Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Perception of fricatives in an AX paradigm by children 3-4 1/2 years old Barker, Dorothy Kathleen 1982

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-UBC_1982_A6_7 B36_7.pdf [ 3.35MB ]
JSON: 831-1.0095179.json
JSON-LD: 831-1.0095179-ld.json
RDF/XML (Pretty): 831-1.0095179-rdf.xml
RDF/JSON: 831-1.0095179-rdf.json
Turtle: 831-1.0095179-turtle.txt
N-Triples: 831-1.0095179-rdf-ntriples.txt
Original Record: 831-1.0095179-source.json
Full Text

Full Text

PERCEPTION OF FRICATIVES IN AN AX PARADIGM BY CHILDREN 3-4 1/2 YEARS OLD by DOROTHY KATHLEEN BARKER B.A., University of Br i t i s h Columbia, 1976 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE THE SCHOOL OF AUDIOLOGY AND SPEECH SCIENCES We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA October, 1982 in In presenting t h i s thesis i n p a r t i a l f u l f i l m e n t of the requirements for an advanced degree at the University of B r i t i s h Columbia, I agree that the Library s h a l l make i t f r e e l y available for reference and study. I further agree that permission for extensive copying of t h i s thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. I t i s understood that copying or publication of t h i s thesis for f i n a n c i a l gain s h a l l not be allowed without my written permission. The University of B r i t i s h Columbia 1956 Main Mall Vancouver, Canada V6T 1Y3 DE-6 (3/81) - i i -ABSTRACT In the f i e l d of child language acquisition, i n particular phono-logical acquisition, many have queried the role played by perception. The present study was undertaken to exairiine the perception of some speech sounds by children i n the process of phonological development. Perception of the group of sounds known as fricatives was examined i n a group of subjects aged 3;0 to 4;11. Pairs of nonsense syllables were presented to eight sub-jects i n an AX paradigm. Results were examined for each fricative pair i n terms of mean error rate. Some discussion of individual subjects was also included. Results showed that children find i t more d i f f i c u l t to discriminate between some pairs of fricatives than others. In particular, the three voiced/voiceless minimal pairs: [s»-z9, ^*~y* r fd-ve] were found to be significantly more d i f f i c u l t to discrjurdnate than other pairs of fricatives. Findings were for the most part i n agreement with the results of other similar studies. In addition, methodological problems inherent to the nature of the investigation were encountered and discussed. - i i i -TABLE OF CONTENTS Tit l e Page ABSTRACT i i TABLE OF CONTENTS i i i LIST OF TABLES V ACKNOWLEDGEMENT v i 1. INTRODUCTION 1 2. LITERATURE REVIEW 4 2.1 Speech perception - theory 4 2.2 Speech perception - research 6 2.3 Theories of phonological development 10 2.4 Speech sound-discrimination studies 14 2.5 Discussion 26 3. OBJECTIVES 31 3.1 Original Objectives 31 3.2 Revised Objectives 32 4. METHOD 34 4.1 Subjects 34 4.2 Stimuli 34 4.3 Procedure 35 5. RESULTS 42 5.1 Group Results 42 5.2 Results for individual subjects 43 - i v -6, DISCUSSION 53 6.1 Summary of Results 53 6.2 Error Rate 53 6.3 Distinctive Features 55 6.4 Experimental Paradigm 56 6.5 Conclusion 58 - v -LIST OF TABLES Table Page 1. Fricative pairs employed i n the listening test 38 2. The listening test - AX paradigm 39 3. The listening test - ABX paradigm 40 4. Training items used 41 5. Mean number of errors for each of the fricative pairs expressed in percentage 47 6. Collapsed results for 105 t-tests 48 7. Fricative pairs which most often produced st a t i s -t i c a l l y significant greater number of errors shown with pairs that they were not significantly different from ...49 8. Sutrmary of results of t-tests comparing S pairs and D pairs 50 9. Groups of fricative pairs i n terms of distinctive features 51 10. Overall error rate for each subject 52 A(^OWEiFIX3F2ffiNT I would like to extend my sincerest thanks to a l l those who helped i n the completion of this thesis. My delightful subjects for their participation. John Gilbert for his i n f i n i t e patience. Laura, Karen, and mom for typing. Doug, my family and friends for their continued support and encouragement. CHAPTER 1 1.0 INTRODUCTION Perception i s an action by which the mind refers i t s sensations to an external object as the cause. While sensation takes place at the peripheral end organs, perception implies higher level processing, and integration of sensations with past experience. When investigating the child's acquisition of language, both auditory perception and linguistic perception are important. Auditory perception refers to the processing of any auditory stimulus, whether speech or non-speech, while linguistic per-ception implies the realization that the auditory stimulus originated i n the human vocal tract and carries meaning i n that individual's language system. This study w i l l be concerned primarily with phonological percep-tion, defined by Barton (1976) as "the classification of speech into the minimal units which signify meaning differences" (p. 2), i n the child i n the process of language acquisition. In the f i e l d of child language acquisition i n general, and phono-logical acquisition i n particular, many have queried the role of perception in the acquisition of phonology. Questions have been raised about the relationship between perception and production. Does perception f a c i l i t a t e production or vice versa? Is the child's phonological system complete be-fore he begins to say his f i r s t words, or does perception follow a develop-mental pattern similar to production? What i s the nature of the child- si: repre-sentational system and how does i t relate to the adult's representational system? Phonology i s the study of the sound system of a language. Any phonological system i s composed of two distinct levels, the phonetic level - 2 -- referring to the actual production of speech sounds, and the phonemic level - referring to native speaker's underlying representation of those elements i n the phonetic output which convey meaning. Not a l l phonetic output conveys meaning, speech i s highly redundant. The phonetic level and the phonemic level of a language's sound system, are related by rules re allowable sequences, assimilations, redundancies, etc. The child acquiring the sound system of his language not only learns to produce speech sounds, but also develops an underlying representation, in other words, acquires both phonetics and phonology. "Learning to produce and recognize a wide range of sounds, i s not the same as learning the contrasts between sounds which convey d i f f e r -ences i n meaning" (Moskowitz 1975 p. 141) As Moskowitz suggests, learning contrasts between sounds which produce differences i n meaning i s one aspect of the development of phonology. Phonology emerges i n conjunction with other aspects of developing language grammar, syntax, and above a l l semantics; and language emerges in conjunc-tion with other aspects of the developing c h i l d — sensory, motor, physiolog-i c a l , and cognitive. As the child interacts with his environment, accrues experience and reaches certain maturational levels i n terms of physiological, social and cognitive functioning, his phonological system w i l l develop i n a reg-ular progression toward that of the adult. Research i n child development, speech perception and phonological development, adds to our understanding of the role of perception i n the acquisition of phonology. Studies concerned specifically with the child's development of the perception of phonological segments are fewer, and often more questions are raised rather than answered. The present study hopes to - 3 -investigate further the role of perception i n the acquisition of phonol-ogy. 4 -CHAPTER 2 2.0 LITEPATUPE REVIEW 2.1 SPEECH PERCEPTION - THEORY A theory of speech perception should ultimately account for the recognition of certain auditory signals as speech, the processing of those signals from an i n i t i a l frequency intensity duration analysis to the psycho-logical reality of the phoneme, and should include some reference to the acquisition of speech perception by the child. "The general objectives of a theory of speech perception are to describe the process whereby an acoustic signal i s decoded i n linguistic units." (Stevens and House 1972 p. 9) Various models of speech perception have been proposed, which view speech perception i n different ways and may have different goals i n inind. Most would agree that a l l auditory signals undergo i n i t i a l periph- eral processing i n terms of frequency, intensity and duration, and that this i n i t i a l processing i s the same for any auditory signal, whether i t i s speech or nonspeech. Stevens and House (1972) suggest that evidence from studies on the sensory systems of various animals shows that " f a i r l y complex processing takes place peripherally" (p.48). Disagreement arises as to the output of peripheral analysis and the nature of further analysis. An acoustic theory of speech perception proposed by Fant (1962), views speech perception as a passive process. More recent theories such as the motor theory of speech perception, and a refinement of that theory; analy-sis by synthesis, regard speech perception as an active process. The motor theory evolved out of work at Haskins Labs, dealing -5-mainly with the perception of stop consonants, and the finding that they are perceived categorically. Subjects could only discriminate as many items as they could identify. Discrimination of stop consonants varying only i n voice onset time, was found to be very good i f the stJUTtuli were identified as two different phonemes, but poor i f the stimuli were identi-fied as belonging to one phoneme. In most sensory systems, perception takes place along a continuum, and subjects can discriminate many more items than they can identify. The i n i t i a l discovery of categorical perception was therefore thought to be speech specific. The point at which the phoneme boundary occurred when stop consonants were perceived categorically corres-ponded to a point of change i n the production of the sound. The motor theory defines speech perception physiologically; perception i s dependent on articu-latory rules. Speech perception i s possible by recourse to the set of commands to the articulators necessary to produce the sound perceived. The analysis by synthesis model of speech perception (Stevens and House 1972) i s a more refined version of the motor theory. Analysis by synthesis was originally developed as a speech synthesis model. Analysis by synthesis has i t s roots i n distinctive feature theory. The model " i s based on the premise that there exist close ties between the processes of speech production and speech perception, and that there are components or operations that are common to both processes" (Stevens and House 1972 p. 51). In order to identify speech sounds the listener must have knowledge of acoustic^articulatory correlates, i.e. must have knowledge of articulatory commands necessary to produce a certain sound i n order to perceive that sound. The listener must have some sort of "catalogue" of the acoustic-articulatory - 6 -correlates. Stevens and House maintain that the child adds to his/her catalogue, as he/she begins to utter sounds and associate the acoustic output with articulatory commands. They would argue therefore that production fa c i l i t a t e s perception, the child learns to perceive by being able to produce. Research examining the child's a b i l i t y to produce and perceive speech sounds suggests that perceptual a b i l i t i e s develop i n advance of productive a b i l i t i e s . Christine Thorpe (1977) suggests that "the impli-cation i s that children must be able to perceive phonetic information i n a segment without relying upon their a b i l i t y to articulate the same seg-ments. Therefore the nul l hypothesis i s that the analysis by synthesis model of speech perception cannot be valid for the child" (..pp. 18,19). Recent research i n the f i e l d of speech perception reveals evidence that may offer an explanation of the child's use of the analysis by synthesis model. 2.2 SPEECH PERCEPTION - RESEARCH The discovery of categorical perception of stop consonants by researchers at Haskins Lab was originally thought to be speech specific, and therefore to be a major breakthrough i n understanding speech perception. Replication of this finding with infants was originally thought by some re-searchers to lend greater support to the hypotheses that categorical percep-tion was speech specific and perhaps even innate. Using a high amplitude non-nutritive sucking paradigm, Eimas, Siqueland, Jusczyk and Vigorito (1971) exairuned discrimination of stop consonants /b/ and /p/ i n infants, one to four months of age. Synthetic - 7 -speech sounds were tested. A baseline rate of high amplitude sucking was established, and an.increase i n rate upon presentation of a novel stimulus was taken as evidence of <±Lscrimination. The two stimuli to be discriminated either lay on opposite sides of an adult phoneme boundary, or both were within a phonemic category. Results showed that the infants discriminated the stop consonants categorically, with a phoneme boundary similar to the adults. "The implication of these findings i s that the means by which the categorical perception of speech, that i s perception i n a linguistic mode, i s accomplished may well be part of the biological makeup of the organism and moreover that these means must be operative at an unexpectedly early age." ( Eimas et aL 1971 p. 91) The authors assume f i r s t , that categorical perception i s speech specific, and second, that i t i s perception in a linguistic mode. More recent re-search has revealed evidence against these two assumptions. Categorical perception has been demonstrated for nonspeech stimuli, and categorical perception has been demonstrated i n animals. Miller, Weir, Pastore, Kelly and Dooling (1974) investigated "the question of whether categorical perception i s a unique perceptual mode used by humans when listening to speech or whether categorical perception may be a general property of sensory behaviour " (p. 410). Eight subjects were tested on their a b i l i t y to discriminate noise-buzz stimuli varying i n noise-lead times, and previously labelled by the subjects as "noise" and "no-noise". The discriinination task was presented in an oddity paradigm. Stimuli were presented i n groups of three, two identical and one different, and subjects chose which was the odd stimulus. Results were found to be similar to those obtained for stop consonants vary-- 8 -ing i n voice onset time. Based on the criteria, necessary for categorical perception; "(1) There should be "peaks" or regions of high discriminabil-i t y i n the discrimination function; (2) There should be "troughs" or reg-ions where discriminability i s near chance; and (3) the peaks and troughs should correspond to the shape of the identification functions, peaks should occur at the identification boundaries and troughs within categor-ies" (Studdert-Kennedy et al. 1970 as cited by Miller et aL 1974 p. 412), Miller et al . concluded that their noise/buzz sequences were perceived categorically. Miller et al. suggest that categorical perception i s re-lated to interactions between the constant part of the stimulus and the component that i s varied. While emphasizing the major importance of the stimulus configuration i n categorical perception, Miller et al. do not rule out effects of training, experience, memory etc. "For example, i t i s known that the boundary between voiced and voiceless categories of plosive conson-ants depends on the listener's language and the details of synthesis" (Abramson and Lisker, 1970, Williams, 1974, Liberman et al . 1958, cited by Miller et a l . 1974 p. 416). They concluded that "categorical perception of sound i s not unique to speech and ... i t may be a general property of sensory behaviour" (p. 410). Other researchers came to similar conclusions at the same time as Miller et a l . Cutting and Rosner (1974) demonstrated i n two experiments that rise time i n both speech and music can produce categorical perception, and that "plucked" and "bowed" notes are perceived categorically. These results suggest "that certain aspects of phonetic coding may be intimately related to the coding of naturally occurring nonlinguistic sounds" (Cutting and Rosner 1974 p. 569). Categorical perception of speech sounds was demonstrated i n - 9 -chinchillas, by Kuhl and Miller (1975). I t was decided to study the chinchilla, because i t s auditory system functions i n a way similar to man's, but i t would have no "phylogenetic history of phonetic know-ledge" (p.69). Two experiments were performed. In the f i r s t , 4 chinchillas were trained to respond differentially to / t / /d/ i n a CV context, with a variety of V s and a variety of speakers. The animals were able to generalize their learning to new vowel contexts, and to synthesized speech. In a second experiment, synthesized stimuli were varied i n voice onset time and presented for identification as either /ta/ or /da/. The stimuli were presented to chinchillas, and to English speaking adults, and remarkably similar "phonetic boundaries" resulted. Kuhl and Miller con-cluded that "speech-sound oppostions were selected to be highly distinc-tive to the auditory system" (p. 72). Categorical perception of nonspeech stimuli, and speech perception by animals, suggest that categorical perception i s a property of sensory systems i n general and has a psychophysical basis, and that phonemes were so classed, because of properties of the auditory system. I t was previously suggested that the child acquiring a phonolog-i c a l system could not be using a process such as analysis by synthesis in speech perception, because there was evidence to suggest perceptual ca-pabilities were in advance of productive capabilities. Speech perception research indicates that the auditory system i s innately equipped to handle complex processing of both speech and nonspeech stimuli. Prior to complete development of his phonological system, the child could be relying on i n -formation supplied by the auditory system. Infant speech perception i s described by David Ingram (1976a) as a nonlinguistic sensorimotor task. - 10 -When the child moves to linguistic perception, the perception of contrasts with meaning attached, the child goes through what Piaget has called "decal-age", ... the child has to relearn the a b i l i t y at the new level of develop-ment" (Ingram 1976a p. 22). 2.3 THEORIES OF PHONOLOGICAL DEVELOPMENT A theory of phonological development must meet several c r i t e r i a . Every language has a phonological system made up of a series of distinctive oppositions. A theory of phonological development must account for the u l -timate acquisition by the child, of a l l characteristics of the adult system. The theory must be compatible with more encompassing theories of language acquisition, and general development. The development of both the percep-tion and production of phonetic and phonological distinctions must be con-sidered. ( Ferguson and Garnica 1975). It i s important to clearly define the difference between the phonetic level and the phonological level. Phonetic refers to speech sounds, and phonological refers to the integration of those speech sounds into a linguistic system. "The phonological system of language imposes structure on the phonetic continuum" (Langacker 1968 p. 153). The phono-logical (or phonemic) level refers to an internalized representation speakers have of the sounds in their language. "Since each time a speaker pronounces the sound [p] i t i s acoustically never quite the same as the last [p] the speaker must have internalized an image or idealized picture of the sound, a target which he tr i e s to approximate." (Hyman 1975 p. 72) A child acquiring the sound system of language i s faced with a phonetic - 11— continuum, i n the form of adult speech, from which he must draw conclusions about the structure of language. A major controversy exists over the nature of the child's phono-logical system. Does the child have a system of his own, separate from the adult's, or i s the child's internal representation the same as the adult's, with the phonetic output limited by articulatory constraints? Current theories of phonological development have been reviewed by Ferguson and Garnica (1975) and divided into four classifications: behaviourist, structuralist, naturalist, prosodic. Their review i s sum-marized here. The major proponent of the behaviourist theory i s Mowrer. The goal of the behaviourist theory i s to integrate phonological development with with more general learning theories. Interaction between caretaker and child, and selective reinforcement are important factors. The development of the child's phonological system i s broken down into four steps. In step one, the child identifies with the caretaker. In step two, vocalizations of the caretaker are associated with some primary reinforcement such as food. The child's vocalizations because of their similarity to the care-taker **s w i l l acquire reinforcement value i n step three. Step four shows the caretaker reinforcing the child's utterances i f they are identified by the caretaker as a close approximation, Mowrer supports the hypothesis that the child has perceptual limitations which affect his representation of adult speech. Olmsted, i n a modification of Mowrer's theory, postulates a hier-archy of "ease of perception", based on Mi l l e r and Nicely's confusion matrix. Olmsted states that the child's listening conditions vary con-stantly with, different amounts of masking noise and interruptions. This interference "gives an advantage (higher probability of being learned ^ 1.2 r-earlier) to those phones whose components are more discernable" (Olmsted 1966 p. 333), Evidence from Miller and Nicely's experiment suggests that voicing and nasality are most easily discrlminabie, followed by f r i c -tion and duration, and last, place of articulation. Olmsted hypothesizes therefore that the child w i l l acquire the features voice and nasality prior to place etc. A second theory of phonological development i s the structuralist, or linguistic feature theory, derived, by Jakobson, who attempts to account for- the acquisition of phonology with linguistic universals. Jakobson (1968) hypothesized that the child learns phonological oppositions i n an invariant order, which i s universally valid and starts with an optimal consonant and an optimal vowel. Jakobson divides the acquisition process into two stages— 1) prelinguistic babbling and 2) acquisition of language— and maintains that there i s a discontinuity between the two stages. The acquisition of phonology takes place i n regular invariant stages, governed by "laws of irreversible solidarity" (a universal hierarchy of structural laws). Jakobson would support the hypothesis that the child has his own phonological system. Moskowitz extends and modifies the basic theory by adding gen-erative components, and claims that the syllable (rather than the phoneme) i s the important unit. Moskowitz (1970) comments on the interaction between phonetics and phonology. In early stages of phonological development a child l's phonetic a b i l i t i e s w i l l be more advanced than his utterances suggest, a limited phonological system constraining output. In later stages of development, constraints w i l l be reversed. The phonological system may contain contrasts which cannot be produced, due to d i f f i c u l t y of phonetic segments Moskowitz suggests (in agreement with Mowrer) that the child has perceptual l a c t a t i o n s which affect his internal representations. Stampers Natural Phonology claims that the child makes use of - 13 -an innate, and univeral set of phonological processes, that act to "collapse potential oppositions into that member which least trie s the restrictions of the human speech capacity" ( Stamps 1971 cited by Ferguson & Garnica p. 36). In other words, there i s a hierarchy of ease of production, as opposed to Olmsted's hierarchy of ease perception. The child produces a "simplified" version of an internalized adult model. Stampe feels that the child's per-ceptual development i s either complete, or far i n advance of his productive development. In the course of acquisition the child progressively struct-ures his utterances u n t i l they match the adult model. A fourth theory of phonological development i s the Prosodic Theory, proposed by Natalie Waterson, i n which emphasis i s placed on indiv-idual differences, i n contrast to the structuralist theory, which empha-sizes linguistic universals. The individual child's specific input, and the selective role of perception are important factors. The child f i r s t attends to an utterance as a whole, and picks out highly salient features. He builds a skeleton, composed of those features in the adult model.; that have high perceptual saliency. Phonological development takes place, as the child moves from gross to fine distinctions, i n both perception and prod-uction. I n i t i a l l y then, the child's phonological system i s limited by per-ception. David Ingram (1976) discusses the acquisition of phonology i n re-lation to cognitive development. Stages of language acquisition are best seen within the broader framework of the child's cognitive a b i l i t y (p. 7). Ingram suggests the following phonological stages: 1) Preverbal vocalization and perception (birth - 1;0) 2) Phonology of the f i r s t f i f t y words (1;0 - 1;6) 3) Phonology of simple morphemes (1;6 - 4;0) 4) Completion of phonetic inventory (4;0 - 7;0) 5) Morphophonemic development (7;0 - 12;0) Stage one includes babbling and infant speech perception. Ingram defines infant speech perception as a nonlinguistic sensorimotor task. Stage two i s described as qualitatively different compared to later stages of devel-opment. Contrary to Jakobson's belief that a system of contrasts emerges right away, Ingram questions whether different forms are actually being used contrastively. During this stage there are many "correct" forms which are simplified at later stages of development. A large amount of the child's phonetic inventory i s acquired i n stage three, and a large number of phonological processes are used. Completion of the phonetic inventory occurs i n stage four. Fewer phonological processes are used. Acquisition of morphophonemic rules i s begun. Stage five involves acquisition of com-plex rrarphophonemic rules which require the cognitive development of re-versible operations. An example i s the use of contrastive stress in Eng-l i s h , e.g. blackboard blackboard. Ingram's stages of phonological development show, clearly that phonetic development i s complete long before phonological development. Ingram raises the question again re the nature of the child's phonological system; does the child have a system of his own separate from the adult's, or i s his system the same as the adult's, or a subset of the adult's, mismatches i n phonetic output resulting from perceptual and motor constraints? How and why does the child use the phonological processess suggested by Ingram? 2.3 SPEECH SOUND DISCRIMINATION STUDIES Studies concerned with the role of perception i n the acquisition of phonology are for the most part speech sound discrimination (SSD) tasks. Ferguson and Garnica (1976) have suggested that when discussing the child's acquisition of phonology the question "at what point does the child's lan-guage show the distinctive function of sound differences and how does the child acquire the f u l l inventory of adult oppositions" (p. 7) i s more im-portant than examining what sounds are acquired i n what order. The current trend i s toward a cognitive model of phonological acquisition, but contributions of earlier studies investigating an order of acquisition should not be overlooked. A classic study was that of Shvachkin (1948), who studied eighteen Russian children to determine an order of acquisition for the perception of Russian speech sounds. The children were followed over a period of time, while they were acquiring their phonology. The children were quite young, ranging i n age from 10 months to a year and one half. No one since Shvachkin has been able to obtain results with children so young. The children i n Shvachkin's study were taught monosyllabic CVC nonsense names for objects. The nonsense names followed rules of Russian phonology. Objects were introduced individually, and the c h i l -dren were "taught" the name. In testing a phonemic opposition, there would be three CVC words, two being a minimal pair and a third different i n a l l three segments, e.g. mak bak zub. Once the child had been taught the names, and could discriminate i n a non-rtiinimal situation, a l l three objects were brought together, Oppositions which did not meet the^criterion for "perceived" were tested further, u n t i l a l l oppositions were established. Shvachkin postulates a sequence of twelve stages in the acquis-.... i t i o n of Russian speech sounds, based on the results of the speech sound discrimLnation. Every child went through the sequences with l i t t l e varia-- 16 -tion i n the order, and a l l discriminations could be made by age 2. The main criticism of Shvachkin's work, i s that he did not clearly state his methodology or his criterion for determining whether or not the child could make a discrjurdnation. Shvachkin feels that i n child language acquisition phone-mic development i s determined by semantic development. "Under the i n -fluence of semantic change the child moves toward the phonemic perception which i s connected with a radical reconstruction of both articulation and speech perception" (p. 96). In early speech and language development, the child derives a name for an object from i t s function, and a "general sound picture" of a word, based on intonation and rhythm, differentiates meaning rather than the phoneme. Shvachkin c a l l s this "prephonemic and prosodic speech". During this stage a child may have a "correct" pronunciation of a word due to "involuntary articulation" (learning a word as a sound pic-ture) . In later development when the child no longer relies on function to derive meaning, the phoneme takes on greater significance. A word which was articulatorily correct at an earlier stage may seem to regress and be-come less l i k e the adult model. Shvachkin's order of acquisition paralleled Jakobson's (1968) closely, although the two apparently worked independently, lending support to Jakobson's notion of a universal hierarchy of structural laws. While Shvachkin's emphasis was on perception, and Jakobson's emphasis was on production, both authors support a theory of phonological development i n which the child's system i s i n i t i a l l y limited by perception. In 1971, a p i l o t study was conducted by Olga Garnica, using Shvachkin's technique. The purpose of the p i l o t was three-fold: 1) to refine the methodology for testing phonological oppositions (as Shvachkin - 17 -did not report f u l l y his methodology), 2) to determine a preliminary ordering of English speech sounds (as opposed to Russian), 3) to det-ermine the best age to begin testing. Twelve children aged 1;2 to 2;5 were subjects for the p i l o t study. Painted wooden blocks were given nonsense names "MR. CVC". I n i t i a l C's were nu\nimal pairs, V was either t l ] , [A] or [V], and f i n a l C was a voiceless stop. The blocks were named for the child, and a training session using maximally different CVC's was carried out. After perforjxdng to criterion i n the training session (the number of training sessions varied with the child) the children went on to the task. Two blocks were placed i n front of the child, who was then asked to perform various tasks with one or the other of the blocks. From the results, i t was deterirujied whether or not a discrimination could be made, and an order of acquisition was suggested. Garnica's main study (1973) was carried out with the modi- . fications indicated by the p i l o t study. Sixteen children aged 1;5 to 1;10 were tested as i n the p i l o t , except that more attention was paid to teaching the words during the sessions. The children were tested three times a week for four weeks, on pairs determined by the order of acquisi-tion established i n the p i l o t study. Garnica found more individual variation i n the main study com-pared to the p i l o t , and suggested that while there may be general trends i n the acquisition of phonemic perception of speech sounds, there i s no universal order of acquisition. Edwards (1974), using a technique similar to that of Shvachkin and Garnica, attempted, to relate perception to production. Several hypothe-- 18 -ses were investigated, including the notion that perception precedes production. Edwards was also interested i n an order of acquisition for certain classes of sound. In general, results seemed to suggest perception does precede production. Edwards, l i k e Garnica, found a great deal of individual variation, so only general trends were suggested i n an order of ac-quisition. Shvachkin, Garnica, Edwards and others investigated an order of acquisition for phonemic oppositions. Other investigators have focussed on a distinctive feature hierarchy i n the learning of speech sound oppositions. David Barton (1976) reveals several inadequacies i n what he ca l l s "the Shvachkin-Garnica technique" for investigating speech sound discrimination a b i l i t i e s . In particular, Barton points out that the s t a t i s t i c a l analysis used i n Garnica's study was inadequate i n two. ways; 1) only those t r i a l s including and following the f i r s t c t r i a l on which a correct choice was made were considered and, 2) i f a child failed to reach criterion i n one session the session was re-peated. Both of these inadequacies bias the results to give fewer errors than actually occurred. Barton also c r i t i c i z e s Garnica for not allowing the pos-s i b i l i t y that a discrimination i s i n the process of being acquired. When recording results, Garnica assumed a discrimination was either acquired or not acquired, there was no room for par t i a l discrimination. Barton (1976)>"investigates the hypothesis that children can perceive the majority of the phonological discriminations of their lan-guage at an early stage i n their speech development" (p. 2). Unlike - 19 -Shvachkin and Garnica, Barton was not interested i n postulating an order of acquisition. I t i s generally accepted that the phoneme has some "psycholog-i c a l reality", and Barton assumes that distinctive features also have some psychological reality. Although the phoneme and distinctive features are not necessarily the units of perception and production Barton suggests that they "mediate storage i n some way". In Barton's f i r s t experiment, twenty children 2; 3 to 2;11 were tested on twenty oppositions. The stimuli consisted of monosyllabic ixiinimal pairs, names of objects that the children would be familar with. Prelim-inary testing showed that knowledge of the words used was"an important var-iable. The children were better at making a discrimination when they knew both words. This was incorporated i n the main study. Children were tested i n two or more sessions of variable length. Children were f i r s t required to identify the pictured objects/^discrimina-tion was not attempted u n t i l the child could name a l l the pictures. In the discrimination task, two pictures of objects with names forming a mini-mal pair were placed 'in front of the child. A card was fed to a modified Language Master, and an adult female speaker named one of the objects using the carrier phrase "point to the ". Barton found an overall error rate of 20% using the number of errors i n the f i r s t five t r i a l s . This error rate i s much lower than that found by Garnica (1973). Garnica's overall error rate was 34%; recal l that Garnica's s t a t i s t i c a l treatment biased the results to give a lower number of errors than actually occurred. Error rate i n Barton's study was affected by the children's knowledge of the words used. There was a much higher error rate for those words which the children could name without any - 20 -prior teaching. Results showed a great variation between children. The over-a l l error rate was very low; when pairs which had been taught were ex-cluded from results, children failed to make a discrimination less than 5% of the time. No claims could be made re int r i n s i c d i f f i c u l t y , order of acquisition, features etc. due to the small number of errors. A second experiment, investigating SSD of younger children, was also carried out. Ten children aged 1;8 to 2;0 served as subjects, the second experiment followed from the f i r s t , but more time was spent teaching the words, more production data was obtained, and only two contrasts were investigated. PreljLminary testing showed that with younger children, l i v e voice presentation was preferable to use of stimulus presentation via the Language Master. The minimal pairs used were "goat/coat" and "bear/pear". The words were taught f i r s t i n a non-minimal situation, i n which the children were asked to put each object i n a bag, then take i t out again. Since this was a teaching situation, children were told of their errors. Once the children had learned the words, the contrasts were introduced f i r s t i n object pairs, followed by presentation of a l l four objects together. Results showed that; 1) six of the children could make both discriiiiLnations consistently, 2) one child could make the discr irrigations several weeks after starting the experiment, 3) one child could make one of the discrimLnations. Productive data showed that; 1) two children used correct voicing, 2) one child distinguished the pairs but not with correct voicing, - 21 -and 3) three children did not produce the contrasts, but sometimes distinguished them when imitating. The two contrasts tested with the younger children, support the hypothesis that a phonological contrast w i l l be acquired perceptive-l y , prior to productive use. Barton concludes that phonological discriiriination i s complete when children begin to speak. He views perception as an active process, and supports a perceptually based theory of phonological development. Graham and House (1971) investigated "the characteristics of the organization the child imposes upon what he hears" (p. 560), i n terms of Sound Pattern of English features. Thirty g i r l s aged 3;0 to 4,-6 were trained to give a same/ different response. A total of 240 contrast pairs were tested each consonant was paired with every other consonant i n both orders, e.g. [p t ] , [t p], and each consonant was also paired with i t s e l f , the ratio of different pairs to same pairs was 3:1. Consonants were tested i n a nonsense utterance; /k^C^d^A simulating a multi-syllabic word. Pairs were presented l i v e voice, and several response ac t i v i t i e s (such as throw-ing blocks i n a bucket) were used to maintain children's interest. Child-ren went through a training session, responding same/different to progressive-l y more abstract items. Testing took place during 4 sessions; 168 items were tested each session. Graham and House found an overall error rate of 14% on contrast items, with a range from 2% to 79%, Individual subjects had error rates varying from 3% to 37%, Eleven of the contrast pairs had an error rate of 20% or more. In terms of features, a significantly higher error rate was found for contrast pairs differing by only 1 feature, compared to those pairs differing by 2-6 features. Pairs differing by the feature coronal, anter-io r , continuant or voice, had significantly higher error rates than pairs differing by the feature round, nasal or strident. Graham and House used a multidimensional scaling system i n studying the perceptual significance of the features. They concluded that for the children studied, SPE features had no psychological reality. They found that children's perceptual behaviour was similar to adults', but that children make more errors. Graham and House feel that their data does not support Jakobson's theory of phonological acquisition. Lewis (1974) examined perception and production of selected consonants i n terms of distinctive features confusions. Lewis studied 50 children: one group of 25 children described as having articulation problems, and a second group of 25 children with normal articulatory development. The children had to complete an articulation task and a dis-crimination task. The articulation task consisted of 18 pictures from the Templin Darley Test of Articulation. The perceptual task involved a same/different judgement of VC and CV nonsense syllables i n a signal: noise ratio of 38 dB SPL. Results showed differences between the two groups of children. Children with articulation problems had greater d i f f i c u l t y discrjuriLnating pairs separated by only 1 feature, compared to pairs separated by 2 or more features. Graham and House also found a significantly higher error rate for a l l normal subjects for contrast pairs differing by one feature com-pared to pairs differing by 2 or more features. Perceptual errors were not found to be predictive of productive errors, i n terms of distinctive features. Several studies have chosen to examine the child's acquisition of SSD, i n terms of only one class of sounds. The group "fricatives" are of-ten chosen, evidence suggesting this sound class i s "a major source of a r t i -culation and discrimination problems i n kindergarten and f i r s t grade" (Abbs and Minifie 1969 p. 1515). Moskowitz collected data from several children, and also used data previously collected by other authors. The children's corpora were analyzed i n terms of both the phonetic and phonological acquisition of fricatives. Moskowitz's criterion for stating a phoneme was learned was as follows: "A phoneme X can be said to have been acquired when the pattern of phonetic realization of X i s consistently distinct from the pattern of phonetic realization of any other phoneme Y." p. 146 Moskowitz found that the voiceless fricatives / s f o c c u r r e d phonetically before a l l voiced fricatives. The fourth fricative of the voiceless group, /0/, was found to be a problem phonetically, most often realized as [ f ] , long after a l l other fricatives were acquired phonetically. Based on Moskowitz's definition of> when a phoneme can be said to be acquired /0/ occurred phonologically very early. Phonologically then, a l l voiceless fricatives are acquired before voiced fricatives. The development of /z/ was found to be different from that of any of the other voiced phonemes. Moskowitz feels that this i s probably related to English morphology: alternation between /s/ and /z/ for plural, possessive, and 3rd person present tense. Abbs and Minifie (1969) investigated acoustic cues i n fricatives, i n terms of their effect on perceptual confusions i n children. Two acoustic cues have been previously identified as being impor-tant i n the discriitiination of fricatives: 1) frequency of the noise produced by the point of constriction and, 2) 2nd and 3rd formant transition to adja-cent segments. While frequency of f r i c t i o n was found to be important for discrjurdjiating [s] and [^], formant transitions were more important for discriminating [f] and [0] (Harris, cited by Abbs and Minifie p. 1435). Duration and intensity may also play a role i n fricative iden-t i f i c a t i o n . I t has been suggested that length may differentiate sibilants from the other fricatives (Miller and Nicely, Parmenter and Trevino, cited by Abbs and Minifie p. 1435). Abbs and Minifie tested seventeen children aged 3:0 to 5:1, on their a b i l i t y to discriminate fricative pairs. The stimuli used consisted of nonsense syllables, created by pairing each of the six fricatives [s, z, f, v, with each of the vowels [a, i,3*»]. The syllables were ordered both as VC and CV. Each pair was tested six times, resulting i n ninety pairs. Six randomly ordered l i s t s were prepared and recorded by an adult male speaker. An additional seven syllable pairs were recorded as a con-t r o l . These pairs were designed to be maximally different, e.g. 6&] -[ki], i n order to test the child's a b i l i t y to do the task. Stimuli described above were presented via earphones. Each auditory event was paired with a visual stimulus i n an ABX paradigm. Visual stimuli were "pseudo Seuss" pictures of objects and animals, dis-played on panels. One panel would light up as the child heard stimulus A, the other panel would light up as the child heard stumulus B, both panels - 25 -would light up as the child heard "who said X". Children could take as long as they liked to respond. A green light would go on i f the response was correct, the light on the inappropriate visual stimulus would go out, and the child would hear "I said X". When an incorrect response was given, a red light would go on, and the child would have time to correct the error. A mean error rate of 13% was found for control pairs, sug-gesting that the task was d i f f i c u l t . Mean error rate for fricative pairs was 28%. In general, VC pairs were found to be easier to discriminate than CV pairs, although /v - z/ /f - s/ /v -%/, contrary to general trends, showed fewer errors as CV pairs. No significant difference was found for different vowels used. Consonant pairs / f / - /&/ and /v/ - /%/ were found to be the most d i f f i c u l t to discriminate, and /&/ - /z/ and /s/ -/z/ showed the lowest mean error rate. Voiced/voiceless contrasts showed a lower mean error rate than voiced/voiced or voiceless/voiceless contrasts. Abbs and Minifie carried out a second experiment, acoustic analysis of the stimulus tape of fricative pairs. The stimuli were analyzed i n terms of duration, intensity and spectrum. In general, voiceless fricatives were found to be longer than voiced fricatives. The fricatives /s/ and /z/ were found to be more intense than the other fricatives. No significant differences were found i n the intensity of voiced fricatives compared to voiceless fricatives. In terms of spectrum, the fricatives /s/ and /z/ were again set apart from a l l other fricatives by having higher center resonance frequencies and shorter bandwidths. Abbs and Minifie conclude that "the voicing dimension, and the frequency - bandwidth - intensity dimension appear to provide salient cues for distinguishing between fricatives" (p. 1542). - 26 -2.5 DISCUSSION - NATURE OF THE CHILD'S PHONOLOGICAL SYSTEM A major controversy becomes apparent after reviewing the literature, regarding the nature of the child's phonological system: • Does the child have his own system separate from the adult system, or i s the child's system the same as the adult system, either i n whole, or a proper subset? Smith and Stampe would argue against the hypothesis and that the child has his own system. "...I agree largely with Stampe (1974 p. 4) that 'no evidence whatsoever has been ad-vanced to support this assumption that the child has a phonemic.system of his own -NVS' ..." (p.37) Jakobson argues i n favour of the hypothesis, and Ingram suggests that there i s not enough supporting evidence. The relationship between perception and production becomes important in examining the above hypothesis. There i s evidence to sug-gest that the child's perceptual a b i l i t i e s are i n advance of his productive a b i l i t i e s . Barton, Stampe, and Smith suggest that perceptual development i s either complete when productive development begins, or at least far i n advance of productive development. Jakobson, Waterson, Ingram, Mowrer and Moskowitz feel that the child has perceptual limitations which affect his representation of adult speech. Kornfeld (1971) exarnines the issue of the nature of the child's phonological system. She clearly states the hypotheses: HO: The adult system of phonological distinction determines the child's system. Mismatches between adult and child output are recognized and assumed to be due to the child's motor constraints. HA: The child's phonological system does not equal the adult's system. The child i s viewed as having better motor control than HO attests; phonetic mismatches result from a different set of distinctive features. Kornfeld breaks down the two major hypothe-ses further. HO can be stated as HO-^  constraints on the motor system result i n distortion, child phonology = adult phonology + phonetic mistakes, and H02: motor constraints li m i t production only i n certain ways, child phonology = adult phonology + readjustment rules. HA can be stated as HA^ : the child's perceptual system i s a proper subset of the adult system and HA^ the child's system i s not a subset of the adult system, the child may hear speech i n a different way, and mark things the adult would not mark. These hypotheses clearly state the two dichoto-mies: motor constraints vs. perceptual constraints, and child's phonolog-i c a l system a subset of the adult system vs. child's system a system of his own. ' Kornfeld feels HO-^  can be eliminated immediately, because i t does not account for regularities found i n child phonology. While H02 handles regularities, i t cannot account for exceptions to the rules. If motor constraints li m i t the child's output only i n certain ways, how i s i t that the child may produce regular alternations of a segment i n three different words, and produce the segment correctly i n the fourth, word. Kornfeld states that H02 i s the position taken i n most studies; how-ever she feels that HA accounts for data that H02 can't explain. Kornfeld studied the production of i n i t i a l clusters i n 13 children aged If - 2| years. At this age, single segments are mastered, but many adult/child mismatches are seen for i n i t i a l clusters. The children played i n a playroom for 40 minutes once every three weeks. Play sessions were taped, and transcriptions of spontaneous speech were - 28 -done both during the play session, and later from the tape. Kornfeld was interested i n comparing production of i n i t i a l clusters to produc-tion of single C's i n the same environment e.g. truck-tuck/rub. Utterances used i n the studies were spliced out of the tape and anal-yzed spectrographically. A l l subjects reduced adult clusters but did not simply eliminate or assimilate one of the C's. The word "blue" might be produced by two different subjects as [bu:] or [baiu:], with [bu:] spectrographical-ly different than [bu:] for "boo". Child productions of clusters with liquids and glides, e.g. glass and grass, were transcribed by the adult as [gwats]. Spectro-graphic analysis again revealed two distinct different productions. "... show observable acoustic differences for the segment, yet these distinctions were collapsed by adult listeners." (p. 462) Kornfeld also found that a l l subjects had more trouble with clusters containing a strident. She speculates re a hierarchy for ac-quisition of features. "If order of acquisition corresponds to a place in hierarchy, stridency was the last distinction to be acquired by our subjects, as well as by those i n other studies (Smith 1970)." (p. 463) Kornfeld alternatively explains the; mismatches of the s t r i -dent member of the cluster, i n terms of predictability. In English, the i n i t i a l segment i n a cluster i s always [s] i f followed by a plosive; therefore the [s] i s highly predictable, and learned later. Kornfeld concludes that her data supports HA; definetly HA^  and possibly H^. She feels that the child has his own phonological system, and marks some distinctions which are not marked by the adult. - 29 -Aitchinson (1977), cited by Catherine Browman (1979 p. 2), studied malapropisms in children and adults. Adult malapropisms were found to be most similar to the target word i n i n i t i a l and f i n a l con-sonant, and less similar for stress pattern, number of syllables and stressed vowel. The opposite was found for the children's malapropisms. This suggests that children are processing speech i n a different way; that there are certain features with higher perceptual saliency (as suggested by Olmsted, Waterson and others) that the children recal l i n the malapropisms. Barton (1976) found that knowledge of words used i n his SSD study was an important variable; discrimination of learned words was much better than discrimination of words taught during testing. I t i s possible that while auditorily the child can make the discriminations, because he i s i n the process of language acquisition, linguistic infor-mation not f u l l y acquired i s interfering. Barton's findings also suggest that being able to produce a constrast results i n a more firmly established underlying representation; therefore the child makes fewer discrimination errors. Kornfeld hypothesized that the child produces What he perceives, and does so according to a system more abstract than the adult system. Does this allow for the child to progress toward the adult system? I t has been shown that even young infants can make complex auditory discriminations. The child developing his phonological system may be responding to some auditory cues i n a linguistic mode, prior to the other cues especially i f there are conflicting auditory cues i n a stimulus. Researchers often seem to be arguing at cross purposes. The - 30 -two notions 1) perception precedes production and 2) production i s limited by incomplete perceptual a b i l i t i e s , can be compatible. Devel-opment of perception may be in advance of development of production, but may not be complete before the child begins to produce, and therefore may affect productions. If the psychological reality of the phoneme i s accepted, then ultimately a developmental view of perception must be accepted. If Piaget's scheme of cognitive development i s followed, the child w i l l not have an abstract underlying representation u n t i l approximately age 1;8, when the sensorimotor period ends. - 31 -CHAPTER 3 3.0 OBJECTIVES 3.1 ORIGINAL OBJECTIVES The original aim of this study was to exaniine the relationship between perception and production in the child's developing phonological system. Review of the literature revealed several aspects of previous studies which deserved closer investigation, such as experimental para-digm, speaker, nature of the linguistic task, and these were to be inves-tigated. I t was hypothesized that as a child performed a series of per-ceptual tasks, from acoustic discrimination through increasingly complic-ated linguistic discrimination, his discrimination score would change, e.g. sounds in isolation, nonsense syllables, monosyllabic words, multi-syllabic words, sentences. I t was also hypothesized that scores would change for the same items, depending on the speaker, e.g. child listening to him/herself, another child, adult male, adult female, synthesized speech. Experimental paradigm used could also be a factor affecting results i n previous studies. Some studies used an AX paradigm involving a same/different judgement. Subjects heard two stimuli arranged as AA or AB, and judged them to be the same or different. Other studies used an ABX paradigm. Subjects heard three stimuli, and judged X to be the same as A or the same as B. In studies where objects were used, the child would hear only one stimulus, and would determine i f i t belonged with object A or object B. I t has been suggested that an AX paradigm permits direct comparison of the auditory stimuli, using echoic memory. When an ABX paradigm i s employed, i t i s suggested that some sort of coding (possibly phonetic coding) must take place, as three items would produce an echoic memory overload. An AX paradigm may be only a task i n acoustic percep-tion, an ABX paradigm may involve phonetic perception, and one stimulus with real objects involves linguistic perception. Fricatives were chosen as the speech sound stimuli, because previous research suggested that they are acquired late. This would allow the use of older subjects, s t i l l i n the process of acquisition, but old enough to alleviate experimental constraints. I t was decided to use approximately 20 subjects aged 3 to 5 years. 3.2 REVISED OBJECTIVES Problems arose immediately trying to find real word minimal pairs of monosyllabic and multisyllabic words containing fricatives. Combinations of position of fricative i n word and position of word i n sentence were found to be endless. I t was therefore decided to use only nonsense syllables with different experimental paradigms and different speakers. Attempts to record the children were also unsuccessful. Productions varied depending on whether the productions were spontaneous or imitated. The decibel level of the recordings and the interstimulus interval could not be controlled. I t was therefore decided to record only one speaker, i n two experimental paradigms, AX paradigm and ABX paradigm. - 33 -Recruiting sufficient subjects was another serious obstacle encountered. Approximately twenty responses were obtained after i n i t i a l notices were sent out; however only ten subjects followed through and made a v i s i t for screening purposes. One subject failed the screening tests, another subject dropped out, and a third subject completed only 1 t r i a l and was therefore not included i n the f i n a l results. This l e f t a small N of seven. CHAPTER 4 4.0 METHOD 4.1 SUBJECTS Eight children, seven female and one male, ranging i n age from 3;0 to 4;9, served as subjects i n this study. Children were recruited primarily through a day care center, associated with a graduate student housing development, at the University of B.C. Criteria considered when choosing subjects included: 1) Age - 3-5 years 2) Socioeconomic Background - middle class 3) Language Background - parents native English Speakers - English only spoken at home 4) Hearing - Normal, defined as: thresholds for a i r conduc-tion pure tones, 250 HZ - 8000HZ, 20 dBHL or better bil a t e r a l l y 5) Speech and Language - Normal 4.2 STIMULI The stimuli used; consisted of six CV syllables, produced by combining each of six fricatives; /s z f v | ^ / ' with the neutral vowel schwa The vowel / 3 / was chosen to produce only nonsense syllables. I t has been reported by Abbs and Minifie (1969) that no significant d i f -ferences are noted when comparing results obtained using three different vowels i n CV contexts. A test was constructed i n which each fricative was paired with - 3 5 -i t s e l f and with every other fricative, giving a total of six same (S) CV pairs and 15 different (D) CV pairs (Table 1). A l i s t of sixty CV s y l -lable, pairs was generated, i n which thirty S pairs, and thirty D pairs were randomized. Thirty S pairs consisted of each of the six fricatives, paired with i t s e l f five times. Thirty D pairs were generated using the fifteen different pairs tested twice, each member of the pair being pres-ented once i n i n i t i a l position. The same l i s t of sixty pairs was tested three times. (Table 2). A second test was constructed using the thirty D pairs arranged i n an ABX paradigm. A total of sixty test items were generated (Table 3), by testing each of the items i n the l i s t of thirty twice, once as ABA, and once as ABB. Two short l i s t s of training items were complied, using four different animal sounds. The sounds were arranged in AX paradigm and ABX paradigm. The two l i s t s of training items were used to familiarize the children with the task. (Table 4). A l l stimuli were recorded by an adult female native speaker of English, from the Vancouver area of B.C. Items were recorded with an i n -terstimulus interval of two seconds, and an inter item interval of four seconds. The inter item interval varied during actual testing; the tape was stopped i f the child wasn't paying attention. 4.3 PROCEDURE The children were seen four times, for sessions of approx-imately 1/2 hour i n length. Time elapsed between the f i r s t session and the last session was no longer than five weeks. The purpose of the f i r s t v i s i t was to introduce the child to the testing environment, and for screening purposes. After a period of informal chatting, during which a short language sample was recorded, three screening tests were administered: the Arizona Articulation Test, Peabody Picture Voc-abulary Test, and a hearing test. Hearing was f e l t to be especially important for this study. Testing was done i n a soundproof room, using a Maico 22 C l i n i c a l audio-meter and TDH39 earphones. Equipment was calibrated to ANSI '69 prior to beginning the study. Thresholds for pure tones from 250HZ to 8000HZ were obtained using play audiometry. Speech reception thresholds were obtained using children's spondees, to support the r e l i a b i l i t y of pure tone thresholds. The Arizona Articulation Test • and the Peabody Picture Voc-abulary Test were chosen to screen speech and language because test scores can be converted to MA and because of ease of administration. One child was excluded from the study after f a i l i n g the screening tests, as previously mentioned. During the second session, training items were presented to determine whether or not children could do the task, and to establish a response mode. Some children answered verbally same/different, others performed an activity such as throwing a toy in a bucket or placing a piece i n a puzzle, others responded i n both ways. A l l children but one had established the concept same/different and responded correctly for a l l training items. The one subject who could not do the task was trained on progressively more abstract stimuli u n t i l she could respond correctly on a l l training items. After the children finished the training session, they pro-ceeded to the listening test. A short p i l o t study was run with several of the older subjects exhibiting the best attention span, to determine whether or not i t would be possible to have subjects take both l i s -tening tests. The listening test-ABX paradigm.was found to be too d i f f i c u l t , thus only the listening test-AX paradigm was run with a l l subjects. The listening test of sixty items was administered three times over the next sessions. A session was terminated when the end of the child's attention span was reached, even i f the child had not completed one whole l i s t of items. The next session would carry on from the point at which the previous session ended. Subject number 2, the subject who had not established the concept same/different, was only able to complete the listening test once, and therefore was not included i n the f i n a l results. - 38 -TABLE 1 Fricative Pairs Employed i n the Listening Test. Different Pairs Same Pairs 1. s» za ' 1. sd S3 2. S3 f3 2. za za 3. S3 v* 3. fa fa 4. sd fa 4. va va 5. sa 5. S3 j » 6. zA fa 7. z» va 8. z3 fa 9. z© J3 io. fa va i i . fa Ja 12. fa }3 13. va Ja 14. va 33 15. Ja 5a - 39 -TABLE 2 THE LISTENING TEST - AX 1. V2 ^3 2. fa -fa 3. fa fa 4. v3 v3 5. Z3 S3 6. 39 fa 7. -fa fa 8. S3 3-3 9. J 3 S3 10. j 3 ^ 11.33 y 12. V3 Z-3 13. Z3 f 3 14. 33 "J3 15 .33 39 16. S3 za 17. S3 f3 18. S3 J3 19. S3 S3 20. S3 S3 21. $3 ^3 22. fd Z3 23. va va 24. za 3a 25. fa J3 26. 27. 28. 29. $3 f3 Z3 J3 35 S3 53 ^ 30. fa 33 31 32 S3 S3 za za 33. va va 34. va sa 35. J 3 $ 3 36. va va 37. [3 S3 38. f3 £3 39. Z3 Z3 40. Y3 V3 41. sa va 42. za za 43. za 44. 39 Z-3 45. S3 sa 46. £3 V3 47. $Q 33 57. Z 3 V 3 58. 33 33 59. £3 £3 60. S3 S3 - 40 -TABLE 3 THE.LISTENING TEST - ABX 3. y 4. za 5. sa 6. v3 7. fa 8. f3 9. fa io. sa n . va 1 2 . sa 13. zd 14. Ja is. va 16. 17. if 18. v3 19. va 2 0 . za sa sa sa sa v3 s3 )* 3 a v9 v3 3a 33 sa s3 fa fa fa va va va sa za fa fa fa fa s3 s© va a za 21. \» v3 va 41. J» z3 J9 22. v y 42. za fa f d 23. f<? ? fa 43. y sa j3 24. sa 53 3 a 44. zd va za 25. sa fa sa 45. za sa sa 26. S3 sa 46. fa 33 27. sa sa 47. fa za za 28. va 48. f3 va fa 29. V za 49. va S d S 3 30. za va va 50. 3 3 z3 3a 31. va y va 51. \> v3 J3 32. fd fa 52. sa fa 33. fa ? J» 53. z3 z3 34. \> fa i» 54. 35. va za za 55. va (? v3 36. fa za fa 56. y \9 33 37. y za za 57. sd za za 38. za y 3 s 58. y va va 39. za fa zd 59. sa za sa 40. va zd va 60. ffafS - 41 -TABLE 4 TRAINING ITEMS USED AX 1 meow meow 2 moo moo 3 woof meow 4 quack quack 5 quack moo 6 moo quack 7 meow woof 8 woof woof ABX 1 meow woof woof 2 quack moo moo 3 woof meow meow 4 quack moo quack 5 moo quack quack 6 moo quack moo 7 woof meow woof 8 meow woof meow 5.0 RESULTS 5.1 GROUP RESULTS The mean number of errors for each of the different (D) pairs and same (S) pairs of fricatives was calculated. These results are presented i n Table 5 as percentage of errors. The highest mean error rate was found for the D pair [ - ^ 3 ], the lowest for the D pairs [Z3 - f<3 ] and [ ^ 9 - v3 ] . The only pair for which every subject made at least one error was the pair [fd - v9 ]. Prelimin-ary examination of results suggested that the three D pairs differing only in the feature voicing; [^ 9 -^3 ], [s9 - zg ], [f£ - v3 ] produced more errors than the other pairs. A series of dependent t-tests was computed to determine i f each D pair was or was not significantly different from each of the other D pairs. This resulted i n a tot a l of 105 t-tests. The number of times that each pair was found to be significantly different from another pair, i s presented i n Table 6. Preliminary examination of the data suggested that the pairs -v&,' ] produced more errors than the other pairs. This was supported by the result of the t-tests. The pair [^ d w a s f o u n d t° ke significantly harder i n 11 of 14 t-tests, [SO - ] and [f3 - v8 ] i n 9 of 14 t-tests. Other pairs which pro-duced significant findings were: [za -J3 ] 6/14 [za -V© ] 2/14 [j6 -fd ] 2/14 The fricative pairs which were found to be significantly more - 43 -d i f f i c u l t are presented i n Table 7, with those pairs from which they were not significantly different. The three pairs [sa - za ] [ 3^ -^3 ] [f9 - vg ] were not found to be significantly different from each other, suggesting that they f a l l into a natural group. In terms of distinctive features, each of these i s a voiced/voiceless rninimal pair. Each of the same (S) fricative pairs was compared to each of the other S pairs, resulting i n 15 t-tests. None of the S pairs were found to produce significantly more errors than any of the other S pairs. A third series of t-tests was generated, to determine whether or not D pairs were significantly different from each of the two S pairs used to produce the D pair. A suirinary i s presented i n Table 8. Examination of each of the fricative pairs individually, suggested that three of the pairs could be grouped' together... Fouri groups were generated i n terms of distinctive features. These are shown i n Table 9. Results were analyzed for the four groups, f i r s t i n a series of 6 t-tests, then i n a one way repeated measures ANOVA. Group three, consisting of the three pairs differing only i n voicing, was found to be significantly more d i f f i c u l t than the other three groups. 5.2 RESULTS FOR INDIvTDUAL SUBJECTS Overall error rate was calculated for each individual sub-ject, and i s presented i n Table 10. Subject #1 - At the time of testing S. 's chronological age was 4;4. - 44 -She scored 4;7 on the Peabody Picture Vocabulary Test (PPVT), and 11;0 on the Arizona Articulation Test. She substituted [s] for [^ ] i n word f i n a l position. Her scores for the pairs [ sd - z3] and [^ 3 -^9 ] on the listening test were much poorer than for any of the other pairs. Subject #2 - This child was the youngest i n the study, C.A. 3;0. She scored 3;9 on the PPVT, and 11;0 on the Arizona. She substituted [s] for [z] i n word i n i t i a l position on the Arizona, but this seemed to be an isolated error, since the finding was not repeated i n her spon-taneous speech sample. This subject did not have the concept same/dif-ferent at the beginning of the study. Several sessions were spent on training same/different, u n t i l she responded correctly for a l l training items. On the actual listening test however, she only responded for 38 of 60 items, so testing was discontinued. Subject#3 - C.A. 3;7 at time of testing. This subject scored 4;11 on the PPVT. He was the only subject to score near his C.A. on the Arizona Articulation Test, scoring 4;0. A l l other subjects scored much higher, making very few errors. Errors on the Arizona for sub-ject #3 were as follows: [s] substituted for [f] word f i n a l [$] substituted for [tj] word i n i t i a l [s] substituted for [0] word i n i t i a l [s] substituted for [v] word medial omit [v] word f i n a l omit [s] word f i n a l Note that a l l productive errors involve fricatives. Subject #3 had the highest number of errors for the listening test, 21.2%. There appears to be a correlation berween perception and production fot this subject; his attention span was very short however, and may have contri-- 45 -buted to his high error rate. Subject #4 - This subject was 4,-2 at the time of testing. She scored 5;4 on the PPVT and 10;0 on the Arizona. Errors for the fricative pairs were distributed evenly across a l l items; there was not one particular pair which she found more d i f f i c u l t than any other. Subject #5 - This subject was the oldest of a l l those tested, C.A. 4;8. She scored 9;8 on the PPVT and 11;0 on the Arizona. This sub-ject made very few errors for D pairs; she had the lowest error rate for D pairs, and was the only subject to make more errors for S pairs than for D pairs. Subject #6 - This subject was 4;3 at the time of testing. She scored 4;8 on the PPVT and 10;0 on the Arizona, substituting [f] for [8] in word f i n a l position. She had the lowest overall error rate of a l l subjects. Subject #7 - C.A. 4;0 at time of testing. PPVT score 5;11 and Arizona score 5;5. She distorted ] i n word i n i t i a l position. This subject had the best attention span; she listened carefully and repeated each syllable after the recording. On repeating items on the listening test she consistenly substituted [63] for [f3 ]. This subject was the only one to score 100% correct for the S pairs. She had one of the highest error rates for D pairs, resulting primarily from pairs 1 and 10, the pairs [sd - z3 ] and [^ 9 -J9 ], for which a l l were i n -correct. Subject #8 - C.A. 3;7 at time of testing. She scored 6;3 on the PPVT and 10;0 on the Arizona. She substituted [f] for [6 ] and [0] for [z] in word i n i t i a l position. Greatest number of errors occurred for the D pairs [s 9 - z 9 ]. - 46 -The following i s a summary of the characteristics of the subjects tested. (1) C.A. : 3;0 to 4;8. (2) PPVT scores: 3;9 to 9;8, most S scoring only slightly higher than their C.A. (3) Arizona Articulation Test scores: 4;0 to 11;0 most S g scoring very high. (4) A l l subjects made productive errors for fricatives, but these were not predictive of perceptual errors. (5) Pattern of errors for S 4,5,6 was distributed evenly across a l l fricative pairs. S 1,3,7,8 had a greater number of errors for the pairs [S3 - za] and [J3 -y* ]. - 47 -TABLE 5 Mean number of errors for each of the fricative pairs expressed i n percentage. Fricative Pairs Mean Number of Errors 2. s a j a 4 2 - 9 % 3. si 3a 4 > 8 J 4. sa va 14.3% 5. sa fa 9 - 5 % 6. za fa 9.5% 7. 2 a l a 4 - 8 % 8. za fa 1 9 - ° * 9. za va 0 % 10. a 33 9 - 5 % 11. (a fa 5 0-°% 12. ja v a 9 - 5 % 13. W f3 0 * 14. id v a H- 9% 9.5% 15. T3 va 28.61 1. sa sa 12 42. za za 3.8% 3. Ja ja 7.6% 4. 333a 4.8% 5. f3 f3 8.6% 6. va va 8.6% - 48 -TABLE 6 Collapsed results for 105 t-tests. Fricative Pairs Number of Significant t-tests 1 Sd Z3 9 2 S3 0 3 sa 3 9 0 4 sa va 0 5 S3 0 6 za 0 7 za 3 3 6 8 za 0 9 za V3 2 10 3 3 11 12 5 3 V3 0 13 fa 2 14 va 0 15 f a va 9 Note: based on one tailed t-test p < .05 - 49 -TABLE 7 Fricative pairs which most often Fricative pairs which were not showed significantly greater error significantly different from rates. pair l i s t e d at l e f t . S3 za f3 va sa fa sa va S8 z3 J« y f3 z3 3* z3 v3 , f3 fd v3 za z3 y z3 va f3 j f3 va * )3 33 S H O W E D S3 z3 significantly p a , " o r e errors than S3 v3 J ] € f3 3)3 f3 33 f3 J3 v3 - 50 -TABLE 8 Summary of results of t-tests comparing S pairs and D pairs D pairs with signficantly more errors.than S pairs 3 ^3 significantly S3 Z3 significantly fd V3 significantly Z3 "^ 3 significantly more errors than more errors than more errors than more errors than $3 ^ and ^3 S3 S3 and Z3 Z3 f<3 f 3 and va v3 Z3 Z3 and y ^3 S pairs with significantly more errors than D pairs 3 ^3 significantly more errors than va * f3 f3 significantly more errors than Z3 f 3 * * anomalies possibly resulting form the fact that 13 VP and Z3 fd were the 2 D pairs with 0% error rate. J - 51 -TABLE 9 Groups of fricative pairs i n terms of distinctive features. GROUP 1 - VOICELESS PAIRS Pairs differing i n 1 feature, place of articulation GROUP 2 - VOICED PAIRS Pairs differing i n 1 feature, place of articulation sa se fd fa z3 ^3 za va y va GROUP 3 - VOICED VOICELESS PAIRS GROUP 4 - VOICED VOICELESS PAIRS Pairs differing i n 1 feature, Pairs differing i n 2 features, voicing voicing and place of articulation S 3 fa y za va sa S3 Z3 3 3 V3 1» za J » y fa va fa - 52 -TABLE 10 Overall error rate for each subject Subject Error Rate 1 14.4% 3 21.2% 4 8.9% 5 7.8% 6 6.7% 7 11.7% 8 9.5% - 53 -CHAPTER 6 6.0 DISCUSSION 6.1 SUMMARY OF RESULTS 1) ; Results must be interpreted with caution due to; a) low N of seven, and b) high number of t-tests involved. 2) Effects of memory, attention span, and level of cognitive functioning become particularly important when dealing with younger subjects. 3) Overall error rate was found to be f a i r l y low. 4) Highest mean error rate for D fricative pairs was 50% for the pair [ 33 ]. 5) Lowest mean error rate for D fricative pairs was 0% for the pairs [23 f 3 ] and [ Ja V3 ]. 6) A significantly higher mean error rate was seen for the three voiced/voiceless irdnimal pairs \<d sa za, fa va 7) No comment can be made relating perception to production, due to the limited amount of productive data obtained. 6.2 ERROR RATE Examination of the literature shows great variation i n error rate obtained i n SSD studies. Graham and House (1971) using nonsense syllables simulating multisyllabic words, and an AX paradigm, found an overall error rate of 14% for contrast items, with a range of 2% to 79%. David Barton (1976) found an overall error rate of 20%. Barton used real word pairs, and the child heard only one stimulus and had to choose between two objects: when making a discrimination. Error rate was affected by knowledge of words used; when "taught" words were excluded from the analysis, error rate dropped to less than 5%. Abbs and Minifie (1969) studied only fricative sounds. They-j-used an ABX paradigm, and nonsense syllables paired with a visual stimulus to make the syllables more li n g u i s t i c a l l y relevant. They found a some-what higher error rate than had been found i n previous studies. Over-a l l error rate was 28%; however a high error rate of 13% on control items suggested that the task was d i f f i c u l t . Olga Garnica (1973) found one of the highest error rates - 34%. A detailed discussion by Barton (1976) shows that Garnica's s t a t i s t i c a l treatment biased results to give a lower error rate than actually occurred. The present study found an overall mean error rate for contrast pairs of 15.1%, with a range from;0% to 50%. One of the highest error rates 42.9%, was found for the pair [s5 - zd ]• This finding i s i n agreement with several other studies looking at a whole range of contrast pairs; e.g. Graham and House found a 52% error rate for S-Z, however Abbs and Minifie (1969), the only SSD study discussed which examined only f r i c -atives found the lowest mean error rate for /s/ /Z/. The spectrographic analysis done by Abbs and Minifie i n the second part of their study, showed that /s'/ and Ml were set apart from the other fricatives by 1) greater intensity, 2) higher center resonance frequency and, 3) shorter bandwidth. These findings suggest that /si or /zV paired with any other fricative would be easier to discriminate than /S/ paired with /2/, as acoustically they were most similar. This was not the finding i n Abbs and Minifie's SSD study. Moskowitz (1975) i n her study of the phonetic and phonological development of fricatives, found the development of /z/ to be different - 55 -from the other fricatives. She f e l t that the importance of the sounds [s], [z] i n English morphology; alternation of [s] ; [z] [Sz] for English plural, possessive, and 3rd person present tense, offer a possible explana-tion for this. Caution should be exercised when comparing error rate across different studies. Error rate appears to be affected by several factors including 1) experimental paradigm, 2) s t a t i s t i c a l analysis, 3) stimulus items, 4) attention span and 5) cognitive level of subjects. 6.3'DISTINCTIVE FEATURES Many experimenters chose to examine results of SSD studies i n terms of distinctive features. Barton (1976) suggested that while dis-tinctive features have some psychological re a l i t y , and may mediate storage i n some way, they are not necessarily the units involved i n perception and production. Error rate for Barton's study was so low that he could make no comment concerning the role of distinctive features. Many studies (of both adults and children i n both perception and production) have demonstrated a significantly higher error rate for pairs of phonemes separated by only one feature compared to pairs separated by two or more features (Graham and House 1971, Lewis 1974, Koenigsknecht and Lee 1968, Tikofsky and Mclnish 1968). Graham and House (1971) found more errors for the features coronal, anterior, continuant and voice. Koenigsknecht and Lee (1968) found the highest error rate for place of articulation and voicing. A l l errors were either place of articulation or voicing i n Tikofsky and Mclnish's 1968 study. Moskowitz (1975) found that a l l voiceless fricatives were - 56 -acquired phonologically prior to the voiced fricatives. Voiceless fricatives with the exception of /0/ were also acquired phonetic- a l l y prior to voiced counterparts. A high error rate for the feature "voicing" appears to be a consistent finding among most studies. Abbs and Minifie (1969) i n contrast to the general trend, found a lower mean error rate for voiced/voiceless pairs, compared to voiced/voiced or voiceless/ voiceless pairs. Results of the present study indicate a higher mean error rate for voiced/voiceless pairs separted only by the feature "voicing". 6.4 EXPERIMENTAL PARADIGM AX and ABX.. are the two most common experimental paradigms in SSD studies. Less often, experimenters use real objects or pictures, and present one stimulus, the subject being required to choose between the objects. I t was suggested previously that an AX paradigm, because i t allows direct comparison of the acoustic stimuli, examines only acoustic perception not linguistic perception. An ABX paradigm exa-mines something more than acoustic perception; because short term memory undoubtedly constrains children's responses, some form of coding must take place i n order that items can be stored for later comparison. I t i s assumed that discrimination of real word pairs must require some form of l i n g u i s i t i c processing. Studies of infant speech perception have shown that the aud-itory system i s innately equipped to make fine discriminations. If an AX experimental paradigm examines only acoustic perception, one would - 57 -expect to find almost perfect results i n an SSD study employing an AX paradigm. Review of the literature and findings i n the present study.•• do not support this expectation. Whilst i t was observed that overall error rate was for the most part f a i r l y low, error rate varied con-siderably depending on the phoneme pair, some pairs showing extremely high error rates. It i s assumed that the infant relies almost exclusively on auditory information when making speech sound discriminations, up to a certain period i n development. As i t s phonology develops, however, some factor must enter into the discrimination process which adds to the auditory information to give a phonological or li n g u i s t i c a l l y meaningful percept. If the act;, of discrimination was purely psycho-acoustic, i.e. based on DL's for frequency intensity and duration, then responses should be error free. I t i s learning the complex language code which now begins to contaminate the response data. Thus, varying error rates for different contrast pairs i n an AX paradigm suggest that the child i s making phonological discrim-inations, incomplete phonological perception resulting i n errors for certain contrasts, (depending on the stage of development) which might hot have occurred i f the child was relying s t r i c t l y on auditory i n -formation. In experimental paradigms employing real word pairs and objects, the child has semantic information available to him, which again changes the nature of the task. - 58 -6.5 CONCLUSION The present study was designed to investigate the perception of fricative sounds by children aged 3;0 to 4;11. Findings were, for the most part, i n agreement with the findings of other SSD studies. The present study uncovered methodological problems encountered when dealing with younger subjects. Research should be interpreted with caution, and controls, such as the control pair used by Abbs and Minifie (1969) should be used whenever possible. The role of perception i n the acquisition of phonology i s s t i l l unclear, due i n part to d i f f i c u l t i e s inherent i n investigation. I t i s also s t i l l unclear whether speech perception i s necessarily re-lated to language processing. - 59 -BIBLIOGRAPHY Abbs, M.S., and Minifie, F.D. Effect of acoustic cues i n fricatives on perceptual confusion i n preschool children. Journal of  The Acoustical Society of America, 1969, 40, 1535-1542. Barton, David "The Role of Perception i n The Acquisition of Phonology". Ph. D. Thesis, University College, London, 1976. Browman, Catherine P. "Tip of The Tongue and Slip of The Ear: Impli-cations for Language Processing" UCLA Working Papers i n Phone-t i c s . University of California, Los Angeles, July 1978. Cutting, J.E., and Rosner, B.S. Categories and boundaries i n speech and music. Perception and Psychophysics, 1974, 16, 564-570. Dale, Philip S. Language Development Structure and Function. Holt, Rinehart, and Winston, copyright 1976. Edwards, M.L. Perception and production i n chil d phonology: The test-ing of four hypotheses. Journal of Child Language, 1974, 2, 205-219. Eimas, P.D., Siqueland, E.R., Jusczyk, P., and Vigorito, J.M. Speech perception i n infants. Science, 1971, 171, 303-306. Ferguson, Charles A., and Garnica, Olga K. Theories of phonological development. In Eric H. Lenneberg and Elizabeth Lenneberg (Eds.) Foundations of Language Development A Multidisciplinary Approach  Volume I. New York, San Francisco, London: Academic Press, 1975. Garnica, Olga K. The development of phonemic speech perception. In T.E. Moore (Ed.) Cognitive Development and The Acquisition of Lan- guage. New York: Academic Press, 1973, pp. 215-223. Graham, A.W., and House, L.W. Phonological oppositions i n children: A perceptual study. Journal of The Acoustical Society of America, 1971, 49, 559-566. Ingram, David Issues i n child phonology. In Donald M. Morehead and Ann E. Morehead (Eds.) Normal and Deficient Child Language. Baltimore, London, Tokyo: University Park Press, copyright 1976. Ingram, David The relationship between comprehension and production. In Richard L. Schiefelbusch and Lyle L. Lloyd (Eds.) Language  Perspectives - Acquisition Retardation and Intervention. Baltimore, London, Tokyo: University Park Press, copyright 1974. Jakobson, Roman The sound laws of child language and their place i n general phonology. In Aaron Bar-Adon and Werner F. Leopold (Eds.) Child Language A Book of Readings. Englewood C l i f f s , New Jersey: Prentice Hall Inc., copyright 1971. - 60 -Karnil, Michael L., and Rudegeau, Robert E. Methodological improvements in the assessment of phonological discrimination in children. Child Development, September 1972, Vol. 43, Number 3, 1087-1091. Klein, Harriet "The Relationship between Perceptual Strategies and Productive Strategies in learning the Phonology of Early Lexical Items". Ph. D. Thesis Columbia University, 1977. Kornfeld, J.R. "Theoretical Issues in Child Phonology". Papers from the Seventh Regional Meeting, Chicago Linguistics Society, 1971, 454-468. Kuhl, P.K., and Miller, J.D. Speech perception by the chinchilla: Voiced-voiceless distinction in alveolar plosive consonants. Science, 1975, 190, 69-72. Menyuk, Paula The acquisition and development of language. In John C. Wright (Ed.) The Prentice Hall Series in Developmental Psycho- logy. Englewood Cliffs, New Jersey: Prentice Hall Inc., copy-right 1971. Miller, G.A., and Nicely, P.E. An analysis of perceptual confusions among english consonants. Journal of The Acoustical Society of America, 1955, 27, 338-352. Miller, J.D., Weir, C.C., Pastore, R.E.,. Kelly, W.J., and Dooling, R.T. Discrimination and labelling of noise-buzz sequences with varying noise lead times: An example of categorical perception. Journal of The Acoustical Society of America, 1976, 60_, 410-417. Morse, P.A. The discriniination of speech and nonspeech stimuli in early infancy. Journal of Experimental Child Psychology, 1972, 14, 477-492. Moskowitz, Breyne Arlene The acquisition of fricatives: A study in phonetics and phonology. Journal of Phonetics, 1975, 3_, 141-150. Palermo, David S., and Molfese, Dennis L. Language acquisition from age five onward. In Freda Rebelsky and Lynn Dorman (Eds.) Child Develop- ment and Behaviour. Alfred A. Knopf, Inc., copyright 1974. Shvachkin, N.K. The development of phonemic speech perception in early childhood. In Charles A. Ferguson and Daniel I. Slobin (Eds.) Studies in Child Language Development. Holt, Rinehart and Winston, Inc., copyright 1973. Smith, Neilson V. The Acquisition of Phonology A Case Study. Cambridge University Press, 1973. Stevens, Kenneth N., and House, Arthur S. Speech perception. In Jerry V. Tobias (Ed.) Foundations of Modern Auditory Theory Volume II. N^ew York and London: Academic Press, Inc., copyright 1972. - 61 -Templin, Mildred C. Certain Language S k i l l s i n Children. Minneapolis: The University of Minnesota Press, copyright 1957. Thibodeau, Linda M., and Sussman, Harvey M. Performance on a test of categorical perception of speech i n normal and communication disordered children. Journal of Phonetics, 1979, 375-391. Thorpe, Christine Ann "Use of The Analysis by Synthesis Model of Speech Perception by Children Acquiring The Sound System of Language". M. Sc. Thesis University of B r i t i s h Columbia, 1977. 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items