S.D. (S.D. (S.D. (SD.
k + f + +
15.6) 11.4) 8.8) 9.0)
1.8)
65 53
(SD. + 10.85) (SD. +- 12.57)
38
(SD. kO.62)
Test is hierarchical in nature and is comprised of a series of subtests (detection, association, discrimination, sequencing, rate processing and serial memory) that build one upon the other. In brief, subjects are trained, using an operantly conditioned two-alternative forced-choice procedure, to press the bottom panel of a response box in response to stimulus 1 and the top panel in response to stimulus 2 (discrimination subtest). Next. binary series of these two stimuli are presented in increasingly longer series at various rates of presentations. Subjects respond to what they perceive by pressing the panels on the response box corresponding to the order of presentation (sequencing and serial memory subtests). In the present study auditory, visual and cross-modal (auditory and visual) nonverbal stimuli of various durations were investigated. These stimuli have been described in detail elsewhere [26]. In addition, a series of computer synthesized minimal pair speech sounds were also used as stimuli m the Repetition Test. These included the pairs /ba/ vs Ida/, /da/ vs /tat, /ei vs /ae/, /dab/ vs /daeb/, /sa,l vs lstai and isa/ vs /Sal. These stimuli also have been described in detail in a previous publication [24]. In addition to the Repetition Test a battery of sensory, perceptual and motor neurodevelopmental “soft sign” tests were given. This battery included tests of motor control and coordination, balance and station, tactile perception and laterality relationships. These procedures, adapted from those used routinely by neurologists to assess neurological “soft signs” have been described in detail in a previous publication [7]. Demographic, social, family and medical history data were collected via a questionnaire. Receptiw
language assessment
In order that relationships between these experimental variables and receptive language abilities might be estimated, it was necessary to assess the level of receptive language functioning of each subject. Receptive language ages were derived from standardized tests used routinely by speech pathologists to assess receptive language. These included the Token Test [3]. the Test of Auditory Comprehension of Language [?I, the receptive portion of the Northwest Syntax Screening Test [lo] and the Auditory Reception and Auditory Association Subtests from the Illinois Test of Psycholinguistic Abilities [S]. These tests are designed to measure receptive semantic and syntactic linguistic abilities. Mulfitariate
analyses
Multivariate analyses were carried out to discover the extent to which these experimental variables might be correlated with or “predict” the receptive language abilities of the dysphasic children. These multivariate analysis techniques took the form of a stepwise forward multiple regression procedure. In this procedure, experimental variables were ranked according to their contribution to the prediction of receptive language functioning. The first variable selected in the course of this procedure was the one with the largest simple correlation coefficient. All other variables were then examined in tandem with the first variable. That additional variable which was found to reduce the variance of the “prediction” or estimation of level of receptive language to the greatest extent then entered the prediction equation together with the first variable, and so on. Three different approaches were applied in deciding whether or not variables should be included in the prediction equation. These included: (1) An F-test of the extent to which the error in the fitted regression line using N variables was reduced when N + 1 variables were entered; (2) examination of the reduction in the standard error; and (3) examination of the increase in the resulting multiple regression coefficient. It should be emphasized that the variables identified in this manner made a contribution to prediction over and above the contribution made by variables already included in the prediction equation. Thus, variables that are found to be highly significant in univariate analysis do not necessarily make a significant contribution in multivariate analysis. If they do not, it is because they have been pre-empted by other, more powerful predictors with which they are highly correlated.
530
PAULA TALLAL, RACHEL E. STARK and DAVID MELLITS
RESULTS The results of the multivariate analysis indicated that of the experimental perceptual, motor and demographic variables studied, auditory nonspeech and speech perception variables proved most highly correlated with level of receptive language in dysphasic children. The results obtained from the multivariate analysis are shown in Table 2. In this table, all variables that entered the prediction equation are shown, together with the multiple correlation coefficients for each of these variables. Four experimental variables entered the equation for predicting level of receptive language of dysphasic children. All four variables were obtained from Tallal’s Repetition Test. The first three variables were derived from tests incorporating synthetic stop consonant -vowel syllables with various formant transition durations as stimuli. The final variable entering the prediction equation incorporated nonverbal auditory tones. The four variables taken in combination yielded a multiple correlation coefficient (multiple R) of 0.85. That is, 7294 of the variance related to the level of receptive language of the dysphasic children in this study could be accounted for by their ability to discriminate and sequence nonverbal acoustic tones and stop consonanttvowel syllables with various duration formant transitions. This is not to suggest that other variables (i.e. sequencing rapidly presented nonverbal auditory, visual and cross-modal stimuli or sequencing ,ba/ vs /da/ with 40 msec duration formant transitions or /da/ vs /ta/) that were highly significant discriminators between groups in univariate analysis were not correlated with level of receptive language. Rather, it must be kept in mind that in stepwise multiple regression the most powerful variables enter the equation and thus preempt less powerful predictors with which they are highly correlated.
DISCUSSION Of all the perceptual, motor and demographic variables that were assessed in this study including auditory, visual, tactile and cross-modal sensory and perceptual functionings, the only variables entering the multiple regression equation predicting level of receptive language in dysphasic children were acoustic perceptual variables. Furthermore, of the variety of acoustic nonverbal and speech perceptual variables assessed in this study, only the ability to discriminate and sequence stop consonant -vowel stimuli with various duration formant transitions and the ability to discriminate brief duration nonverbal auditory tones entered the prediction equation for level of receptive language. It is of interest to note that these are precisely the same stimuli that we have shown in previous studies to most significantly differentiate the perceptual abilities of dysphasic and normal children, a finding which initially led us to hypothesize a direct relationship between deficits in rapid temporal analysis and disordered language development. The results of the present study support this hypothesis for the first time with empirical data by demonstrating a highly significant (P
synthetic
CV syllables /ba/ and /da/ with
80 msec transitions, 10 ms ISI, (No. of errors) Discriminating synthetic CV syllables /ba/ and /da/, 40 ms formant transitions (No. of trials to criterion) Discriminating synthetic CV syllables /ba/ and /da/, 80 ms formant transitions (No. of errors) Discriminating 300 Hz vs 100 Hz tone (No. of errors)
Sequencing
Variable name
-0.82
0.83
1.08
-3.31/0.001
- 3.79/0.001
4.97/0.00 1
-4.13/0.001
72.70 -2.52
regression Computed c-test P-value
multiple
Regression equation constants
Table 2. Final composite
-0.54
-0.59
0.69
- 0.62
10.93/0.001
10.02/0.001
6.19/0.02
17.08/0.001
Final F-test/ P-value
for the SLD children
Partial correlation coefficient
equation
6.10
7.10
8.13
8.80
Standard error
17.77
14.80
11.11
17.08
Step F overall
0.85
0.78
0.69
0.60
Multiple R
532
PAULATALLAL,RACHEL
E. STARKand DAVIDMELLITS
of receptive language in dysphasic children but rather, the pattern of the variables entering the prediction equation. It is of significance that although the perception of a variety of speech sound contrasts were assessed in this study, the only speech variables that entered the prediction equation for receptive language were those incorporating acoustic spectra that changed rapidly in succession. The ability to process speech sounds that incorporate acoustic spectra not characterized by brief temporal cues [24] (such as /E/ vs /ae/, /daeb/ vs /dab/, Isa/ vs /sta/ /sa/ vs /la/), and longer duration nonverbal tones (250 msec) did not enter the equation predicting level of receptive language. It is important to note that although discrimination and sequencing of stop consonant-vowel syllables incorporating rapidly changing formant transitions have been shown in our previous studies to be grossly impaired in dysphasic children, these same stimuli are amongst the easiest for normally developing children of the same age to discriminate and seque‘nce [24]. In fact, these stop consonant-vowel syllables have been shown to be readily discriminated accurately even by young infants [4,5,9, 131. They are also amongst the first to appear in the speech production of normally developing children (as any proud “da-da” can attest to). Thus, it cannot be argued that these stimuli are merely the most difficult for all children to discriminate and therefore, preempted other speech and nonspeech stimuli from entering the prediction equation. It would also be incorrect to suggest that these data support a theory of delayed development, with the dysphasics just representing the lower end of the normal distribution of speech perceptual development. The pattern of deficits in speech perception demonstrated by this group of children does not seem to be similar to that reported for younger normal children. Our interest in investigating the sensory, perceptual and motor abilities of dysphasic children has been to increase our understanding of the neurological mechanisms underlying language development and disorders. Our now large series of studies with dysphasic children have consistently demonstrated that these children are specifically impaired in their ability to both produce and perceive stimuli that change rapidly in succession, regardless of whether these stimuli are verbal or nonverbal. In a recent study we reported that performance on perception and production tasks (auditory, visual, tactile and cross-modal) alone correctly classified 98:i of subjects as normal or language impaired in discriminant function analysis [27]. Recently, OJEMANNand MATEER [14] have derived a model of the organization of language from electrical stimulation mapping studies. Central to this model is a common peri-Sylvian cortex for language functions in the dominant hemisphere, including sites common to both sequencing of nonverbal movement and identification of phonemes incorporating rapidly changing acoustic spectra. Ojemann suggests that “a single mechanism in this cortex, such as precise timing, may underlie both processes”. It is of considerable interest that the speech stimuli shown by Ojemann and Mateer to be disrupted by peri-Sylvian stimulation in the dominant hemisphere are the same stimuli found in our research to he most difficult to both perceive and produce by dysphasic children. Furthermore, it was these same stimuli that proved to be most highly correlated with the degree of receptive language deficit of developmentally dysphasic children in the present study. In speculating about the possible ontogeny of the peri-Sylvian cortex, Ojemann hypothesized that during the course of human development, the role of this common cortex for language production and understanding initially may be for decoding acoustic cues. Later, during language acquisition, this function may be utilized for identifying those acoustic cues significant to speech. Ojemann further suggests that the same cortical area later
AUDITORYTEMPORALANALYSISAND
RECEPTIVE
LANGUAGE
DEVELOPMENT
533
may develop the patterns of motor output to produce these significant speech sounds. Our data pertaining to the nonverbal perceptual and production abilities of language impaired children, as they relate to decoding and phoneme perception and production and subsequently to receptive language development, provide empirical data directly in support of Ojemann’s speculation. Similarly, Ojemann and Mateer’s data from electrical stimulation mapping studies provide strong anatomical support for our hypothesis concerning a precise timing mechanism underlying hemispheric specialization (the right-ear advantage on dichotic listening tasks) for speech perception in normal adults [ 151 as well as for a deficient timing mechanism which may underlie the receptive language deficits of developmentally dysphasic children. Acknowledgements--This
work was supported
by a contract
from NINCDS
(Project
NS-25353).
REFERENCES aphasia and brain damage. Cortex I, 4c-52, 1964. 1. BENTON, A. L. Developmental 2. CARROW, E. Test for Auditory Comprehension of Language. Urban Research Group, Austin, Texas, 1973. in aphasics. 3. DE RENZI, E. and VIGNOTI, L. A. The Token Test: A sensitive test to detect receptive disturbances Brain 85, 665-678, 1962. changes in speech discrimination in infants. J. 4. EILERS, R. E., WILSON, W. R. and MOORE, J. M. Developmental Speech Hear. Res. 20, 766-788, 1911. 5. EIMAS, P. D., SIQUELAND, E. R.. JUSZYCK, P. and VIGOROTO, J. Speech perception in infants. Science 171, 303-306. 1971. 1957. 6. HOLLINGSHEAD, A. B. Two Factor Index of Social Position. New Haven, Connecticut, and normal I. JOHNSTON, R., STARK, R., MELLITS, E. D. and TALLAL, P. Neurological status oflanguage-impaired children. Ann. Neurol. 10, 159-163. 8. KIRK, S. A., MCCARTHY, J. J. and KIRK, W. D. Illinois Test o~Psycholingui.rtic Abilities. University of Illinois Press, Urbana, Illinois, 1968. categories. In Hearing and 9. KUHL, P. K. Speech perception in early infancy: The acquisition of speech-sound Dacis: Essays Honoring HolloweN Davis, S. K. HIRSH, D. H. ELDREDGE, I. J. HIRSH and S. R. SILVERMAN (Editors). Washington University Press, St. Louis, 1976. University Press, Northwestern University, 10. LEE, L. L. Northwestern Syntax Screening Test. Northwestern Evanston, Illinois, 1969. of deviant child language. Word 29, 139-147, 1978. II. LEONARD, L. The phonology 12. MOREHEAD, D. and INGRAM, D. The development of base syntax in normal and linguistically deviant children. J. Speech Hear. Res. 16, 330-352, 1973. of speech and nonspeech stimuli in early infancy. J. exp. Child Psychol. 14, 13. MORSE, P. A. The discrimination 477492, 1912. 14. OJEMANN, G. and MATEER, C. Human language cortex: Localization of memory, syntax, and sequential motor-phoneme identification systems. Science 205, 1401-1403, 1979. 15. SCHWARTZ, J. and TALLAL, P. Rate of acoustic change may underlie hemispheric specialization for speech perception. Science 207, 138&1381, 1980. 16. STARK, R. and TALLAL, P. Analysis of stop-consonant production errors in developmentally dysphasic children. J. acoust. Sot. Am. 66, 1703-1712. 1979. 17. STARK, R. and TALLAL, P. Selection of children with specific language deficits. J. Speech Hear. Disorders 46, 114-122, 1981. 18. TALLAL, P. Perceptual requisites for language. In Non-Speech Languyge and Communication, R. SCHIEFELBUSCH,(Editor), pp. 449-467. University Park Press, Baltimore. 1980. 19. TALLAL, P. Neuropsychological foundations of specific developmental disorders of speech and language: Implications for theories of hemispheric specialization. Submitted to Psychiatry 1984. 20 TALLAL, P., CURTISS, S. and KAPLAN, R. Unpublished results, 1984. 21. TALLAL, P. and PIERCY, M. Defects of non-verbal auditory perception in children with developmental aphasia. Nature 241, 468-469. 1973. 22. TALLAL, P. and PIERCY, M. Developmental aphasia: Rate of auditory processing and selective impairment of consonant perception. Neuropsyrhologia 12, 83-93, 1974. 23. TALLAL, P. and PIERCY. M. Developmental aphasia: The perception of brief vowels and extended stop consonants. Neuropsychologia 13, 69-14, 1975.
534
PAULA TALLAL, RACHEL E. STARK and DAVID MELLITS
24. TALLAL, P. and STARK, R. Speech acoustic-cue discrimination abilities of normally developing and language-impaired children. J. UCOUS~.Sot. Am. 69, 568-574. 198 1. 25. TALLAL, P., STARK, R. and CURTISS, B. Relation between speech perception and speech production impairment in children with developmental dysphasia. Bruin Lang. 3, 305-317. 1976. 26. TALLAL, P., STARK, R., KALLMAN. C. and MELLITS, E. D. A re-examination of some nonverbal perceptual abilities of language-impaired and normal children as a function of age and sensory modality. J. Speech Hear. Rrs. 24, 351-357, 1981. 27. TALLAL, P.. STARK, R., and MELLITS, E. D. Identification of language impaired children on the basis of rapid perception and production skills. Brain Larz,q. (in press). 28. WECHSLER, D. Wechsler Intelligence Scale for Children--Revised (WISC-R). The Psychological Corporation, New York, 1974. Production errors in developmentally dysphasic children. /. ucous~. Sot. Am. 66, I703 17 12, 1979.
i
002X-3932/85 $3M)+0.00 1985 Pergamon Press Ltd.
THE RELATIONSHIP BETWEEN AUDITORY TEMPORAL ANALYSIS AND RECEPTIVE LANGUAGE DEVELOPMENT: EVIDENCE FROM STUDIES OF DEVELOPMENTAL LANGUAGE DISORDER PAULA Department
of Psychiatry,
University
RACHEL Department
of Neurology,
Johns Hopkins
TALLAL
of California,
San Diego, La Jolla, CA 92093, U.S.A.
E. STARK
School of Medicine 21205, U.S.A.
and J. F. Kennedy
Institute,
Baltimore,
MD
and DAVID Department
of Biostatistics
and Pediatrics,
MELLITS
Johns Hopkins
School of Medicine,
(Accepted 9 November
Baltimore,
MD 21205, U.S.A.
1984)
Abstract-The relationship between sensory, perceptual and motor abilities and receptive language abilities was studied in developmental dysphasic children. The tests administered included experimental auditory, visual and cross-modal perceptual tests. In addition, a battery of neurodevelopmental “soft sign” sensory, perceptual and motor tests were also given. Demographic and case history data were collected. Receptive language was derived based upon a battery of standardized language tests. Multivariate analyses were employed to examine the relationship between receptive language and sensory, perceptual and motor abilities. Results demonstrated that auditory perceptual variables, specifically those requiring rapid temporal analysis, were most highly correlated with the degree of receptive language deficit of the dysphasic children.
INTRODUCTION the diagnosis of specific developmental language delay (developmental dysphasia) is made by a process of elimination. That is, having determined that a child is a year or more delayed in receptive and/or expressive language development, further assessment is based on ruling out deficits of hearing, nonverbal intelligence, motor development, oral structural development, middle ear functioning, language environment and emotional development. A child who demonstrates a significant delay in language development in the absence of all of these factors is diagnosed as having a specific developmental language delay or disorder known as developmental dysphasia [ 11. In an attempt to understand better the nature of developmental language disabilities, research studies have focused on either a psycholinguistic approach [ 11, 12, 201 or on investigating the basic perceptual, motor and cognitive mechanisms that are presumed to subserve language [19]. In general, these two research approaches have progressed in isolation from each other. That is, studies evaluating the linguistic abilities of developmentally dysphasic children have not simultaneously investigated the more primary perceptual, motor and cognitive abilities of the same children. Similarly, although studies of the development of perceptual, motor and cognitive abilities of dysphasic children have
CLINICALLY,
521
528
PAULA TALLAL, RACHEL E. STARK and DAVID MELLITS
hypothesized a theoretical relationship between these abilities and the language disorders of these children, direct experimental assessment of this theorized relationship has not been attempted. The purpose of the present study was to determine the extent to which basic perceptual and motor deficits correlate with the degree of receptive language deficit in developmentally dysphasic children. More specifically, in our research we have previously demonstrated that dysphasic children are primarily impaired in their ability to discriminate and sequence both rapidly presented nonverbal signals and also speech sounds that incorporate rapidly changing acoustic spectra, such as stop consonant-vowel syllables [21, 22, 241. These subjects were unimpaired in their ability to respond correctly to the same nonverbal signals when they were presented more slowly and were also unimpaired in responding to speech sounds that were either steady state in nature or had been computer synthesized in such a way as to slow down the rate of change of formant transitions within stop consonant--vowel syllables [23, 241. These studies demonstrated that children with developmental language deficits are not equally impaired in processing all speech sounds. Rather, they have specific difficulty processing those speech sounds that are characterized by rapidly changing spectra which are critical for their discrimination. Further studies [16,25] demonstrated that these subjects’ errors in speech production were similar to their errors in speech perception. In these studies, it was found that those speech sounds that rely on brief temporal cues for their discrimination were not only most often misperceived by dysphasic children, but were also most frequently misproduced or omitted in their speech output. It was these results that led us to hypothesize that deficits in temporal acoustic analysis may preclude or delay the development of normal speech perception and production, and subsequently result in delayed or disordered language development, specifically receptive language. The present study was designed to investigate the hypothesis that deficits in temporal analysis are specifically related to the degree of receptive language disorder in dysphasic children.
METHODS Subjects
Twenty-six developmental dysphasic children between the ages of 5 and 9 yr participated as subjects in this study. Potential subjects had been previously diagnosed clinically as language impaired and were re-examined for inclusion in this study using standardized clinical tests. All subjects had a performance IQ as measured by the Wechsler Intelligence Scale for children [28] of at least 85 or above. The mean performance IQ for the dysphasics was 100 (range 87 123). The mean socioeconomic status was 3.4 based on the Hollingshead Scale [6]. All subjects had normal absolute thresholds for hearing, normal oral peripheral structure and oral motor abilities and none had frank neurological “hard signs”. Each subject was given a large battery of receptive and expressive language tests assessing phonology, semantic and syntactic linguistic functions, an articulation test, and for subjects 7 yr and older, a reading test assessing vocabulary, comprehension and phonetic decoding skills. Details of this test battery have been reported elsewhere [17] and receptive language assessment is described below. In order for a child to be included as a developmentally dysphasic subject, his or her receptive language age had to be at least six months behind performance mental age (MAP) and chronological age (CA). Expressive language had to be at least 12 months behind MAP and CA and composite language age (expressive and receptive language ages combined) had to be at least 1 yr behind MAP md CA. Articulation age and reading age for all subjects was within six months of expressive language abilities. The characteristics of the dysphasic children are shown in Table 1.
PROCEDURES The main procedures used in this study to assess central processing abilities are essentially the same as those that have been used in our previous studies with language impaired and normally developing subjects. These procedures, designated the Repetition Method, have been described in detail in previous publications [ 18, 211. The Repetition
AUDITORY
TEMPORAL
Table
ANALYSIS
1. Description
Number of subjects Mean CA (months) Verbal IQ Performance IQ Full Scale IQ SES (Hollingshead Scale) Language “Age” (months) Receptive Expressive Speech articulation Raw score (Templin Darley)
AND
RECEPTIVE
LANGUAGE
of demographic
DEVELOPMENT
529
variables 26 78 82 100 90 3.4
(S.D. (S.D. (S.D. (S.D. (SD.
k + f + +
15.6) 11.4) 8.8) 9.0)
1.8)
65 53
(SD. + 10.85) (SD. +- 12.57)
38
(SD. kO.62)
Test is hierarchical in nature and is comprised of a series of subtests (detection, association, discrimination, sequencing, rate processing and serial memory) that build one upon the other. In brief, subjects are trained, using an operantly conditioned two-alternative forced-choice procedure, to press the bottom panel of a response box in response to stimulus 1 and the top panel in response to stimulus 2 (discrimination subtest). Next. binary series of these two stimuli are presented in increasingly longer series at various rates of presentations. Subjects respond to what they perceive by pressing the panels on the response box corresponding to the order of presentation (sequencing and serial memory subtests). In the present study auditory, visual and cross-modal (auditory and visual) nonverbal stimuli of various durations were investigated. These stimuli have been described in detail elsewhere [26]. In addition, a series of computer synthesized minimal pair speech sounds were also used as stimuli m the Repetition Test. These included the pairs /ba/ vs Ida/, /da/ vs /tat, /ei vs /ae/, /dab/ vs /daeb/, /sa,l vs lstai and isa/ vs /Sal. These stimuli also have been described in detail in a previous publication [24]. In addition to the Repetition Test a battery of sensory, perceptual and motor neurodevelopmental “soft sign” tests were given. This battery included tests of motor control and coordination, balance and station, tactile perception and laterality relationships. These procedures, adapted from those used routinely by neurologists to assess neurological “soft signs” have been described in detail in a previous publication [7]. Demographic, social, family and medical history data were collected via a questionnaire. Receptiw
language assessment
In order that relationships between these experimental variables and receptive language abilities might be estimated, it was necessary to assess the level of receptive language functioning of each subject. Receptive language ages were derived from standardized tests used routinely by speech pathologists to assess receptive language. These included the Token Test [3]. the Test of Auditory Comprehension of Language [?I, the receptive portion of the Northwest Syntax Screening Test [lo] and the Auditory Reception and Auditory Association Subtests from the Illinois Test of Psycholinguistic Abilities [S]. These tests are designed to measure receptive semantic and syntactic linguistic abilities. Mulfitariate
analyses
Multivariate analyses were carried out to discover the extent to which these experimental variables might be correlated with or “predict” the receptive language abilities of the dysphasic children. These multivariate analysis techniques took the form of a stepwise forward multiple regression procedure. In this procedure, experimental variables were ranked according to their contribution to the prediction of receptive language functioning. The first variable selected in the course of this procedure was the one with the largest simple correlation coefficient. All other variables were then examined in tandem with the first variable. That additional variable which was found to reduce the variance of the “prediction” or estimation of level of receptive language to the greatest extent then entered the prediction equation together with the first variable, and so on. Three different approaches were applied in deciding whether or not variables should be included in the prediction equation. These included: (1) An F-test of the extent to which the error in the fitted regression line using N variables was reduced when N + 1 variables were entered; (2) examination of the reduction in the standard error; and (3) examination of the increase in the resulting multiple regression coefficient. It should be emphasized that the variables identified in this manner made a contribution to prediction over and above the contribution made by variables already included in the prediction equation. Thus, variables that are found to be highly significant in univariate analysis do not necessarily make a significant contribution in multivariate analysis. If they do not, it is because they have been pre-empted by other, more powerful predictors with which they are highly correlated.
530
PAULA TALLAL, RACHEL E. STARK and DAVID MELLITS
RESULTS The results of the multivariate analysis indicated that of the experimental perceptual, motor and demographic variables studied, auditory nonspeech and speech perception variables proved most highly correlated with level of receptive language in dysphasic children. The results obtained from the multivariate analysis are shown in Table 2. In this table, all variables that entered the prediction equation are shown, together with the multiple correlation coefficients for each of these variables. Four experimental variables entered the equation for predicting level of receptive language of dysphasic children. All four variables were obtained from Tallal’s Repetition Test. The first three variables were derived from tests incorporating synthetic stop consonant -vowel syllables with various formant transition durations as stimuli. The final variable entering the prediction equation incorporated nonverbal auditory tones. The four variables taken in combination yielded a multiple correlation coefficient (multiple R) of 0.85. That is, 7294 of the variance related to the level of receptive language of the dysphasic children in this study could be accounted for by their ability to discriminate and sequence nonverbal acoustic tones and stop consonanttvowel syllables with various duration formant transitions. This is not to suggest that other variables (i.e. sequencing rapidly presented nonverbal auditory, visual and cross-modal stimuli or sequencing ,ba/ vs /da/ with 40 msec duration formant transitions or /da/ vs /ta/) that were highly significant discriminators between groups in univariate analysis were not correlated with level of receptive language. Rather, it must be kept in mind that in stepwise multiple regression the most powerful variables enter the equation and thus preempt less powerful predictors with which they are highly correlated.
DISCUSSION Of all the perceptual, motor and demographic variables that were assessed in this study including auditory, visual, tactile and cross-modal sensory and perceptual functionings, the only variables entering the multiple regression equation predicting level of receptive language in dysphasic children were acoustic perceptual variables. Furthermore, of the variety of acoustic nonverbal and speech perceptual variables assessed in this study, only the ability to discriminate and sequence stop consonant -vowel stimuli with various duration formant transitions and the ability to discriminate brief duration nonverbal auditory tones entered the prediction equation for level of receptive language. It is of interest to note that these are precisely the same stimuli that we have shown in previous studies to most significantly differentiate the perceptual abilities of dysphasic and normal children, a finding which initially led us to hypothesize a direct relationship between deficits in rapid temporal analysis and disordered language development. The results of the present study support this hypothesis for the first time with empirical data by demonstrating a highly significant (P
synthetic
CV syllables /ba/ and /da/ with
80 msec transitions, 10 ms ISI, (No. of errors) Discriminating synthetic CV syllables /ba/ and /da/, 40 ms formant transitions (No. of trials to criterion) Discriminating synthetic CV syllables /ba/ and /da/, 80 ms formant transitions (No. of errors) Discriminating 300 Hz vs 100 Hz tone (No. of errors)
Sequencing
Variable name
-0.82
0.83
1.08
-3.31/0.001
- 3.79/0.001
4.97/0.00 1
-4.13/0.001
72.70 -2.52
regression Computed c-test P-value
multiple
Regression equation constants
Table 2. Final composite
-0.54
-0.59
0.69
- 0.62
10.93/0.001
10.02/0.001
6.19/0.02
17.08/0.001
Final F-test/ P-value
for the SLD children
Partial correlation coefficient
equation
6.10
7.10
8.13
8.80
Standard error
17.77
14.80
11.11
17.08
Step F overall
0.85
0.78
0.69
0.60
Multiple R
532
PAULATALLAL,RACHEL
E. STARKand DAVIDMELLITS
of receptive language in dysphasic children but rather, the pattern of the variables entering the prediction equation. It is of significance that although the perception of a variety of speech sound contrasts were assessed in this study, the only speech variables that entered the prediction equation for receptive language were those incorporating acoustic spectra that changed rapidly in succession. The ability to process speech sounds that incorporate acoustic spectra not characterized by brief temporal cues [24] (such as /E/ vs /ae/, /daeb/ vs /dab/, Isa/ vs /sta/ /sa/ vs /la/), and longer duration nonverbal tones (250 msec) did not enter the equation predicting level of receptive language. It is important to note that although discrimination and sequencing of stop consonant-vowel syllables incorporating rapidly changing formant transitions have been shown in our previous studies to be grossly impaired in dysphasic children, these same stimuli are amongst the easiest for normally developing children of the same age to discriminate and seque‘nce [24]. In fact, these stop consonant-vowel syllables have been shown to be readily discriminated accurately even by young infants [4,5,9, 131. They are also amongst the first to appear in the speech production of normally developing children (as any proud “da-da” can attest to). Thus, it cannot be argued that these stimuli are merely the most difficult for all children to discriminate and therefore, preempted other speech and nonspeech stimuli from entering the prediction equation. It would also be incorrect to suggest that these data support a theory of delayed development, with the dysphasics just representing the lower end of the normal distribution of speech perceptual development. The pattern of deficits in speech perception demonstrated by this group of children does not seem to be similar to that reported for younger normal children. Our interest in investigating the sensory, perceptual and motor abilities of dysphasic children has been to increase our understanding of the neurological mechanisms underlying language development and disorders. Our now large series of studies with dysphasic children have consistently demonstrated that these children are specifically impaired in their ability to both produce and perceive stimuli that change rapidly in succession, regardless of whether these stimuli are verbal or nonverbal. In a recent study we reported that performance on perception and production tasks (auditory, visual, tactile and cross-modal) alone correctly classified 98:i of subjects as normal or language impaired in discriminant function analysis [27]. Recently, OJEMANNand MATEER [14] have derived a model of the organization of language from electrical stimulation mapping studies. Central to this model is a common peri-Sylvian cortex for language functions in the dominant hemisphere, including sites common to both sequencing of nonverbal movement and identification of phonemes incorporating rapidly changing acoustic spectra. Ojemann suggests that “a single mechanism in this cortex, such as precise timing, may underlie both processes”. It is of considerable interest that the speech stimuli shown by Ojemann and Mateer to be disrupted by peri-Sylvian stimulation in the dominant hemisphere are the same stimuli found in our research to he most difficult to both perceive and produce by dysphasic children. Furthermore, it was these same stimuli that proved to be most highly correlated with the degree of receptive language deficit of developmentally dysphasic children in the present study. In speculating about the possible ontogeny of the peri-Sylvian cortex, Ojemann hypothesized that during the course of human development, the role of this common cortex for language production and understanding initially may be for decoding acoustic cues. Later, during language acquisition, this function may be utilized for identifying those acoustic cues significant to speech. Ojemann further suggests that the same cortical area later
AUDITORYTEMPORALANALYSISAND
RECEPTIVE
LANGUAGE
DEVELOPMENT
533
may develop the patterns of motor output to produce these significant speech sounds. Our data pertaining to the nonverbal perceptual and production abilities of language impaired children, as they relate to decoding and phoneme perception and production and subsequently to receptive language development, provide empirical data directly in support of Ojemann’s speculation. Similarly, Ojemann and Mateer’s data from electrical stimulation mapping studies provide strong anatomical support for our hypothesis concerning a precise timing mechanism underlying hemispheric specialization (the right-ear advantage on dichotic listening tasks) for speech perception in normal adults [ 151 as well as for a deficient timing mechanism which may underlie the receptive language deficits of developmentally dysphasic children. Acknowledgements--This
work was supported
by a contract
from NINCDS
(Project
NS-25353).
REFERENCES aphasia and brain damage. Cortex I, 4c-52, 1964. 1. BENTON, A. L. Developmental 2. CARROW, E. Test for Auditory Comprehension of Language. Urban Research Group, Austin, Texas, 1973. in aphasics. 3. DE RENZI, E. and VIGNOTI, L. A. The Token Test: A sensitive test to detect receptive disturbances Brain 85, 665-678, 1962. changes in speech discrimination in infants. J. 4. EILERS, R. E., WILSON, W. R. and MOORE, J. M. Developmental Speech Hear. Res. 20, 766-788, 1911. 5. EIMAS, P. D., SIQUELAND, E. R.. JUSZYCK, P. and VIGOROTO, J. Speech perception in infants. Science 171, 303-306. 1971. 1957. 6. HOLLINGSHEAD, A. B. Two Factor Index of Social Position. New Haven, Connecticut, and normal I. JOHNSTON, R., STARK, R., MELLITS, E. D. and TALLAL, P. Neurological status oflanguage-impaired children. Ann. Neurol. 10, 159-163. 8. KIRK, S. A., MCCARTHY, J. J. and KIRK, W. D. Illinois Test o~Psycholingui.rtic Abilities. University of Illinois Press, Urbana, Illinois, 1968. categories. In Hearing and 9. KUHL, P. K. Speech perception in early infancy: The acquisition of speech-sound Dacis: Essays Honoring HolloweN Davis, S. K. HIRSH, D. H. ELDREDGE, I. J. HIRSH and S. R. SILVERMAN (Editors). Washington University Press, St. Louis, 1976. University Press, Northwestern University, 10. LEE, L. L. Northwestern Syntax Screening Test. Northwestern Evanston, Illinois, 1969. of deviant child language. Word 29, 139-147, 1978. II. LEONARD, L. The phonology 12. MOREHEAD, D. and INGRAM, D. The development of base syntax in normal and linguistically deviant children. J. Speech Hear. Res. 16, 330-352, 1973. of speech and nonspeech stimuli in early infancy. J. exp. Child Psychol. 14, 13. MORSE, P. A. The discrimination 477492, 1912. 14. OJEMANN, G. and MATEER, C. Human language cortex: Localization of memory, syntax, and sequential motor-phoneme identification systems. Science 205, 1401-1403, 1979. 15. SCHWARTZ, J. and TALLAL, P. Rate of acoustic change may underlie hemispheric specialization for speech perception. Science 207, 138&1381, 1980. 16. STARK, R. and TALLAL, P. Analysis of stop-consonant production errors in developmentally dysphasic children. J. acoust. Sot. Am. 66, 1703-1712. 1979. 17. STARK, R. and TALLAL, P. Selection of children with specific language deficits. J. Speech Hear. Disorders 46, 114-122, 1981. 18. TALLAL, P. Perceptual requisites for language. In Non-Speech Languyge and Communication, R. SCHIEFELBUSCH,(Editor), pp. 449-467. University Park Press, Baltimore. 1980. 19. TALLAL, P. Neuropsychological foundations of specific developmental disorders of speech and language: Implications for theories of hemispheric specialization. Submitted to Psychiatry 1984. 20 TALLAL, P., CURTISS, S. and KAPLAN, R. Unpublished results, 1984. 21. TALLAL, P. and PIERCY, M. Defects of non-verbal auditory perception in children with developmental aphasia. Nature 241, 468-469. 1973. 22. TALLAL, P. and PIERCY, M. Developmental aphasia: Rate of auditory processing and selective impairment of consonant perception. Neuropsyrhologia 12, 83-93, 1974. 23. TALLAL, P. and PIERCY. M. Developmental aphasia: The perception of brief vowels and extended stop consonants. Neuropsychologia 13, 69-14, 1975.
534
PAULA TALLAL, RACHEL E. STARK and DAVID MELLITS
24. TALLAL, P. and STARK, R. Speech acoustic-cue discrimination abilities of normally developing and language-impaired children. J. UCOUS~.Sot. Am. 69, 568-574. 198 1. 25. TALLAL, P., STARK, R. and CURTISS, B. Relation between speech perception and speech production impairment in children with developmental dysphasia. Bruin Lang. 3, 305-317. 1976. 26. TALLAL, P., STARK, R., KALLMAN. C. and MELLITS, E. D. A re-examination of some nonverbal perceptual abilities of language-impaired and normal children as a function of age and sensory modality. J. Speech Hear. Rrs. 24, 351-357, 1981. 27. TALLAL, P.. STARK, R., and MELLITS, E. D. Identification of language impaired children on the basis of rapid perception and production skills. Brain Larz,q. (in press). 28. WECHSLER, D. Wechsler Intelligence Scale for Children--Revised (WISC-R). The Psychological Corporation, New York, 1974. Production errors in developmentally dysphasic children. /. ucous~. Sot. Am. 66, I703 17 12, 1979.