"Medicine, Faculty of"@en . "Audiology and Speech Sciences, School of"@en . "DSpace"@en . "UBCV"@en . "Tsui, Vicki Chi Ki"@en . "2009-07-09T18:56:06Z"@en . "2000"@en . "Master of Science - MSc"@en . "University of British Columbia"@en . "It is a well-established fact that infants are born with the ability to discriminate a\r\nuniversal set of phonetic categories. However, the ability to consciously perceive phonetic\r\ncontrasts which are not relevant in one's native language gradually decreases within the first year\r\nof life. The objective of this investigation is to determine whether non-native category changes\r\nare still recognized by adults at early levels of auditory processing even though they find it\r\ndifficult, or impossible, to discriminate them behaviourally.\r\nElectrophysiological and behavioural responses to native (/ba/ vs. /da/) and non-native\r\ncontrasts (/da/ vs. /Da/) were obtained from native English listeners. The component of the\r\nevent-related potential (ERP) of particular interest is the mismatch negativity (MMN), generated\r\nin the auditory cortex. The MMN is a waveform difference which occurs, independent of\r\nattention, in response to a discriminable change in a repeated sound stimulus. Therefore, if the\r\nability to perceive the non-native category change is preserved at the level of the primary\r\nauditory cortex, we would expect to find an MMN response to that change. For the passive ERP\r\nmeasures, stimuli were presented in an oddball paradigm, with conditions counterbalanced for\r\norder. The behavioural condition consisted of a forced-choice task in which participants were\r\npresented with various stimulus pairs and asked to decide whether the stimuli within the pair\r\nwere \"same\" or \"different\".\r\nBehavioural results confirm that the majority of the English listeners could not\r\ndiscriminate the non-native category change. Performance in the native category was\r\nsignificantly better. Surprisingly, ERP results not only indicate that there is a distinct MMN in\r\nresponse to the non-native category change, but that this response actually tends to be larger in\r\namplitude than the response to the native category change. The order of presentation of the\r\nstimuli also has a significant effect on the amplitude of the MMN. The MMN response appears\r\nto be stronger when /da/ is the standard stimuli and when /Da/ is the deviant. It is clear from\r\nthese results that the auditory cortex is indeed able to recognize non-native phonetic category\r\nchanges even when they are not consciously perceived by the listener. These findings support\r\nthe theory that the change in categorical perceptual ability from infancy to adulthood involves a\r\nfunctional reorganization rather than a permanent loss of ability."@en . "https://circle.library.ubc.ca/rest/handle/2429/10486?expand=metadata"@en . "4180994 bytes"@en . "application/pdf"@en . "M I S M A T C H NEGATIVITY CORTICAL E V E N T - R E L A T E D POTENTIAL MEASURES OF CROSS-LINGUISTIC PHONEME PERCEPTION by VICKI C.K. TSUI B.Sc , The University of British Columbia, 1997 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF M A S T E R OF SCIENCE in THE F A C U L T Y OF G R A D U A T E STUDIES (School of Audiology and Speech Sciences) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH C O L U M B I A April 2000 \u00C2\u00A9 Vicki C. K. Tsui, 2000 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. T5 The University of British Columbia Vancouver, Canada Date A^\\ ^ ( ^ C C Q DE-6 (2/88) 11 A B S T R A C T It is a well-established fact that infants are born with the ability to discriminate a universal set of phonetic categories. However, the ability to consciously perceive phonetic contrasts which are not relevant in one's native language gradually decreases within the first year of life. The objective of this investigation is to determine whether non-native category changes are still recognized by adults at early levels of auditory processing even though they find it difficult, or impossible, to discriminate them behaviourally. Electrophysiological and behavioural responses to native (/ba/ vs. /da/) and non-native contrasts (/da/ vs. /Da/) were obtained from native English listeners. The component of the event-related potential (ERP) of particular interest is the mismatch negativity (MMN), generated in the auditory cortex. The M M N is a waveform difference which occurs, independent of attention, in response to a discriminable change in a repeated sound stimulus. Therefore, if the ability to perceive the non-native category change is preserved at the level of the primary auditory cortex, we would expect to find an M M N response to that change. For the passive ERP measures, stimuli were presented in an oddball paradigm, with conditions counterbalanced for order. The behavioural condition consisted of a forced-choice task in which participants were presented with various stimulus pairs and asked to decide whether the stimuli within the pair were \"same\" or \"different\". Behavioural results confirm that the majority of the English listeners could not discriminate the non-native category change. Performance in the native category was significantly better. Surprisingly, ERP results not only indicate that there is a distinct M M N in response to the non-native category change, but that this response actually tends to be larger in amplitude than the response to the native category change. The order of presentation of the Ill stimuli also has a significant effect on the amplitude of the M M N . The M M N response appears to be stronger when /da/ is the standard stimuli and when /Da/ is the deviant. It is clear from these results that the auditory cortex is indeed able to recognize non-native phonetic category changes even when they are not consciously perceived by the listener. These findings support the theory that the change in categorical perceptual ability from infancy to adulthood involves a functional reorganization rather than a permanent loss of ability. iv T A B L E OF CONTENTS A B S T R A C T ii T A B L E OF CONTENTS iv LIST OF FIGURES . . . viii PREFACE ix A C K N O W L E D G E M E N T S x DEDICATION xi CHAPTER 1: CROSS-LINGUISTIC SPEECH PERCEPTION 1 C A T E G O R I C A L PERCEPTION OF SPEECH SOUNDS 2 CROSS-LINGUISTIC PHONEME PERCEPTION 4 Adult Perception of Non-Native Phonemes 4 Infant Speech Perception 6 Second Language Learners 8 Adult Training Studies 10 THEORIES OF D E V E L O P M E N T A L CHANGES IN PHONEME PERCEPTION . . 12 Maintenance/Loss Model 12 Native Language Magnet Theory 13 Perceptual Assimilation Model 14 Functional Reorganization Model 15 S U M M A R Y 16 V CHAPTER 2: M I S M A T C H NEGATIVITY CORTICAL E V E N T - R E L A T E D POTENTIAL 18 RELATIONSHIP TO OTHER ERP COMPONENTS 21 APPLICATION OF M M N IN P H O N E M E PERCEPTION STUDIES 22 M M N STUDIES OF P H O N E M E PERCEPTION 23 THESIS STUDY 26 CHAPTER 3: M I S M A T C H NEGATIVITY R E V E A L S CORTICAL DISCRIMINATION OF NON-NATIVE CONSONANT CONTRASTS IN A D U L T S 28 INTRODUCTION 29 METHOD 33 Participants 33 Stimuli 34 EEG Recording Sessions 35 Behavioural Discrimination Sessions 37 Procedure 38 Data Analysis 38 RESULTS '. 40 Behavioural Discrimination Results ; 40 Native vs. Non-Native 40 Control vs. Native 40 Control vs. Non-Native 41 M M N Results 41 Native vs. Non-Native 43 Control vs. Native 43 Control vs. Non-Native 43 Late Negativities 44 DISCUSSION 44 Behavioural Discrimination 44 M M N Recordings 45 Late Negativities 51 vi Relationships Between M M N and Behavioural Discrimination 51 Conclusion 52 REFERENCES 54 FOOTNOTES 65 APPENDIX A H A N D U S A G E QUESTIONNAIRE 77 APPENDIX B INDIVIDUAL SUBJECTS' B E H A V I O U R A L RESPONSE A C C U R A C Y A N D REACTION TIME RESULTS 79 APPENDIX C INDIVIDUAL SUBJECTS' M I S M A T C H NEGATIVITY MEASURES AT Fz 81 APPENDIX D A L T E R N A T I V E DIFFERENCE W A V E CALCULATIONS 84 APPENDIX E M I S M A T C H NEGATIVITY MEASURES OF A WITHIN-CATEGORY CONTRAST: /ba/ #2 vs. /ba/ #5 92 vii LIST OF T A B L E S Table 1: Formant Frequency Values of Stimuli '. 66 Table 2: ERP Recording Conditions 67 Table 3: Behavioural Response Accuracy and Reaction Times 68 Table 4: Results of Repeated Measures A N O V A s for Behavioural Response Comparisons 69 Table 5: M M N Peak-amplitude and Latency Measures at Fz 70 Table 6: Results of Repeated Measures A N O V A s for M M N at Fz 71 viii LIST OF FIGURES Figure 1: Control Condition: Grand Mean Difference Waves 73 Figure 2: Native Condition: Grand Mean Difference Waves 74 Figure 3: Non-Native Condition: Grand Mean Difference Waves 75 Figure 4: Grand Mean Difference Waves at Fz: A l l Conditions 76 ix P R E F A C E This thesis consists of three chapters: (1) a review of the literature on cross-linguistic phoneme perception, (2) a review of the literature on mismatch negativity, and (3) a research paper in which mismatch negativity is used as a tool to study cross-linguistic phoneme perception. The research paper is intended to be submitted for publication. There is necessarily overlap between the research paper in Chapter 3 and the information in Chapters 1 and 2. A C K N O W L E D G E M E N T S x I would like to express my deepest gratitude to David R. Stapells, without whom this thesis would never have been completed. Thank you for your guidance, constant encouragement, and the endless hours of editing, re-editing, and then some more editing. Also, a special thanks to Michelle McCaughran for putting up with all the long meetings. Your patience and understanding were very much appreciated. Janet F. Werker and Rushen Shi deserve many thanks for their contributions to this paper. Their ideas and suggestions have been invaluable. Thanks also, Janet, for generously providing the stimuli for the project. Thank you to Chengyuan Wu for frequently helping me with many technical aspects of carrying out and analyzing the data for a mismatch negativity experiment and Lisa Tremblay for her assistance in the recruitment, recording, and analysis of the last few subjects. Many thanks to the subjects who endured hours of ERP recordings and button-pressing. I am forever grateful to my parents for their unconditional love and support. Most of all, I would like to thank Tom Ly for sharing with me the good times and giving me the strength to get through the bad. Words cannot express how much your love and encouragement mean to me. This research was made possible by NSERC, which provided financial assistance for the project through a research grant to D.R. Stapells and a post-graduate scholarship to myself. XI DEDICATION This work is dedicated to my mother, Ada M.C . Tsui. CHAPTER 1: CROSS-LINGUISTIC SPEECH PERCEPTION 2 C A T E G O R I C A L PERCEPTION OF SPEECH SOUNDS Human speech consists of a continuous stream of rapidly changing acoustic cues. Speech perception involves splicing this stream into individual sound units and accurately identifying the sounds that the speaker intended in order to interpret meaning. This is quite a remarkable feat given the lack of invariance within the acoustic signal and the fact that it can be accomplished at rates as fast as 30 speech sounds per second (Liberman, Cooper, Shankweiler, & Studdert-Kennedy, 1967). It has been demonstrated that the same phoneme (speech sound) produced in different contexts or by different speakers can be highly variable in its acoustic properties, and yet, the identification of phonemes is highly consistent across listeners (Borden, Harris, & Raphael, 1994, pp. 184-233). This phenomenon is known as perceptual constancy. Perceptual constancy in the categorization of phonemes can be compared to colour constancy in the visual domain, in which colours of objects are perceived consistently even though the wavelengths of light reflected off of them vary under different lighting conditions (Land, 1977). Acoustic properties of phonemes are influenced by their sound environments due to the co-articulation of sounds in running speech (Borden et al., 1994, pp. 136-138; Ohman, 1966). There is also a great deal of speaker-to-speaker variability in the production, and therefore the acoustic properties, of phonemes (Kent & Read, 1992; Liberman et al., 1967). Thus, there is no one-to-one relationship between the acoustic signal and phonemic interpretation of that signal. It is now known that the identification of phonemes involves categorical perception, in which phonemic category boundaries are perceptually imposed onto an acoustic continuum (Liberman et al., 1967; Liberman, Harris, Hoffman, & Griffith, 1957). Two types of behavioural tasks have been used to document the categorical perception of speech sounds: identification 3 tasks and discrimination tasks. In identification tasks, subjects are asked to assign phonetic labels (provided by the researchers) to an acoustic continuum of minimally contrastive speech stimuli which differ in a single acoustic parameter in step-wise fashion. For example, a voice-onset time (VOT) continuum of /ba/ to /pa/ would hold all acoustic variables constant except VOT which gradually increases in equal increments from one token to the next. The identification function shows that although acoustic changes within the continuum are gradual, the perceptual changes from one category to the next are not. That is, there is an abrupt perceptual boundary at which the identification of the stimuli changes from one phonetic label to the other (Liberman et al., 1967; Liberman et al., 1957). For discrimination tasks, subjects are asked to detect changes in speech stimuli. These tasks involve the same acoustic continua used in the identification tasks. Listeners' ability to discriminate minimally contrastive sound pairs separated by equal acoustic differences is measured. Discrimination tasks reveal that there is a large increase in the ability to discriminate a small acoustic change when the contrasting pair come from either side of an identification boundary (i.e., across-category contrasts), whereas discrimination of contrasts labelled as the same phonetic segment (i.e., within-category contrasts) is poor (Liberman et al., 1967; Liberman etal., 1957). It has been been suggested that the categorical perception of speech sounds provides evidence that humans have a specialized speech processor (Liberman et al., 1967; Liberman & Mattingly, 1989). Others, however, have found that the phenomenon of categorical perception is not unique to speech stimuli (Burns & Ward, 1973; Pisoni, 1977), nor is the ability unique to humans (Kuhl & Miller, 1978). 4 CROSS-LINGUISTIC PHONEME PERCEPTION Although there is a limited set of sounds that can be used in speech, most languages differ in the set of sounds, or phonemes, that are used to distinguish meaning. Every language in the world has its own phonemic inventory consisting of a subset of the universal inventory. In addition to sounds that phonemic contrasts, phonetic differences may also occur within a language^ Phonetic variations which differ in acoustic properties but have the same underlying phonemic representation are allophones. Allophones occur in different sound environments but are not used contrastively to signify meaning. Adult Perception of Non-Native Phonemes Studies have shown that adults typically have a great deal of difficulty discriminating phonemic contrasts which do not exist in their native languages (Werker, Gilbert, Humphrey, & Tees, 1981; Werker & Lalonde, 1988; Werker & Tees, 1984b). Using paired discrimination tasks, one study showed that English-speaking listeners were unable to discriminate the glottalized velar vs. glottalized uvular place contrast in /k7 and /q'/ which occur in the Salish language, Nthlakampx, but not in English (Werker & Tees, 1984b). Similarly, another study showed that the perceptual discrimination of Hindi contrasts which are not phonemic in English such as retroflex versus dental \"t\" by English speakers was poorer than by Hindi speakers (Werker et al., 1981). The above does not imply that it is impossible for adults to discriminate any sounds that are not found in their native language. Werker and Logan (1985) showed that adult discrimination of non-native contrasts can be improved by altering the conditions of the task, for example, shortening the inter-stimulus interval (ISI). Best and McRoberts (1988) showed that English-speaking adults (and infants) are able to distinguish the difference between apical and 5 lateral clicks, even though neither of these two sounds are present in English. However, it can be argued that linguistic input has not led to language-specific perception in this case because the clicks are not actually speech sounds. Another question of interest is whether or not the amount or variability of linguistic input will affect one's ability to categorically perceive sounds which one has not been exposed to. If it did, we would expect that listeners who are multi-lingual would perform better in non-native phoneme discrimination tasks than those who are mono-lingual. Werker compared the performance of monolingual, bilingual, and trilingual adults in a perceptual discrimination task (Werker, 1986). The stimuli for the task were taken from languages that none of subjects had been exposed to: Hindi and Nthlakampx. The reasoning behind the study was that if the subjects who spoke more languages performed better on the task, it would suggest that increased linguistic experience, although not language specific, would lead to an increase in \"perceptual flexibility\". The results of the study showed no benefit of additional linguistic knowledge in the discrimination task, leading to the conclusion that the ability to discriminate phonetic contrasts does require linguistic experience of that specific language. Aside from requiring specific experience of sound contrasts, a recent study has shown that passive exposure to speech sounds alone is not enough to maintain the ability to discriminate them (Pegg & Werker, 1997). In this study, English listeners in three age groups: (1) 6-8 months, (2) 10-12 months, and (3) adults were tested on their ability to discriminate the phonetic contrast of [da] vs. [t\"a]. These two sounds represent allophonic variations which exist in English but are not used contrastively, i.e., they are not a phonemic distinction. Theoretically, if discrimination abilities can be maintained with passive exposure to sounds, we would expect that all three age groups would be able to discriminate the contrast. However, if it is necessary for 6 the sounds to have a phonological role in the native language by being actively used to contrast meaning, we would expect a decline in the ability to discriminate the allophones in the two older groups. The results of the study showed that the younger infants were able to discriminate the contrast but the older infants could not. The adult listeners, while able to discriminate the sounds at better than chance levels, did not discriminate them as well as would be expected for native phonemic contrasts. Infant Speech Perception Using cleverly designed behavioural paradigms for infants, the ability to perceive phonemes categorically has been demonstrated in early infancy, long before a child produces his first words (Aslin, Pisoni, Hennessy, & Perey, 1981; Eimas, Siqueland, Jusczyk, & Vigorito, 1971; Streeter, 1976; Trehub, 1976). Eimas and colleagues found that infants aged one to four months are able to discriminate small differences in voice onset time if they cross the voiced-voiceless category boundary, but are unable to discriminate differences of equal acoustic differences if they occur within the same phonemic boundary (Eimas et al., 1971). This finding was further developed to show that not only are infants capable of categorical perception, but they are even capable of perceiving phonetic contrasts which do not exist in their native language (Aslin et al., 1981; Streeter, 1976). Werker and colleagues compared the abilities of English adults, Hindi adults, and 7-month-old English infants to discriminate consonant contrasts that are found in Hindi but not English (Werker et al., 1981). They tested the infants using a conditioned head-turn paradigm and found that infants demonstrated a remarkable ability to discriminate speech sounds belonging to different phonetic categories even if the categories did not exist in their native language. From this, and later studies exploring different languages (Werker & Tees, 1984b), 7 they concluded that infants are born with the ability to perceive and discriminate all the phonemes which exist in all of the world's languages. That is, humans are born \"universal listeners\" and therefore capable of learning any language (Eimas, 1975; Strange & Jenkins, 1978). The knowledge that adults often have difficulty in discriminating phonemic contrasts which do not exist in their native language leads to the question of when this perceptual change from universal listener to language-specific listener might occur. More studies of non-native discrimination ability were conducted, using children of different age levels. Finally, it was concluded that the change occurs within the first year of life (Werker & Tees, 1984b). In their study, Werker and Tees tested groups of children: 6-8 months, 8-10 months, and 10-12 months in their ability to discriminate non-native consonant contrasts. They found that performance gradually declined with age and the oldest group performed at the same level as the English speaking adults. Therefore, the strong influences of linguistic experience take effect by 10-12 months of age. Similar research has been conducted using the conditioned head-turn procedure to test the ability of young infants to discriminate non-native vowel contrasts (Kuhl, Williams, Lacerda, Stevens, & Lindblom, 1992; Polka & Werker, 1994; Trehub, 1976). The findings of these studies indicate that the change from language-general to language-specific perception occurs even earlier for vowels, by approximately six months of age. Another study has shown that infants 5.5 to 6.5 months of age are capable of sorting vowels into their phonemic categories even when confusing information such as change of speaker or pitch is introduced into the stimuli (Kuhl, 1979). It has been posited that the earlier acquisition of vowels may be due to 8 their greater perceptual salience (Werker & Polka, 1993). They have higher intensities and longer durations than consonants in running speech and they carry intonation. Another line of research has shown that the effects of early linguistic experience go beyond the level of individual phonemes. Languages differ in their phonetic and phonotactic patterns and constraints. These are the rules that govern the ordering of sounds in a speech stream. Jusczyk and colleagues tested infant listening preferences by using unfamiliar words that either followed or violated the phonotactic constraints of the native language of their subjects (Jusczyk, Friederici, Wessels, Svenkerud, & Jusczyk, 1993). They found that 6-month-old infants did not show any preferences but 9-month-old infants would listen longer to words which fit into the sound patterns of their native language. In addition to this, both groups of infants preferred listening to words which were prosodically patterned like their native language (Jusczyk et al., 1993). Therefore, phonology and prosody are also influenced by early language input. Second Language Learners Older children or adults learning a second language (L2) not only have difficulties in production, they also have problems perceiving the phonemes in L2. As indicated by adult cross-linguistic phoneme perception studies, phonemes or linguistic contrasts in L2 either may not exist in the person's first language (Ll) or may have a different prototype in L l . For example, both English and Spanish listeners divide the VOT continuum into two categories: voiced and voiceless, but the boundary separating the category is in a different place for the two languages. In English, the boundary occurs at approximately +20 ms to +40 ms VOT, whereas Spanish listeners set the boundary at about -5 ms VOT (Abramson & Lisker, 1970; Williams, 1977). Rvachew and Jamieson (1995) identify ways in which a person's native language system would 9 interfere with their perception of L2 contrasts. For example, contrasting sounds in L2 may be collapsed into a single category in L l . Such is the case of Hindi retroflex IT I and alveolar Ixl being perceived as the same \"t\" by English listeners (Werker et al., 1981). Alternatively, contrasting pairs of L2 phonemes which do not exist in L l may be assimilated as opposing members of an L l contrast. An example of this would be the assimilation of the Ethiopian ejective contrast /p'/ and /t'/ into contrastive categories in English /p/ and Ixl by English listeners (Best, 1990). Thus, the phonemes of L2 for second language learners are often \"accented\" in both perception and production. These findings form the basis of Best's Perceptual Assimilation Model (see below) (Best, 1995). Rochet (1995) found evidence suggesting that accented L2 speech in second language learners is often a result of perceptual errors rather than articulatory difficulties. He suggests that the targeted L2 phonemes are mistakenly assigned to existing L l phonetic categories. Rvachew and Jamieson (1995) also agree that there is a relationship between perception and production of L2 phoneemes. They propose that one must be able to make the perceptual distinction before correct production is possible but perceptual mastery does not guarantee unaccented production. However, there have been cases of L2 learners who can consistently produce but not perceive correctly, such as Japanese speakers learning the English \"1\" and \"r\" contrast (Sheldon & Strange, 1982). Tees and Werker (1984) explored different factors which might facilitate L2 acquisition. They compared three groups of English-speaking adults with varying degrees of exposure to Hindi in their ability to discriminate the Hindi-specific voicing and place contrasts. Subjects who had studied the language for five years or more were able to discriminate both linguistic contrasts. Those who had studied Hindi as a second language for a year had improved 10 discrimination of the voicing contrast compared to those with no experience at all. But this latter group was unable the discriminate the place contrast. The study's most significant finding was that subjects who had had very early exposure to the Hindi contrasts, but then had no further input of this kind into adulthood, were able to discriminate both contrasts, even though they had not studied Hindi as a second language. Adult Training Studies There is a body of research which examines the efficacy of training subjects to discriminate non-native contrasts. These studies provide evidence for or against the existing models used to explain the developmental changes we see in phonemic perception (see below). Using training studies, we can explore how malleable, or flexible, the linguistic perceptual system is. There can also be a practical application of any findings from these experiments. They allow us to trial different techniques in perceptual training so that we may increase training efficacy and help reduce the difficulty of learning a second language. Tees and Werker designed a training program in which subjects listened to sequences of examples of one non-native sound and then listened to examples of the contrasting categories (Tees & Werker, 1984). Short-term training was successful in teaching English-speaking adults to discriminate the Hindi voicing contrast. The subjects were also able to maintain the improvement that they made in discriminating this sound pair. However, training efficacy was much poorer for the Hindi place of articulation contrast. Fewer people were able to learn this contrast and, of those who did, even fewer were able to maintain it. Jamieson and Morosan (1986) suggested that it is possible to successfully train listeners to discriminate non-native phonemes if three criteria are met in the training task. First, the training stimuli must have the sound to be learned embedded in an appropriate acoustic and 11 phonetic context so that the subject gets an idea of how it is used in regular speech. Secondly, an identification task with feedback should be used instead of a simple discrimination task. This . was believed to promote phonetic categorization. Finally, the training stimuli should gradually shift from extreme categorical exemplars, where attention is on the critical acoustic parameters, towards more variable stimuli with increased acoustic uncertainty. This should facilitate the generalization of the training to natural speech. Using their technique, these researchers were successful in training French speakers to discriminate the voiced vs. voiceless contrast of the English dental fricatives. However, though general improvement in perception following completion of the training tasks was noted, there was always a wide range of individual variability which could not be explained by the amount of linguistic exposure alone. It is possible that some people inherently have a higher capacity for learning different speech sounds. These phoneme training studies prove that it is possible to train adult listeners to perceive and discriminate non-native sounds that previously they were unable to. The methods that have been developed may help improve L2 learners' perception, and therefore, comprehension of their second language. But does the perceptual training affect their articulation of non-native speech sounds? It has been shown that perceptual learning does indeed improve speech production as well (Bradlow & Pisoni, 1997). In this study, Japanese speakers learning English as L2 were trained in the perceptual contrast between Ixl and IM. After the training, the perceptual learning of the phonemic contrast was generalized to the production of the sounds. This provides evidence for the hypothesis that there is a strong link between the ability to perceive and produce L2 phones (Rvachew & Jamieson, 1995). 12 THEORIES OF D E V E L O P M E N T A L C H A N G E S FN P H O N E M E PERCEPTION Maintenance/Loss Model This model was hypothesized as a result of early findings that there is a developmental change in the ability to discriminate non-native phonemic contrasts (Abramson & Lisker, 1970; Aslin et al., 1981; Lasky, Syrdal-Lasky, & Klein, 1975; Lisker & Abramson, 1970; Werker et al.\u00E2\u0080\u009E 1981). The Maintenance or Loss theory of how this change occurs is based on experimental findings that a critical period for sensory exposure exists in the visual system, (Wiesel, 1982; Wiesel & Hubel, 1963). These experiments showed monocular sensory deprivation causes the active synapses from the remaining eye to gradually displace the inactive ones of the deprived eye. If visual stimulation is reestablished to the deprived eye after a critical period has passed, its sensitivity to visual stimuli is lost because its synapses to the visual cortex have been displaced by the other eye. In other words, experience, in the form of sensory stimulation, is required during the critical period in order to maintain ability or sensitivity. The Maintenance or Loss model states that, similar to the visual system, humans are born with the ability to perceive all kinds of speech contrasts but experience with a phoneme is required in order to maintain the ability to discriminate it (Eimas, 1975; Strange & Jenkins, 1978). If the exposure to a sound does not come within a certain critical period, our abilities to perceive it would be gone completely and permanently. This model assumes that the change from language-general to language-specific perception is on a sensory level (Eimas, 1975; Strange & Jenkins, 1978) but the data now suggests that the change is actually on an attentional or perceptual level (Werker, 1994; Werker & Tees, 1984b; Werker & Tees, 1992). Existing data on the ability of adults to learn a second language, especially with appropriate perceptual training (Jamieson & Morosan, 1986; Lively, Pisoni, Yamada, Tohkura, & Yamada, 1994; Tees & 13 Werker, 1984), have largely discredited this theory. In addition to this, it has been demonstrated that adult listeners sensitivity to non-native contrasts can be improved without training by altering testing conditions, for example, by shortening ISI (Werker & Logan, 1985; Werker & Tees, 1984a). Finally, some non-native sounds, like the Zulu click, are easily discriminated by adults with no previous exposure whatsoever (Best & McRoberts, 1988). Native Language Magnet Theory In a number of studies, Kuhl and colleagues have noted the existence of a \"perceptual magnet effect\", whereby a prototypical exemplar of a phonemic sound category has a tendency to pull other, less ideal, members of that category towards itself perceptually (Kuhl, 1993; Kuhl & Iverson, 1995; Kuhl et al., 1992). Less prototypical exemplars fail to have the same effect on their fellow category members, hi perceptual discrimination tasks using an oddball paradigm (where subjects respond when the category changes from the background sound to a different one), it was found that if prototypical examples are played as the standard background stimulus, less prototypical examples which followed tended to blend in with the background (Grieser & Kuhl, 1989; Kuhl, 1991). That is, subjects, both adults and six-month-old infants, did not discriminate them. When the ordering of presentation was changed, with less prototypical exemplars presented as the standard stimulus, the subjects were more likely to make the discrimination (Grieser & Kuhl, 1989). Kuhl and colleagues further proposed that these prototypical magnets are language specific. Only those prototypes found in the native language of subjects show the magnet effect because it is linguistic experience that causes the magnet effect to develop (Kuhl & Iverson, 1995). In this theory, humans are born with general auditory processing mechanisms which include an innate ability to divide the acoustic space for speech sounds into phonetic categories. Sensitivity to these natural auditory boundaries in the acoustic 14 space has also been demonstrated in non-humans (Kuhl & Miller, 1978). What is unique to humans is the development towards language specificity of the perceptual boundaries with linguistic experience. By the time infants are six months of age, linguistic experience begins to influence phonetic categories and their prototypes. Phonetic boundaries that are irrelevant to the native language are then erased (Kuhl, 1993; Kuhl et al., 1992). This model is able to account for Rvachew and Jamieson's (1995) findings that L2 phonemes will sometimes be assimilated into L l categories. So far, this theory appears to be quite strong, but much more research is required using other languages to prove the robustness of the model. One weakness of this theory is that it fails to account for non-native consonants that can be easily discriminated by people who have never been exposed to them, such as Zulu clicks (Best & McRoberts, 1988). Perceptual Assimilation Model The perceptual assimilation model states that non-native phonemic contrasts that cannot be assimilated into native sound categories are unaffected by language-specific changes in phonemic perception (Best, 1995). This would explain why English speakers have no trouble discriminating the two clicks in Zulu because there is nothing similar to these sounds in the English language, either in acoustic or articulatory properties. According to this model, these sounds would be designated as \"non-assimilable\". In contrast, any phonemes which can be assimilated into existing native language categories will be assimilated. Best (1995) even provides patterns in which assimilation of L2 contrasts into L l can occur. For example, Two-Category Assimilation represents a situation where two non-native contrastive phonemes are each assimilated into a different contrastive native category, so that the discrimination of the contrast should be perfect even though the production of the non-native sounds is accented. Category-Goodness Difference is the case where both sounds are placed in the same native 15 category, but one is more acceptable than the other. In this situation, discrimination should still be fair or good. Single-Category Assimilation is when both non-native sounds are grouped into the same native category, neither of which are ideal exemplars. In this case, discrimination would be poor. This theory seems to incorporate some ideas of the Motor Theory of speech perception (Liberman et al., 1967) in that it assumes that perception is based on articulatory gestures. That is, in speech perception, sounds are assigned to different categories according to how they might be produced. However, this theory fails.to explain the instances in which subjects are capable of perceiving but not producing particular L2 phonemes (Rvachew & Jamieson, 1995). Functional Reorganization Model Although there does appear to be a critical period for linguistic experience to maintain perceptual contrasts, it has been proven possible, with enough training, to learn non-native phonemic contrasts after this period of maximal sensitivity is long past. The Functional Reorganization Model is a revision of the Maintenance or Loss Model which accounts for this finding. In this revised model, the decrease in ability to perceive non-native phoneme contrasts viewed not as a loss but as a change in ability. With linguistic experience, human phonemic perception develops from a state of general capabilities (universal listeners in infancy) to more efficient, language-specific processing (Werker & Pegg, 1992). Rather than assuming that there is a fundamental rewiring of the auditory system, this theory postulates that the change actually occurs in selective attention or the perception of sounds (Werker, 1995). In other words, there is a change in what the child is sensitive to. Sounds which the child is not exposed to early on (i.e., not in the native language) are functionally \"ignored\" because they are irrelevant to language learning. The change in the sensitivities of the perceptual system reflects a need in a developing 16 linguistic system to concentrate only on the sounds which distinguish meaning in the language to be acquired. This would lighten the cognitive load of the first language learner and allow him/her to ignore irrelevant acoustic cues. These non-native phones are not permanently lost. They simply lie dormant until they are stimulated enough to resurface perceptually. This would explain how adults can learn to perceive L2 sounds with training. This model also fails to account for all of the data on cross-linguistic phoneme perception. For example, if we assume that perceptual flexibility is maintained, we cannot account for the fact that there is a critical period, if not for learning language then at least for the ease of learning a second language. For now, a detailed theory that accounts for all the data on the effects of linguistic experience on the development of phonemic perception has yet to be developed. S U M M A R Y This review represents only a small portion of the wealth of existing information regarding the effects of linguistic input on our perception of speech sounds. These behavioural studies document performance ability in listeners with a variety of linguistic experience and using a variety of tasks such as discrimination, categorization, and identification. In summary, it has been shown that during the course of development, there is a functional change in phonemic perceptual ability. This change occurs early in life and is strongly influenced by linguistic experience. Infants are born \"universal listeners\" but their phonemic perception becomes language-specific within the first year of life. After about 10-12 months of age,, the ability to discriminate sound contrasts that do not carry functional meaning in the native language greatly decreases. A number of theories have been suggested to account for this change. It is now fairly 17 evident that these changes do not involve a permanent loss of ability to perceive non-native phonemic contrasts. However, the neurophysiological nature of this change is still unclear. At which level of auditory processing does this change occur? Is there a fundamental rewiring of the auditory system so that we are no longer sensitive to phonemic contrasts that are irrelevant to speech processing in the native language? Or is it possible that adult listeners are able to perceive the non-native contrasts at early, pre-attentive levels of auditory processing even though they are unable to access the information on a conscious level? In hopes of answering these questions, another direction of research is just beginning to be explored. This involves the use of electrophysiological measures in the study of cross-linguistic phonemic perception. Using auditory event-related potentials (ERPs), it may be possible to obtain a better understanding of the level of perception at which developmental changes in speech perception occur. That is, whether the loss of the ability to perceive non-native sound contrasts occurs early, at the more sensory levels of auditory processing, or whether they do in fact occur later, at a subconsciously controlled attentional level as hypothesized (Werker & Tees, 1992). If this technique can be perfected, it may also be a simple and elegant way of studying perceptual abilities of infants without having to question their attention during the experiment. The following chapter will discuss how the mismatch negativity (MMN), a component of ERPs, may be used as a tool in phoneme perception research. > CHAPTER 2: MISMATCH NEGATIVITY CORTICAL EVENT-RELATED POTENTIAL 19 The M M N is a component of cortical ERPs that denotes an automatic, pre-attentive change detection generated in response to an acoustic change in a repetitive stimulus sequence (Naatanen, Gaillard, & Mantysalo, 1978; Naatanen, Simpson, & Loveless, 1982). M M N is elicited using an oddball paradigm in which an infrequent stimulus (the \"deviant\") is presented within a series of repeated \"standard\" stimuli (Naatanen et al., 1978). It manifests as a frontal negative wave measured in the deviant-minus-standard difference waveform. It typically peaks at 100-200 ms following the onset of the stimulus (Naatanen, 1995). The M M N has a fronto-central scalp distribution and shows an inversion in polarity at the mastoids when recorded using a nose-reference (Picton, Alain, Otten, Ritter, & Achim, in press). A number of M M N characteristics show that it is indeed a specific response to a change in acoustic parameters rather than simply a response caused by the activation of a different set of afferents to a new stimulus. M M N only occurs when there is a change from a standard to a different stimulus, not when a stimulus is presented alone at long ISIs (Naatanen, Paavilainen, Alho, Reinikainen, & Sams, 1987; Picton et al., in press). Similarly, M M N is not evoked by the initial stimulus within a series, even though neurons are not yet in a refractory state (Cowan, Winkler, Teder, & Naatanen, 1993). Also, the M M N latency is sensitive to the size of the difference between stimuli which cannot be explained by refractory effects (Picton et al., in press). Finally, M M N may be elicited by either decreases in intensity or changes in duration and even omission of a component of the standard stimulus (Naatanen, Paavilainen, Tiitinen, Jiang, & Alho, 1993a; Yabe, Tervaniemi, Reinikainen, & Naatanen, 1997). This provides further evidence that M M N must be a response to change rather than a stimulation of new afferents. The M M N is mainly generated in the primary auditory although other brain structures, such as the frontal lobe, also contribute in generating it (Alho, 1995; Scherg, Vajsar, & Picton, 20 1989). It is believed that it reflects change-detection via the comparison of a stimulus with the sensory memory trace left by the preceding one (Naatanen et al., 1978). Repetition of the standard stimulus leads to the build-up of a stronger memory trace. A strong relationship between the M M N response and auditory short-term memory is demonstrated by the fact that M M N amplitude generally increases with decreases in ISI. M M N quickly decays at ISIs longer than 2 seconds and no M M N is recorded at ISIs exceeding 10 seconds (Naatanen et al., 1987). However, at very short ISIs, M M N to speech stimuli may deteriorate (Kraus, McGee, Carrell, & Sharma, 1995; Lang et al., 1995). A large number of trials (upwards of 200 deviant stimuli) are necessary in order to record a M M N response (Kraus et al., 1995; Lang et al., 1995; Martin, Kurtzberg, & Stapells, 1999; Naatanen, 1995). Increasing the number of trials will improve the M M N recording by increasing the signal-to-noise ratio. The proportion of standard to deviant trials will also affect the M M N . The less frequent the occurrence of deviant stimuli, the larger will be the response. In order to ensure a robust M M N , probability of deviant occurrence should be between .1 and .2 (Lang et al., 1995). Also, stimuli should be presented in a sequence that prevents the occurrence of two consecutive deviants (Sams, Alho, & Naatanen, 1983) but need not be random (Scherg et al., 1989). The M M N is sensitive to small acoustic changes in auditory stimuli such as intensity (Naatanen et al., 1978), frequency (Sams, Paavilainen, Alho, & Naatanen, 1985), duration (Naatanen, Paavilainen, & Reinikainen, 1989), and location (Paavilainen, Karlsson, Reinikainen, & Naatanen, 1989). The amplitude and latency of the M M N response are directly related to the size of the acoustic change from the standard to the deviant stimulus (Lang et al., 1990; Sams et al., 1985; Tiitinen, May, Reinikainen, & Naatanen, 1994). M M N amplitude increases and 21 latency decreases as the acoustic distance between the standard and deviant stimuli increases, suggesting that larger changes are detected earlier and elicit a larger change-detection response. This effect has also been demonstrated in complex stimuli such as speech syllables (Aaltonen, Eerola, Lang, Uusipaikka, & Tuomainen, 1994; Maiste, Wiehs, Hunt, Scherg, & Picton, 1995; Naatanen, 1992, p. 137; Sharma, Kraus, McGee, Carrell, & Nicol, 1993). Finally, studies using non-speech (Naatanen, Schroger, Karakas, Tervaniemi, & Paavilainen, 1993b) and speech (Tremblay, Kraus, Carrell, & McGee, 1997) stimuli have demonstrated effects of training on the M M N , and thus plasticity of the central auditory change-detection system. RELATIONSHIP TO OTHER ERP COMPONENTS Because the M M N component of the ERP is relatively small, it is easily confounded or obscured by other ERP waves such as NI and N2b (Kraus et al., 1995). The NI is a stimulus-specific response which occurs before the M M N . The possibility of a differential wave due to stimulus differences (NI effects; Butler, 1968) can be controlled for by using alternative methods of subtraction to obtain the difference wave (Kraus, McGee, Sharma, Carrell, & Nicol, 1992; Stapells, in press). Instead of subtracting the standard from the deviant waveform within a condition (Maiste et al., 1995), an additional baseline recording of the deviant stimulus presented alone may be subtracted from the deviant waveform (Kraus et al., 1995; Kraus et al., 1992). Alternatively, directions of stimulus presentation can be counterbalanced and the response to a stimulus presented as a standard can then be subtracted from the response to the same stimulus as a deviant (Martin et al., 1999). Both of these other methods would control for stimulus differences, thereby isolating the change-detection response (Kraus et al., 1995). Decreasing the 22 ISI is another method of isolating the M M N from the N I . NI amplitude is reduced with shorter ISI whereas M M N can be enhanced by shorter ISI (Naatanen et al., 1987). The N2b-P3b complex is an attention-dependent response which closely corresponds to behavioural discrimination of changes in auditory stimuli (Picton, 1992). It is elicited when subjects actively attend to the stimuli. The negative component, N2b, overlaps in latency with M M N and is much larger in amplitude and easier to record than the M M N and can therefore easily obscure the M M N when the two are superimposed (Naatanen, 1988; Naatanen et al., 1982). Because the M M N is independent of attention and N2b, the easiest way to isolate M M N from N2b is to redirect the subject's attention away from the test stimuli. M M N and N2b differ in scalp topography with the maximum amplitude of N2b occurring more centrally. Also, unlike the M M N , N2b does not invert at the mastoid in nose-referenced recordings (Sams, Aulanko, Aaltonen, & Naatanen, 1990) because it is not generated in the auditory cortex. Recording multiple-channel EEGs allow monitoring for the presence of the P3b as indicated by its centro-parietal distribution and lack of polarity inversion at the mastoids (Stapells, in press). Blocks showing P3b presence should be rejected because they indicate that the subjects are paying attention to the test stimuli. APPLICATION OF M M N IN PHONEME PERCEPTION STUDIES M M N is an especially useful tool for studying developmental changes in phoneme perception because it represents discrimination of sounds made by the central auditory system. Because the M M N is a pre-perceptual measure of auditory processing, it could demonstrate a physiological response to a change in phonemes even when there is no measurable behavioural response (Alho, Sams, Paavilainen, Reinikainen, & Naatanen, 1989; Naatanen, 1992, p. 137-23 138). Therefore, it may be possible to determine whether the central auditory system is actually detecting changes in sound stimulus, when listeners are not able to discriminate them on a conscious level. The M M N response pattern can help determine whether the decrease in ability to discriminate non-native sounds is a physiological or a functional loss. If, in fact, there is a reorganization of perceptual ability at or before the level of the primary auditory cortex which makes adults unable to discriminate non-native phonemes, we would expect to find an M M N to contrasts which cross a native phonemic boundary but not to contrasts which cross non-native phonemic boundaries (i.e., contrasts which belong to the same category according to the listener). Conversely, if the ability to discriminate non-native contrasts is retained at lower levels of auditory processing, there should be no difference in the M M N in response to native vs. non-native phoneme changes, regardless of surface ability. M M N STUDIES OF PHONEME PERCEPTION M M N studies of the categorical perception of consonants have measured M M N responses to speech stimuli which involve across-category changes and within-category changes. The acoustic distance in each type of contrast is typically held constant. Across-category changes represent phonemic contrasts while the within-category changes represent purely acoustic contrasts. Behavioural studies typically show that listeners are better at discriminating across-category changes (Liberman et al., 1967; Liberman et al., 1957). The premise of these studies is that if the M M N response reflects the categorical perception of speech sounds, (i.e., phonemic processing) it should result in a larger M M N response to the across-category contrasts which exist in the listener's native language. Existing studies using these methods show that the 24 ' M M N response reflects acoustic, rather than phonemic processing (Maiste et al., 1995; Sharma etal., 1993). In a comparison of M M N to acoustically equidistant stimulus changes across and within articulatory place categories of /da/ through I gal, Sharma and colleagues reported the M M N was present in response to both types of contrasts, with no significant differences between contrasts (Sharma et al., 1993). Similarly, Maiste and colleagues measured behavioural discrimination and ERP recordings to stimulus contrasts from a /ba/ to /da/ continuum. Although behaviourally, subjects demonstrated categorical perception of the stimulus continuum, the ERP showed that M M N amplitude was dependent on acoustic changes alone and was not affected by whether or not a categorical boundary had been crossed (Maiste et al., 1995). Aaltonen and colleagues also found that M M N was dependent on acoustic changes with vowel contrasts (Aaltonen et al., 1994). In contrast to the above within-language studies, M M N studies of cross-linguistic vowel perception have shown that M M N amplitude does reflect categorical perception capabilities measured in behavioural tasks (Cheour et al., 1998b; Naatanen et al., 1997). Naatanen and colleagues recorded M M N responses of Finnish and Estonian listeners to vowel changes from Id to 161,161, or lol. These vowels are all prototypical in Estonian but only Id, 161, and lot are native to the Finnish language. This study found M M N to be present to both native and non-native vowel contrasts in both subject groups. In general, M M N amplitude increased as the F2 frequency difference between the vowels increased but, for the Finnish listeners, the M M N to a phonemic contrast involving two native prototypes was larger in amplitude than the M M N to the change from the native prototype, Id, to the non-prototype (i.e., non-native vowel), 161, even though the acoustic distance in the latter contrast is greater (Naatanen et al., 1997). Winkler and 25 colleagues also found that M M N reflected the effects of linguistic experience on behavioural discrimination (Winkler et al., 1999). In this study, ERPs to a Finnish vowel contrast which does not exist in Hungarian were recorded in three subject groups: native Finnish listeners, Hungarian listeners who had acquired Finnish as a second language, and Hungarian listeners with no Finnish experience. M M N was present in the first two subject groups but not in the subject group who had no experience with the Finnish vowel contrast, suggesting that M M N does indeed reflect the categorical perception capabilities measured behaviourally (Winkler et al., 1999). In contrast to the responses of adults, six-month old infants' M M N amplitude is dependent only upon the size of the acoustic difference between stimuli and not on its linguistic relevance in the native language (Cheour et al., 1998a). However, by one year of age, the M M N response pattern mimics that of adults, in that M M N reflects the categorization of native phonemes (Cheour et al., 1998a). The results of these cross-linguistic vowel studies suggest that M M N reflects phonemic categorization abilities. Results of studies measuring M M N in cross-linguistic consonant perception are less clear in their interpretation. In a study comparing the auditory processing of native vs. non-native phonemic contrasts, Dehaene-Lambertz recorded ERPs of French-speaking listeners to native, /ba/ vs. /da/, and non-native, /da/ vs. retroflex /Da/ (Dehaene-Eambertz, 1997). Dehaene-Lambertz claims that the French listeners showed an M M N response to native category changes but not to non-native ones. However, the study involved an active paradigm which typically evokes an N2b-P3b complex that will overlap the M M N (Alain & Woods, 1997; Naatanen, 1988; Naatanen et al., 1982; Picton, 1992). Therefore, the negativity measured is more likely N2b rather than M M N . In a subsequent study recording ERPs to the same stimulus contrasts in infants, Dehaene-Lambertz and Baillet (1998) report apositivity at about 350-400 ms in response 26 to contrasts that cross native and non-native boundaries. Both of these studies likely did not study the M M N . Therefore, a definitive comparison of the M M N in response to native vs. non-native consonant contrasts is necessary in order to determine whether non-native category changes which are poorly discriminated in behavioural tasks can be discriminated at pre-attentive levels of auditory processing. THESIS STUDY The present study is designed to determine whether the change in ability to discriminate non-native phonemic contrasts occurs at pre-attentive levels of auditory processing or whether it occurs later, at higher levels of processing. In this experiment, ERPs to both native and non-native consonant contrasts are recorded in order to determine whether there is a difference in the M M N response to these contrasts. The phonemic contrasts studied involve changes in place of articulation. The /ba/-/da/ contrast which represents a phonemic distinction in both English and Hindi will be used for the \"native\" contrasts. The stimulus pair for the \"non-native\" contrast will be the /da/-/Da/ contrast which exists in Hindi but is not found in English. In behavioural tasks, English listeners are easily able to discriminate the native phonemic contrast /ba/ vs. /da/ but have difficulty discriminating the non-native contrast /da/ vs. /Da/ (Werker & Lalonde, 1988). In the current study, attention is controlled for during the ERP recordings by: (1) asking subjects to ignore the sounds, and (2) having subjects watch closed-captioned movies. A separate behavioural discrimination task is included at the end of the experiment to establish a correspondence between the conscious ability to discriminate phonemes and the change detection at pre-attentive levels of processing represented by the M M N . 27 If the decrease in ability to discriminate non-native category changes occurs at early, pre-attentive level of processing, we would expect the M M N response to reflect behavioural patterns of discrimination. That is, M M N to the non-native contrast would likely be absent or smaller in amplitude than M M N to the native contrast. Conversely, if the ability to detect non-native contrasts is retained in adulthood at pre-attentive levels of processing, M M N s of equal amplitude should be present to both the native and non-native contrasts. CHAPTER 3: MISMATCH NEGATIVITY REVEALS CORTICAL DISCRIMINATION OF NON-NATIVE CONSONANT CONTRASTS IN ADULTS 29 INTRODUCTION Humans demonstrate the ability to consistently categorize a continuum of speech sounds into phonemes, the basic sound units of language, despite the lack of a one-to-one signal to phoneme correspondence (Borden et al., 1994, pp. 184-233). A single phoneme has many different acoustic representations which vary with context. This \"categorical\" perception capability is quite remarkable given the huge variability of productions of a single phoneme both within and across speakers. This ability to separate speech sounds into distinct phonemic categories has also been demonstrated in young infants (Eimas et al., 1971; Streeter, 1976; Trehub, 1976). Furthermore, evidence suggests that infants are born with the capability to distinguish phonemic categories of all the world's languages, making them \"universal listeners\" (Werker et al., 1981). However, infants lose the acoustic distinctions that are irrelevant to their native language by about 10 months of age (Werker & Tees, 1984b). That is, category boundaries that are not necessary in their language are not retained. This finding suggests that the speech processing mechanism begins with the natural ability to perceive and distinguish all phonemes but that there is a reorganization of this system early in life which causes the loss of the ability to discriminate sounds that we are not exposed to. Adult listeners are typically unable to distinguish sound contrasts that do not exist in their native language. For example, the voiced alveolar stop, /d/, and the voiced retroflex stop, /D/ represent two distinct sound categories in the Hindi language, but not in English . To English listeners, the two sounds are perceived as members of the same category, Id/ (Werker et al., 1981). Although it is a well-established fact that the ability to consciously perceive non-native phonemic distinctions decreases with linguistic experience, the existing behavioural data does not allow us to determine the level of auditory processing at which the change occurs (Aslin et 30 al., 1981; Streeter, 1976; Werker et al., 1981; Werker & Lalonde, 1988; Werker & Tees, 1984b). It is possible that a fundamental rewiring of the auditory system occurs with increased language exposure, causing us to be completely insensitive to these acoustic differences. However, this hypothesis cannot account for research showing that the ability to discriminate non-native contrasts is not completely or permanently lost. By altering task demands, inter-stimulus interval (ISI), and number of trials, Werker and Logan found that it is possible for adult listeners to discriminate non-native'consonant contrasts (Werker & Logan, 1985). Researchers have also shown that it is possible to train adult listeners to improve their ability to discriminate non-native phonetic changes (Jamieson & Morosan, 1986; Lively et al., 1994; Pisoni, Aslin, Perey, & Hennessy, 1982; Tees & Werker, 1984; Tremblay et al., 1997). A more plausible explanation might be that non-native discriminations still occur, well into adulthood, but only on a subconscious level. A theory of functional reorganization suggests that the change in discrimination capabilities actually reflects a change in selective attention (Werker, 1995). The assumption is that the primary auditory cortex \"detects\" the category changes but higher cognitive processes do not or cannot access the information that is not relevant to speech processing. Sound changes that did not contain linguistic information in the listener's past experience are ignored. Recently, electrophysiological measures have been applied as a tool to explore categorical perception. The component of the event-related potentials (ERP) of interest is the mismatch negativity (MMN). The M M N is a waveform difference which occurs in response to a discriminable change in a repeated sound stimulus (Naatanen et al., 1978; Naatanen et al., 1982). The key here is that the M M N is elicited only when the acoustic change is detected in the primary auditory cortex. Because the M M N is an early, pre-attentive measure of auditory 31 processing, it could demonstrate a physiological response to a change in phonemes even when there is no measurable behavioural response (Alho et al., 1989; Naatanen, 1992, pp. 137-138). Therefore, it may be possible to determine whether the auditory system, at the level of the auditory cortex, is actually detecting changes in sound stimulus when listeners are not able to discriminate them behaviourally. If, in fact, there is a reorganization of perceptual ability at or before the level of the primary auditory cortex which makes adults unable to discriminate non-native phonemes, we would expect to find an M M N to contrasts which cross a native phonemic boundary but not to contrasts which cross non-native phonemic boundaries (i.e., contrasts which belong to the same category according to the listener). Conversely, if the ability to discriminate non-native contrasts is retained at lower levels of auditory processing, there should be no difference in the M M N in response to native vs. non-native phoneme changes, regardless of surface ability. Previous studies of M M N in phonemic perception have focussed on determining whether there are any differences in M M N to stimulus changes which are (1) \"within-category\", i.e., do not cross any phonemic boundaries, (2) across native categories, i.e., cross phonemic category boundaries within the native language of the listeners, and (3) across non-native categories, i.e., cross phonemic boundaries that are irrelevant in the native language of the listeners. Results reported in the existing literature are mixed. Early M M N studies of categorical perception of consonants suggest that the M M N response reflects acoustic, rather than phonemic processing (Maiste et al., 1995; Sharma et al., 1993). Sharma and colleagues compared M M N to acoustically equidistant stimulus changes across and within articulatory place categories of /da/ through /ga/ and found no significant differences (Sharma et al., 1993). In a study which measured behavioural discrimination and 32 ERP recordings to stimulus changes from a /ba/ to /da/ continuum, Maiste and colleagues found that M M N amplitude was dependent on acoustic changes alone and was not affected by whether or not a categorical boundary had been crossed (even though behaviourally, subjects demonstrated categorical perception) (Maiste et al., 1995). This same pattern of M M N response was shown using vowel instead of consonant contrasts (Aaltonen et al., 1994). Conversely, in more recent studies involving cross-linguistic vowel perception, a larger M M N has been shown to reflect behavioural categorical perception capabilities (Cheour et al., 1998b; Naatanen et al., 1997). A study of adult listeners found M M N present to both native and non-native vowel category changes, but with significantly larger M M N amplitude to native category changes (Naatanen et al., 1997). In a study involving infants, the M M N results to native and non-native vowel changes were shown to change over time (Cheour et al., 1998b). At six months of age, the M M N amplitude reflects the acoustic differences between stimuli, but by age 12 months, M M N amplitude is larger to native category changes and smaller to non-native ones (Cheour et al., 1998b). Existing data from studies measuring M M N in cross-linguistic perception of consonants are less clear. Dehaene-Lambertz (1997) recorded ERPs of French-speaking listeners to native, /ba/ vs. /da/, and non-native, /da/ vs. retroflex /Da/. The French listeners showed a significantly larger negative peak in response to the native contrast than to the non-native one. Although Dehaene-Lambertz reported results for ERP waves: N I , M M N and P3, the study involved an active paradigm, and the negativity measured is more likely N2b, and attention-dependent negativity, rather than M M N (Naatanen, 1988; Naatanen, 1995; Naatanen et al., 1982). In a subsequent study recording ERPs to the same stimulus contrasts in infants, Dehaene-Lambertz 33 and Baillet (1998) report a positivity at about 350-400 ms in response to contrasts that cross native and non-native boundaries. Both of these studies likely did not study the M M N . The present study is an electrophysiological parallel to the behavioural study conducted by Werker et al. (1988). Using an oddball paradigm, this study measures M M N responses to both native and non-native consonant contrasts in order to determine whether there is a difference in the M M N response to the contrasts. The /ba/-/da/ contrast, which represents a phonemic distinction in both English and Hindi, will be used for the \"native\" contrasts. The stimulus pair for the \"non-native\" contrast will be the /da/-/Da/ contrast which exists in Hindi but is not found in English. In behavioural tasks, English listeners easily discriminate the native phonemic contrast /ba/ vs. /da/ but have difficulty discriminating the non-native contrast /da/ vs. /Da/ (Werker & Lalonde, 1988). An absent or smaller M M N to the non-native contrast and a larger M M N to the native contrast would suggest that the decrease in ability to detect non-native categorical changes occurs at an early level of auditory processing (e.g., primary auditory cortex). If, however, MMNs of equal amplitude are found to both the native and non-native contrasts, it would indicate that the primary auditory cortex has retained the ability to discriminate non-native contrasts, and the \"loss\" in ability must occur at a higher level of linguistic processing. METHOD Participants Thirteen adult native English speakers (5 male, 8 female) with no history of neurological disorder participated in this study. Two of the subjects were subsequently rejected, one due to non-completion of the study and the other due to failure to meet the M M N criterion (see below). The 11 accepted subjects ranged in age from 20-30 years, with a mean age of 23.9 years. A l l had 34 pure-tone hearing thresholds bilaterally of 20 dB H L (ANSI, 1996) or better from 250-8000 Hz. Acoustic immitance testing was performed prior to each session to ensure normal middle-ear functioning bilaterally, as defined by normal middle-ear compliance with a single peak between \u00C2\u00B150 daPa and ipsilateral acoustic reflexes present at 500 Hz and 1000 Hz. Each participant was determined to be right-handed using the self-report Chapman and Chapman Measurement of Handedness (Chapman & Chapman, 1987; see Appendix A). In order to qualify for the study, listeners were required to have English as their first acquired language. Participants were also asked to report any other languages that they spoke or understood. Potential listeners who had more than minimal exposure to Hindi, or any other language in which a dental-retroflex contrast exists, were excluded from this study. Other bilingual or multilingual participants were accepted if English was their strongest language and English language exposure began at infancy. Previous research has shown that the diversity of linguistic experience gained by exposure to a second or third language does not result in an increased, generalized, perceptual flexibility (Werker, 1986). That is, specific linguistic experience is required in order for phonetic discrimination abilities to be maintained. One final subject criterion was added after extensive piloting indicated that a clear M M N does not manifest in all subjects. Thus, as a control, participants who failed to demonstrate M M N to a control condition in which the comparison stimuli (/ba/ and /Da/) were separated by a large acoustic change were excluded. Data from one 20-year-old male subject were discarded due to failure to meet this criterion. Signed consent was obtained from the participants after they were fully informed of the nature of the study. Subjects were recruited from the University of British Columbia campus and paid an honorarium for their participation. Stimuli This experiment involved a number of contrasts using three speech syllables: labial /ba/, dental /da/, and retroflex /Da/. These syllables were selected from the same synthesized continuum of 16 syllables, ranging from /ba/ (#1 - #6), through /da/ (#7 - #10), to /Da/ (#11 -#16) used by Werker and Lalonde (1988). The stimuli were edited to exactly 275 ms in duration.1 Because of difficulty obtaining clear M M N results using more acoustically similar stimuli (e.g., /ba/ #5 vs. /da/ #9) during the piloting of this study, stimuli #2, #9, and #16 were selected as the respective representatives for the /ba/, /da/ and /Da/ phonetic categories. These tokens allow for the largest possible acoustic distance between stimuli (and should theoretically yield the largest M M N ) while maintaining equidistant comparisons. Each stimulus consists of a 10-ms noise burst, followed by a 50-ms formant transition, and finally, 215 ms of steady-state vowel. The key differentiating features of the stimuli are the starting frequencies of the second and third formants, F2 and F3. FI , F2, and F3 of the stimuli are listed in Table 1. For all three stimuli, fundamental frequency was steady at 100 Hz for the first 100 ms. It gradually rose to 120 Hz from 100 ms to 275 ms. F4 and F5 were held constant at 3500 and 4000 Hz respectively. The vowel steady-state portion of the three test stimuli are identical in amplitude, duration, and formant frequencies. Schematic spectrograms and detailed description of acoustic parameters are presented in Werker and Lalonde (1988). INSERT T A B L E 1 A B O U T HERE Stimuli were delivered using STIM software (NeuroScan) and presented via a single loudspeaker. The speaker was centred in front of the chair at an inclining angle of 23.6 degrees. The distance from the speaker to the subject's head was 220 cm. Each syllable was calibrated daily to 73 dB peak SPL at the position of the subject's head. E E G Recording Sessions Stimuli were presented in blocks of 500 trials, with a stimulus onset asynchrony (SOA) of 700 ms. Each block contrasted two of the three test syllables in a fixed oddball paradigm, with one syllable as the standard (probability of occurrence = .9) and the other as the deviant stimulus 36 (probability of occurrence = .1). In this paradigm, a series of standard syllables were played with every 10 th stimulus being the deviant, (e.g., ba...ba...ba da...ba). The experiment consisted of six different test conditions to allow for a bi-directional comparison of three stimulus contrasts, \"native\", \"non-native\", and \"control\", as shown in Table 2. Four replications of each test condition were obtained, yielding a total of 2000 trials (200 deviants) in each direction for each stimulus contrast.2 Conditions were presented in a pseudo-random order over the two recording sessions. Breaks were given between blocks at the subject's request or after about an hour of recording. INSERT T A B L E 2 A B O U T HERE Continuous recordings of 30 E E G channels referenced to the nose were obtained in S C A N (NeuroScan) using a 32-channel electrode cap. The channels recorded represent a subset of the International 10-20 System (American EEG Society, 1994): Fz, FCz, Cz, Cpz, Pz, Oz, FP1, FP2, F3, F4, F7, F8, FC3, FC4, FT7, FT8, C3, C4, T7, T8, CP3, CP4, TP7, TP8, P3, P4, O l , 02, M l , and M2. Additionally, electrodes placed above and below the left eye (vertical electrooculogram, VEOG) were used to monitor vertical eye movements such as blinking. Horizontal eye movements were monitored using electrodes placed near the outer canthi of the eyes (horizontal electrooculogram, HEOG). Electrode impedances were maintained below 15 kQ. The signal was amplified (gain = 5000), digitally filtered (passband = 0.1-30 Hz), and digitized (rate = 500 Hz) using a NeuroScan SynAmps amplifier. EEG recordings were processed offline. V E O G and HEOG artifacts were subtracted from the continuous EEG with the application of an ocular artifact reduction algorithm (Semlitsch, Anderer, Schuster, & Presslich, 1986). After ocular artifact reduction, continuous E E G recordings were epoched into windows from 100 ms pre-stimulus to 700 ms post-stimulus. The epoch files were then baseline corrected and any trials with amplitude exceeding \u00C2\u00B1100 / / V still remaining in any of the EEG channels were discarded. The first two trials of each block and 37 the first standard following each deviant trial (e.g., #11,21, 31...) were also rejected. The remaining epoch windows were averaged separately for standard and deviant trials. A l l four replications of each condition were then averaged together. For each participant, difference waves were obtained by subtracting the average waveforms to the standard stimuli from the responses to the deviant stimuli for each condition. This method of subtraction allowed a comparison of the recordings obtained within the same block (Maiste et al., 1995; Stapells, in press).3 Grand mean difference waves were created by averaging together the difference waves of all eleven subjects. Behavioural Discrimination Sessions Previous studies using the same stimulus set have shown that English listeners can accurately and reliably discriminate the native phonetic contrast, /ba/ vs. /da/, but experience difficulty in discriminating the non-native /da/ vs. /Da/ contrast (Dehaene-Lambertz, 1997; Werker & Lalonde, 1988). In this study, behavioural discrimination measures were obtained separately from the ERP recordings. These measures allowed a within-subject comparison of behavioural phonemic discrimination capabilities with M M N measures of change detection. The behavioural measures involved a forced-choice discrimination task in which participants decided whether two speech syllables were \"same\" or \"different\". Stimuli were presented in pairs, with a 580-ms ISI separating the two syllables within a pair and a 2000-ms SOA between successive pairs. There were nine possible combinations of stimulus pairings: /ba/-/ba/, /da/-/da/, /Da/-/Da/, /ba/-/Da/, /Da/-/ba/, /ba/-/da/, /da/-/ba/, /da/-/Da/, and /Da/-/da/. Subjects were instructed to press a button if the two stimuli within a pair sounded \"different\" and do nothing if they sounded the same. There was a .4 probability that the pair would be the same (divided equally among the first three contrasts) and a .6 probability that the pair would be different (divided equally among the last six contrasts). However, the expected percentage of \"different\" responses was approximately .4 due to the results of previous studies which showed that English listeners cannot discriminate, or have a great deal of difficulty 38 discriminating, the non-native differences (Dehaene-Lambertz, 1997; Werker & Lalonde, 1988). In order to reduce subject fatigue, the behavioural discrimination task was separated into four blocks of 100 trials (i.e., 100 stimulus pairs) per block. A short training run of approximately 10 to 20 stimulus pairs was given to ensure that the subjects understood the task. Procedure A l l sessions were carried-out in a sound-attenuated booth. Subjects sat in a comfortable reclining chair. Due to the lengthy duration of the study, recordings for each subject required two or three sessions for completion, with a separate behavioural discrimination condition recorded at the end of the final session. During electrophysiological recording, subjects were instructed to relax and watched a closed-captioned movie of their choice with the sound off while the speech stimuli were played via loudspeaker. The importance of remaining alert and ignoring the speech stimuli was emphasized to them. During the behavioural portion of the study, subjects attended to the stimulus contrasts and responded by button-press when they perceived stimuli to differ. Data Analysis M M N response measurement for statistical analysis focussed on the difference waveforms at Fz because this recording site showed the largest M M N amplitude in the grand mean waveforms. Measures were obtained from the waveforms of individual subjects. Using the M M N in the grand mean difference waveforms across conditions, a latency window that would encompass responses from each individual was determined for peak measurements. The largest (\"peak\") amplitude (relative to the pre-stimulus baseline) within this window (133-233 ms) and its latency were measured from each of the individual subjects' waveforms at the Fz electrode site. There were four questions of primary interest in the analyses of the ERP data: (i) For each of the conditions, is a mismatch negativity present? (ii) Is there a significant difference in M M N amplitude or latency dependent upon the stimulus contrast (native vs. non-native)? (iii) Is 39 there a significant difference of M M N amplitude or latency dependent upon the direction of the category change (e.g., [ba...ba...ba...da...] vs. [da...da...da...ba...])? (iv) Finally, the behavioural data were analysed to compare these results with those for the M M N . To answer the first question, point-by-point t-tests were completed on the grand mean difference waveforms of each experimental condition at the Fz electrode. To be considered present, at least five consecutive Fz negative points (10ms) within the 133 to 233 ms M M N window had to be significantly different from zero (p < .05). Six separate 2-way repeated-measures analyses of variance (ANOVAs) were performed on the amplitude and latency measures in order to determine whether there were effects of contrast or direction on the M M N . \"Contrast\" was defined by the two stimuli which were compared within each condition, i.e., whether the change was \"native\", \"non-native\", or \"control\". Each contrast had two possible \"directions\", dependent upon which stimulus was presented as the standard and which was presented as the deviant. There were three comparisons: (i) native conditions (\"ba to da\" and \"da to ba\") vs. non-native conditions (\"Da to da\" and \"da to Da\"), (ii) control conditions (\"ba to Da\" and \"Da to ba\") vs. native conditions (\"ba to da\" and \"da to ba\"), and (iii) control conditions (\"ba to Da\" and \"Da to ba\") vs. non-native conditions (\"da to Da\" and \"Da to da\"). Newman-Keuls post-hoc analyses were carried-out where significant interactions were found. Results of all analyses were considered significant ifp<.05. For analyses of the behavioural discrimination data, the percentage of correct responses to each of the nine stimulus pairings, as well as median reaction times were calculated for each subject. Two-way repeated measures A N O V A s comparing results across the native and non-native stimulus contrasts were used to determine if there were effects of contrast condition or direction of stimulus presentation on accuracy or reaction time. Direction was coded by which stimulus was presented first within the pair. Results were considered significant if p < .05. 40 RESULTS Behavioural Discrimination Results Table 3 shows mean percent correct and reaction times for the different stimulus pairings. In the catch trials, the average percent correct is 94.4%. The mean percent correct for the control and native change trials is also 94.4%. In contrast, the non-native change conditions resulted in only 33.8% correct. The average reaction time is shorter for the discrimination of control and native contrasts (mean =1114 ms) and longer for the non-native contrasts (mean = 1197 ms). Note that the mean reaction times reported in Table 3 represent the average over all 11 subjects. Only three of the subjects, however, achieved accuracy scores of 50% or better in the two non-native conditions. The mean reaction times of these three subjects are 1081 ms for the control conditions, 1100 ms for the native conditions, and 1091 ms for the non-native conditions. INSERT T A B L E 3 A B O U T HERE Native vs. Non-Native. The results of two-way A N O V A s (contrast, direction) of percent correct and reaction time data across stimulus pairings are shown in Table 4. Response accuracy to the non-native category difference is significantly lower than to the native one. There is no significant effect of direction on the accuracy across these conditions. For reaction time, there is no significant difference between native and non-native contrasts. There is a non-significant trend for direction, such that when /da/ is the deviant stimulus, reaction time is faster. The fact that this did not reach statistical significance may be due to the variability of the reaction times in the non-native contrast, which for most subjects were based on less than 50% correct. Control vs. Native. The A N O V A of percent correct for the \"control\" vs. \"native\" contrasts reveals there is no significant effect of contrast or direction on percent correct, but that there is a significant contrast x direction effect. Post hoc analyses indicate the \"ba-Da\" pairs result in significantly higher percent correct scores than \"Da-ba\", \"ba-da\", or \"da-ba\" pairs. 41 Further, the latter three pairings are not significantly different from each other. The maximum difference in percent correct between \"ba-Da\" and the other pairings, however, is only 3.2%. Analysis of reaction time between the \"control\" and \"native\" stimulus pairings indicate no difference between contrasts, but a highly significant effect of contrast direction. Reaction time is significantly slower when /ba/ is the deviant stimulus. Control vs. Non-Native. As with the \"native\" versus \"non-native\" comparison, a significant effect of contrast is seen for the \"control\" vs. \"non-native\" comparison, such that percent correct for discriminating non-native phoneme changes is significantly lower than for discriminating the large changes, which also cross a native phoneme boundary, found in the \"control\" pairings. There is no significant effect of direction on accuracy in this comparison. Although they did not reach statistical significance, there is a trend for reaction times to \"control\" contrasts to be shorter than reaction times to \" non-native\" contrasts. Similarly, there is also a trend towards a contrast x direction interaction, such that in the \"control\" contrasts, reaction time tends to be faster when /Da/ is the deviant stimulus, but not for \"non-native\" contrasts. Again, statistical significance may not have been obtained due to variability of the reaction times in the non-native contrasts. INSERT T A B L E 4 A B O U T HERE M M N Results Figures 1, 2, and 3 depict the effects of contrast and direction on the grand mean difference waves for all 30 scalp sites. A negative peak occurs within the 133-233 ms latency region, with largest negative amplitudes occurring in the fronto-central electrodes. This peak inverts in polarity at the mastoids. These findings are all features of the M M N (Picton et al., in press). In the \"ba to Da\" control condition (Figure 1), a relatively large M M N is immediately followed by a frontal-central positivity at approximately 275 ms which is likely the ERP wave P3a. In the \"ba to da\" direction of the \"native\" contrast, shown in Figure 2, a negative peak does 42 occur in the 130 - 230 ms latency region but this peak is smaller fronto-centrally, only barely crossing into the negative region at Fz. However, this peak does invert at the mastoid electrodes, and the amplitude of this peak at the mastoids is equal to those in the other native change condition (\"da to ba\"). Thus, it is likely that a M M N is present for both directions of the native condition. INSERT FIGURES 1-3 A B O U T HERE The M M N can be seen more clearly in the grand mean difference waveforms of Figure 4, which shows all six conditions at electrode Fz. Fz difference waveforms to the \"native\" contrasts show a definite M M N in one direction, when /da/ is the standard and /ba/ is the deviant, but less clearly in the reverse direction, when /da/ is the deviant. A definite M M N is visible in both directions of the \"non-native\" contrast, with a slightly larger amplitude in the \"da to Da\" direction. M M N is also evident in both \"control\" conditions. The amplitude of the M M N to \"ba to Da\" is relatively large compared to all other conditions. Also illustrated in Figure 4 are the results of point-by-point t-tests performed on these waveforms. According to the pre-determined acceptance criterion, a significant negativity was required to be present for at least five points (10 ms) in order to be considered a present M M N . Statistical analysis of grand mean waveforms at Fz again shows M M N to be present in all but the \"ba to da\" native contrast condition, where /ba/ is the standard and /da/ is the deviant stimulus. However, point-by-point t-tests of the difference waveforms for this latter condition at the mastoids indicate that there are significant positive peaks (i.e., inverted) at both mastoids. INSERT FIGURE 4 A B O U T HERE Table 5 lists the means and standard deviations of M M N peak amplitude and latency measures taken from each individual's difference waveforms at Fz. Results of the 2-way A N O V A s are shown in Table 6. These analyses focus on two questions: (1) whether there is a 43 significant difference in the M M N dependent upon the two stimuli being compared (i.e., \"contrast\"), and (2) whether there is a significant difference in the M M N dependent on the direction of presentation of the stimuli (i.e., which stimulus is the standard and which is the deviant). Native vs. Non-Native. A 2-way A N O V A of the M M N peak amplitudes, comparing the native contrast conditions, \"ba to da\" and \"da to ba\", with the non-native contrast conditions, \"Da to da\" and \"da to Da\", indicates there is no significant difference between M M N amplitude to the \"native\" vs. \"non-native\" contrasts (Table 6). There is a significant effect of direction, such that M M N amplitude is significantly larger when /da/ is presented as the standard stimulus. There are no significant differences between the contrasts or directions on M M N latency (Table 6). Because only seven of the subjects showed clear M M N peaks in the \"ba to da\" condition, and thus M M N latencies may be unreliable, a 1-way A N O V A of direction on M M N latencies in the non-native contrast only was carried out. The results of this 1-way A N O V A confirm that there is no significant effect of direction on M M N latency [F(l ,8) = 3.10, p = .117]. Control vs. Native. M M N amplitudes and latencies to the control conditions with the large acoustic changes (\"ba to Da\" and \"Da to ba\") and the native phonemic change conditions (\"ba to da\" and \"da to ba\") were compared next. Peak M M N amplitudes of the control conditions are significantly larger than those of the native change conditions. However, contrast interacts with direction in that opposite effects of direction occur, depending on the stimulus contrast.4 In the control condition, M M N amplitude is larger when /ba/ is the standard stimulus and smaller when /ba/ is the deviant. In the native condition, M M N amplitude is smaller when /ba/ is the standard and larger when the direction is reversed. No significant effects of contrast or direction on M M N latency are found in the control vs. native contrasts Control vs. Non-Native. Finally, A N O V A s comparing M M N in the control conditions to those obtained in the non-native conditions were completed. The larger M M N amplitudes for the control conditions compared to the non-native conditions did not quite reach significance 44 (p= .054). There is, however, a significant direction effect, such that M M N peak amplitudes are significantly larger when /Da/ is presented as the deviant stimulus. This was seen in both native and control conditions. The A N O V A of M M N latency indicates no effects of contrast or direction on M M N latency. INSERT TABLES 5 & 6 A B O U T HERE Late Negativities Inspection of Figure 4 (as well as Figures 1, 2, and 3) shows that in addition to the negative peaks found in the M M N latency region, additional negativities are present later in the ERP waveform, from about 300 ms and later. A late fronto-central negativity is present in the \"control\" conditions but questionable in the \"non-native\" conditions. This late fronto-central negativity is not seen in the waveforms for \"native\" conditions, but a late centro-parietal negativity is present in the waveform for the \"da to ba\" direction of the native contrast. This centro-parietal negativity is also present in the \"control\" conditions but absent in the \"non-native\" conditions. These late negativities are well outside of the M M N latency window and show no evidence of inversion at the mastoids. They are, therefore, different from the M M N , and likely have different underlying sources. DISCUSSION Behavioural Discrimination The overall pattern of performance in the discrimination task is as expected for an English-listener subject group with no Hindi experience. Accuracy, as measured in percent of Correct responses, shows that the subjects are able to discriminate the \"control\" and \"native\" contrasts which cross a native phonemic boundary, but have difficulty discriminating non-native phonemic contrasts. These results are consistent with those reported by previous studies which used the same synthesized continuum (Dehaene-Lambertz, 1997; Werker & Lalonde, 1988). 45 Though not significantly different, discrimination of \"control\" contrasts also tends to be faster than the non-native condition. There are no differences in reaction time, however, between the \"control\" and \"native\" conditions. Comparison of performance between the \"control\" and \"native\" conditions shows that accuracy is slightly, but significantly, higher to one direction of the control contrast pair (/ba/ -/Da/). Better performance in the \"control\" conditions might be expected simply because they involve a larger acoustic contrast. Although significant, this advantage appears only minimal (3.2%) in the behavioural paradigm used in the current study, likely due to ceiling effects in discrimination of both \"native\" and \"control\" contrasts. A significant effect of the direction of stimulus change on reaction time is seen in the \"control\" and \"native\" contrasts. Reaction time is slower when /ba/ is presented as the second stimulus. Some trends which fall short of reaching statistically significant levels are also seen in comparisons of reaction time across different directions of stimulus changes. In both the \"native\" and \"non-native\" contrasts, reaction time tends to be faster when /da/ is presented as the second stimulus and slower when it is presented first. Also, reaction time for discriminating the control contrast tends to be faster when /Da/ is presented as the second stimulus. This effect of direction in the \"control\" contrasts is opposite of the direction effect in the \"non-native\" ones. Effects of direction on performance have been found in previous studies involving phoneme discrimination tasks (Whiting, Martin, & Stapells, 1998). Whiting and colleagues report larger d' values for /ba/ (as the deviant stimulus) than for /da/ in a change detection task. They also found that broadband noise masking produced longer reaction times to /da/ compared to /ba/. M M N Recordings The original questions that this study attempted to answer were: Is there a M M N to the non-native phonetic contrast, and, if so, is there a difference in its amplitude or latency from the M M N to the native phonemic contrast? The results show that there is indeed a M M N to the non-native phonetic contrast. In fact, M M N was present to all three contrasts tested: native, non-46 native, and control. There is no statistically significant difference in M M N amplitude or latency between the native and non-native contrasts, although the mean amplitude of M M N to the non-native contrast is actually slightly larger than M M N to the native contrast. These results support the view that the developmental change in ability to discriminate phonetic contrasts (from universal capabilities at birth, to language-specific abilities in adulthood) does not involve a permanent loss of ability at the level of the auditory cortex. The M M N measures indicate that the auditory cortex discriminates non-native phonetic contrasts with the same \"strength\" as native contrasts, even though conscious behavioural discrimination of non-native contrasts is much poorer. Thus, at pre-attentive levels of auditory processing, the ability to discriminate non-native contrasts is retained long after the ability to demonstrate this in behavioural ability is apparently lost. The change in auditory stimulus is detected by the auditory cortex but the information is not readily accessible at higher, more conscious levels of linguistic processing. This likely indicates that, for adult listeners, attention plays a critical role in the categorical perception of phonemes. Some aspects of the results from the current study are consistent with previous findings suggesting that the M M N response reflects primarily acoustic, rather than phonetic, processing. Sharma and colleagues found similar M M N amplitude and latency values to acoustically equidistant stimulus changes across and within articulatory place categories (i.e., /da/ vs. /ga/) (Sharma et al., 1993). Similarly, Maiste and colleagues concluded that the M M N was sensitive to the detection of acoustic changes but does not reflect categorical perception (Maiste et al., 1995). The equal (indeed, non-significantly larger) amplitude of the M M N to non-native category changes compared to native category changes is in contrast with some results of previous cross-linguistic M M N studies. Naatanen and colleagues also report M M N present to both native and non-native category changes \u00E2\u0080\u0094 in this case vowels ~ but they found the M M N amplitude to be significantly larger in response to native category changes (Naatanen et al., 47 1997). Winkler and colleagues recorded M M N to a Finnish vowel contrast which does not exist in Hungarian for three subject groups: native Finnish listeners, Hungarian listeners who had acquired Finnish as a second language, and Hungarian listeners with no Finnish experience. They found that M M N was present in the first two subject groups but not in the subject group who had no experience with the Finnish vowel contrast, suggesting a definite relationship between M M N and behavioural response (Winkler et al., 1999). Finally, Cheour and colleagues, studying infants, report that at six months of age, the M M N amplitude appears to reflect only acoustic differences between vowel stimulus changes, with larger M M N amplitude in response to larger acoustic differences. By age 12 months, however, M M N amplitude becomes larger to native category changes and smaller to non-native ones, similar to the results of adults reported above (Cheour et al., 1998b). A number of possible explanations exist to explain the conflicting results of the present study with the findings of these previous studies (Cheour et al., 1998b; Naatanen et al., 1997; Winkler et al., 1999). The development of phonemic processing may be different for vowels and consonants. For example, it has been shown that vowels and consonants differ in their developmental time course, in that language-specificity occurs earlier for vowels than for consonants (Kuhl et al., 1992; Werker & Polka, 1993). This may be due, in part, to the acoustic . differences between vowels and consonants which make vowels more salient perceptually. Also, in adult behavioural perception, consonants and vowels have been considered different by some researchers. For example, Liberman and colleagues postulate different processing mechanisms for consonants and vowels (Liberman et al., 1967). Alternatively, the pattern of results of the present study could be unique to the place of articulation continuum (from labials, through alveolar, to retroflex). Manner of articulation or voicing may exhibit different results. Finally, it is possible that the non-native exemplar for /Da/ used in this experiment contained acoustic (not linguistic) attributes which cued the M M N . More studies are necessary in order to clarify the source of the discrepancy. 48 The peak amplitudes of M M N in the \"control\" contrast conditions are significantly larger than those in the \"native\" conditions, and are nearly significantly larger (p = .054) than those in the \"non-native\" conditions. This difference in amplitude might be expected due to the fact that amplitude of the M M N response is directly related to the size of the acoustic change from the standard to the deviant stimulus, for non-speech (e.g., Lang et al., 1990; Sams et al., 1985) and speech stimuli (Naatanen, 1992; Picton et al., in press). According to this, we would expect to see a larger M M N to the larger acoustic changes in the control condition. In support of this, a frontal P3a wave, often seen following the M M N to large stimulus differences, is clearly present in the ERP to the \"ba to Da\" contrast (Figure 1). One would also expect, however, M M N latency in response to the control contrasts to be shorter. This is not the case. It is possible that some acoustic rather than linguistic characteristic about the /Da/ exemplar in this continuum causes the larger mismatch to be evoked by those contrasts in which it is included. Perhaps there is an acoustic cue in our/Da/ exemplar (stimulus #16 from Werker & Lalonde, 1988) that is unrelated to speech and/or language that causes a large M M N . The additional subtraction methods3 used in the data analysis rule out the possibility of an N l effect (Butler, 1968; Naatanen & Picton, 1987) but it is possible that some acoustic cue in stimulus #16 is causing a stronger change detection in the auditory cortex. Spectral analyses of the stimuli do not support this suggestion (Werker & Lalonde, 1988) and no differences in the three stimuli have been found to account for this. Another possibility is that there is something inherently special to English listeners about the phoneme /Da/. According to the quantal theory of speech, certain articulatory regions produce more stable acoustic output (Stevens, 1989). These regions are particularly favoured for distinctive features underlying phonemes in human languages. The retroflex stop, /D/, is produced in one such region. In the acoustic spectrum of ID/, the onset of F3 and F4 converge, resulting in a \"spectral prominence\" (Stevens, 1989). Because these two formants are lower in frequency than the spectral peak of /d/, the prominence for /D/ is higher in energy level. The spectral peak for /d/ occurs with the convergence of higher formants, which 49 are lower in energy. This makes it perceptually more salient for ID/ than /d/ or /b/ (which does not show a spectral prominence at all). Thus, it may be possible that M M N to contrasts involving /Da/ are larger because it is acoustically more salient. There are significant effects of direction on the amplitude of the M M N . The size of the response is highly dependent upon which stimulus is presented as the standard and which is presented as the deviant. Similarly, some trends towards directional effects can be seen in the behavioural results of this study. These results cannot be compared with previous behavioural studies involving across-category contrasts because they have typically counterbalanced direction and reported overall effects of contrast (e.g., Werker et al., 1981; Werker & Lalonde, 1988; Werker & Tees, 1984b). The effect of directionality on performance has only been explored for within-category discriminations. For example, Kuhl has shown that the ability to discriminate equal acoustic differences within a vowel category can be affected by whether the vowel token presented first (as a standard) is a prototype or a non-prototype of that vowel category (Kuhl, 1991). A definite M M N was found in one direction of the native condition, \"da to ba\", but not clearly to the other, \"ba to da\", the native phonemic contrast in which /ba/ was the standard stimulus and /da/ was the deviant. Although point-by-point t-tests show that the negative peak at Fz does not reach significance, there are other indications that there is a M M N in this condition. For example, both mastoids show an inversion of polarity, with positive peaks that reach statistically significant levels. The amplitude of these mastoid peaks in this condition are equivalent to those of the counterbalanced condition, \"da to ba\". Nevertheless, at Fz, the M M N was higher in amplitude in response to the \"da to ba\" direction of change. It should be noted that our specification of \"direction\" in the A N O V A s is arbitrary. We defined direction by holding constant the stimulus which was the standard. Another method would have been to define direction as moving up or down the stimulus continuum. Changing to this latter definition changes A N O V A \"direction\" effects to \"contrast x direction\" interactions 50 and vice versa. Either way, one may conclude that the direction in which stimulus change occurs impacts the amplitude of the M M N : M M N amplitude is larger when /Da/ is the deviant stimulus or when /da/ is the standard stimulus. If we consider the \"native\" condition alone, it would also appear that M M N amplitude is smaller when /ba/ is the standard. The fact that this directional effect is reversed in the \"control\" condition ( M M N larger when /ba/ is the standard) may be due to the fact that /Da/ as a deviant creates such a large M M N that it overshadows the effect of /ba/ as a standard. Results indicating strong directional effects on the M M N response have been previously reported. In a study involving /ba/ vs. /da/ contrasts, Maiste and colleagues reported that M M N is present when /ba/ is the standard stimulus but is small or absent when /da/ is the standard stimulus (Maiste et al., 1995). Similarly, Martin and colleagues found that M M N amplitude is larger to a /ba/ vs. /da/ contrast when /ba/ is the standard (Martin et al., 1999). These results are opposite to the current study, which shows a larger M M N amplitude to /ba/ as a deviant when contrasted with /da/. Whiting and colleagues showed that broadband noise masking has a greater effect on /da/ (as the deviant stimulus) than on /ba/ discriminability, measured electrophysiologically and behaviourally, as indicated by longer P3b latencies and reaction times to /da/ with masking (Whiting et al., 1998). Recently, Sanders and Neville also reported directional effects on M M N measures of phonetic discrimination, with M M N present to one direction of presentation but not the other of behaviourally discriminable stimuli (Sanders & Neville, 2000). The implications of these directional effects on theories of phoneme perception are unclear at this time. There are various explanations as to why the M M N to the \"ba to da\" is low in amplitude and did not reach significance. This finding is likely related to the underlying cause of the directional effects noted above. It is possible that the /ba/ token chosen from the continuum (stimulus #2 from Werker & Lalonde, 1988) is a poor exemplar because it is near the end of the synthetic continuum and therefore less stable or less speech-like. Another possibility is that there 51 is something about the /ba/ phoneme, at least the exemplars in this synthesized continuum, that forms a poor echoic memory trace. Finally, another alternative to the lack of a significant M M N in response to the \"ba to da\" condition could be simply due to a lack of power, which could be remedied by increasing the number of trials per subject or the number of subjects. Late Negativities A late fronto-central negative peak occurs about 300 ms and later in the difference waveforms of conditions which include /Da/ as part of the contrast. A late centro-parietal peak also occurs at about 280 ms for the \"da to ba\" condition and at about 380 ms for the control conditions. These negativities do not have the characteristics of the M M N (Picton et al., in press). Their latencies are much later, falling outside of the regular M M N latency window, and it fails to invert in polarity at the mastoids. Similar late negative waves have been reported in other M M N experiments, but what this response might represent remains unknown (Alain & Woods, 1997; Escera, Alho, Winkler, & Naatanen, 1998; Schroger & Wolff, 1998). The fact that the late fronto-central negativity, preceded by a P3a, only reaches significant levels in conditions which include /Da/ is further evidence that there is something salient about this stimulus. Relationships Between M M N and Behavioural Discrimination The large M M N amplitude in the \"ba to Da\" condition corresponds to the significantly higher accuracy in discriminating the \"ba-Da\" stimulus pair. The effects of direction on M M N amplitude, shown in Table 6, are also reflected in the differences in accuracy and reaction time, which are dependent upon the order of presentation, shown in Table 4. The larger M M N s seen in conditions where /Da/ is the deviant is realized behaviourally by faster reaction times when /Da/ is presented as the second stimulus. One of the more unexpected findings of this study was the small M M N in one of the native change conditions (\"ba to da\"). Given that subjects were able to discriminate the two phonemes /ba/ and /da/ with a high degree of accuracy, regardless of the order of their presentation, the presence of a M M N would be expected in both native change conditions. 52 However, as discussed above, previous research has shown that it is not unusual to find an M M N to one direction of presentation but not the other, even when the contrast is discriminable behaviourally in both directions (Maiste et al., 1995). Also, studies have shown examples of subjects with absent M M N when contrasts, either speech or non-speech, are easily discriminated behaviourally (Sanders & Neville, 2000). This was also seen in the present study when one subject was rejected due to the absence of M M N response to the large, easily discriminable acoustic change in the control condition. In the present study, three of the subjects were able to discriminate the non-native contrasts at accuracy levels of 67.5% or better, but their M M N patterns did not differ from the rest of the subjects who were unable to discriminate the non-native contrasts. Thus, the M M N in these three subjects was independent of behavioural discrimination capabilities. This is further evidence to support the view that M M N more reflects acoustic than phonetic/categorical differences (Maiste et al., 1995; Sharma et al., 1993). Conclusion Through the measurement of M M N , this study suggests that the auditory cortex responds equally to stimulus changes which cross native and non-native phonetic boundaries. Thus, at pre-attentive levels of auditory processing, adults retain the ability to discriminate non-native phonetic contrasts even though they are unable to do so in conscious behavioural tasks. This suggests that the decrease in ability to discriminate non-native phonetic contrasts occurs at higher levels of linguistic processing. Another interesting, though perplexing, finding of this experiment is that the amplitude of the M M N response is significantly affected by the direction of the stimulus presentation. Although effects of direction on M M N responses to speech stimuli are not uncommon, their theoretical implication are not understood at this time. It would be useful to repeat the experiment on a group of native Hindi listeners to determine whether they exhibit the same M M N patterns. If Hindi listeners also show a larger amplitude M M N to the \"da to Da\" and \"Da to da\" conditions compared to the \"ba to da\" and \"da 53 to ba\" conditions, it would support the idea that there is indeed something special about stimulus contrasts involving /Da/ because, to a Hindi listener, the \"ba vs da\" contrasts should be equivalent to the \"da vs Da\" contrasts in terms of phonlogical \"weight\" (Werker et al., 1981; Werker & Lalonde, 1988). An electrophysiological measure of phonemic discrimination in which listeners are actively attending to stimuli in a discrimination task would also further our understanding of the processes involved in phoneme discrimination and how they differ for native vs. non-native sound contrasts. This would, perhaps, help explain the discrepancy in ability to discriminate non-native contrasts at the level of the primary auditory cortex with behavioural response capabilities. This study would focus on the N2b-P3b response recorded during discrimination tasks involving native and non-native phonemic contrasts. The N2b-P3b complex is thought to be closely related to overt discrimination abilities as measured through behavioural responses (Hillyard & Picton, 1987; Novak, Ritter, Vaughan, & Wiznitzer, 1990; Whiting et al., 1998). Such a study using the same stimulus contrasts is currently underway in our laboratory. 54 REFERENCES Aaltonen, O., Eerola, O., Lang, A . FL, Uusipaikka, E., & Tuomainen, J. (1994). Automatic discrimination of phonetically relevant and irrelevant vowel parameters as reflected by mismatch negativity. Journal of the Acoustical Society of America. 93, 1489-1493. Abramson, A . S., & Lisker, L . (1970). Discriminability along the voicing continuum: Cross-language tests. Paper presented at the Proceedings of the Sixth International Congress of Phonetic Sciences, Prague. Alain, C , & Woods, D. (1997). Attention modulates auditory pattern memory as indexed by event-related brain potentials. Psychophysiology, 34(534-546). Alho, K . (1995). Cerebral generators of mismatch negativity (MMN) and its magnetic counterpart (MMNm) elicited by sound changes. Ear and Hearing. 16. 38-50. Alho, K. , Sams, M . , Paavilainen, P., Reinikainen, K. , & Naatanen, R. (1989). Event-related brain potentials reflecting processing of relevant and irrelevant stimuli during selective listening. Electroencephalography and Clinical Neurophysiology. 77, 151-155. American E E G Society (1994). Guidelines for standard electrode position nomenclature. Journal of Clinical Neurophysiology, 8, 200-202. American National Standards Institute (ANSI, 1996). American National Standard Specification for Audiometers (ANSI S3.6-1996). New York: ANSI. Aslin, E. N . , Pisoni, D. B., Hennessy, B. L. , & Perey, A . J. (1981). Discrimination of voice onset time by human infants: New findings and implications for the effects of early experience. Child Development, 52,1135-1145. Best, C. T. (1990). Adult perception of nonnative contrasts differing in assimilation to native phonological categories. Journal of the Acoustical Society of America. 88, SI77. Best, C. T. (1995). A direct realist view of cross-language speech perception. In W. Strange (Ed.), Speech Perception and Linguistic Experience: Issues in Cross-Language Research (pp. 171-204). Baltimore, Md: York Press. 55 Best, C. T., & McRoberts, G. W. (1988). Examination of perceptual reorganization for nonnative speech contrasts: Zulu click discrimination by English-speaking adults and infants. Journal of Experimental Psychology: Human Perception & Performance. 14. 345-360. Borden, G. J., Harris, K . S., & Raphael, L . J. (1994). Speech Science Primer: Physiology. Acoustics, and Perception of Speech. Baltimore, Maryland: Williams & Wilkins. Bradlow, A . R., & Pisoni, D. B. (1997). Training Japanese listeners to identify English Ixl and Some effects of perceptual learning on speech production. Journal of the Acoustical Society of America. 101. 2299-2310. Burns, E. M . , & Ward, W. D. (1973). Categorical perception of musical intervals. 86th Meeting of the Acoustical Society of America. JJ1. 96 (A). Butler, R. A . (1968). Effect of changes in stimulus frequency and intensity on habituation of the human vertex potential. Journal of the Acoustical Society of America. 44, 945-950. Chapman, L. J., & Chapman, J. P. (1987). The measurement of handedness. Brain and Cognition. 6. 175-183. Cheour, M . , Alho, K., Ceponiene, R., Reinikainen, k., Sainio, K. , Pohjavuori, M . , Aaltonen, O., & Naatanen, R. (1998a). Maturation of mismatch negativity in infants. International Journal of Psychophysiology. 29. 217-226. Cheour, M . , Ceponiene, R., Lehtokoski, A . , Luuk, A. , All ik, J., Alho, K . , & Naatanen, R. (1998b). Development of language-specific phoneme representations in the infant brain. Nature Neuroscience, 1(5), 351-353. Cowan, N . , Winkler, I., Teder, W., & Naatanen, R. (1993). Memory prerequisites of mismatch negativity in the auditory event-related potentials (ERPs). Journal of Experimental Psychology: Learning, Memory, and Cognition, 19, 909-921. Dehaene-Lambertz, G. (1997). Electrophysiological correlates of categorical phoneme perception in adults. Neuroreport. 8. 919-924. 56 Dehaene-Lambertz, G. & Baillet, S. (1998). A phonological representation in the infant brain. Neuroreport. 9.1885-1888. Eimas, P. D. (1975). Developmental studies in speech perception. In L. B. Cohen & P. Salapatek (Eds.), Infant Perception: From Sensation to Cognition . New York: Academic Press. Eimas, P. D., Siqueland, E. R., Jusczyk, P. W., & Vigorito, J. (1971). Speech perception in infants. Science. 171. 303-306. Escera, C , Alho, K. , Winkler, I., & Naatanen, R. (1998). Neural mechanisms of involuntary attention to acoustic novelty and change. Journal of Cognitive Neuroscience. 10. 590-604. Grieser, D., & Kuhl, P. K . (1989). Categorization of speech by infants: Support for speech sound prototypes. Developmental Psychology, 25. 577-588. Hillyard, S. A. , & Picton, T. W. (1987). Electrophysiology of cognition. In F. Plum (Ed.), Handbook of Physiology. Section 1: The Nervous System, Volume V-Higher Functions of the Nervous System (pp. 519-584). Bethesda, MD: American Physiological Society. Jamieson, D. G., & Morosan, D. E. (1986). Training non-native speech contrasts in adults: Acquisition of the English 161 - /0/ contrast by francophones. Perception & Psvchophvsics. 40. 205-215. Jusczyk, P., Friederici, A . D., Wessels, J. M . , Svenkerud, V . Y . , & Jusczyk, A. M . (1993). Infants' sensitivity to the sound patterns of native language words. Journal of Memory and Language. 32. 402-420. Kent, R. D., & Read, C. (1992). The Acoustic Analysis of Speech. San Diego, CA: Singular Publishing Group, Inc. Kraus, N . , McGee, T., Carrell, T. D., & Sharma, A . (1995). Neurophysiologic bases of speech discrimination. Ear and Hearing. 16. 19-37. Kraus, N . , McGee, T., Sharma, A. , Carrell, T., & Nicol, T. (1992). Mismatch negativity event-related potential elicited by speech stimuli. Ear and Hearing, 13. 158-164. 57 Kuhl, P. K . (1979). Speech perception in early infancy: Perceptual constancy for spectrally dissimilar vowel categories. Journal of the Acoustical Society of America. 66(6). 1668-1679. Kuhl, P. K . (1991). Human adults and human infants show a \"perceptual magnet effect\" for the prototypes of speech categories, monkeys do not. Perception & Psychophysics, 50. 93-107. Kuhl, P. K. (1993). Infant speech perception: A window on psycholinguistic development. International Journal of Psvcholinguistics. 9. 33-56. Kuhl, P. K. , & Iverson, P. (1995). Linguistic experience and the \"perceptual magnet effect\". In W. Strange (Ed.), Speech Perception and Linguistic Experience: Issues in Cross-Language Research (pp. 121-154). Baltimore, Md: York Press. Kuhl, P. K. , & Miller, J. D. (1978). Speech perception by the chinchilla: Identification functions for synthetic VOT stimuli. Journal of the Acoustical Society of Amaerica, 63. 905-917. Kuhl, P. K. , Williams, K. A. , Lacerda, F., Stevens, K. N . , & Lindblom, B. (1992). Linguistic experience alters phonetic perception in infants by 6 months of age. Science. 255. 606-608. Land, E. H. (1977). The retinex theory of color vision. Scientific American, 237, 108-128. Lang, A . H. , Eerola, O., Korpilahti, P., Holopainen, I., Salo, S., & Aaltonen, O. (1995). Practical issues in the clinical application of mismatch negativity. Ear & Hearing. 16. 118-130. Lang, A . H. , Nyrke, T., Ek, M . , Aaltonen, O., Raimo, I., & Naatanen, R. (1990). Pitch discrimination performance and auditory event-related potentials. In C. H. M . Brunia, A . W. K. Gaillard, & A . Kok (Eds.), Psychophysiological Brain Research. Vol . 1 (Vol. 294-298,). Tilburg: Tilburg University Press. 58 Lasky, R. E., Syrdal-Lasky, A. , & Klein, R. E. (1975). V O T discrimination by four to six and a half month old infants from Spanish environments. Journal of Experimental Child Psychology. 20.215-225. Liberman, A . M . , Cooper, F. S., Shankweiler, D. P., & Studdert-Kennedy, M . (1967). Perception of the speech code. Psychological Review. 74. 431-46J. Liberman, A . M . , Harris, K . S., Hoffman, H . S., & Griffith, B. C. (1957). The discrimination of speech sounds within and across phoneme boundaries. Journal of Experimental Psychology. 54. 358-368. Liberman, A . M . , & Mattingly, I. G. (1989). A specialization for speech perception. Science. 243. 489-494. Lisker, L. , & Abramson, A . S. (1970). The voicing dimension: Some experiments in comparative phonetics. Paper presented at the Proceedings of the 6th international congress of phonetic sciences, Prague, 1967, Prague. Lively, S. E., Pisoni, D. B., Yamada, R. A. , Tohkura, Y . I., & Yamada, T. (1994). Training Japanese listeners to identify English Ixl and IV: UI. Long-term retention of new phonetic categories. Journal of the Acoustical Society of America. 96, 2076-2087. Maiste, A. , Wiens, A . S., Hunt, M . J., Scherg, M . , & Picton, T. W. (1995). Event-related potentials and the categorical perception of speech sounds. Ear and Hearing. 16. 68-90. Martin, B. A. , Kurtzberg, D., & Stapells, D. R. (1999). The effects of decreased audibility produced by high-pass noise masking on N l and the mismatch negativity to speech sounds /ba/ and /da/. Journal of Speech. Language, and Hearing Research. 42. 271-286. Naatanen, R. (1988). Implications of ERP data for psychological theories of attention. Biological Psychology. 26. 117-163. Naatanen, R. (1992). Attention and Brain Function. Hillsdale, NJ: Lawrence Erlbaum. Naatanen, R. (1995). The mismatch negativity: A powerful tool for cognitive neuroscience. Ear & Hearing. 16(1). 6-15. 59 Naatanen, R., Gaillard, A . W., & Mantysalo, S. (1978). Early selective-attention effect on evoked potential reinterpreted. Acta Psvchologica, 42. 313-329. Naatanen, R., Lehtokoski, A. , Lennes, M . , Cheour, M . , Huotliainen, M . , Uvonen, A. , Vainio, M . , Alku, P., Umoniemi, R., Luuk, A. , All ik, J., Sinkkonen, J., & Alho, K . (1997). Language-specific phoneme representations revealed by electric and magnetic brain responses. Nature. 385(30). 432-434. Naatanen, R., Paavilainen, P., Alho, K. , Reinikainen, K. , & Sams, M . (1987). Interstimulus interval and the mismatch negativity. In C. Barber & T. Blum (Eds.), Evoked Potentials ITJ (pp. 392-397). London: Butterworths. Naatanen, R., Paavilainen, P., & Reinikainen, K . (1989). Do event-related potentials to infrequent decrements in duration of auditory stimuli demonstrate a memory trace in man? Neuroscience Letters. 107, 347-352. Naatanen, R., Paavilainen, P., Tiitinen, H. , Jiang, D., & Alho, K. (1993a). Attention and mismatch negativity. Psychophvsiologv, 30. 436-450. Naatanen, R., & Picton, T. (1987). The N l wave of the human electric and magnetic response to sound: A review and an analysis of the component structure. Psychophvsiologv. 24. 375-424. Naatanen, R., Schroger, E., Karakas, S., Tervaniemi, M . , & Paavilainen, P. (1993b). Development of memory trace for a complex sound in the human brain. Neuroreport, 4, 503-506. Naatanen, R., Simpson, M . , & Loveless, N . E. (1982). Stimulus deviance and evoked potentials. Biological Psychology. 14. 53-98. Novak, G. P., Ritter, W., Vaughan, H . G., & Wiznitzer, M . L . (1990). Differentiation of negative event-related potentials in an auditory discrimination task. Electroencephalography and Clinical Neurophysiology, 75, 225-275. Ohman, S. E. G. (1966). Coarticulation in V C V utterances: Spectrographic measurements. Journal of the Acoustical Society of America, 39, 151-168. 60 Paavilainen, P., Karlsson, M. -L . , Reinikainen, K. , & Naatanen, R. (1989). Mismatch negativity to change in spatial location of an auditory stimulus. Electroencephalography and Clinical Neurophysiology. 73,129-141. Pegg, J. E., & Werker, J. F. (1997). Adult and infant perception of two English phones. Journal of the Acoustical Society of America. 102(6). 3742-3753. Picton, T. (1992). The P300 wave of the human event-related potential. Journal of Clinical Neurophysiology. 9. 456-479. Picton, T. W., Alain, C , Otten, L. , Ritter, W., & Achim, A . (in press). Mismatch negativity: Different water in the same river. Audiologv and Neuro-Otology. Pisoni, D. B. (1977). Identification and discrimination of the relative onset time of two component tones: Implications for voicing perception in stops. Journal of the Acoustical Society of America. 61. 1352-1361. Pisoni, D. B., Aslin, R. N . , Perey, A . J., & Hennessy, B. L. (1982). Some effects of laboratory training on identification and discrimination of voicing contrasts in stop consonants. Journal of Experimental Psychology: Human Perception and Performance. 8(297-314). Polka, L. , & Werker, J. F. (1994). Developmental changes in perception of nonnative vowel contrasts. Journal of Experimental Psychology: Human Perception and Performance. 20(2), 421-435. Rochet, B.L. (1995). Speech perception and linguistic experience: Issues in cross-language research. In W. Strange (Ed.), Speech Perception and Linguistic Experience: Issues in Cross-Language Research (pp. 171-204). Baltimore, Md: York Press. Rvachew, S., & Jamieson, D. G. (1995). Learning new speech contrasts: Evidence from adults learning a second language and children with speech disorders. In W. Strange (Ed.), Speech Perception and Linguistic Experience: Issues in Cross-Language Research (pp. 411 -432). Baltimore, Md: York Press. 61 Sams, M . , Alho, K., & Naatanen, R. (1983). Sequential effects in the ERP in discriminating two stimuli. Biological Psychology. 17, 41-58. Sams, M . , Aulanko, R., Aaltonen, O., & Naatanen, R. (1990). Event-related potentials to infrequent changes in synthesized phonetic continuum. Journal of Cognitive Neuroscience. 2, 344-357. Sams, M . , Paavilainen, P., Alho, K. , & Naatanen, R. (1985). Auditory frequency discrimination and event-related potentials. Electroencephalography and Clinical Neurophysiology. 62. 437-448. Sanders, L . D., & Neville, H . J. (2000). Fast temporal processing of linguistic and non-linguistic stimuli: A mismatch-negativity study. Cognitive Neuroscience Society Annual Meeting Program 2000. 48. Scherg, M . , Vajsar, J., & Picton, T. W. (1989). A source analysis for the late human auditory evoked potentials. Journal of Cognitive Neuroscience. 1. 336-355. Schroger, E., & Wolff, E. (1998). Behavioral and electrophysiological effects of task-irrelevant sound change: a new distraction paradigm. Cognition and Brain Research. 7. 71-87. Semlitsch, H. V. , Anderer, P., Schuster, P., & Presslich, O. (1986). A solution for reliable and valid reduction of ocular artifacts applied to the P300 ERP. Psychophysiology, 23, 695-703. Sharma, A. , Kraus, N . , McGee, T., Carrell, T., & Nicol, T. (1993). Acoustic versus phonetic representation of speech as reflected by the mismatch negativity event-related potential. Electroencephalography and clinical Neurophysiology, 88. 64-71. Sheldon, A. , & Strange, W. (1982). The acquisition of /r/ and IV by Japanese learners of English: Evidence that speech production can precede speech perception. Applied Psycholinguistics, 3, 243-261. Stapells, D. R. (in press). Cortical event-related potentials to auditory stimuli. In J. Katz (Ed.), The Handbook of Clinical Audiology, 5th Edition. Baltimore: Lippingcott, Williams & Wilkins. 62 Stevens, K . N . (1989). On the quantal nature of speech. Journal of Phonetics. 17. 3-45. Strange, W., & Jenkins, J. (1978). Role of linguistic experience in the perception of speech. In R. D. Walk & H . L . Pick (Eds.), Perception and Experience . New York: Plenum Press. Streeter, L . A . (1976). Language perception of two-month old infants shows effects of both innate mechanisms and experience. Nature. 259. 39-41. Tees, R. C , & Werker, J. F. (1984). Perceptual flexibility: Maintenance or recovery of the ability to discriminate non-native speech sounds. Canadian Journal of Psychology. 38. 579-590. Tiitinen, H. , May, P., Reinikainen, K. , & Naatanen, R. (1994). Attentive novelty detection in humans is governed by pre-attentive sensory memory. Nature. 372, 90-92. Trehub, S. E. (1976). The discrimination of foreign speech contrasts by infants and adults. Child Development, 47. 466-472. Tremblay, K. , Kraus, N . , Carrell, T., & McGee, T. D. (1997). Central auditory system plasticity: Generalization to novel stimuli following listening training. Journal of the Acoustical Society of America. 102. 3762-3773. Werker, J. F. (1986). The effect of multilingualism on phonetic perceptual flexibility. Applied Psvcholinguistics, 7. 141-156. Werker, J. F. (1994). Cross-language speech perception: Development change does not involve loss. In J. C. Goodman & H. C. Nusbaum (Eds.), The Development of Speech Perception: The Transition from Speech Sounds to Spoken Words (pp. 93-120). Cambridge, Mass.: MIT Press. Werker, J. F. (1995). Exploring developmental changes in cross-language speech perception. In R. Gleitman & M . Liberman (Eds.), Language (Vol. 1, pp. 87-106). Werker, J. F., Gilbert, J. H., Humphrey, K. , & Tees, R. C. (1981). Developmental aspects of cross-language speech perception. Child Development. 52. 349-355. 63 Werker, J. F., & Lalonde, C. E. (1988). Cross-language speech perception: Initial capabilities and developmental change. Developmental Psychology. 24(5), 672-683. Werker, J. F., & Logan, J. S. (1985). Cross-language evidence for three factors in speech perception. Perception & Psvchophysics. 37(1). 35-44. Werker, J. F., & Pegg, J. E. (1992). Infant speech perception and phonological acquisition. In C. Ferguson, L . Menn, & C. Stoel-Gammon (Eds.), Phonological Development: Models. Research, and Implications (pp. pp 285-311). Timonium, M D : York Press. Werker, J. F., & Polka, L . (1993). The ontogeny and developmental significance of language-specific phonetic perception. In B. de Boysson-Bardies, S. de Schonen, P. Jusczyk, P. McNeilage, & J. Morton (Eds.), Developmental Neurocognition: Speech and Face Processing in the First Year of Life . The Netherlands: Kluwer Academic Publishers. Werker, J. F., & Tees, R., C. (1984a). Phonemic and phonetic factors in adult cross-language speech perception. Journal of the Acoustical Society of America, 75. 1866-1878. Werker, J. F., & Tees, R. C. (1984b). Cross-language speech perception: Evidence for perceptual reorganization during the first year of life. Infant Behavior and Development. 7. 49-63. Werker, J. F., & Tees, R. C. (1992). The organization and reorganization of human speech perception. Annual Review of Neuroscience, 15. 377-402. Whiting, K. A. , Martin, B. A. , & Stapells, D. R. (1998). The effects of broadband noise masking on cortical event-related potentials to speech sounds /ba/ and /da/. Ear & Hearing. 19. 218-231. Wiesel, T. N . (1982). Postnatal development of the visual cortex and the influence of environment. Nature. 299. 583-591. Wiesel, T. N . , & Hubel, D. H. (1963). Single cell responses in striate cortex of kittens deprived of vision in one eye. Journal of Neurophysiology. 26. 1003-1017. Williams, L. (1977). The voicing contrast in Spanish. Journal of Phonetics, 5 ,169-184. 64 Winkler, I., Kujala, T., Tiitinen, H. , Sivonen, P., Alku, P., Lehtokoski, A . , Czigler, I., Csepe, V . , Hmoniemi, R. J., & Naatanen, R. (1999). Brain responses reveal the learning of foreign language phonemes. Psychophysiology. 36. 638-642. Yabe, H., Tervaniemi, M . , Reinikainen, K. , & Naatanen, R. (1997). Temporal window of integration revealed by M M N to sound Omission. Neuroreport. 8. 1971-1974. 65 FOOTNOTES 1. As some of the tokens were slightly longer than 275 ms, stimuli were edited by removing traces (0-7 ms) of the vowel portion beyond 275 ms. Even though the energy of the vowel was quite low toward the end, a Hanning window was applied to the edited end of the vowel to ensure the natural fall of intensity. 2. Three baseline conditions, \"ba\", \"da\", and \"Da\", were also included in order to obtain a measure of the electrophysiological response to each stimulus presented alone. Baseline conditions consisted of 100 trials in which the same stimulus was presented repeatedly with an SOA of 700 ms. ERP recordings from these conditions were used in the supplementary subtraction method described below. 3. In order to control for possible stimulus-specific effects, two other methods of subtraction were assessed. One was the subtraction of an average waveform from the baseline condition of a stimulus from the waveform of the same stimulus presented as a deviant (Kraus et al., 1995; Kraus et al., 1992). The other method involved a comparison of the same stimulus across the counterbalanced directions of the conditions, allowing the subtraction of a stimulus when it was the standard from the same stimulus as a deviant (Martin et al., 1999; Whiting et al., 1998). Because all three analysis methods yielded the same pattern of results, only the difference waves from the original subtraction method (i.e., deviants minus standards within the same block) will be presented in this paper. 4. See Discussion concerning coding of \"direction\". Table 1 Formant Frequency Values of Stimuli Starting Steady-State Stimulus Formant Frequency (Hz)1 Frequency (Hz) 2 /ba/ stimulus #2 F l 250 500 F2 950 1090 F3 2288 2440 /da/ stimulus #9 F l 250 500 F2 1300 1090 F3 2624 2440 /Da/ stimulus #16 F l 250 500 F2 1650 1090 F3 2960 2440 1 att= 10 ms 2att = 60 ms-275 ms Table 2 ERP Recording Conditions 67 Contrast Condition Description Total Number of Trials Native Contrasts' \"ba to da\" /ba/ = standard, /da/ = deviant 1800 standards, 200 deviants \"da to ba\" /da/ = standard, /ba/ = deviant 1800 standards, 200 deviants Non-native Contrasts2 \"da to Da\" /da/ = standard, /Da/ = deviant 1800 standards, 200 deviants \"Da to da\" /Da/ = standard, /da/ = deviant 1800 standards, 200 deviants Control \"ba to Da\" /ba/ = standard, /Da/ = deviant 1800 standards, 200 deviants (large acoustic contrast) \"Da to ba\" /Da/ = standard, /ba/ = deviant 1800 standards, 200 deviants Vba/ = stimulus #2; /da/ = stimulus # 9 2/Da/ = stimulus #16 68 Table 3 Behavioural Response Accuracy and Reaction Times (n=l 1) Catch Trials1 Control Native Non-native ba-ba da-da Da-Da ba-Da Da-ba ba-da da-ba da-Da Da-da Percent Correct M SD 96.4 93.6 93.2 , 3.0 8.9 8.8 96.6 5.4 93.6 10.1 93.4 7.5 94.1 11.9 31.1 37.3 36.5 38.4 Reaction M 1090 1130 1099 1137 1257 1138 Time 2 (ms) SD 79 74 79 82 181 263 No reaction times are reported for the catch trials because correct trials are those in which there is no response. Reaction times reported include the full duration of the initial 275-ms syllable followed by a 580-ms ISI before the onset of the second 275-ms syllable. \"3-co C o C O '5 D s o u u C o o cu CO u. 3 O ' > CO -C CJ CQ co < > o z < C O u UH 3 CO CO D a CO CD Q u C M O C O \"3 co CJ 0* CD > c o Z C O > C O U \u00E2\u0080\u0094 ON O O T l - (N o r - \u00E2\u0080\u0094 T3 r - \u00E2\u0080\u0094 ON < \u00E2\u0080\u0094'\u00E2\u0080\u00A2 O m O N oo CN O fN \u00E2\u0080\u0094 r- o \u00E2\u0080\u0094 m o _> CO Z C O > C O o -a \u00C2\u00BB CO Z I c o Z co > CJ _> z u co u . 3 CJ CJ < CJ fc o o cn o oo o r--o fN o m r -CN O N O m CN fN *r> O r -O O N r- fN m N O N O CN o fN o \u00C2\u00A9 N O \u00C2\u00A9 fN V 00 r- ' N O O N o rn o fN N O fN o N O m O fN o o o o o O N O o ON CN \u00E2\u0080\u0094 O N ON' \u00C2\u00A9 \u00E2\u0080\u0094 CN ,\u00E2\u0080\u0094.' (N C O on C O on tra o Q CO u Q on 111 X on i rp ii v X U Q U u Q U c o u CJ CO P cu .Js 69 c o CJ c c o CJ CJ \u00C2\u00BB co CO u. C O cj Q x U 70 Table 5 M M N Peak-amplitude and Latency Measures at Fz Control Native Non-native ba to Da Da to ba ba to da da to ba da to Da Da to da Peak M -2.04 -1.37 -0.51 -1.28 -1.63 -0.99 Amplitude (uV) SD 0.95 0.64 1.30 0.967 0.87 0.75 Peak M 195 201 197 184 191 203 Latency (ms) SD 27 36 24 36 35 17 n = 11 subjects N Z C O < > o z < co i i 3 \u00E2\u0080\u00A2o 4) \u00E2\u0080\u00A2*-\u00C2\u00BB ca u Q u O C O 3 co a! u Z \u00E2\u0080\u00A2 a o > c o U > z co > a o U z I n o Z C O > 1) _> 08 z C M T3 rj- O NO \u00E2\u0080\u0094 so O O O N OO l/-> 1 1 ( ON CN o o o CN cn \u00C2\u00A9 0 > \u00C2\u00A9 CN \u00C2\u00A9 00 r- r- m ON CN m o 00 o o m o o o o o o o oo m O N mom C N o oo co CO k* c o CJ c C J cu Q x U ^4 C U 3 U-) \u00E2\u0080\u0094 CN \u00E2\u0080\u00A2\"3- N O 00 O N \u00E2\u0080\u0094 i n O N cn N O O N o \u00C2\u00A9 \u00C2\u00A9 N O \u00C2\u00A9 cn N O O r-~ ^ r-~ cn \u00E2\u0080\u0094> O N N O NO o 00 o 00 co 03 C o U c CD Q x U 72 FIGURE CAPTIONS Figure 1. Deviant minus standard grand mean difference waveforms for all EEG electrodes for the \"control\" conditions /ba/ vs. /Da/. The two difference waves in each plot represent the two directions of stimulus presentation (i) when /ba/ is the standard stimulus and /Da/ is the deviant, and (thick line) (ii) when /Da/ is the standard stimulus and /ba/ is the deviant (thin line). Negativity is represented as an upwards deflection (referenced to the nose) in this and all subsequent waveform figures. Figure 2. Deviant minus standard grand mean difference waveforms for all E E G electrodes for the \"native\" conditions (/ba/ vs. /da/). The two difference waves in each plot represent the two directions of stimulus presentation (i) when /da/ is the standard stimulus and /ba/ is the deviant (thick line), and (ii) when /ba/ is the standard stimulus and /da/ is the deviant (thin line). Figure 3. Deviant minus standard grand mean difference waveforms for all EEG electrodes for the \"non-native\" conditions (/da/ vs. /Da/). The two difference waves in each plot represent the two directions of stimulus presentation (i) when /Da/ is the standard stimulus and /da/ is the deviant (thick line), and (ii) when /da/ is the standard stimulus and /Da/ is the deviant (thin line). i Figure 4. Deviant minus standard grand mean difference waveforms at Fz for all six conditions. Top: \"native change\" conditions, comparing /ba/ and /da/. Middle: \"non-native\" change conditions, comparing /da/ and /Da/. Bottom: large change \"control\" conditions, comparing /ba/ and /Da/. Points reaching p < .05 and p < .01 levels of significance in point-by-point t-tests are indicated by bar graphs accompanying each waveform. 73 Figure 4 Native M M N /da/ /ba/ /ba/to/da/ /da/to/ba/ p<.05 Non-Native Control -100 ms /Da/ /ba/ 7 _l_ [ p < .05 / p < .01 1.6 nV APPENDIX A HAND USAGE QUESTIONNAIRE HAND USAGE QUESTIONNAIRE 78 Please indicate below which hand you ordinarily use for each activity. With which hand do you: 1. Draw? Left Right Either 2. Write? Left Right Either 3. Use a bottle opener? Left Right Either 4. Throw a snowball to hit a tree? Left Right Either 5. Use a hammer? Left Right Either 6. Use a toothbrush? Left Right Either 7. Use a screwdriver? Left Right Either 8. Use an eraser on paper? Left Right Either 9. Use a tennis racket? Left Right Either 10. Use scissors? Left Right Either 11. Hold a match when striking it? Left Right Either 12. Stir a can of paint? Left Right Either 13. On which shoulder do you rest a bat before swinging? Left Right Either Scoring is as follows: Left = 3 points Right = 1 point Either = 2 points To be termed Right-handed a subject must have a score between 13 and 17. APPENDIX B INDIVIDUAL SUBJECTS' BEHAVIOURAL RESPONSE ACCURACY AND REACTION TIME RESULTS 80 W E H e o > o 2 3 O o \u00C2\u00AB 1/1 B o Q cn s 3 \"> ed 4> CQ o u S 3 \"3 TJ C la 9 cn C o lu t> 13 \u00C2\u00AB5 i la \u00E2\u0080\u00A2o 13 =5 la 13 cn i 5 o ICJ .2 'C I H JS o ea CJ H e* fc o O N? cn B H oi, fc o U 0 s*. ~c3 Q I 13 cn e H Od fc o CJ \\u00C2\u00B0 13 i 13 Q 13 I la 13 \"a \u00E2\u0080\u009E CN __ CN r - c n m CN oo r n f- 0 0 \u00C2\u00AB\u00E2\u0080\u0094 T t o oo m SO c n O Os CN NO CN m CN i n i n \u00C2\u00A9 m \u00C2\u00A9 \u00C2\u00A9 i n \u00C2\u00A9 m \u00C2\u00A9 CN o r~ i n CN \u00C2\u00A9 i n o i n r n 0 0 CM r~ Os ON CN CN m c n \u00C2\u00AB n CN Os m ON m T t m r - r -c n 0 0 ND ON NO 0 0 0 0 r - Os m 0 0 m o ON CN CN T t i n CN p l/~l \u00C2\u00A9 \u00C2\u00A9 i n \u00C2\u00A9 i n \u00C2\u00A9 \u00C2\u00A9 i n i n m i n O i n CN i n \u00C2\u00A9 CN ,\u00E2\u0080\u0094'. m NO \u00C2\u00A9 OS CN m c n NO o NO ON SO NO ON T t c n c n CN T t T t NO m T t T t CN ON i n m oo CN Os CN \u00C2\u00A9 o o \u00C2\u00A9 \u00C2\u00A9 i n \u00C2\u00A9 \u00C2\u00A9 i n i n m \u00C2\u00A9 ON o \u00C2\u00A9 \u00C2\u00A9' i n \u00C2\u00A9 \u00C2\u00A9 t~ \u00C2\u00A9' T t o \u00C2\u00A9 \u00C2\u00A9 ON oo \u00C2\u00A9 o ON Os ON NO ON *\u00E2\u0080\u00A2\" NO Os m \u00C2\u00A9 T t i n CN c n 0 0 T t oo Os r~ m T t ON m ON \u00C2\u00A9 ON t\u00E2\u0080\u0094 \u00C2\u00A9 Os CN \u00C2\u00A9 ON \u00C2\u00A9 o o \u00C2\u00A9 i n \u00C2\u00A9 \u00C2\u00A9 i n \u00C2\u00A9 \u00C2\u00A9 m \u00C2\u00A9 \u00C2\u00A9 T t i n i n o i n i n r~ \u00C2\u00A9 i n i n r n Os \u00C2\u00A9 Os Os 0 0 ON Os ON ON \u00C2\u00A9 r~ ON T t 0 0 \u00C2\u00A9 m i n NO T t T t SO CN c n \u00C2\u00A9 m 0 0 s\u00C2\u00A3> CN m CN 0 0 NO OO oo m \u00C2\u00A9 ON CN \u00C2\u00A9 CN \u00C2\u00A9 O i n \u00C2\u00A9 \u00C2\u00A9 i n \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 i n m \u00C2\u00A9 NO \u00C2\u00A9 \u00C2\u00A9 iri r-' \u00C2\u00A9 i n i n i n c n \u00C2\u00A9 o Os \u00C2\u00A9 Os 0 0 \u00C2\u00A9 ON ON Os Os SO Os Os r - NO \u00C2\u00A9 _ NO ON \u00C2\u00A9 ON O S T t ND \u00C2\u00A9 Os c n m T t c n i n NO ON r~ o \u00C2\u00A9 Os CN \u00C2\u00A9 Os \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 i n m \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 i n NO T t \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 r~ \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 SO i n \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 ON 0 0 \u00C2\u00A9 ON \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 0 0 ON p i n i n i n \u00C2\u00A9 \u00C2\u00A9 i n \u00C2\u00A9 \u00C2\u00A9 i n i n CN 0 0 i n CN CN i n o \u00C2\u00A9 \u00C2\u00A9 r n 0 0 0 0 r~ ON Os 0 0 \u00C2\u00A9 ON \u00C2\u00A9 \u00C2\u00A9 Os ON ON c n oo m \u00C2\u00A9 \u00C2\u00A9 i n oo c n 0 0 \u00C2\u00A9 c n NO Os , ! c n CN i r i \u00C2\u00A9 0 0 NO 0 0 \u00C2\u00A9 so c n 0 0 r - Os 0 0 Os \u00C2\u00A9 ON ON Os Os \u00C2\u00A9 ON Os i n \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 i n \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 \u00C2\u00A9 T t \u00C2\u00A9 CN \u00C2\u00A9 i n \u00C2\u00A9 CN \u00C2\u00A9 i n i n \u00C2\u00A9 i n i n NO r n O N \u00C2\u00A9 ON \u00C2\u00A9 ON \u00C2\u00A9 ON Os \u00C2\u00A9 ON ON ON \u00C2\u00A9 Q \u00E2\u0080\u0094 CN c n m NO r~ 0 0 Os \u00E2\u0080\u00A2~ X tV3 APPENDIX C INDIVIDUAL SUBJECTS' MISMATCH NEGATIVITY MEASURES AT Fz Individual Subjects' MMN Peak-Amplitude Measures (microvolts) at Fz Subject /ba/-/Da/ /Da/-/ba/ /ba/-/da/ /da/-/ba/ /da/-/Da/ /Da/-/da/ 1 -2.77 -0.63 -0.37 -1.19 -0.52 -0.50 2 -1.52 -0.88 -2.00 -1.19 -1.09 -0.29 3 -2.92 -1.01 0.29 -0.82 -1.58 -1.09 . 4 -3.08 -1.53 0.06 -1.89 -1.18 -0.92 5 0.14 -0.66 -0.16 0.42 -1.42 0.38 6 -1.78 -2.08 -0.07 -2.95 -0.69 -2.24 7 -1.42 -1.44 -1.95 -1.06 -0.85 -1.36 8 -2.88 -1.79 1.33 -0.76 -2.71 -1.96 9 -1.90 -0.82 1.20 -0.32 -2.60 -0.76 10 -1.69 -1.64 -1.43 -1.76 -2.74 -1.45 11 -2.61 -2.61 -2.51 -2.56 -2.60 -0.67 MEAN , -2.04 -1.37 -0.51 -1.28 -1.63 -0.99 SD 0.95 0.64 1.29 0.97 0.87 0.75 83 Individual Subjects' MMN Latency Measures (ms) at Fz Subject IbaJ-IDal IDaJ-lbaJ /ba/-/da/ /da/-/ba/ /da/-/Da/ /Da/-/da/ 1 210 232 214 230 208 218 2 166 180 224 140 132 192 3 216 224 196 222 208 190 4 206 228 158 180 132 206 5 136 140 192 154 212 192 6 196 198 150 148 202 232 7 206 232 220 200 220 224 8 226 220 222 232 228 210 9 212 198 204 168 222 206 10 204 232 186 212 180 194 11 162 132 196 136 160 174 MEAN 194.5 201.5 196.5 183.8 191.3 203.5 SD 27.7 36.7 24.7 36.9 35.1 17.0 /ba/ = #2 /da/= #9 /Da/ = #16 APPENDIX D ALTERNATIVE DIFFERENCE WAVE CALCULATIONS 85 The M M N results presented in the preceding chapter were derived from difference waves, \"z\", obtained by subtracting the waveform to the standard stimulus from the waveform to the deviant stimulus within the same condition. In order to control for possible stimulus-specific effects, two alternative methods of calculating the difference waves were examined. The first method, \"x\", was the subtraction of an average waveform from the baseline condition of a stimulus from the waveform of the same stimulus presented as a deviant (Kraus et al., 1995; Kraus et al., 1992). The second method, \"y\", involved a comparison of the same stimulus across the counterbalanced directions of the conditions, allowing the subtraction of a stimulus when it was the standard from the same stimulus as a deviant (Martin et al., 1999; Whiting et al., 1998). Because all three analysis methods yielded similar patterns of M M N results, only the difference waves from the original subtraction method were presented in this paper. z = deviant waveform - standard waveform (within-block comparison) x = deviant waveform - baseline waveform y = deviant waveform (Direction 1) - standard waveform (Direction 2) The following figures show grand mean (n = 11 subjects) difference waveforms for all EEG recording channels to all six experimental conditions (three contrasts; 2 directions per contrast). Each figure illustrates both the \"x\" and \"y\" difference waves, with a separate figure per condition. The \"-x\" or \"-y\" at the end of each file name denotes the type of subtraction. The numbers preceding it code for the conditions. The order of presentation of the figures and the legend for the file names of the conditions are as follows: 1. 2_16 = /ba/to/Da/ 2. 16_2 = /Da/to/ba/ 3. 2_9 = /ba/ to /da/ 4. 9_2 = /da/ to /ba/ 5. 9_16 = /da/to/Da/ 6. 16_9 = /Da/to/ba/ Alternative Calculations: Grand Mean Difference Waves Alternative Calculations: Grand Mean Difference Waves Alternative Calculations: Grand Mean Difference Waves 88 Condition: /ba/ to /da/ Alternative Calculations: Grand Mean Difference Waves Condition: IdalXofDal Alternative Calculations: Grand Mean Difference Waves APPENDIX E MISMATCH NEGATIVITY MEASURES OF A WITHIN-CATEGORY CONTRAST: /ba/#2 vs./ba/#5 93 After beginning the study, it was discovered that the /ba/ (#2) exemplar chosen from the synthetic continuum differed slightly from the rest of the continuum in the frequency of the noise burst. The frequency of this burst is necessarily slightly lower in the first two /ba/ tokens than in the rest of the continuum (occurring between 2500 - 3500 Hz rather than 3000 - 4000 Hz) in order to maintain a more natural, speech-like quality. Due to the acoustic difference that would therefore occur in the native but not the non-native contrasts, a pilot study was conducted in order to assess a possible confound of the study. This was deemed necessary because if the results of the study were to show that the M M N in response to the native contrast is larger than to the non-native contrast, it might possibly be due to this acoustic difference rather than to linguistic factors. In fact, the necessity of this pilot study was subsequently negated by the result that the M M N to the native contrast is not larger in amplitude, and, indeed, tends to be smaller than that to the non-native contrast. The pilot consisted of recording M M N in response to a within-category contrast: /ba/ #2 vs. /ba/ #5, in both directions of presentation. If the difference in noise burst frequency was salient enough to cause a large M M N to this, otherwise smaller acoustic difference, it would suggest that the noise burst frequency is a possible confound. Six subjects participated in each direction. A l l other recording variables (e.g., ISI, number of trials, intensity) were the same as the main study. The M M N results to the within-category change are similar to, though smaller in amplitude than, those to the native /ba/ vs. /da/ of the main study. There is a questionable M M N present to the /ba/ #2 to /ba/ #5 condition, where #2 is the standard, and a relatively small M M N to the opposite condition, where #5 is the standard. This could be due to the differences between burst frequencies and/or F2/F3 frequencies. The peak amplitude and latency values for each direction are reported on the following page. M M N Measures of a Within-Category Condition: /ba/ #2 vs. /ba/ #5 94 /ba/ #2 to /ba/ #5 /ba/ #5 to /ba/ #2 Subject Peak Amplitude (pV) Latency (ms) Peak Amplitude (pV) Latency (ms) 1 0.33 200 0.05 224 2 -0.37 160 DNT DNT 3 DNT DNT -0.90 204 4 -1.78 174 DNT DNT 5 DNT DNT -0.49 132 6 -0.98 158 DNT DNT 7 -0.73 228 DNT DNT 8 DNT DNT -.60 204 9 DNT DNT -1.36 134 10 1.20 158 DNT DNT 11 DNT DNT -1.53 132 Mean -0.39 180 -0.81 172 SD 1.05 29 0.58 43 DNT = Did not test "@en . "Thesis/Dissertation"@en . "2000-05"@en . "10.14288/1.0089470"@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 . "Mismatch negativity cortical event-related potential measures of cross-linguistic phoneme perception"@en . "Text"@en . "http://hdl.handle.net/2429/10486"@en .