"Medicine, Faculty of"@en . "Audiology and Speech Sciences, School of"@en . "DSpace"@en . "UBCV"@en . "Wyman, Virginia Jean"@en . "2011-05-09T21:02:49Z"@en . "1971"@en . "Master of Science - MSc"@en . "University of British Columbia"@en . "Fifteen predominantly English-speaking, and nine predominantly French-speaking kindergarten and grade one children served as subjects in an experiment designed to investigate discrimination learning of distinctive features. Stimuli were presented aurally and consisted of (a)meaningful non-linguistic sounds, (b)four Non-Nasalized Vowels, (c)Nasalized equivalents of the Non-Nasalized Vowels. Subjects were required to press one of four response buttons on each trial. Feedback was provided. Testing continued until all subjects had achieved asymptote across all Tasks.\r\nMean probability of a correct response by session was compared for (a) linguistic versus non-linguistic Tasks, (b)Non-Nasalized versus Nasalized Vowels, and (c)the distinctive features characterizing the vowels. Confusion matrices were obtained for inter-vowel confusions.\r\nResults shewed that: (a)linguistic stimuli were not as well learned as non-linguistic stimuli, (b)Nasalized Vowels were learned significantly less well than their Non-Nasalized equivalents, and (c)only the distinctive feature \u00B1nasal appeared to correlate with discrimination learning scores. The other features,\u00B1round and \u00B1back , did not appear to operate independently in vowel perception. No evidence was obtained to support the hypothesis that one feature is more easily learned than other features within a discrimination learning task, or that a hierarchy governing the perception of distinctive features exists. The youngest children, however, were observed to perform linguistic tasks significantly less well than the older children. The mains effect for linguistic background of subjects was not significant, although significant interactions between this variable and particular tasks were obtained.\r\nVowels best learned by the children were found to be those that elicited the largest number of verbally mediated responses; these were usually onomatopoeic in nature. Analysis of the discrimination learning results was found to be more amenable to interpretation in terms of traditional parameters of the vocal tract than in terms of distinctive features. No one theoretical framework, however, served consistently to explain the perceptual results obtained."@en . "https://circle.library.ubc.ca/rest/handle/2429/34374?expand=metadata"@en . "DISCRIMINATION LEARNING OF NASALIZED AND NON-NASALIZED VOWELS BY FIVE-, SIX-,AND SEVEN-YEAR-OLD CHILDREN by VIRGINIA JEAN WYMAN B.A., Simon Fraser University, 1969 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF . MASTER OF SCIENCE in the Department of Paediatrics We accept this thesis as conforiiiing to the required standard. THE UNIVERSITY OF BRITISH COLUMBIA September, 1971 In p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t o f the r e q u i r e m e n t s f o r an advanced deg ree a t the U n i v e r s i t y o f B r i t i s h C o l u m b i a , I a g r e e t h a t t h e L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and s t u d y . I f u r t h e r a g r e e t h a t p e r m i s s i o n f o r e x t e n s i v e c o p y i n g o f t h i s t h e s i s f o r s c h o l a r l y p u r p o s e s may be g r a n t e d by t he Head o f my Depar tment o r by h i s r e p r e s e n t a t i v e s . I t i s u n d e r s t o o d t h a t c o p y i n g o r p u b l i c a t i o n o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l not be a l l o w e d w i t h o u t my w r i t t e n p e r m i s s i o n . Depar tment o f / dfsAtyi//IJSA D i v i s i o n o f Audiology- and Speech Sciences The U n i v e r s i t y o f B r i t i s h C o l u m b i a V a n c o u v e r 8, Canada P a t . /t/M- /fW - i i -ABSTRACT Fifteen predominantly English-speaking, and nine predominantly French-speaking kindergarten and grade one children served as subjects in an experiment designed to investigate discrimination learning of distinctive features. Stimuli were presented aurally and consisted of (a)meaningful non-linguistic sounds, (b)four Non-Nasalized Vowels, ('c)Nasalized equivalents of the Non-Nasalized Vowels. Subjects were required to press one of four response buttons on each t r i a l . Feedback was provided. Testing continued until a l l subjects had achieved asymptote across a l l Tasks. Mean probability of a correct response by session was compared for (a) linguistic versus non-linguistic Tasks, (b)Non-Nasalized versus Nasalized Vowels, and (c)the distinctive features characterizing the vowels. Confusion matrices were obtained for inter-vowel confusions. Results shewed that: (a)linguistic stimuli were not as well learned as non-linguistic stimuli, (b)Nasalized Vcwels were learned significantly less well than their Non-Nasalized equivalents, and (c)only the distinctive feature +nasal appeared to correlate with discrimination learning scores. The other features,+round and +back , did not appear to operate independently in vowel perception. No evidence was obtained to support the hypothesis that one feature is more easily learned than other features within a discrimination learning task, or that a hierarchy governing the perception of distinctive features exists. The youngest children, however, were observed to perform linguistic tasks significantly less well than the older children. The mains effect for linguistic background of subjects was not significant,although significant interactions between this variable and particular tasks were obtained. Vcwels best learned by the children were found to be those that elicited the largest number of verbally mediated responses; these were usually onomatopoeic in nature. - i i i -Analysis of the discrimination learning results was found to be more amenable to interpretation in terms of traditional parameters of the vocal tract than in terms of distinctive features. No one theoretical framework, however, served consistently to explain the perceptual results obtained. - i v -TABLE OF CONTENTS Page ABSTRACT i i LIST OF TABLES . , v i LIST OF FIGURES 4 . . . \u00E2\u0080\u00A2 v i i AOKNOWLEIDGMENT . . v i i i Chapter 1. INTRODUCTION 1 L i t e ra tu re Review 6 D iscr i in inat ion learn ing 18 Re l a t i ona l Versus Absolute D i scr iminat ing Learning 20 Advantages of D i scr iminat ion E a r n i n g Methodology 21 Speech Sound Di scr iminat ion Test ing. 23 Conclusion 25 Chapter 2. STATEMENT OF PROBLEM 27 Method 27 Experimental design 28 S t imu l i 28 Recording of S t imu l i 30 Test Sessions 34 Chapter 3. RESULTS. 38 Chapter H. DISCUSSION. 56 L imi tat ions . . 56 Performance on cont ro l versus experimental tasks. . . ' 57 Language background of subjects 59 Age 6 1 -V-Page Vowel 6 5 D i s t i n c t i v e Feature 67 Chapter 5 CONCLUSIONS . 70 Implications f o r future research . . . 71 BIBLIOGRAPHY 72 APPENDIX 77 - v i -LIST OF TABLES Table Page 1. Group mean 10 scores as measured by PPVT 29 2. Distinctive features of non-nasalized and nasalized vcwels. 31 3. Formant frequencies of non-nasalized and nasalized vcwels in Hertz 32 4. Analysis of Variance: mean probability of correct response for last session 44 5. Individual Means Comparisons for linguistic background, age and task: mean probability of correct response for last session 49 6. Analysis of Variance-.mean probability of correct response across vcwels during final session 50 7. Individual Means Comparisons of vowels:final session a l l subjects 51 8. Confusion matrix of vcwels: final session 5 year olds represented by mean proportion correct response 52 9. Confusion matrix of vowels: final session 6 year olds represented by mean proportion correct response 53 10. Confusion matrix of vcwels: final session 7 year olds represented by mean proportion correct response 54 11. Rank order of intervowel confusions: final session 55 - v i i -LIST OF FIGURES Figure Page 1. Stimulus generalization gradient for the response I' [\u00C2\u00A3\ 1. 5 \u00E2\u0080\u00A2 2. Spectrograms and amplitude displays of the eight linguistic stimuli 33 3. Block diagram of experimental apparatus 36 4. Mean probability of a correct response, by session 3g 5. C-A 5-years: mean probability of a correct response by session 41 6. C-A 6-years: mean probability of a correct response by session. 42 7. C-A 7-years: mean probability of a correct response by session. 43 8. C-A 5-years: mean probability of a correct response by vowel during final session . 46 9. C-A 6-years: mean probability of a correct response by vowel during final session 47 10. C-A 7-years: mean probability of a correct response by vowel during final session 48 - v i i i -ACKNOWLEDGMENT To the following, who guided and assisted me i n the preparation of this thesis-, I wish to express my sincere ' THANKS: . . .To Dr. John Gilbert for suggesting the topic, inspiring i t s development, and teaching his graduate student the art of the possible . . .To John Hayhurst and Sister Therese, principals of Alderson Elementary and Notre Dame de Fatima schools, as well as the teachers and children at these schools, for providing a l i t t l e room and many l i t t l e ears . . .To John Delack and Chris Tragakis f o r their patient decisions and revisions as Committee members . . .To Andre Pierre Benguerel for recording the stimuli and replaying many of my words to me . . .To Donna Cumming for cheerfully typing my manuscript . . .To Dan and Mrs. McDonald for their' kind provision of a house on campus and f i n a l l y , . . .To Heather and Tom, without whom nothing. CHAPTER 1 1. INTRODUCTION At present, l i t t l e i s known about the d i sc r iminat ion learn ing of phonemes during primary language acqu i s i t i on . The time required to c o l l e c t l ong i tud ina l data, together w i th d i f f i c u l t i e s in \" developing evaluat ion techniques su i ted to i n fan t s and young ch i ld ren are, i n pa r t , responsible f o r t h i s lack of knowledge. Information perta in ing to ch i l d ren ' s speech perception i s of t heo re t i c a l and p r a c t i c a l i n t e r e s t f o r members of many d i s c i p l i n e s , f o r example (a) l i n g u i s t s , psychologists or b io log ists , who may be in teres ted i n the unfolding perceptual a b i l i t y of the c h i l d as holding poss ib le clues t o s p e c i f i c cen t ra l nervous system mechanisms which make poss ib le t h i s achievement; (b) audiologists/speech pathologists confronted w i th c l i n i c a l problems of auditory perception (See Fr iedlander 1970); (c) e l e c t r i c a l engineers attempting to simulate speech recogn i t ion by mechanical means. Studies of the emergence and i n te rac t i on of auditory perceptual mechanisms may even be prerequis i tes to fu r ther advances i n research. For example, Menyuk (1968) considers that : The major lack , a& present [ i n attempts to evaluate the r o l e of d i s t i n c t i v e features i n the developing phonological system ] i s that no data i s ava i l ab le on the perceptual d i s t i n c t i on s ch i ld ren make during the developmental per iod of morpheme construct ion . . . . [p. 145] Eisenberg (1970) speculates that questions concerning the emergence of perceptual mechanisms and t h e i r organizat ion during develo la-ment \" may be cen t ra l to our understanding of language learning and perhaps even of i n t e l l i g e n t human behavior \" (p. 453). Two long range object ives f o r invest igators of developing auditory 2. perceptual ability have been proposed by Eisenberg (1970); they are (a) the development of a cohesive theoretical framework within which, to ask meaningful questions about the organization of auditory behaviour and, (b) the establishment of baselines for longitudinal research by tracking functional correlates of physical signals during newborn l i f e (pp. 453-4-). Progress has been made toward achieving proposal (b) through, for example, studies of infant preferences for various sound inputs (Friedlander, 1970), and measurements of gross physiological responses to physical signals differing minimally (Eisenberg, 1970). Relatively l i t t l e investigation has, however, been undertaken in terms of a well-defined theoretical framework. The Distinctive Features Theory {Jakobson et a l . ,1957 thereafter \"DF'T\"] would seem to provide a framework within which to study developing auditory perceptual ability . DF describes speech sounds in terms of a finite set of \"features\" specifiable in terms of both articulatory and acoustic phonetic terms. A feature i s presumed to operate on a binary basis; thus each speech sound is characterized by a two-valued matrix specifying presence or absence (\"+\" or \"-\") of each feature. The process of speech perception is viewed as- a succession of binary judgments supposedly made on the speech wave in order to transform the quasi-continuous physical signal into a succession of linguistic units. Distinctive features theory assumes the binary scale to be the \"pivotal principle\" (DFT p. 9) of linguistic structure, since binary oppositions are imposed upon the incoming speech wave. The originator of a structural approach to- language acquisition through, the use of distinctive features was Roman Jakobson. Prior to 3. Jakobsori's structural approach., attempts by-both laymen and phoneticians alike to describe the order in which, children acquire phonemes had proved ineffective. Some persons had attended to individual sounds while some others had \" based their accounts of sound learning on i n i t i a l letters in standard spelling: the 'c's 'of English, 'cat,' cellar 'and 'church', made strange bedfellows \"(Leopold 1953,p. \u00E2\u0080\u00A2352). Jakobson claimed that the sequence of sound categories and the relative chronology of distinctive feature usage were universally predictable, despite enormous individual differences in the speed of acquisition of individual speech sounds. It appears logical to assume that a child, attending to his sound environment, w i l l at f i r s t be able to distinguish in what he hears and articulates only the coarser or optimal contrast e. gr, between sounds of long versus, short duration or great versus weak energy. Analogously, in second language learning, as well as in clinical attempts to establish or recondition discriminations between phonemes, a gross contrast which the subject is capable of perceiving must be established, then successively refined (Lane, 1962; Winitz, 1969). An interpretation of the speech sound discrimination process in terms of several learning theories is presented by Winitz (1969),who adopts the notion of \"generalization gradients\" from Spence (19M-2), and extends an analogy to linguistic stimuli. Within this context, he is able to describe phonemic generalization gradients which he presumes are specifiable in terms of DFT. In extending his observations of generalization gradients to therapeutic situations, Winitz proposes a procedure called \"stimulus pre training\" based on predicting the ability of a subject to discriminate I 4. between phonetic variations, intermediate between a defective articulatory response and the correct production. For example, i n order to examine whether stimulus generalization can be demonstrated with phonemes, Winitz and Bellerose (1963) required f i r s t and second grade children to listen to, and repeat, the pre training stimulus /JV t/. for ten successive tr i a l s . On the eleventh, t r i a l , one of four stimuli were presented: /tJ3f t/,/JVt/ /Qjf.X/ and / t 3 f t / , with i n i t i a l consonants scaled for distinctive feature differences using the analysis of Miller and Nicely (1955) ( / t J / 4 a phoneme not included by Miller and Nicely, was rated as having a difference 1 of 0). The generalization gradient for the mean number of /JsT t/ responses i s shown in Figure 1. It may be observed that the discrimination becomes more di f f i c u l t , and thus the number of /Jjft/ responses increases in the presence of a non-/J3f t/ stimulus, as the sounds to be aascriminated approach each other along the gradient, i.e. , the strength of the generalization is', for- the most part, inversely related to the distinctive difference between the training phoneme and the test phoneme (Winitz, 1969, p. 104). Figure 1 about here 1. A generalization gradient predicts the likelihood of a subject's performing a learned response when any one of a set of .stimuli to be discriminated i s presented. Through reinforcement, the positive stimulus elic i t s a specific response and \"excitatory response tendencies\" spread to similar' stimuli to yield an \"excitatory response gradient\". Similarly, an\"inhibitory response gradient\" describes the spread of \"inhibitory response tendencies\" around the non-reinforced or negative stimulus. The algebraic summation of these two gradients describes the likelihood of the subject's responding in the presence of any one of the set of stimuli to be discriminated. c 5. 1.00. w 0 CO ro 4H O CD 0.85A 0.50 0 Scaled Difference Between Training Stimulus and Test Stimuli Fig. 1: Stimulus Generalization Gradient for the Response Note: Test stimuli were / tJsTt /, / sfX. / 9 /eft /, arid /t/t /, and appear as 0, 1, 2 and 3. SOURCE: Winitz , 1969, p.104 6. The process of hearing includes both sensation and perception and is generally appreciated to be hot a unitary function. It is a . hierarchy of functions having different kinds and degrees of adaptive significance; and i t reflects a hierarchy of mechanisms for processing physical properties of sound'(Eisenherg, 1970, p. 453).\" Naturally occuring speech has been shown to be quasi-continuous; contrariwise, perception of the speech signal appears to be discrete (Liberman et al . , 1967-; Lieberman, 1965; Fodor \u00C2\u00A3 Bever, 1965). Segmentation into linguistic units, i.e. into units which serve . a differentiating function within the pattern of language is thus, apparently, a perceptual process. A biologically based argument in favour of an innate capacity of the human infant to search out and recognize abstract features in the acoustic signal during speech perception is presented by Lenneberg (1967). The following literature review represents an attempt to assess what is known or assumed about innate linguistic capacity and it s relationship to speech sound perceptual development. Literature Review Present understanding of a child's articulatory development greatly exceeds present understanding of the same child's perception of phonological entities. It can easily be observed, for example, that a normal two-year-old child i s able to perform speech-like approximations to the labels applied by adults to items in the child's sensory environment. The two-year-old begins to join items from his repertoire into two or three word utterances. (Brown and Bellugi 5 1964). From these paralinguistic approximations of two-year-olds i t is possible to hypothesize a certain level of linguistic competence for both the phonological and syntactic systems of his native language. It is 7. unfortunate, however, that apart from Friedlander's (.1970) review of gross physiological response measurements to changes in various forms of sound stimulation, and Eisenberg's (.1970) summary- of studies of frequency, intensity and pattern discriminations of infants, our knowledge of the ability of children^ below the age of three years- to discriminate between phonemes is almost non-existent. Eisenberg summarizes the infant studies of frequency, intensity and dimensionality discrimination with, the observation that a new born's auditory- behaviour follows differential rules for high-frequency signals and for constant versus patterned signals, and expresses her belief that this must reflect differential coding mechanisms. For example, patterned stimuli are reported to e l i c i t much more discriminative responses such as pupil dilation, onset or cessation of crying than do constant stimuli. Eisenberg's model for the intrinsic organization of hearing mechanisms consists of the following neuronal networks linked by feedback: Frequency Gating, Parameter Analysis, Ensemble Analysis, and Serial Order Analysis. These circuits, assumed to be innate, are subsequently elaborated from simple \"grading\" mechanisms into well-organized\"subceptor systems\"for the adaptive process of speech perception. Eisenberg assumes that these auditory mechanisms have counterparts in motor specializations required for language operations - (p. 467). The parameter analysis circuit in her model is of great interest because i t serves as a species-specific adaptor of the signal input. For example, selective tuning to language carrier frequencies (500, 1,000 S 2,000 Hz.) is presumably accounted for at this stage, as is the graded response to intensity which permits the developing child to orient himself towards socially significant sounds, i.e. those of speech. The significant 8. perceptual elements of speech, sounds are refined through successive recoding operations, the insignificant elements being filtered through non-specific pathways. Aside from any theoretical interest her model may have for studies of developmental'organization of perception, Eisenberg leaves the reader\u00E2\u0080\u00A2 ~ -'\u00E2\u0080\u00A2 with the valuable caveat that, though ease of control may appear to recommend constant or fixed-dimensional stimuli, such stimuli are rarely found in nature and thus: degree of potency for individual signals in either class [i.e., constant vs patterned l.w.ould vary according to their bearing on verbal communication (p. 460) . . The next step in studying infant pattern discrimination concerns differential response to actual speech stimuli. An experiment in which Russian infants from eleven months to one year eleven months were trained to discriminate a l l of the phonemes of Russian-, is reported by Shvarchkin ( Smith and Miller, 1966}- - I n n i s study Shvarchkin posited a dironological sequence in the receptive development of Russian phonemes. Evidence for a hierarchy of appearance of the distinctive features in the productions of children was presented by Menyuk (1968). Menyuk correlated this hierarchical ordering with resistance to perceptual confusions of children. Like Shvarchkin, Jakobson and Halle (1956) viewed phoneme learning as essentially a discrimination task. They maintained that the physical stimuli most distinctive in a language are learned f i r s t , and that discrimina-tions- involving finer and finer distinctions are gradually acquired. \"Distinctions\" between stimuli are defined in terms of distinctive feature differences. Jakobson and Halle postulate that a distinctive contrast, e.g., 9. grave/acute, once learned w i l l continue to be introduced into successive discriminations. Empirical evidence in support of an assumed innate capacity of a child to selectively- respond to abstract features in the acoustic signal is presented by Eimas ,,et a l . , 1971 and is further elaborated by. Stevens,. 1971. . The assumption that a child acquires a language by learning which features of the physical signal are distinctive has received further theoretical support from Miller (1965) and Chomsky (1965). Chomsky-describes the acquisition process as the adoption of a strategy for selecting a grammar of the appropriate kind that is compatible with the primary linguistic data, i.e., the physical signals of the sound environment. Whereas Eisenberg (1970) poses the question of innate versus acquired . special sensory mechanisms for speech sound perception and production, Chomsky assumes the innateness of a specialized language learning capacity. If the assumption of innateness is accepted, The important question i s : What are the i n i t i a l assumptions concerning the nature of language that the child brings to language learning, and how detailed and specific is the innate schema (the general definition of 'grammar.') that gradually becomes more explicit and differentiated as the child learns the language?(1965, p.-27) Chomsky divides this task into the search for substantive universals, to provide a vocabulary for description of language, and formal universals to describe rules of the grammar and their inter-relationships. Dp is interpreted by Chomsky (1965) to be \"making an assertion about substantive universals with respect to the phonological component of a generative grammar\" (p.28), i.e., the output of the phonological component is assumed to consist of elements characterized in terms of fifteen to twenty- universal phonetic features, each specifiable by language-independent articulatory and acoustic characteristics. . i 10. The material thus far reported i s insufficient to support a complete discussion of how the discriminative ability of a child progresses. Moreover, recent arguments in the literature have concerned the feasibility of collecting any meaningful data on early speech perception. The validity of theoretical constructs such as DF in the analysis of phonetic development is questioned by Winitz (1969), particularly i f biological interpretations are to be made from such analyses. Specifically, Winitz criticizes Menyuk's (1968) thesis of universal order in the development of phonetic features. Winitz advises that some formal criteria must be developed that are inherently meaningful. Most likely these w i l l include statements about the physiological mechanism as a supplement to behavioural observations. Otherwise we may find ourselves trying to explain confusing bits of data such as the fricatives appearing late in two different languages although specific sounds or phonemes may be unordered . . . \" (1969, p.62) . He also suggests that a careful distinction be made and maintained between phonetic development prior to and following actual word learning. Criticisms of Menyuk's findings are made largely on methodological grounds. One of her experimental group subjects were at a level of performance where phonetic features were actualized in words, whereas a second group were at a more \"primitive\" levels Winitz assumes that two different processes are operating at these two linguistic levels, i.e. response integration versus response association. As he points out the problem of analysis and comparison among languages when the measure is appropriate\" usage.(of distinctive features) and not phonemic contrasting or the abstract ordering of phonological rules, i s a sticky one (1969,p.62). He suggests that distinctive features may be correlated differently for different languages, and that comparative data are needed. In his review of studies by Wellman C1931), Poole (1934) and Templin (1957), Winitz 11. finds no evidence for an orderly sequence of phonemic development (in the sense that presence of one phoneme i s a necessary prerequisite for appearance of another) or for an order i n the sequential appearance of phonetic features, although: . ..certain features .are used.more appropriately than others ... a greater proportion of the children at the early age levels. For example, nasals and stops- show- a. greater frequency of appropriate usage at an early age than fricatives and glides (Templin, 1957 and Menyuk, 1968). \" ( 1969, pp. 60-61) For Winitz i t seems clear that the bulk of phonemic learning, at least on the perceptual side, i s acquired prior to or not very much later than three years of age. After age three the production of sound units may be influenced by a number of. factors: CD morphological development., -(2) - sound acquisition i n words, and (3) mastery and refinement of certain motor units\" (1969, p. 74). The underlying structure for the a b i l i t y of this articulatory development to take place i n the three-year-old Winitz attributes to \"some not very well understood behavioral and physiological mechanisms, whose origins no doubt antedate the age of three\"(1969, p. 76) . Excitatory and inhibitory phonemic gradients,which Winitz proposes as specifiable i n terms of DF constitute his only solutions for these unknowns. He concludes that discrimination pre- training represents \" a s h i f t ' i n . our approach from that of an attempt to describe 'underlying' auditory deficiencies to \"that of describing auditory training procedures \"(p.285) ., although, as he points out, i f two or more sounds are equally appropriate to be the stimuli for pre training then further research i s necessary i n order to determine which one of the two or more sounds w i l l most often f a c i l i t a t e - generalization. For example, for defective /// sounds, either the /s/ sound or the If I sound might be used. Both of these sounds, according to Saporta's table. . .,have a distinctive feature difference of three. Which of the two sounds, assuming both are i n the child's repertoire, do we select i n i t i a l l y for stimulus-response generalization learning ? (1969, p.291) ' 12. Some observations concerning Winitz1s:theory of phonemic gradients are in order for i t would appear that results of these discrimination pre-training exercises provide a further.set of unknowns, such as follow: (a) Why is one sound consistently more able to facilitate generalization to the desired sound than others, equally separate in. terms of DF differences from the desired sound ? (b) Why is sound discrimination pre-trairiing a necessary but not a sufficient condition for sound learning in the examples reported by Winitz? It appears that while Winitz rejects a l l existing evidence for a hierarchy of feature acquisition, he i s , at the same time, looking for similar rules governing phonetic similarity and/or ease of acquisition to explain results of his experiments. If there.are indeed excitatory and inhibitory gradients specifiable in terms of distinctive feature systems (such as those of Saporta,i955 and Miller and Nicely, 1955), then i t would seem that this indicates some thing of the \"not very well understood b\u00C2\u00A9>--i6rai and physiological mechanisms\" (Winitz, 1969, p.76) . A presentation by Winitz of the comprehensive data on phonemic gradients derivable from this approach would be useful, but is not presented in his work. A model for the developmental course of phoneme acquisition based on the hypothesis that direction of acquisition is from segments least 2 'marked' in terms of distinctive features to progressively more marked _ According to the theory of Markedness (Chomsky and Halle, 1968), matrix entries are i n i t i a l l y specified by the symbols u for'unmarked'and m for 'marked.' These symbols are systematically replaced by + and - according to \"universal .. interpretive conventions [which 1 w i l l assign to i t [the lexical entry ] a particular phonological structure in terms of + and -entries. The complexity of the lexical item w i l l depend on the number of features that are not left unmarked in i t s matrix representation; each such marked entry w i l l distinguish the item from the 'neutral \u00E2\u0080\u00A2> simplest lexical item\"(p.403) . 13. segments i s presented by Menyuk. (.1968). The purpose of her study was. to analyze in terms of distinctive features the following data: (a)\" correct usage of consonants during the Tiiorpheme construction period\"(Powers, 1957; Nakazima et a l . , 1961 as reported by Menyuk, 1968; (b) substitutions from spontaneous utterances of children with normal versus- defective articulation (c) confusions in short-term recall of consonants by adults (Wickelgren,1966). The outcome of - Menyuk's study should theoretically have included a definitive ordering in the acquisition of distinctive attributes of speech sounds, some information on cues used in perception and production of consonants by children, and an explanation of the observed order in terms of the capacity of the human organism to produce and perceive speech sounds. Distinctive Features investigated by Menyuk included gravity, diffuseness, stridency, nasality, continuancy and voicing. Developmental data were analyzed in terms of percentage of sounds containing a feature used correctly at various ages. For example, i f 3/10 speech sounds marked +grave are correctly used or mastered at any age, then thirty percent usage of this feature at this age was recorded. Menyuk inferred from her data that children above six years CA show complete mastery of consonants (consonant clusters not included). Children of 2^ -5 O P l - 3 years, in two disparate linguistic environments (American English and Japanese respectively) were a l l found to be using consonants containing the features +nasal, +grave and +voice more correctly, and mastering them at an earlier age than consonants containing the features +diffuse, +continuant and. + strident. For American cMldren, this contradicts the rank order of occurrence of features in the \"primary linguistic data\".available to the child for Denes (1963) has shown that the rank order of feature usage by adults i s : diffuse, nasal, voiced,continuant, 14. strident and grave. In children with articulation disorders of sound substitution, the rank order of features used differed from that of the normal usage, i.e. (a) normal: voice, nasal,.strident, continuant, grave, diffuse; (b) articu-lation disorders : nasal, voice, gra\e, continuant, diffuse, strident. Both groups differ considerably from the rank order of feature maintenance in the short-term recall of consonants by adults: strident, voice, nasal, continuant, grave, diffuse. A chi-square evaluation yielded no significant differences between adults and children developing language normally when calculated in percentage maintenance of features in sound substitutions. Children with articulatory disorders maintained a l l features significantly less well (p.=.01) in their consonant substitutions than did children developing language normally, the only exceptions being the feature +nasal ., where there was no significant difference, and the feature +grave (p=.05). Menyuk interpreted her findings to mean that distinctive features play differing roles in production and perception of speech sounds. She indicated that a distinctive features hierarchy for each of these processes may be a linguistic universal reflecting the developing competence of the child. In Menyuk's analysis, the features +nasal and +grave, appear easiest to perceive, recall and produce. Relative ordering of +_ strident and +_ continuant features depends upon task, age and status (i.e. normal versus defective articulation) variables. Adults recall consonants marked +_ strident most readily. Normal children maintain these features best after +_ voice' and +_ nasality while articulatory defective children maintain +_ strident features least well; for this discrepancy, Menyuk offers the following interpretation: other features of consonants which are maintained better 15. and acquired and mastered sooner seem to represent articulatory gestures which, have on-o.f f characteristics or a maximal degree of difference. The vocal cords vibrate or they do not (t voice); the sound is emitted through the nasal passages or i t is not' (t nasal). The attributes +continuant and -grave seem to represent a varying degree of difference. It is on for a somewhat longer time than a burst (+continuant) and i t i s produced somewhere other than the periphery of the vocal mechanism (.-grave)\" (1968,- p. 143). Menyuk's data on a hierarchy of feature acquisition do not correspond with results from Sharvchkin's (1966) study nor with results from a study of the \"acoustic space\" for the perception of consonants by five and six year olds (Peters, 1967). Both Shvarchkin and Peters show their data to be most readily analyzed in terms of features descriptive of the traditional parameters: vocal tract excitation, obstruction and shape. Data obtained from studies where major variables were short term memory', signal detection and shadowing (with adults) also appeared to be more easily analyzed in terms of traditional parameters (Wickelgren, 1965, 1966; Cole et a l . , 1968; Miller and Nicely, 1955; Kozhevnikov and Chistovich, 1965). It should be pointed out that Kozhevnikov and Chistovich indicate that differentiating features may act as intermediaries between the processes of re-synthesis and perception of speech sounds. A consonant, i f not fully perceived, can s t i l l be assigned to membership in a certain group of phonemes dependent on certain perceived features. The authors assume that perception occurs when the last \"inner state\" of the automat is transferred to the outside. They also suggest solutions to two problems that have concerned theorists who wish to maintain the phoneme as the unit for speech perception. These two problems, i.e. (a) the time which would be required for an ongoing selection of phonemes to occur in speech perception and, (b) the errors that would invariably result from decisions based on a very brief exposure to a speech signal (Miller, 1962), elicited 16. the following solutions by Kozhevnikoy and Chistovich; (a) The formation process of a phoneme i n shadowing Can analysis by synthesis procedure), does not include the operation of choice for i t has been shown experimentally that reactions of imitation are not reactions of selection (Chistovich, 1962).. Thus Mi l l e r ' s (.19.62) objections to phoneme re-synthesis based on time requirements are overcome, (b) Since Analysis by re-synthesis theorists consider entirely possible procedures i n which the units of decision correspond to phonemes, but i n the reaching of these decisions i s u t i l i z e d information distributed over the syllable portion. In such a case the minimum carriers of. useful information i n the syllable may be > considered to be the acoustic effects occuring on transition of the speech tract from one state into another (Kozhevnikov . et a l . , .1965, pp. 253-4) For Kozhevnikova and Chistovich., objections based on probability of error i n a phoneme-based analysis are answered by a proposed analyser which takes advantage of inherent redundancy i n the speech signal. Reviewing the failure of analysis schemes based on i n i t i a l segmentation into l i n g u i s t i c units followed by an identification of the units thus derived, Halle and Stevens C1962) concluded that the problem of speech perception should be redefined as an attempt to \"devise a procedure which W i l l transform the continuously-changing speech signal into a discrete output without depending crucially on segmentation\"(p. 156) Recognition i n such an analysis by synthesis scheme occurs by means of a feedback loop connecting the signal (input) to a set of spectral patterns internally generated according to a sequence of rules. Halle and Stevens propose a\"two-stage output, the f i r s t stage being the set of phonetic parameters, which describe the pertinent motions and excitations of the vocal tract and the time-governed rules for their operation, the second stage then being the sequence of discrete entities (phonemes) representing 17. the message perceived by the.listener. In 1962, Fant described the.limitations in the original theory of DF as a function of theory- formulation explicitly for the benefit of linguistic theory, rather than for engineering or phonetic applications. He therefore -proposed a classification of sound segments and their features which would be \" detailed enough to provide correlates to any category- of interest thus not only to phonemic units \"(Fant, 1962, p. 11). For Fant s a sound segment is of the dimension of a speech sound or smaller and there may occur several successive sound segments within the time interval of the speech wave traditionally assigned to the phoneme (1962, p. 11) . Fant (1962, p. 12) categorizes\" segments as : (a) segment type features (manner of production), a set of features referred to by speech production terminology and considered to be binary in nature and, (b) segment pattern features that are essentially characterizations of the segments in terms of place of articulation. Relationships between physiological parameters of speech and corresponding acoustic speech wave characteristics are then summarized in the form of a table. ,' It may be said that although there is considerable debate and some dissatisfaction with attempts to find practical application for the original theory of DF, Menyuk, Winitz and others have found DF a convenient framework within which to examine the processes of speech perception. Approaches to the study of discriminative ability by means of distinctive features have so far yielded l i t t l e useful data and many conflicting interpretations. No comprehensive developmental information or theoretical formulations of development can be found in the speech perception literature. In the following section, a discrimination learning, approach to analyses of distinctive features as structural elements in speech perception w i l l be examined. 1 8 . Discrimination Learning Discrimination learning in a laboratory setting may be viewed as prototypic to everyday learning situations. Distinctive attributes of environmental stimuli are considered to become, through conditioning, cues for an adaptive response on the part of the organism. For example, acquired distinctiveness of letters in the alphabet enables the recipient of an orthographically coded message to associate speech sounds with the various letters and to thus decode the message. The \"world view\" attributed to a child is also a result of aoiftiingencies of reinforcement. The child learns to discern and recognize familiar items from his complex environment. The developmental course of discrimination learning is the object of much of the research in child development, at present. : Discrimination learning studies are divided into three groups (Reese and Lipsitt, 1970): a) perceptual discrimination learning research seeks to discover i f the organism is capable of perceiving a difference between the' stimuli presented; b) Conceptual discrimination learning research poses questions concerning the ability of subjects to use rules or concepts to govern their responses, c) Dimensional discrimination learning research seeks to discover answers to questions based on the assumption that subjects can find the rule(s) governing differences between stimuli presented; this type of paradigm therefore focuses on the learning process per se. The central characteristic of a l l three groups is in the presentation of two or more stimuli and the subsequent conditioning of an appropriate response to each. In the present context the dimensional paradigm w i l l be considered since i t appears to be the most desirable alternative to speech sound discrimination tests.employed at present. To be adequate, criteria for cUscrirnination learning must be differential, I 19. that i s , behaviour must be shown to change with a change i n stimulation, and the s p e c i f i c i t y of the behavioural response to stimulus a l t e r a t i o n must be measurable (Gibson and Olum, 1960). The paradigm usually invoked f o r dimensional discrimination learning meets both requirements. For each stimulus i n a set to be discriminated, a d i f f e r e n t i a l response i s learned through successive reinforcement of the correct response and nonreinforcement of i r r e l e v a n t responses. Response s p e c i f i c i t y i s measured by deriving the generalization gradient of i r r e l e v a n t responses and by an examination of the learning curve to see how quickly the subject extinguishes i r r e l e v a n t responses. In a discrimination learning paradigm, there may be a greater number of s t i m u l i i n the ensemble to be aascrijiinated than there are response al t e r n a t i v e s . Thus i f a subject emits the conditioned response i n the presence of the p o s i t i v e or discriminative stimulus (S+) reward w i l l follow. I f he makes the same response when a negative stimulus , CS-), i s presented , ; he w i l l not be rewarded. I t follows that i n every discrimina-t i o n learning s i t u a t i o n , some relationships are being strengthened while others are being weakened. A l t e r n a t i v e l y , i n a discrimination learning paradigm where the number of s t i m u l i i s equated with, the number of response a l t e r n a t i v e s , each S+ w i l l come to e l i c i t i t s own a r b i t r a r i l y selected response (.e.g. \u00E2\u0080\u00A2 motor or verbal). Thus, t h i s experimental design has sometimes been termed \"paired associates\" or \" i d e n t i f i c a t i o n learning\" (Gibson and Olum, I960; I'imble \"., 1961). Both foregoing paradigms, however, f i t under the general r u b r i c \"discriiriination learning\". When verbal labels are applied to stimulus d i s t i n c t i o n s , discrimination learning i s assumed to be mediated'by these labe l s . House et a l (1962) found that synthesized s t i m u l i least speech-like were more e a s i l y learned i n 20. an eight-stimulus discrimination learning task. They interpret these results as indicative of a probable mediational labelling of these simple stimuli in terms of \"natural\" parameters; for sound perception, i.e., frequency (high-low;), intensity (loud-soft) and time (.long-short). More speech-like stimuli were -synthesized by- alterations.-, in these basic parameters as well, but their time varying values were assumed by the authors to be too complex for this mediation process to be effective. When the stimuli were linguistically meaningful items, acquisition of the discrimination was most easily achieved, suggesting .that labelling was again operating to accelerate the discrimination. Relational versus Absolute Discrimination Learning Learning theorists are in disagreement as to the nature of the items discriminated. Thus relational theorists maintain that the development of a disqrimi.nati.on involves an active comparison, either between stimuli which are physically present, or between one stimulus which is present and the memory trace of another. (Hilg.ard, 1961, p. 392) On the other hand, the absolutest theories maintain that absolute properties of a stimulus provide the cues for a response. Thus ease of acquiring a discrimination should \" vary directly with the psychophysical differences between the stimuli to be discriminated \u00E2\u0080\u00A2\" (Kimble, 1961, p.392 )\". Predictions associated with the discrimination learning process, and described by both groups are as follows:(a. ) The occurrence of transposition, or transfer of the response to pairs of stimuli not involved in the original discrimination but standing in the same relationship to each ether,and (b; the adoption by the organism of a series of hypotheses which are successively discarded until the correct one is h i t upon (Lawson, 1960, pp. 392-3). Empirical validation of these predictions would be very- relevant to current questions concerning speech sound perception. The following 21. questions might be posed: (3.) Is. the phonemic fractionization process (postulated as a basis for phoneme differentiation by- Jakobson and Halle, 1956) comparable to the transposition observable in carry-over of discrimination 1iraining?(b) Is i t possible to state the nature of the hypotheses that the human successively adopts and discards while i n i t i a l l y learning the distinctive linguistic elements in his native language?(c) How does a child learn to select the distinctive portions of the physical input sounds and eventually- assign allophones to their various phoneme bins? Advantages of.Discrimination^Learning Methodology: The advantages, of a discrimination- learning paradigm for a study of . children's speech sound perceptual ability are the following: 1. There is a clearvcut discrete response to each stimulus. 2. Previously existing, connections in stimulus-response pairings are avoided: i n i t i a l pairs are assigned arbitrarily and may be varied from ensemble to ensemble. . 3. Nonsense items may be the stimuli thus overcoming contaniination of the results by subject familiarity or by presence of the items in the child's active or passive repertoire. 4. The embarrassing variable, ability of the experimenter to interpret the overt response of the subject ( i f verbal) is avoided. (Winitz, 1969, reports the confusions resulting even with tape-recorded responses). 5. The experimenter can alter the task to suit \"the age level of subjects by varying the number of stimuli in a set, the inter-stimulus similarity and the inter-stimulus interval. ' The usual response measure in discrimination learning is number of correctly identified stimuli per block of tr i a l s . The similarity between stimulus items is operationally defined by the shape of the learning curve. The slope and relative smoothness, of the learning curves also provide data about the difficulty- of discrimination for subjects, and about relative ease of acquisition of various items. 22. Lawson (1960) points, out that no phenomenon, such as stimulus generalization, can he described independently of the operations by which, the' phenomenon is measured* (p.219) and regards: i t as unfortunate that wb do . not as yet'have,' as has- often been' assumed, a firm empirical basis for deciding whether discrimination learning determines generalization effects or vice versa\". (pp. 217-219). Summary-The work of Paylov has been interpreted as presuming that generalization precedes discrimination, (Lav/son, 1960)> Spence (1936, 1937), however 5 proposed the opposite to be true; that is;, he depicted generalization gradients of excitation and inhibition based on the dimensions for the original discrimination.. ( Kimble - ,\u00E2\u0080\u00A2 1S61 ) As reported earlier, Winitz's (196 7, 196 9) attempts to facilitate acquisition of a defective or unavailable articulatory discrimination by auditory pretraining are based on Spence's generalization gradients. This basis for the prediction of which discriminations w i l l be relatively more easily acquired has not been shown to have predictive validity. Thus, Lawson (1960) feels that: The failure to cliscover^the relevant stimulus:,variables in \u00E2\u0080\u00A2 discrimination learning via... . the direct, generalization-oriented approach has caused some psychologists to try turning the problem around. They have employed well-established discrimination learning techniques as a method of determining how subi ects do analyze their environments. .They seek to answer.such Questions .as. Are some stimulus differences \"!'more naturally -1 responded to than . others? 'Just . how complicated can the positive and negative stimulus differences be?' (p.210) . In summary, a discrimination learning paradigm would seem to be advantageous for an investigation.of the relative ease of acquisition of some selected members of a set of distinctive features of a set of phonemes (e.g., Jakobson, Fant and Halle, 1957). In addition, given such a methodological paradigm, theoretical matrices, models of distinctive features similarities, (e.g., Saporta, 1955; Peters, 1967; and Greenberg g Jenkins, 19 23. and hierarchies, of distinctive feature discriminations (cf. Shvarchkin, 19 66; Jakobson and Halle, 1956; Menyuk, 1968) may be tested. Although a discrimination learning task with. appropriate button presses to auditory- stimuli is probably beyond the abilities of children under three years . (the age by which. Winitz assumes a l l phonemes are at least perceptually available to the child), i t has been used to provide some information on the discrirnination ability of four-year-old children (Gilbert, 1970b). The collection of comprehensive data on discrimination learning of distinctive features in older children, e.g., five to seven years old, i s possible. Normative data across these age groups should provide information relative to questions of hierarchical ordering in dinstinctive feature learning. Speech. Sound Discrimination Testing, Existing analyses of speech sound discrimination'development and tests of relative perceptual status, (e.g., of children with perceptual difficulty as compared to normal children have presented difficulties which preclude establishment of comprehensive norms. A lack of data has not, however, interfered with the employment of speech, discrimination tests as clinical tools for the assessment of what are generally labelled \"perceptual\" problems. The major untested hypothesis in such tests concerns, the nature of the items being Discriminated. For example: Ca) Is the child hearing and responding appropriately- to speech sound differences between two words (usually minimal pairs), and i f he i s , what constitute the cues for this differential response? (b) Is the child responding to Gestalten, i.e., to general overall' properties of stimulus items ? In traditional discrimination tests, stimuli are generally pictorially represented. Thus, the examiner' cannot differentiate response to a visual 24. cue as. opposed to an auditory.cue, i t may he that the child is responding, either at random or consistently-, to the picture with which, he is most familiar. Within the format of one speech, sound discrimination test (Schiefelbusch fet al*., 1958); ..subjects\" are .asked to, say which two out of three pictures have names which sound alike, by- naming the picture silently and then pointing appropriately. Although this paradigm gives scores similar to those obtained from more traditional methods (where the examiner names the pictures) the nature of item discrimination remains unknown. Visual cuing of auditory' stimuli, presents further problems by limiting the number of sounds, and sound combinations which may be tested, since i t i s difficult to find minimal pairs for a l l combinations of speech sounds the examiner may- wish to test. In addition, mental age and vocabulary- level of subjects may either confound or limit results obtained from such tests, Since speech, signals are complex entities, psychophysical methods have not been used to study speech sound discriirination by young children (Gibson and Olum, 1960). Pronovost efr al*>, 1953), -however, developed.a quasi, \"methods of adjustments\" procedure; their test ejuploys a paradigm1 in which aural stimuli are presented with, a visual cue. A method of obtaining data of theoretical interest from children with disorders in speech sound production is to conduct comparative studies of substitutions in speech, sounds and of the syntagmatie speech sound contexts in which these substitutions occur. Menyuk (1968) has suggested such an approach, indicating that i t may serve to provide information about which, distinctive features are most often and universally substituted, and of the feature context in which these substitutions occur. It is 25. possible that information pertaining to the use of distinctive features in the acquisition of phonology in two or more different languages, (and for children at varying levels of linguistic competence), w i l l provide information about psychological universals necessary for language acquisition (e.g. auditory memory-span) as well as possible individual differences in underlying psychological mechanisms. Conclusion This review of the literature has served to assess what is known or assumed ahout an innate linguistic capacity- and i t s implication in speech sound perceptual development. The purpose of this assessment was to outline research, to date and to suggest areas where further research, was indicated. Since Winitz (1969) assumes that a l l perceptual development of functional speech elements such as phonemes has occurred by the age, of three years, the course of perceptual development, as well as the mechanisms involved are - or should according to, Winitz ~ remain unknown . He expresse s distrust of statements about acquisition processes from which biological interpretations were to be made although his own theory of acquired phonemic gradients does not differ notably in intent from other maturational interpretations. It is apparent that Winitz has not succeeded in resolving the contradictions within his attempt to make a measurable, difference in speech production through, appropriately spaced discrirnination pret^aining. Menyuk (1968), in presenting evidence that distinctive features play differing roles in the process of production versus: discrimination, has provided some hypotheses to explain the operation of distinctive features as cues in speech sound perception, but \"-'has e'iprfeg.sed... .necessity for empirical 26. validation. . Distinctive features theory- provides., a useful theoretical framework within which to question innate mechanisms designed to search, for and recognize abstract elements in the speech signal. Redundancy of speech sound description is. reduced and patterns among stimuli, are exposed. Available evidence suggests distinctive features involvement in speech perception (Wickelgren, 1966; Menyuk, 1968). A discrimination-learning paradigm,in which Ca) arbitrary- stimulus response associations are assured by the use of nonsense syllables- as stimuli, and (b) information on relative ease of acquisition of various distinctions i s obtainable from measures: of the percentage of correct response for each stimulus (as well as from the shape and smoothness of the curve representing acquisition of the response),would provide information pertaining to the proposed questions. CHAPTER 2 27. STATEMENT OF PROBLEM 1. If distinctive features are considered'the' structural units in speech perception, and i f each feature is assumed to carry as i t s only message i t s distinctiveness, are some distinctive features more easily learned than others - in a discrimination learning task? 2. What combinations of distinctive features cause some sounds to be more often confused than others? 3. Are the most easily learned distinctive features the same for children at different age levels? Method Twenty-three kindergarten and grade one children from two linguistic backgrounds and three chronological age levels served as subjects in the present experiment. Nine, predominantly French speaking subjects [three each at CA's 5-3, 6^-3 and 7-1 years ' (+_ 1 month), one male, eight females ] came from homes in the French language community- of Maillardville, British Columbia, and attended schools where eighty percent of their instruction was conducted in French. Extent of bilingualism was partially assessed for these subjects by noting language of response when subjects were stimulated to provide French names for pictures on the Laradon Articulation Scale CRe-r-vised, B). If an English .response was given, the bilingual experimenter asked in English i f the child could provide the French name. If this was supplied on the second t r i a l , a score of one^-half the usual per item score was given. A l l French subjects achieved high scores on this task (x 84%, s.d. 15%) suggesting only slight English contamination. Fourteen English speaking subjects from monolingual homes in the same 28. geographical area attended classes.. conducted in English.. This group consisted of five subjects, at each, of two CA's: 5T-3 and 7-1 years (+_ 1 month), and four subjects at CA 6^ 3 (+_ 1 month). Seven of these subjects were male and seven were female. A l l subjects demonstrated normal speech sound articulation ability assessed by administration of the Laradon Articulation Scale. Hearing sensitivity for pure tones was demonstrated to be 25dB or . better for octave frequencies between 0.5-8 KHz. ISO (1964) Standard. A Belt one portable audiometer (10-AC) was used to administer hearing tests. The Peabody Picture Vocabulary- Test (PPVT) was administered f i r s t in French to French subjects, followed two months later by a repetition in English. Intelligence quotients, for'French subjects appeared to vary l i t t l e with language of adninistration. Scores as measured by- the PPVT are shown in Table 1. Insert Table 1 about here Experimental Design ' Stimuli In order to assess the effect of difficulty due to perceptual-motor requirements of the task i t s e l f , a control tape consisting of four non-speech stimuli was a<\u00C2\u00B1rinistered to a l l subjects. Stimuli for this task (Task 1) consisted of: (a) white noise of 45Q msec, duration generated by a General Radio Company Random Noise Generator (1382); (b) the buzz from an a r t i f i c i a l larynx (Bell Telephone Laboratories); (c) a bicycle b e l l ; (d) the sound of two long, light sticks- struck together. A l l stimuli were 29. TABLE 1 Group Mean IQ Scores As Measured by PPVT Group X S.D. French (1): i n French 86.44 14.28 French (2): i n Engl i sh 88.22 18.65 Engl i sh 100.79 9.63 30. equated for intensity- by ensuring that peak VU deflections during recording on a Scully- 280 tape-recorder did not exceed' 0 VU. It was anticipated that performance on this relatively simple task would exceed performance under the same conditions when actual speech sounds served as stimulus items. Two other tasks were administered to a l l subj ects: Stimuli for'Task 2 consisted of four Non-Nasalized Vowels in isolation, and for Task 3, the Nasalized equivalents of these vowels. Distinctive Feature matrices for these vowels are shown in Table 2 and formant values are tabulated in Table 3. Figure 2 shows spectrograms, of the eight vowels. Insert Tables 2 & 3 Insert Figure 2 Recording of Stimuli Tasks. 2 and 3 were recorded in the following manner: A bilingual (French/English) male phonetician recorded the stimuli at 1\ ips. on a Scully- 280 tape recorder with associated Bruel and Kjaer condenser microphone (model 4145). During recording, the'experimenter ensured that peak VU deflections did not exceed 0 VU. Twenty recordings were made of each set of eight vowels, sufficient so that a final set of eight could be selected as stimulus items and at the same time meet the following criteria: (a) consistency of intonational pattern within each set of vowels; (b) absence of voice quality differences within a set which might provide discrirnination cues to listeners; Cc) uniformity of intensity and duration 31. TABLE 2 D i s t i n c t i v e Features of Non-Nasalized and Nasalized Vowels Feature Vowel Non-Nasalized e a o ce Nasalized e \u00E2\u0080\u00A2 a o CE back round nasal + + + + + + + + + + + + TABLE 3 Formant Frequencies of Non-Nasalized and Nasal ized Vowels i n Hertz Formant Vowel Non-Nasalized Nasal ized e a o as e d O CE F l 550 775 425 625 900 \u00E2\u0080\u00A2 1025 675 900 F 2 1875 1250 750 1300 1575 1500 850 1275 33. N X f\u00E2\u0080\u00941'\u00E2\u0080\u0094T*1\u00E2\u0080\u0094! I\u00E2\u0080\u0094!\u00E2\u0080\u0094I\u00E2\u0080\u0094I\u00E2\u0080\u0094r o o c q q r i i \u00E2\u0080\u0094 r \u00E2\u0080\u0094 ! \u00E2\u0080\u0094 n o q o o oo oi \u00E2\u0080\u0094 M X o o O O o o 34. of vowels w i t h i n each set. The vowels were judged by three trained phoneticians.' Judges\"were unanimous i n t h e i r agreement of. a c c e p t a b i l i t y of the set of eight vowels f i n a l l y selected. The four s t i m u l i i n each major group (Nasalized/Non-Nasalized) were played from tape loops on a Tandberg 64X tape recorder. St i m u l i were recorded i n a quasi-random order to produce a sixteen item l i s t , w i t h i n l i s t s each, vowel appearing four times. Four, such l i s t s were combined to produce a s i x t y - f o u r item task f o r presentation during one session. The ordering of s t i m u l i w i t h i n a task was constrained to ensure that each item followed i t s e l f and every other item four times. In order to avoid subject f a m i l i a r i z a t i o n with the f i r s t items i n the f i r s t l i s t (of f o u r ) , a second tape was made i n which order of the four l i s t s comprising each tape was reversed. Subjects received t h i s second version, of each of the tasks (2a and 3a) on alternate days from administration of task 2 or 3. Test Sessions Subjects l e f t t h e i r classrooms at approximately the same time each day f o r a period of approximately f i f t e e n minutes of t e s t i n g . The order of presentation of s t i m u l i was balanced to counter morning-afternoon t e s t e f f e c t s . Each group was equated f o r l i n g u i s t i c background and chronological age. Testing continued f o r approximately two months (see r e s u l t s section f o r asymptotic data) and was conducted i n a room with low ambient noise l e v e l ; care was taken to avoid t e s t i n g during recess periods. Day 1. On the day of a subject's f i r s t experimental session, he/she was seated i n front of a response panel i n an area adjacent to the experimental instrumentation and designed to conceal the instrumentation from the subject. A block diagram of the experimental set-up is. shown i n Figure 3. I 35. Figure 3 about here The following instructions were then read to English subjects: (See Appendix A for Trench translation). \"Listen! We are going to play- a game. You w i l l hear a man saying some funny- sounds. Your job i s to find the sound's house. These are the houses for the sounds (Experimenter indicates buttons of response panel3\u00E2\u0080\u00A2) When you hear one of the funny sounds, press a button. Make the light come on over your pressing finger by putting the sound in i t ' s right house. If the light won't come on over your finger, you pressed the wrong button but I w i l l show you which house the sound really lives in. Only one sound lives in each house. Let's try one. Listen f i r s t , then press a button.\" Stimuli were played back on a Tandberg 6UX tape recorder and presented at a comfortable level over TDH-39 headphones set in NAF-1481490-1 cushions. The tape recorder was stopped after each stimulus presentation to allow the subject as much time as required to respond. If a correct response was made, the experimenter activated the light above the correct response button. If an incorrect response was made, the light above the appropriate button was activated. No reinforcement program was provided by the experimenter or the experimental apparatus. Subjects received the linguistic tasks as follows: one half of the subjects received Task 2 followed by Task 3}and one half performed the listening tasks in a reverse order in an attempt to control for possible sequential effects. Testing continued until asymptote across a l l subjects and sessions was achieved (See Figure 4). 3 Four unlabelled lights, with associated response buttons were arranged in a row on rectangular metal panels. Depression of a button by the subject or Experimenter caused the corresponding light on the other response panel to be activated. P O W E R TAPE R E C O R D E R R E S P O N S E B O X E 1^ E H E A D P H O N E S n [RESPONSE B O X S S H E A D P H O N E S F i g . 3: Block Diagram of Experimental Apparatus CO CD . 37. The non - l i n gu i s t i c cont ro l task (Task 1) was adninistered to a l l subjects a f t e r c r i t e r i o n had been reached on Tasks. 2 and 3. Day 2 and Succeeding Sessions. On second and succeeding administrations of e i the r task to a subject, the assignment of button to sound was randomized to con t ro l f o r po s i t i o na l e f f e c t . CHAPTER 3 38. RESULTS The purpose of this- investigation has been to examine discrimination learning of distinctive features by children at three chronological age levels with specific reference to the following questions: (a) Are some distinctive features more easily learned than others?; (b) What combinations of features cause some sounds to be more often confused than others?; Cc) Are the most easily learned distinctive features the same for children at different age levels? Complete data were obtained for a l l subjects across a l l tasks. A preliminary survey of the results revealed no difference in performance due to order of task adninistration, (Nasalized followed by Non-Nasalized Vowels5or vice-versa), therefore data for the two subgroups of subjects performing each task have been combined. Figure 4 about here Mean probability of a correct response per session is presented in Figure 4. Since subjects were forced to choose one of four response alternatives on each t r i a l , chance level of response is indicated at the 0.25 point x Pc. It may be observed that learning of non-linguistic stimuli (Task 1) as demonstrated by the learning curve is much more rapid than learning of linguistic stimuli. (Tasks. 2..and 3). The asymptote f\u00C2\u00B0 r Task 1 appears to have been reached within the f i r s t session whereas for Tasks 2 (Non-Nasalized) and 3 (Nasalized) respectively, asymptote was not achieved until sessions four and five. 1.0-Task 1 Task 2 a Task 3 o x 2&3 T 9 i 10 11 SESSION F i g . Mean P r o b a b i l i t y of a Correct Response, by Session 40. Within l i n g u i s t i c tasks, mean post-asymptotic scores f o r Task 2 exceed those f o r Task 3. Learning curves f o r subjects at each, chronological age l e v e l are presented i n Figures 5 through 7. Figures 5 - 7 about here Language:Table 4 presents the summary- of an unweighted means analysis of variance (Winer, 1962, p.374) performed on the mean p r o b a b i l i t y of a correct response during the f i n a l session f o r Tasks 2 and. 3). I t may be observed that the F ratio: f e l l considerably below the c r i t i c a l value; subject data f o r the two language backgrounds have therefore been combined. Insert Table 4 about here Age: Variance a t t r i b u t a b l e to age yielded a s i g n i f i c a n t F r a t i o (p<.01). Individual means comparisons revealed that fiver year-olds performed s i g n i f i c a n t l y poorer than d i d s i x - o r seven-year-olds, but that six-year-olds d i d not learn s i g n i f i c a n t l y better than severF year-olds. Task: Variance at t r i b u t a b l e to Task (Nasalized vs Non-Nasalized) yielded a s i g n i f i c a n t F value (p(. 05) w i t h o v e r a l l scores f o r Task2 higher than o v e r a l l scores f o r Task 3. The s t a t i s t i c a l l y s i g n i f i c a n t i n t e r a c t i o n between language and Task was further analysed by i n d i v i d u a l means comparisons. Subjects from both language backgrounds performed s i g n i f i c a n t l y better on Task 2 when compared o. IX 1.0-j 0.9-0.8-0.7-0.6 0.5-1 0.4 0.3 0.2 0 .H 0.0-r Non-nasalized a Nasalized e \"T 3 T \" 4 ~T 5 ~T~ 6 T 7 8 9 10 11 SESSION F i g . 5: C.A 5-years: Mean P r o b a b i l i t y of a Correct Response by Session -P i\u00E2\u0080\u00941 1.0' 0.9-0.8-0.7-0.6-0.5-0.4-0.3-1, 0.2-0.1-0.0--0 Non-nasalized \u00C2\u00B0 Nasalized s ~r 2 T\" 3 ~T~ 4 T\" 5 j 6 T 7 8 10 11 SESSION F i g . 6: C.A 6-years: Mean P robab i l i t y of a Correct Response by Session 1.0-0.9-0.8-0.7-0.6-u Q_ 0.5-IX 0.4-0.3-0.2-0.1-0.0-0 ~T~ 2 Non-nasalized a Nasalized \u00C2\u00AE 3 T\" 4 5 6 T\" 7 8 T 9 10 11 SESSION F i g . 7: C.A 7-years: Mean Probability- of a Correct Response by Session CO 44. TABLE 4 Analys is of Variance: Mean P robab i l i t y Correct Response f o r Last Session Source df MS F Language Background (A) 1 38.47 Age (B) 2 2,687.61 10. 75** Task: N vs NN (C) 1 5,444.71 4. 18* A X B 2 278.72 A X C 1 43,327.73 33. 25** B X C 2 76.83 A X B X C 2 21,021.00 16. 13** *p<.05 **p<.01 45. t o performance on Task 3 Cp^. 01) j French, speaking ch i l d ren achieved s i gn i f i can t l y - lower scores on Task. 3 than Eng l i sh ch i l d ren performing Task 2 Cp<.01). For each vowel, mean p r o b a b i l i t y of a correct response across the four blocks of the f i n a l session i s presented f o r a l l subjects by age i n Figures 8 - 1 0 . I t may be observed' that f o r the Non-Nasalized task, the mean, probabi l i ty- of a correct response increases w i th age from a l e v e l of approximately 75% at C.A. 5.3 years to 87% at C.A. 7.1 years. For the Nasal ized task, the'mean l e v e l increases from approximately 43% to 70% f o r the oldest group of subjects. Figures 8 - 1 0 about here An ana lys i s of variance performed on the mean proport ion correct response f o r each vowel during the l a s t session i s summarized i n Table 6. Since the F value f o r vowels exceeded the c r i t i c a l value (F n r -, r _ = 4.00). . 3 b ,-L, b o te s t s on i n d i v i d u a l means comparisons were ca r r i ed out and are summarized fo r . a l l subjects \"in Table- ' 7.< -.-. y.r : - ?.. Within both Tasks and across a l l age l e v e l s , the vowel /o/ was best learned. /ce../ was leas t w e l l learned i n the Non-Nasalized ser ies whereas /a./ -was l ea s t w e l l learned w i t h i n the Nasal ized s e r i e s . Confusion matrices f o r the l a s t session (Tables \"\u00E2\u0080\u00A2 8 11). show leas t confusion of the vowel /of and most confusion f o r the vowels /\u00C2\u00AB\u00C2\u00A3/ and /a/. A rsmk ordering of i n t e r -vowel confusions w i t h i n each Task i s presented i n Table 13. Insert Tables 5 - 1 1 about: here Non-nasalized Nasalized 1.0 : 1 O.o-j 1 1 1 1\u00E2\u0080\u0094 j 1 1 j r 1 2 3 4 1 2 3 4 BLOCKS F i g . 8: C.A 5-years: Mean P r o b a b i l i t y of a Correct Response by Vowel during F i n a l Session -P u Q L IX 1.0-0.9H 0.8 0.7 0.6-0.5-0.4-0.3 0.2-0.1-0.0 Non-nasalized \u00E2\u0080\u00A2\u00E2\u0080\u0094\u00E2\u0080\u00A2 \u00E2\u0080\u0094 O* ,0 \u00E2\u0080\u00A2 \u00E2\u0080\u00A2 6 a o 08 T 2 \"T 3 nr 4 Nasalized a a T 2 e a \u00E2\u0080\u00A2\u00E2\u0080\u00A2 o 55 T 3 BLOCKS F i g . 9: C.A 6-years: Mean P r o b a b i l i t y of a Correct Response by Vowel during F i n a l Session. Non-nasalized u a. IX 1.0-0.9-0.8-0.7-0.6-0.5-0.4-0.3-0.2-0.1-0.0-\u00E2\u0082\u00AC a o ce \"T 2 3 ~T~ 4 Nasalized a a. \u00E2\u0082\u00AC \u00E2\u0080\u00A2 a o ce T\" 2 ~T~ 3 T 4 BLOCKS F i g . 10: C.A 7-years: Mean Probability- of a Correct Response by Vowel during F i n a l Session -p CO TABLE 5 Ind iv idua l Means Comparisons f o r L i n gu i s t i c Background, Age, and Task: Mean Probab i l i t y Correct Response f o r Last Session 25.20 29.67 34.67 40.00 40.75 44.00 44.67 50.20 50.50 53.20 54.00 59.33 4 E/5/N F/5/N F/7/N E/5NN E/6/N F/6/N F/5NN E/7/N E/6NN E/7NN F/6NN F/7NN 25.20 E/5/N ft ft ft ftft ftft ftft ftft ft* 29.67 F/5/N ft ftft ftft ftft ftft 34.67 F/7/N _ * * . * * * * 40.00 E/5NN 40.75 E/6/N * 44.00 F/6/N * 44.67 F/5NN * 50.20 E/7/N 50.50 E/6NN 53.20 E/7NN 54.00 F/6NN 59.33 F/7NN **p:-.. .01 *p .05 E=English speaking F=French speaking 5,6,7=Ages of subjects N=Nasalized Task $JN=Non-Nasalized Task TABLE 6 Analys i s of Variance: Mean P robab i l i t y Correct Response across Vowels during F i n a l Session Source df MS F Age (A) 2 10,970.87 11.06** Vowel (B) 7 93.99 7.04** A X B 14 19.43 * *p <.01 TABLE 7 I n d i v i d u a l Means Comparisons o f Vowels: F i n a l Session A l l Subjects 25.71 31.89 33,27 35.63 36.95 37.80 42.81 ce e ce o e a o 21.41 a A A A A tfh *\u00C2\u00BB *\u00C2\u00BB A A A A A *-\u00E2\u0080\u00A2 A A *\ ** 25.71 ce A A A A A A A A A A A A ** 31.89 e AA A A 1> *\u00C2\u00BB A A A A 33.27 03 - A A A A A A A A 35.63 o V{ ** 36.95 e A A 37.80 a A A /48 **p <.01 TABLE 18 Confusion Matrix of Vowels: Final Session 5 year olds Represented by mean Proportion Correct Response RESPONSE RESPONSE e a o ce e a 6 ae e 0.63 0.15 0.09 0.13 1.00 e 0.43 0.13 0.21 0.23 1.00 a 0.15 0.65 0.08 0.13 1.00 a 0.16 0.33 0.21 0.30 1.00 o o.n 0.09 0.74 0.05 1.00 o 0.16 0.14 0.46 0.23 1.00-09 0.19 0.15 0.08 0.59 1.00 ce 0.20 0.23 0.24 0.34 1.00 DIAGONAL TOTAL: 0.65 DIAGONAL TOTAL: 0.39 TABLE ' 9 Confusion Matrix of Vowels: F i n a l Session .6 year olds Represented by mean Proportion Correct Response RESPONSE N N RESPONSE N e a o To t a l . e a 5 ae T o t a l e 0.79 0.06 0.04 0.11 1.00 e 0.76 0.07 0.03 0.14 1.00 a 0.02 0.84 0.03 0.12 1.00 a 0.14 0.45 0.14 0.27 1.00 CO o 0.03 0.02 0.93 0.03 1.00 o 0.02 0.18 0.75 0.05 1.00 M c e 0.12 0.09 0.10 0.70 1.00 0 3 0.09 0.17 0.07 0.67 1.00 DIAGONAL TOTAL 0.81 -DIAGONAL TOTAL 0.66 Cn CO TABLE 10 Confusion Matrix of Vowels: F i n a l Session 7 year olds Represented by mean Proportion Correct Response RESPONSE RESPONSE fc e a o 03 Total e a 5 ce Total e 0.87 0.02 0.05 0.05 1.00 e 0.80 0.03 0.07 0.09 1.00 a 0.05 0.88 .04 .03 1.00 a 0.11 0.56 0.14 0.19 1.00 0 0.02 0.03 0.93 0.02 1.00 o 0.05 0.08 0.81 0.05 1.00 ce 0.09 0.05 0.05 0.80 1.00 . CE 0.07 0.23 0.09 0.60 1.00 DIAGONAL TOTAL 0.87 DIAGONAL TOTAL 0.69 TABLE 1 1 Rank Order of Intervowel Confusions: F i n a l Session Rank Confusion Contibution to To ta l E r ro r N NN N NN 1 . 2 . C&-Q. 12% 5% 2 . 6 . a-o 7.4% 2% 3. 1 . e-ce 6.9% 6% 4 . 5. oe-o 6.0% 3% I CHAPTER 4 5 6' DISCUSSION This investigation has examined children's auditory discrimination learning of linguistic stimuli varying in number of distinctive features. More specifically, performance\" by children at three chronological ages, and from two linguistic backgrounds, was compared for a Control task of non-linguistic stimuli and two Experimental tasks comprising linguistic stimuli. It was hypothesized that although subjects at a l l age levels would perform the Control task equally well, a hierarchical ordering of linguistic discriminations might emerge during performance of Experimental tasks-, thus by analogy, linguistically more mature subjects (i.e.'those of higher chronological age) might be expected to perform significantly better than linguistically less mature subjects. Performance on the two Experimental tasks (Nasalized versus Non-Nasalized Vowels) was expected to be similar for French speaking subjects since both nasalized and non-nasalized vowels have been shown to exist in their phonemic system. However, i t was anticipated that learning of Nasal-ized vcwels by English speaking subjects might be less effective as these subjects would presumably be unfamiliar with nasalized sounds. No a priori hypotheses were established predicting superior learning of one feature or combination of features, although the feature + nasal was expected to result in slower, more incomplete discrimination learning, for nasalized vcwels have been shewn to appear later in phoneme acquisition than non-nasalized vowels. Limitations The interpretation of results' obtained in this experiment is considered bearing in mind the following limitations: 57. (a) No attempt was made to establish whether perceptual oppositions of nasalized vowels exist for predondnantly French speaking children. The experimenter elicited productions of the phonemes by presenting the stimulus pictures: \"bread\", /[pe'1 \"brown\", [brce ] ., \"white\" , [ ' b I a ], and \"boy\", [ g a r s o . ] to a l l subjects; i t was observed that a l l French speaking subjects were able to maintain these phonemic oppositions despite a tendency to front nasalized back vowels. (b) Only a small number of French speaking subjects were available. The linguistic competence and performance of the subjects was confounded by: 1) relative exposure to spoken English and,a-:2)' fluency in French pronunciation. Thus only tentative conclusions can be drawn about the \u00E2\u0080\u00A2 relative performance of English speaking versus French speaking subjects on the Experimental tasks. (c) No measure,other than the confusion matrix was available with which to judge the number of times a child perceived the stimuli correctly but responded to these stimuli incorrectly, and vice versa. Performance on Control versus Experimental Tasks Within both Control and Experimental Tasks, asymptote (i.e.that point on the learning curve where the subject may be judged to have passed through i n i t i a l learning stages and is performing consistently at or slightly below the highest level he may be expected to achieve ) was judged to have been reached when learning curve scores did not exceed a particular value during five successive presentations of the sixteen item l i s t s (post-asymptotic t r i a l s ) . Asymptote for linguistic tasks was observed to occur after approximately four sessions. For the non-Unguistic task, asymptote occurred during the f i r s t session. Within each experimental session, subjects were presented with four sixteen item l i s t s . 58. Subjects were exposed to the Cont ro l Task only a f te r the Experimental tasks had been learned. I t i s un l i ke l y that be t te r learning of non-l i n g u i s t i c tasks i s a t t r i bu tab le so l e l y to time of t h e i r presentation s ince i t may be observed that mean scores f o r the learning of nasal ized vowels f a i l e d to r i s e above a low asymptotic l e v e l when presentation of these vowels was continued f o r a considerable number of post-asymptotic sessions. I t was seen that motor a b i l i t y of a l l ch i ld ren i n t h i s experiment was s u f f i c i e n t l y developed f o r performance of the required button press response to almost achieve un i ty on the Control Task. That i s , where ch i ld ren were presented with \"meaningful sounds\", such as a b i c yc l e b e l l , they learned rap id l y and to c r i t e r i o n . Thus differences between performance on Control tasks versus performance or Experimental T asks may be a t t r i bu tab le t o the nature of the discr iminat ions required f o r l i n g u i s t i c mater ia l . For example, acovisMo' \" \" spectra of the four non - l i n gu i s t i c s t imu l i were d i s t i n c t and would be expected to provide ea s i l y d iscr iminable cues f o r l i s t ene r s as to t h e i r sources e.g. s t r i k i n g two s t i c k s y i e l d s an impulse- l ike spectrum wi th energy d i s t r i bu ted at a l l frequencies over a very b r i e f durat ion; the laryngeal v i b ra to r produces a continuous sound of low p i t c h and f a i r l y constant i n t e n s i t y ; a b i c yc l e b e l l , i n comparison, has higher frequency components and begins at a f a i r l y high i n t en s i t y , then gradually f a l l s o f f ; white noise has energy d i s t r i bu ted at a l l frequencies but again i s of continuous rather than t rans ient nature. I t was of great i n te re s t to observe that the ch i ldren ass i s ted t h e i r r e c a l l of these s t i m u l i by auditory \"prompts\"; f o r example, white noise was l abe l l ed as the sound of a b u l l pawing or of somebody smel l ing; s t i c k s s t r i k i n g each other r e ca l l ed the sound of a f i r ec racke r or of c lapping; the buzz was described as the sound a horse makes and the b i c yc l e 59. b e l l was ea s i l y i d e n t i f i e d by a l l subjects. The eight l i n g u i s t i c s t imu l i whose spectra are a l l qu ite s i m i l a r were not as amenable to t h i s process of d i sc r iminat ion by use of overt verba l tags ; i .e . ,no \" ve rba l l y mediated responses\" occurred when non-nasalized vowels were presented to subjects, and only a small number of subjects ass i s ted t h e i r d i sc r iminat ion of nasa l ized vowels by l a b e l l i n g these vowels as more f a m i l i a r items from t h e i r environment. \u00E2\u0080\u00A2 I f scores on the Control Task had, f a i l e d to achieve the l e v e l of scores on the Experimental i'-ask, or i f the slope of the learn ing curve f o r Control s t i m u l i had been equal to or less steep than, that f o r l i n g u i s t i c s t i m u l i , then s i m i l a r performance on such d i s s im i l a r tasks would have suggested that the expected response mechanisms were i n f ac t too d i f f i c u l t . With young, naive subjects, i t was v i t a l to e s tab l i sh motor performance capab i l i t y before i n te rp re t i ng d i sc r iminat ive aspects of s t i m u l i presented. Language Background of Subjects Despite f a i l u r e of the main e f f e c t f o r l i n g u i s t i c background to achieve s i gn i f i cance , the i n te rac t i on between t h i s va r iab le and p a r t i c u l a r responses required of subjects was s i g n i f i c a n t at the one percent l e v e l of confidence. I t could be hypothesized that the bet te r performance of French speaking subjects when presented w i th non-nasalized s t i m u l i i s p a r t i a l l y due to t h e i r greater exposure to non-nasalized vowels s ince these are found not on ly -wi th in t h e i r own d i a l e c t , but a l so i n the Engl i sh d i a l e c t to which they are exposed. Better d i sc r iminat ion learning by French speaking subjects of the non-nasalized vowels as compared to the nasa l ized vowels i s perhaps due to the r e l a t i v e l y l a te appearance of nasa l ized vowels w i th i n the developing French phonological system. Gregoire (1933, as reported by Jakobson 1956.) 60. stated that nasalized vowels appear only after a l l remaining vowels, have been acquired, and generally not until the third year. Thus, more recent acquisitions within the phonological system may be discriminated with greater difficulty than primary' acquisitions. Fricatives and other phonemes acquired later in development, are known to cause greater confusions and substitutions than stop consonants which appear very early in the phono-logical system. The opposition nasal-oral for vowels appears rarely among the world's languages. It was observed that the expected superior learning of vowels by French versus English speaking children failed to occur. If poorer discrinination of Nasalized versus Non/Nasalized vowels by- French speaking children is attributed to relatively late integration of the former into the French speaker's phonological system, i t is of interest to note that lack of familiarity of English-speaking subjects with the Nasalized vowels did rot appear to affect their discrimination-learning. Thus the scores of English-speaking subjects -\" wer\"e not. significantly different from scores of French speaking subjects who were assumed to be somewhat more familiar with Nasalized vowels. It must be considered that the stimuli may have been unfamiliar even for French speaking subjects since they were spoken in isolation and outside of a linguistic context (.See Table 5.). French and English speaking subjects exhibited a greater number of verbally mediated responses (such as \"orange\" for the stimulus /of ; \"Duck noise\" or \"Donald Duck\" for /\u00C2\u00A3/ and \"A\" for faf and f&f .), while respond-ing to Nasalized than to Non-Nasalized Vowels. If i t is conjectured that Nasalized Vowels, have been learned as linguistic entities within the phonological system of the French speaking subjects, then i t is to be expected that: they would use fewer verbal 61. mediations f o r comparable s t i m u l i than would Engl i sh subjects f o r whom the s t i m u l i were probably non-recognizable e n t i t i e s . I f a p r e - l i n g u i s t i c s t ructure ex i s t s (Lenneberg, 1967.) and i n f ac t operates on incoming sound s i gna l s , then supposedly the subject w i l l more r ead i l y assign the sound to a phoneme b i n i n accordance with t h i s s t ructure. An operation of t h i s nature would take place without the necess ity f o r verba l mediation between speech sound input and phoneme recogni t ion. I f , however, the subject has a d i f f e r en t phonological system, and incoming sounds are not r ead i l y assignable to phoneme bins i n terms of the subject ' s e x i s t i n g l i n g u i s t i c system, then the subject i s presumed to base decisions on overt phys i ca l propert ies of the s i gna l rather than on an underlying l i n g u i s t i c s t ructure unknown to the subject, and consequently makes comparisons of these acoust ic propert ies w i th whatever standards he has ava i l ab le . This means that i n many cases, when underlying ' l i n g u i s t i c s t ructure of a s i gna l may not be pass ive ly analysed by the subject, verbal mediation i s employed to a s s i s t h i s d i sc r iminat ion and r e c a l l of unfami l ia r s t imu l i . Age: Subjects at a l l three age leve l s performed s i m i l a r l y on the non-l i n g u i s t i c Cont ro l Task during the f i n a l sess ion (x: 59.95/64 range: 58.6761.1)This f i nd ing suggests that the v a r i a b i l i t y observed with age f o r the Experimental asks may r e f l e c t a hierarchy i n ease of a cqu i s i t i on of the required d iscr iminat ions f o r the three age groups. Intersubject v a r i a b i l i t y on the r e l a t i v e l y simple non - l i n gu i s t i c task was greater f o r younger than f o r o lder subject s , ind icat ing that some considerat ion should be given to a d i f f e r e n t i a l mode of response by ch i ld ren w i t h i n the age range studied. I t i s general ly accepted that at some time between the ages of four to 62. seven years CCA) normal children.acquire the ability to use overt Cor internally generated) verbal commands to guide and. control motor response to stimuli. Development of this- \"directive\" function of language has been chiefly investigated' by Russian psychologists and psycholinguists. Luraa 0-968.) for example, cites an experiment in which a child was instructed to press a rubber bulb held in his hand when a red signal appeared, and to refrain from pressing in the presence of a blue signal. A \u00E2\u0080\u00A2\u00E2\u0080\u00A2 M-Sg \u00E2\u0080\u00A2 year old child was seen to be able to perform as instructed and could coordinate his own affirmative verbal command \"Press!\" with appearance of the red signal, and his own inhibitory verbal command \"Don't press!\" when the blue signal appeared. At 35g years, however, Luria found that the child's verbalizations .to the negative blue stimulus failed to inhibit the pressing response and they even have acted in reverse to enhance the response. It was apparent that, at this earlier age, the child was unable to use the semantic aspect of his verbal command to select the required response, i.e. the inhibition of the bulb press. A s t i l l younger child of 2\ failed even to ccorxlinate his verbal commands with the signal (in this case a single flash of light meaning \"Press!\") and was noted to utter continuous ineffective verbal commands, becoming so engrossed in these utterances that he failed to perform the motor response. At this stage neither the response enhancing (i.e., the \"impulsive\" function) nor the response selection Ci.e. the \".semantic\" function of speech) was sufficiently developed to merit consideration as \"directive functions\" of language. Since the youngest subjects in Luria'sexperiment were five year olds, i t was. expected that present performance by subjects at a l l age levels would be similar since they would at or beyond the age level where.language acquires a directive function. 63. Gilbert (1970 b) was able to train four year old children to perform a four choice auditory discrimination learning task requiring a button press response.. Children in Gilbert's study successfully refrained from responding until after they- had listened to the stimulus item. Turnure (1971) has cited 3 years 9 months (CA) as a c r i t i c a l age in the developing ability- of children to orient to a two-choice visual discrimination task, when distracted with an auditory stimuli. Turnure showed that children at or above this age level exhibited far greater task than non-task orienting. He measured this by- incidence and duration of glances away from the task when a tape either of typing noises or of childrens' songs was played continuously- during the testing period. Turnure interpreted his results as placing^ constraints on. .traditional views'of \"'children's distractibility, and as indicating that existing interpretations of the development of attentive abilities in children based on the notion of an increasing attention span are in need of elaboration or reformulation (1971,-p. 16). Eighty four year old children were administered an \"acoustic vigilance\" 4 task by Locke (1970) in which they were required to detect the presence of a 2000 Hz. signal at 15 dB SL and 189 msec, in duration. Locke found that subjects performed similarly to adults^ i.e.,as task length or as the inter-signal interval was increased, the number of correct detections decreased. More false positives, however, were obtained with the young subjects than with adults and Locke reported his subjective impression that subjects were \"playing\" quite often with, the response button. White (1965) has reviewed a large number of studies dealing with, non-auditory as well as: auditory modalities and he too points, to a different mode of learning for five to seven year olds as compared to children below five years. White credits the' use of verbal mediation 64. between stimulus and response by- the older group for their more \"cognitive\" fashion of response in a discrimination learning situation, and compares this to the simple stimulus-response or \"associative\" response mode of younger subjects. He cautions however, that . \" once cognitive functioning is acquired i t does not obviate associational or S-R learning; certain conditions can s t i l l e l i c i t the more primitive learning process \" (p.193). In the present investigation i t was seen that overall mean proportion of correct response for five-year-olds, was considerably below that for six,, and seven-year-olds, although the rank order of vowels best learned was fairly- consistent across agessas was the rank order of intervowel confusions. It i s of interest to note that a l l non-nasalized vcwels were better learned than the. best learned nasalized vowel} ( 10/ )}by five-year-olds, whereas the six-and seven-year-olds had greater relative difficulty with the non-nasalized vowel / 0 3 / than with, the nasalized vowels /\u00C2\u00A9/ and 111. This indicates the possibility that +nasal caused greater confusion for the younger subjects than for linguistically more advanced subjects. The relatively late appearance of nasalized vcwels in production (reported earlier) may have as a concomitant a relatively late stabilization of their perceptual correlates. Whether these are hypothesized to be kinaesthetic images corresponding to motor movements required to produce each vcwel, or acoustic matches (templates) corresponding to the generalized physical signal of the vcwel, would appear to make l i t t l e difference in explanation of the results obtained-that is , (a) physiologically,, -frnasal requires simultaneous, activation of two resonating cavities, and (b) acoustically, i t derives i t s distinctiveness from a reduced anplitucfe: and broadening .\u00C2\u00A9f F]_ iand other ,.' spectral alterations. In terms of the motor Uaeory of speech 65. perception (Liberman.et a l . , 1967), later acquisitions to the speech pro-duction system Would- have ' , , less well established' kinaesthetic correlates. Thus the'observed result, ' that +nasal is more difficult for younger subjects, would be predictable within .this theory. Vowel \"\u00E2\u0080\u00A2 ' Within the non-nasalized series of vowels, /o/ was best learned by a l l subjects, followed by'the vowels /\u00C2\u00A3 /, /a/ and / ce/. Of the nasalized vowels lol was best learned and in addition was. the vowel most heavily loaded with verbal mediation. No verbally-mediated responses to non-nasalized vowels occurred , . probably- because these occur within the phonemic system of a l l subjects and are thus analy-zable in terms- of an existing phonemic structure. It was hypothesized that the vowels /\u00C2\u00A3/ and /o / would be best learned since their distinctive feature configurations are characterized by presence or absence of both features under consideration. Thus /\u00C2\u00B0/ carries the DFs +back- and +round, and /\u00C2\u00A3/ carries the DFs-back and -round. Trubetzkoy (1969) has indicated, however, that every multiclass vowel system must have a i T B x i m a l l y dark and a nsximally clear class of timbre (correlates of either tongue position or l i p rounding or both). Rounded vowels are considered to be darker than unrounded vowels, and front vowels are considered to be clearer than back vowels. Thus, according to Trubetzkoy(1969) \" one should actually not speak of back rounded and front unrounded vowels, but only of maximally dark and maximally clear vowels \"(p. 98). Within the French phonemic system, /o/ and /\u00C2\u00A3/ correspond to Trubetzkoy's descriptions \"maximally- dark\" and \"maximally- clear\" vowels respectively, and thus i t could be predicted that they- would be most readily discriminated within a series of vowels; this was indeed the case. 66. That the vowel I ol was, hest learned w i th i n the Nasal ized ser ies i s probably- a t t r i bu tab le to the great number of verba l mediations, used i n i t s learning which i n turn i s perhaps due to the kLnaesthetic and v i s u a l images t h i s rounded vowel might evoke. I t i s d i f f i c u l t t o i n te rp re t the s p e c i f i c mediations employed by--subjects i n r e c a l l of nasal ized vowels, or to expla in the equal d i f f i c u l t y - associated with the learning of these vowels by both French and English, speaking ch i ld ren. When subjects were questioned concerning mediations f o r a p a r t i c u l a r vowel, t h e i r r ep l i e s were of the va r ie ty \"That i s what the man (p ig , duck . . .) i s saying\". The vowel /e / most r ead i l y e l i c i t e d the response \"That ' s a duck t a l k i n g \" and i s indeed s i m i l a r to the usual English, language onomotopoeia f o r sounds made by-ducks ( i . e . \u00C2\u00A3we? ] ) . Other mediations, such as \"H\",[etJ\" '] o r l j l f o r /6 / are not so easily- explained, although one subject mediated responses to t h i s vowel w i th the overt voca l i z a t i on \" I t says orange\" - perhaps an assoc iat ive response to the sound beginning the f a m i l i a r word \"orange\". \"R\" [ a r ] was the verba l mediation provided by one subject f o r a l l occurences of the vowel / ce /. The noticeable predominance of / r/ mediations f o r nasa l ized vowels may perhaps be due to the s i m i l a r mode of production of l i q u i d s and nasal consonants, i . e . simultaneous presence of an obstruction to a i r f low and a means f o r i t s escape. Since subjects tended to character ize nasa l ized vowels as. s im i l a r to p i g ta l k , (most often described by Eng l i sh speakers as nasal i n q u a l i t y , e.g. [ojrjk] ) ? i t i s poss ib le that the ch i ld ren perceived nasa l ized vowels as. though, they consisted of a vowel fol lowed by- the nasal consonant /n/ , and that they then associated /\"/ wi th the l i q u i d having a s im i l a r place of ' a r t i c u l a t i o n ( i . e . , / r / ) . 67. Distinctive Feature Presence of the distinctive feature tnasal resulted in lower discrimination learning scores for a l l subjects across a l l vowels. Apart from +nasal, no distinctive feature appeared to be better or less well learned by subjects. Within the Non^-Nasalized series (for example, 1\u00C2\u00B0 I and / e /,) the best learned vowels, can be characterized by exactly opposite values for the features back and round..; i.e.,/ 0/ is characterized by.. [+back,- tround] and /.\u00C2\u00A3\u00E2\u0080\u00A2'/\u00E2\u0080\u00A2 is .charcterized-by [-back, -round] . If subjects were able to learn one feature better than others, then i t is to be expected that vowels charcterized by the same value for this feature (i.e., + round) would be more readily learned. Thus within the series of vowels studied i t would be expected that / / would be as well learned as / 0/ because both are characterized by +round. No explanation in terms of better learning of a single distinctive feature appears to be able to account for the f a c i l i t y with which /\u00C2\u00A3>-/ and /o / were discriminated when compared with the discrimination of A / and /03/. Chomsky and Halle (1968) propose a hierarchy to describe the order in which distinctive features are available for marking within a vowel system. Their proposed hierarchy for availability of distinctive features is as follows: i.e. \"No vowel segment can be marked for the feature round unless some vowel segment in the system i s marked for the feature1 high' \" (p. 410). The results of this experiment do not provide evidence for a hierarchy in the perception of distinctive features. Instead, the vowels best back round low 68. learned i.e. AV and /\u00C2\u00A3/ have the same value for the features back and round (/\u00C2\u00B0/ i s +back, tround, and /e I i s -back, -round )5 suggesting that vowels totally- distinctive one from the other in terms of DFg are more easily learned and discriminated from vowels only partially distinctive .in terms of, DFs. -., . .\u00E2\u0080\u00A2 .... The least well learned of the Non-Nasalized vowels, /as /, is characterized by -back, +round, and is therefore only .partially distinctive, in terms of Distinctive Features, from [of and /\u00C2\u00A3/. Further /ce / i s most often confused with /e/ within the Non-nasalized series. Evidence of a frequent confusion of this type might perhaps be presented as supporting a hierarchy of perception of features; i.e.,since /ce / and /e/ are alike in their specification -back, differing only in their opposite values for the feature round, i t might indicate that +back is indeed analysed before +_round. According to Chomsky and Halle, a vowel marked +back, +high,(as are both /-/ and /ce/),may then be marked for the feature \"round\". These two most often confused vowels differ only in their values for the feature round and would therefore, according to a hierarchy theory, be expected to be most often confused. Within the Nasalized series, however, a different result appears: & / and /6g/ are confused most often and yet are characterized by totally different values for the features back and round. The vowel /P/ is[+back, -round]whereas the vowel / is[-back, tround.]- It is not possible, in this case,to speculate as to which feature is best discriminated. In summary, i t would appear that a suitable explanation for the results obtained in this, investigation would be that the best discrimination learning occurs in the presence of a combination of features, or feature \"bundle\".\u00E2\u0080\u00A2 Empirical support has not been provided for the hypothesis that 69. distinctive features correspond to the structural elements in speech perception, nor for the hypothesis, of a hierarchy- in the perception of the distinctive features. Instead, in this study, as well as in-that recently reported by -Graham and House ( 1970), no classification system truly served to distinguish best from least well learned.vcwels. CHAPTER 5 CHAPTER CONCLUSIONS 70. The re su l t s of t h i s experiment ind icate that ch i ld ren of f i v e to seven years C.A. are able to learn a button press response and to perform a four choice auditory d i sc r iminat ion learn ing task. Chi ldren at the age l eve l s studied can learn to associate a p a r t i c u l a r auditory stimulus wi th one of four d i sc rete motor responses,even though ch i ld ren i n the youngest group (C.A. 5.2 years) show greater v a r i a b i l i t y i n response. No analys i s of experimental re su l t s i n terms of D i s t i n c t i v e Features Theory proved success fu l , except that +nasal proved more d i f f i c u l t to learn than d id -nasa l . Best learned vowels appeared to be those ireximally d i s t i n c t i v e from others w i th in the same ser ies i n terms of t h e i r o v e r a l l features matr ix. For example, vowels character ized by one feature, e.g. +round, d id not appear to be bet te r learned than vowels character ized by -round, but a vowel character ized by[+round. +back]was r ead i l y discr ini inated from the vowel wi th opposite feature s pec i f i c a t i on :I -round, -back](/o/ and /e/ 4 r e spec t i ve l y ) . This experiment does not appear to support the p lea f o r a hierarchy of d i s t i n c t i v e features operating i n speech perception at these age l e v e l s . Regardless of age, subjects learned to d iscr iminate ce r ta in vcwels bet ter than others, but d i d not appear to learn to d i scr iminate one feature bet te r than another, although the youngest subjects learned a l l the non-nasalized vowels be t te r than the best learned nasa l ized vowels suggesting that the feature +nasal may r e f l e c t a p a r t i a l hierarchy f o r the perception of d i s t i n c t i v e features. Any hierarchy i n perception would presumably be re l a ted to age of productive mastery of the feature of the c h i l d such that nasa l ized vowels, when present i n a phonemic system, would be expected to be learned much l a t e r than t h e i r non-nasalized equivalents. 71-The non-significant discrimination learning scores for predominantly French versus, predominantly- English-speaking children can be attributed in part to the exposure of French-speaking subjects to the media and prevalence of an English-speaking majority in school and in part to the fact that vowels chosen as stimuli were presented in isolation. The fact that predominantly French-speaking children performed similar-ly to predominantly English-speaking children on non-nasalized vowels is perhaps due to the existence of these vowels in their own phonemic system, as well as in the English phonemic system to which they had been exposed. Implications for Future Research It i s apparent that children of five to seven years C. A. are capable of associating a motor response,such as the button-press employed in this experiment, with variations in an acoustic signal presented to them. Younger subjects,(i.e. ,children of five years C.A) did not provide responses as reliable as those of the older children, although the variation observed was apparently a matter of degree, rather than of kind, since features and vcwels best learned were constant across a l l age groups studied. It is of interest that Nasalized vowels, presented in isolation, were the source of considerable difficulty even for the predominantly French-speaking subjects, who, i t might be expected, would recognize these as linguistic entities within their own phonemic system. Further research using predominantly- French speaking children as subjects should therefore give consideration to presentation of nasalized vowels within a linguistic context. Under such conditions subjects could be expected to exhibit fewer verbally- mediated responses, and use appropriate phoneme \"bins\" for sorting vowels. If vcwels could be presented as minimal pairs, information on discrimination learning of various distinctive features could be obtained and compared with results of the present experiment. 72. BIBLIOGRAPHY Brown, R., and Bellugi, U. (1964) \"Three Processes in the Child's Acquisition of Syntax.\" Harvard Educational Review 34 133-151 Chistovich, L.A. (1962) \"Temporal Course of Speech Sound Perception\" (Paper H.18). Proc. 4th International Congress ori Acoustics. Copenhagen, 1962 Chomsky, N. (1965) Aspects of the Theory of Syntax. M.I.T. Press, Cambridge, Mass. Cole,R.A., Haber,R.N. and Sales,B.D.(1968) \"Mechanisms of Aural Encoding: I. Distinctive Features for Consonants.\" Perception and Psychophysics 3_, 281-284. Denes, P.B. (1963) \"On the Statistics of Spoken English\" J. Acoust. Soc. Amer. 3_5, 892-904. Eimas, P.D., Siqueland, E.R., Jusczyk,P., and Vigorito, J. (1971) \"Speech \u00E2\u0080\u00A2 Perception in Infants.-\" Science 171 , 303-306 Eisenberg,R. (1970) \"The Organization of Auditory Behavior\" J. Speech Hearing Res. 13, 453-471 Edmonston,W. (1963) Laradon Articulation Scale, Revised (Form B). Western Psychological Services, Los Angeles Fant, C.G.M. (1962) \"Descriptive Analysis of the Acoustic Aspects of Speech\" Logos 5_, 3-17. Fodor, J. and Bever, T. (1965) \"The Psychological Reality of Linguistic Segments\" J. Vbl. Learning Vbl. Behav. 4, 414-420. Friedlander, B.Z. (1970) \"Receptive Language Development in Infancy : Issues and Problems.\" Merrill-Palmer Quarterly of Behavior and Development 16, 7-51 Gibson, E.J. , and 01um,V. (1960) \"Experimental Methods of Studying Perception in Children.\" Mussen, P. (Ed.) Handbook of Research Methods in Child Development. Wiley and Sons Inc., N.Y., 311-373 Gilbert. J.H. (1970 a) \"The Learning of Speechlike Stimuli by Children\" J. Exp. Child Psychology, 9_, 1-11. Gilbert, J.H. (1970 b) \"Vowel Productions and Identification by Normal and Language Delayed Children\" J. Exp. Child Psychology, 9_, 12-19. Graham, L.W., and House, A.S. (1971) \"Phonological Oppositions in Children: A Perceptual Study.\" J. Acoust. Soc. Amer. 49, 559-568 73. Greenberg, J.H. and Jenkins, J. (1967) \"Studies i n the Psychological Correlates of the Sound System of American English.\" Jakobovits, L. and Miron, M. (Eds.), Readings i n the Psychol- ogy of Language. Prentice-Hall Inc., Englewood C l i f f s , N.J. Gregoire, A. (1963) \"L'apprentissage deJia parole pendant les deux premieres annees de l'enfance.\" Journ. de Psychol., 30 Halle, M. and Stevens, K. (1962) \"Speech Recognition: A Model and a Program for Research.\" IRE Transactions on Information Theory 17, 155-159 House, A.S., Stevens, K.N., Sandel,T.T. and Arnold,J.B. (1962) \"On the Learning of Speechlike Vocabularies.\" J.Vbl.Learn.Vbl.Behav. 1, 133-143 Jakobson, R. and Halle, M. (1956) Fundamentals of Language. Mouton and Co., The Hague . Jakobson, R. , Fant, C.G.M., and Halle, M. (1957) Preliminaries to Speech . Analysis: The Distinctive Features and Their Correlates. M.I.T. Press, Cambridge, Mass. Jakobson,R. (1968) Child Language, Aphasia and Phonological Universals. Mouton and Co., The Hague ( Translated by A.R. Keiler) Kimble, G.A. (1961) Hilgard and Marquis' Conditioning and Learning, 2nd edition, Appleton-Century-Crofts, New York Kozhevnikov, V.A. and Chistovich, L.A. (1965) Speech, Articulation and Perception . Joint Publications Research Service, Washington, D.C., JPRS 30 543 Lane, H.L. and Moore, D.J. (1962) \"Reconditioning of a Consonant Discrim . ination i n an Aphasic: An Experimental Case History.\" J. Speech Hearing Dis. 27, 232-243. Lane, H. (1965) \"The Motor Theory of Speech Perception: A C r i t i c a l Review.\" Psychol. Rev. 72, 275-309. Lawson, R. (1960) Learning and Behavior. Macmillan Co., New York Lenneberg, E.H. (1967) Biological Foundations of Language. John Wiley and Sons Inc., New York Leopold,W.F. (1953) \"Patterning i n Children's Language Learning.\" Saporta, S. (Ed.), Psycholinguistics- A Book of Readings. , Holt, Rinehart and Winston, New York Liberman, A.M., Cooper, F.S., Shankweiler, D.P. and Studdert-Kennedy,M. (1967) \"Perception of the Speech Code.\" Psychol.-- Rev.74, 431-461 7M-. Lieberman. P. C1965) \"On the Acoustic Basis of the Perception of Intonation by Linguists.\" Word 21, 40-54. Locke, J.L. (1970) \"Acoustic Vigilance Behaviour in Four-Year-Old Children.\" Psychon. Sci. 2, 111-112. Luria, A.R. (1968) \"The Directive Function of Speech in Development and Dissolution, Part I.\" in Oldfield, R.C., and Marshall, J.C. (Eds.) Language, Penguin Books Ltd., Harmondsworth McNeill, D. (1966) \"The Creation of Language.\" i n 01dfield,R.C., and Marshall, J.C. (Eds.) , Language, Penguin Books Ltd., Harmondsworth Menyuk, P. (1968) \"The Role of Distinctive Features in Children's Acquisition of Phonology.\" J.Speech Hearing Res.. 11 138-146. Menyuk, P. (1969) Sentences Children Use. (Research Monograph #52) M.I.T. Press, Cambridge, Mass. Miller, G.A. , and Nicely, P.E. (1955) \"Perceptual Confusions Among Consonants.\" J. Acoust. Soc. Amer. 27, 301-315 Miller, G.A. (1965) \"Some Preliminaries to Psycholinguistics.\" 'Jakobovits, L. and Miron, M. (Eds.) Readings in the Psychology of Language. Prentice-Hall Inc., Englewcod C l i f f s , N.J., 1967,172-179. Miller, G.A. (1962) \"Decision Units in the Perception of Speech.\" IRE Transactions on Information Theory IT-8, 81-83 Nakazima, S.A. (19620.. \"A Comparative' Study of the Speech Development of Japanese and American English in Childhood (1) -a Comparison of the Developments of Voices at the Prelinguistic Period.\" Studia Phonologica 2 27-46 Peters, R.W. (1963) \"Dimensions of Perception for Consonants.\" J.Acoust. Soc. Amer. 35, 1985-1989. Peters, R.W. (1967) \"Perceptual Organization for Speech and Other Auditory Signals.\" Aerospace Medical Research Labs Report. AMRL-TR-68-31 , 1-69 Poole, J. (1934) \"Genetic Development of Articulation of Consonant Sounds in Speech.\" Elementary English Review 11, 159-161 Powers, M.H. (1957) \"Functional Disorders of Articulation-Symptomatology and Etiology.\" in Travis, L.E. (Ed.), Handbook of Speech Pathology, Appleton-Century-Crofts, New York 75. Pronovost, W. ,and Dumbleton, C. (1953) \"A Picture-type Speech Sound Discrijiunation Test.\" J. Speech Hearing Dis. 18, 258-266 \u00E2\u0080\u00A2 Reese, H.W.,and Lipsitt, L.P. (Eds.) Experimental Child Psychology. Academic Press, New York, 151-194 Saporta, S. (1955) \"Frequency of Consonant Clusters.\" Language 31, 25-30 Saporta, S. (Ed.), (1961) Psycholinguistics-A Book of Readings., Holt, Rinehart' and Winston, New York Schiefelbusch, R.L. , and Lindsey ,M.J. (1958) \"A New Test of Sound Discrimination.\" J.Speech Hearing Dis. 23, 153-159 Schvarchkin, N. (1966) \"Development of Phonemic Speech Perception in Early Childhood.\" Smith, F. and Miller, G. (Eds.) (1966) The Genesis of Language-A Psycholinguistic Approach., M.I.T. Press, Cambridge, Mass. Spence, K.W. (1936) \"The Nature of Discrixoination Learning in Animals.\" Psychol. Rev. 43, 427-449 Spence, K.W. (1937) \"The Differential Response in Anemals to Stimuli Varying within a Single Dimension.\" Psychol. Rev. 44, 430-444 \u00E2\u0080\u00A2 Spence, K.W. (1942) \"The Basis of Solution by Chimpanzees of the Intermediate Size Problem.\" J. Exp. Psychol. 31, 257-271 Stevens, K.N. (1971) \"Role of Formant Transitions in the Voice-Voiceless Distinction for Stops.\" Quarterly Progress Report 101, M.I.T. Research Laboratory of Electronics Templin,M.C. (1957) Certain Language Skills in Children-Their Development and Interrelationships. (Institute of Child Welfare Monograph 26.) University of Minnesota Press, Minneapolis Templin, M.C.aiid Darley, F.L. (1960) \"The Templin-Darley Test of Articulation.\" Bureau of Educational Research and Service, Iowa City, Iowa Trubetzkoy, N.W. (1969) Principles of Phonology., University of California Press, Berkeley (Translated from GrundzUge der Phonologie, Vandenhoeck and Ruprecht, Gottingen, 1958, by C.A.M. Baltaxe) Turnure, J.E. (1971) \"Control of Orienting Behaviour in Children Under Five Years of Age.\" Developmental Psych. 4, 16-24 Wellman, B.L., Case, I.M., Mengert, I.G. and Bradbury, D.E., (1931) \"Speech Sounds of Young Children.\" University of Iowa Studies in Child Welfare _5, 2 White, S.H. (1965) \"Evidence for a Hierarchical Arrangement of Learning Processes.\", Lipsitt, L. and Spiker, C. (Eds.) Advances in Child Development and Behavior. 2_, Academic Press, New York, 187-216 76. Wickelgren, W.A. Q965) \"Distinctive Features and Errors i n Short-Term Memory for English Vowels.\" J. Acoust. Soc. Amer. 38, 583-588 Wickelgren, W.A. (1965) \"Acoustic Similarity and Intrusion Errors i n Short-Term Memory.\" J. Exp \u00E2\u0080\u00A2\u00E2\u0080\u00A2.-\u00E2\u0080\u00A2\u00E2\u0080\u00A2Psychol. 70, 102-108 Wickelgren,. W.A. (1966) \"Distinctive Features and Errors i n Short-.. Term Memory for English Consonants.\" J. Acoust. Soc. Amer. 39, 388-398 Winitz, H. and Bellerose, B. (1963) \"Effects of Pretraining on Sound Discrimination Learning.\" J. Speech Hearing Res. 6_ 171-180 Winitz, H. andr'.Preisler, L. \"Effect of Distinctive Feature Pretraining i n Phoneme Discrimination Learning.\" J. Speech Hearing Res 10 515-530 Winitz, H. (1969) Articulatory Acquisition and Behavior. Appleton-Century-Crofts, New York Appendix A Instructions for French Subjects \"Ecoute. Nous allons jouer ensemble.,- Tu vas entendre un homme qui dira des sons etranges. C'est a toi de decouvrir quelle est l a maison de chaque son. Voila. les maisons des sons ' (Experimenter indicates buttons on response panel) Chaque fois que tu entendras un son, tu dois appuyer sur un bouton. Si tu a(ura)s choisi le bon bouton, l a lampe audessus du doigt qui appuie s'allumera. Si elle ne s'allume pas, ca voudra dire que tu auras choisi le mauvais bouton, mais je te ferai voir quelle est la bonne maison du son. I I n'y a qu'un seul son dans chaque maison. Essayons. . Ecoute, puis choisis un bouton\". "@en . "Thesis/Dissertation"@en . "10.14288/1.0101969"@en . "eng"@en . "Audiology and Speech Sciences"@en . "Vancouver : University of British Columbia Library"@en . "University of British Columbia"@en . "For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use."@en . "Graduate"@en . "Discrimination learning of nasalized and non-nasalized vowels by five-, six-, and seven-year-old children."@en . "Text"@en . "http://hdl.handle.net/2429/34374"@en .