MSU LIBRARIES .—:__ RETURNING MATERIALS: PIace in book drop to remove this checkout from your record. FINES wiII be charged if book is returned after the date stamped below. SYLLABIC POSITIGQ AND m OF PERFORMANCE AS FACTORS IN RHONOLOGICAL INTERFERENCE By Erna Jeanne Hentiitte Pluut A THESIS Submitted to Michigan State university in partial fulfillment of the requirements for the degree of MASTER OF ARTS DEPARTMENT OF AUDIOLOGY AND SPEECH SCIENCES 1985 ABSTRACT SYLLABIC POSITIGW AND HIDE OF PERFORMANCE AS FACTORS IN PMOIMICAL INTERFERENCE By Erna Jeanne HenriEtte Pluut This study investigated the syllabic position of phonemes and the nude of performance as factors in phonological interference. 1he study focused on the error rate of initial versus final oral and nasal stops in both production and identification modes of performance for adult bilingual speakers. The focus of the study was threefold. First, the error frequency and pattern of initial versus final position in both nudes of performance were regarded. Second, the relationship between error frequency, and segment class and type was observed in both modes of performance. Third, the relationship between performance on production and identification tasks was considered. Six Mandarin speaking adults of Taiwanese origin, who spoke English as a foreign language, participated in the study. The general design was one in which the subjects read a list of test items, and identified spoken.test iteum by listening to an audio tape. Results supported the conclusion that syllabic position and node of performance function as factors influencing the frequency and pat- tern of consonant errors made in English by native Mandarin speakers. 111e general conclusion was that the phenomenon of interference is more complex than the contrastive analysis approach implies. In sum, comparison of the phonemic inventories of two langauges is insufficient to explain the errors that speakers nuke when learning the sounds of a foreign language. DEDI CAT“)! Dedicated to my parents and my husband and all others who use more than one language for commnication. ii AWS I would like to thank several people for their valued contributions to the carpletion of this paper. First, Dr. Ida Stockman, chairman of the thesis cunnittee, for the considerable amount of time, her advice, and so much appreciated encouragement throughout the “hole study. I also wish to express my appreciation to the members of the thesis conmittee--Dr. Leo V. Deal, Dr. Brad S. Rakerd, and Dr. Su-O Tien—-for their concern and interest as well as for their support during the study. Thanks too are conveyed to Mr. George Ganble for his help and patience in the technical aspects of the research. My thanks are extended to the six subjects for their willingness to participate and the time they spent on behalf of the research. Appreciation and thanks are also expressed to my friends for their support and encouragement and to Mr. George Olen for his help with statistical aspects and his time and energy spent on drawing the graphs. Finally, my sincere gratitude and appreciation are expressed to my parents, who fostered an attitude that enabled living and studying in foreign countries and languages, and to my husband, Owen 010W, who provided the opportunity to begin and conplete a study like this. iii 'EABLEICX? CIIHUEWTS Page CHAPTER I BACKGROUND FOR THE STUDY Introduction. . . . . . . . . . . . . . . . . . . . l Interference as a Phenomenon in Second Language Acquisition. . . . . . . . . . . . . Transfer and Interference Defined. . . . . . . . Phonological Interference. . . . . . . . . . . . Phonological Interference and Speech- Language Pathology. . . . . . . . . . . . . . 5 Contrastive Analysis as an Approach to Phonological Interference. . . . . . . . . 7 Factors Not Taken Into Account by Cbntrastive Analysis. . . . . . . . . . . . . 8 ‘Context as a Factor in Phonological Interference. . . . . . . . . . . . . . . . . 10 Mode of Performance as a Factor in Phonological Interference . . . . . . . . . . 14 SumnaryofIssues................ 17 sz—o Statement of the Problem. . . . . . . . . . . . . . 18 Exanfining Linguistic Cbntext as a Factor in Phonological Interference. . . . . . . . . 19 Examining the Mode of Performance as a Factor in Phonological Interference . . . . . 21 Purpose of the Study. . . . . . . . . . . . . . . . 22 CHAPTER II PROCEDURES Subjects. . . . . . . . . . . . . . . . . . . . . . 24 Selection Criterion. . . . . . . . . . . . . . . 24 Selection Procedure. . . . . . . . . . . . . . . 24 Subject Characteristics. . . . . . . . . . . . . 25 iv Ma‘etialsoosaaaaooooa Overview of Tasks. . . . Data Collection Procedures. Overview of Tasks. . . . The Production Task. . . The Identification Task. Data Analysis . . . . . . . Overview for Both Tasks. The Production Task. . . The Identification Task. CHAPTER III RESULTS The Production Thsk . . . . Error Frequency. . . . . Error Pattern. . . . . . The Identification Task . . Error Frequency. . . . . Error Pattern. . . . . . Description of Stimslus Items. Carrierphrase. . . . . . . . . Rationale for Stinnlus droice. The Relationship Between Production Identification . . . . . and Relationship by Syllabic Position. . . Relationship by Segment Class and Type CHAPTER IV DISCUSSION AND IMPLICATIONS Purpose of the Study. . . . . . . . . . . SunrnaryoftheStudy........... Discussion of the Results . Inconsistency of the Initial Final Effect. Higher Error Frequency for Nasals than for Plosives . . . . . The Relationship Between Production and Identification . . . . Limitations of the Study. . . . . . . . . . Inplications for Practical Purposes . . . . Implications for Future Research. . . . . . 26 26 27 27 27 29 29 29 32 34 34 35 35 36 36 45 46 46 54 55 55 60 62 62 63 63 69 71 73 74 75 Conclusions . . . . . . . . . . . . . . . Appendix A Survey Form for Selecting Subjects . Appendix B Subjects' Reading List for the Production Task . . . . . . . . . Appendix C Judges' List for the Production Task Appendix D Subjects' Written Form for the Identification Task . . . . . . . Appendix E Number of Subjects Participating per Error Pattern in the Production Task. Appendix F Nunber of Subjects Participating per Error Pattern in the Identification Task. Bibl iogt‘phy O O O O O C O O O O O I O O O O O O O O 0 0 vi 77 79 81 84 87 92 93 94 Table Table Table Table Table Table Table Table Tabl e Table Table Table Table 10 ll 12 13 1.! 81' (I TABLES Overview of Subjects' Characteristics . . . . . . Overall Percentage of Error on the ProductionTask............... Percentage of Error by Syllabic Position for Plosive and Nasal Production . . . . . . . . Percentage of Error by Individual Segment for Plosive Production . . . . . . . . . . . . . Percentage of Error by Individual Segment for NasalProduction.............. Percentage of Error by Syllabic Position With and Without /q/ for the Production Task. . . Error Patterns by Class for the Production Task Displayed in Percentage. . . . . . . . . . . Overall Percentage of Error on the Identification Task. . . . . . . . . . . . . Percentage of Error by Syllabic Position for Plosive and Nasal Identification . . . . . . Percentage of Error by Individual Segment for Plosive Identification . . . . . . . . . . . Percentage of Error by Individual Segment for Nasal Identification . . . . . . . . . . . . Percentage of Error by Syllabic Position With and Without /q/ for the Identification Task. Error Pattern by Class for the Identification Task Displayed in Percentage . . . . . . . . vii Page 26 37 39 41 42 44 45 47 48 50 51 53 55 Table 14 Number of Subjects Participating per Error Pattern in the Production Task . . . . . . . . 92 Table 15 Number of Subjects Participating per Error Pattern in the Identification Task . . . . . . 93 viii LIENFIOF'IFIGIIUBS Page Figure 1 The Relationship by Syllabic Position Between the Error Percentage of the Production and the Identification'Task. . . . . . . . . . . 57 ix 1:1. ESL JSI-D ISIQ ELT] IRAL SLA ILIST'CX?.ABBMHTVIAJWIINS Native Language Foreign Language English as a Second Language Journal of Speech and Hearing Disorders Journal of Speech and Ekaring Research English Language Thaching Journal International Research and Linguistics Second Language Acquisition GiAPTER I BACKGRQND FOR TIE STIDY Introduction The sophisticated technology and the increased mobility of the modern world have increased the opportunity for people to reach across the borders of their countries. One of the consequences entailed in this opportunity is cornmnication in a language other than one's own. Considering interlingual conxnunication from a linguistic point of view, one finds that the process of mastering a foreign language (referred to as FL in contradistinction to N1. or native language) involves learning at all levels of a new language system. As described by Akmajian, Demers and Harnish (1979), these levels are phonology, representing the structure of sounds; morphology, which covers the structure of words; syntax, standing for the structure of sentences; and pragmatics, which includes the use of language for conmnication purposes. This implies that the learner of a FL will be faced with new phonological systems, new vocabulary, and different syntax and pragmatics, either totally or partially. I E El . 5 II E . . . Mach has been published on second language acquisition (referred to as SIA henceforth). Of particular interest to scholars studying the process and the results of learning a FL is the phenomenon of transfer. In the broadest sense transfer refers to the use of features or characteristics of one language that can or should be applied to 2 another language. In the literature this phenomenon is referred to with different terminology and definitions. The two most frequently used terms appear to be transfer and interference. WW Albert and Cbler (1978) describe m as imposing structures from (a) previous language(s) on the target language. Laroche (1981) sees transfer as "utterances accepted in both languages conditioned by the native language." Clyne (1975), as cited in Hirschfeld (1983) uses the term transference for "Take-over of elements, characteristics and rules from another language" (p. 51). Hirschfeld (1983) states that generally in linguistics transfer is defined as influences from the mother tongue that do not violate the norm of the FL and stinxrlate the learning process. Bunte and Kendall (1981) add to this, the need of acknowledging and recognizing the existence and importance of transfer at the pragmatic level. Hirschfeld (1983) mentions that the term W is sometimes used in a narrow sense to refer to the interfering influences of the NL language on the acquisition of a FL. Weinreich (1953) explains: The term interference implies the rearrangement of patterns that result from the introduction of foreign elements into the more highly structured domains of language, such as the bulk of the phonemic system, a large part of the morphology and syntax, and some areas of the vocabulary (kinship, color, weather, etc.) (1). 1). 3 In a broader sense, the term interference is used for the situation of language confusion, meaning the process as well as the result of the confusion. In reviewing the definitions on interference and transfer, one finds contradicting use of the same terminology among the different authors. It seems that the distinction between the process and the effect of the phenomenon has been recognized, but not stated clearly (l-Iirschfeld, 1983). Weinreich (1953) describes interference as follows: Those instances of deviation from the norms of either language which occur in the speech of bilinguals as a result of their familiarity with more than one language, i.e. as a result of language contact, will be referred to as mm PM" (o. 1). It follows from this description that the term 11.811131. describes the process whereas the term W indicates the effect of the phenomenon. The term "interference" and its definition as used by Weinreich (1953) will be used in this study, since the goal of the study is to explore the effect and not the process of the phenanenon. We: The aforementioned phenomenon of interference in learning a FL can occur at all levels of the FL. Among those levels the one of phonology is of particular interest. As Strevens (1973) describes, at the phonological level not only are mental processes required, but 4 there is also a great deal of motor involvement. The other levels may function with only mental activities, such as perceiving, decoding, coding. Only the phonological level functions with the actual production and perception of the speech sounds, in which the concepts and thoughts are wrapped. Flege (1980) stresses the existence and inportance of articulatory motor control for the production of speech sounds. He further mentions that the findings of research on articulatory motor learning requirements of second language learning suggest that: . . . establishment of articulatory motor control is itself an important part of second language learning. The language learner, it seems, must acquire complex new sets of highly automatic articulatory gestures or modify existing patterns of phonetic implementation in 351111193, to acquiring control of an abstract, reorganized phonology" (p. 18). In other words the levels of morphology, syntax and pragmatics represent the language system used for exchange of messages and thoughts. The actual realization of the system though occurs through phonology. It follows, then, that phonological interference differs from interference at the other levels in terms of the actual realization of the language system. The results of phonological interference can have a great inpact on the cornmnicative functioning in the FL. It may appear in slight deviances at the allophonic level, such as unaspirated plosives in the initial position of words, that will not influence the cornnrnication 5 in a langauge like English. Phonological interference may appear in more serious forms though, such as deletion, substitution, distortion, addition, and reduction of sounds. It speaks for itself that these forms can lead to confusion, misunderstanding and even complete lack of understanding, which in their turn hinder or even prevent the comnunicative function of the spoken or perceived language. Further it should be realized that phonological interference can penetrate at other linguistic levels. The deletion or addition of a final/s/ at the phonological level, may lead to confusion at the morphological level. So besides the difference of phonology in having one more aspect compared to the other linguistic levels, there is also the very practical aspect of direct conmunication that sets phonology apart and makes phonological interference a topic of special interest. Or, as Rivers states as cited by Ihenacho (1980, p. 245): "Since language is a means of conrmnication, it is not enough for our students to learn words, phrases, granmatical features if they will not be able to produce these in a way which makes their utterance conprehensible to a native speaker of that language." BMW The problems that learners of a FL encounter have not only been recognized from the side of linguistics and teaching of a SL but also from the side of Speech and Hearing Sciences, namely Speech-Language- Pathology. The American Speech-Language-Hearing Association, AS-IA, (1985) states: 6 Researchers and clinicians are only beginning to amass a knowledge base on the characteristics of normal language development in various minority languages, bilingual language learning, second language acquisition, dominance testing, bilingual assessment and remediation of commnicative disorders, and the applications of emerging computer technology for use with minority language groups (p. 29). Further, the position statement continues: However, the speech-language pathologist must have certain competencies to distinguish between dialectal differences (due to interaction from the minority language) and conxnunicative disorders. These cmpetencies include understanding the minority language as a rule-governed system, knowledge of the contrastive phonological, granmatical, semantic, and pragmatic features of the minority language, and knowledge of non-discriminatory testing procedures" (refer to "Social Dialects: a Position Paper," AS-IA, Septenber 1983). There are several reasons for this interest from the side of Speech and Hearing Sciences. The main reason is the fact that nowadays speech-language pathologists are more and more confronted with people for whom English is a FL. This does not only hold for youngsters in schools but also for the older population with disorders such as aphasia, apraxia, dysarthria, voice disorders, hearing disorders. It is very important for the speech-language pathologist to be adequately informed about bilingual language and speech acquisition and 7 possible interference in order to distinguish correctly between pathological patterns in the client's English and errors that result from language contact. The second reason is the specialized competency of the speech-language pathologist for the practical aspects of phonology, such as speech development, production and perception, and the correction of speech errors. Although it can be argued that this competency has been proven for only the NL, it is still such that the field of Speech-Language Pathology can contribute importantly to the knowledge about the practical consequences of phonological interference. C . E l . E I W The literature available on phonological interferences is offered mainly from linguistics and Second Language Teaching. In this literature, the method of contrastive analysis has often been used to predict or account for errors occurring in the FL. In linguistics the term contrastive analysis is used for analyzing the existing differences between the two languages at several linguistic levels that make the two language systems contrastive. At the phonological level, for instance, the phonemic inventories of the languages under study are compared in order to predict pronunciation errors made in the FL. Schumann and Stenson (1974) indicate that contrastive analysis manifests itself in two versions. They cite Wardhaugh (1974) in 8 explaining those two versions. The strong version claims that by means of contrastive analysis the errors that learners of a FL will make in that FL, can be predicted. The weak version uses the differences between two languages to account for observed errors. According to Wardhaugh (1974), the strong version is untenable at the present, since i: would require minimally a linguistic theory, covering linguistic universals of syntax, semantics and phonology, in order to deal adequately with those levels. Although such a linguistic theory is not available at the moment, the strong version is still the version that underlies rmch of the work done on interference. E II II I E I 3 . E I . In the last decade the linguistic theory has started to acknowledge the limitations of contrastive analysis. Wardhaugh (1974) described the weakness of contrastive analysis and concluded: "The contrastive analysis hypothesis has not proved to be workable, at least not in the strong version in which it was originally expressed" (p. 18). He continued: "In its weak version, however, it has proved to be helpful and undoubtedly will continue to be so as linguistic theory develops" (p. 18). Wardhaugh’s (1974) position gains support frean a number of writers who have examined phonological interference directly. Among the studies that concentrate on phonological interference are the ones of Flege (1980), Wilson and Mallergard (1981), and Broselow (1983). Flege (1980), who studied phonetic approximation in English as a FL for Saudi Arabians, concluded that phonetic norms of the NL may carry over 9 to the production of speech in the FL. He suggested that phonetic interference is not static but should be regarded as a part of the interlanguage system of learners of a FL, which would allow for individual phonetic strategies among the learners. He ended with mentioning that the general pattern of phonetic learning in a FL shows similarity with that of a child learning a NL. Wilson and Mollerglrd (1981) found in their study on production errors of the vowel /A/ for Norwegian learners of English, that rarely one simple explanation can be given for a type of error. They considered a range of involved factors, such as different orthography, language background, teaching situation, phonological context and the difference in speech production for a reading task or during spontaneous speech. This position takes side with Chen (1976), who stressed the difference in speech production for single-word-reading and sentence-reading in his study on phonological interference. Broselow (1983) in her study on transfer using epenthesis errors, argued that pronunciation errors "mast be analyzed as belonging to syllable structures which are not permitted in the native language." She further argued that the mispronunciations "represent an attenpt by the language learner to bring second language forms into conformity with first language restrictions defining possible syllables." More writers have mentioned other causes of interference that go beyond contrastive analysis even though their claims were not based on actual studies of interference. These studies are of descriptive 10 nature. The causes of interference that were found in those studies are sunnarized below. The different causes for the found phonological interference vary from lack of proper eartraining for new sounds (Roy, 1975); unfamiliar sounds in the FL (Roy, 1975; Mirhassani, 1983; Reed, et. al., 1949); different spelling for shared sounds (Sunlktu, 1957; Mirhassani, 1983); to unfamiliar sound combinations (Mirhassani, 1983). Saunders (1963) added as reasons for the maintenance of phonological interference lack of awareness, the learner's attitude towards the FL and the psychological factors involved in pronouncing new phonemes that may sound or feel funny. The position follows Sapir, who mentioned in 1927 that speech sounds are not only a matter of articulatory or acoustical image, but also subject to symbolic expression, which involves psychological factors. It appears, then, that there is a lot of support for the inadequacy of contrastive analysis to predict or explain errors in phonological interference. Therefore, further studies need to be done, that go beyond the information which contrastive analysis provides. Two factors that have not been regarded by contrastive analysis are context and mode of performance. In the following sections the importance of these factors will be considered. In reviewing the literature one finds two kinds of context being ignored by the contrastive analysis hypothesis. The first one, 11 situational context, is mentioned by Schunann and Stenson (1974), who state: No theory of contrastive analysis, strong or weak, should be expected to account for all errors of language learning. Much evidence is already available which suggests that many errors are due to target-language rule deviance as well. In addition, there are many errors induced by the classroan situation, but which cannot be considered to be a function of performance (i.e., due to inattention, memory lapse, outside interference, etc.) (p. 2 and 3). Secondly, linguistic context has not been taken into account by the contrastive analysis approach. Trusting contrastive analysis as a predicting or explanatory means for phonological interference inplies acquisition of single sounds, since contrastive analysis often makes use of phonemic charts to contrast the phonological systems of two languages. In this case, context cannot be taken into account. Many authors, though, stress the value of linguistic context. Ingram (1974) described contextual influence as combinatory influences of neighboring speech sounds on production and perception of other speech segments. The contextual importance has been stressed for phonology in the NL as well as for the FL. With respect to the NL it has been shown that context is an importance learning variable in both normal and clinical populations. For normally developing children, Ingram (1974) showed in his study on phonological rules in young children that sounds are produced and 12 perceived in context. Other evidence on normal development is provided by Tomlin (1957) and Olmsted (1971) in their studies on child speech development. They both concluded that the moment of acquisition of a sound depended on syllable or word position. The general conclusion for consonants was that they are acquired in word initial position before they are acquired in word final position. Jakobson (1968) described the similarities between the phonological developnent in children and the phonological systems of the languages in the world. He mentioned the likelihood of a universal tendency towards the development of sounds in CV syllables rather than as single sounds. Mackay (1978) concluded in his study on speech errors inside the syllable by nonhandicapped adult native speakers that the phonological availability for sounds is influenced by the syllabic position of those sounds. For the deviant speech in the NL, Hodson and Paden (1981) found in their study on phonological rules in unintelligble children, the deletion of final consonants as a factor. Blumstein (1980) studied phonological errors in aphasic speech and found syllabic position of phonemes within the syllable of importance for articulatory programing. Often the process of acquiring a FL is believed to have much in cornnon with the acquisition of a first language (Baetens-Beardmore, 1982; Owens, 1984). Ovens (1984) also mentions that "phonological development follows a similar pattern in first and second language. The phonological system from the first language forms a foundation for 13 the second" (9. 321). Similar conclusions are drawn by Flege (1980) in that the general pattern of phonetic learning in a FL for adults resenbles the phonetic learning in the NL for a child. Literature on FL learning has described syllabic position of phonemes as a form of linguistic context. Jackson (1981) who studied pronunciation of English consonants by Indian learners, mentioned the inyortance of considering the whole system of speech sounds in the involved languages. Saunders (1962), Sunslktu (1957), and especially Shen (1954, 1959, 1961, 1962), analyzed the errors observed in English as a Second Language (ESL henceforth) in speakers of different native languages. They stressed the differences in phoneme distribution within word or syllable as a factor of great importance in phonological interference. Sapir (1927) spoke of the influence of phonetic patterning; Amiz (1973) who described ESL for Iraqi students, mentioned the cluster position within a word; and Mirhassani (1983), in describing ESL for Iranian, considered unfamiliar positions for clusters within a word as a factor of phonological interference. All aforementioned studies regarding linguistic context were based on observations, without referral to any methodology for collecting and analyzing the data or specification of the circumstances under which the observations were made. All authors have in cannon in their findings that there seems to be a difference for segments to occur in initial position and in final position. Whether in child speech disorders (Ingram, 1976; Hudson and Paden, 1981), aphasia (Blumstein, 1980), child speech development 14 (Tenplin, 1957; Jakobson, 1968; Olmsted, 1971) or in speech of non- handicapped native speakers (Mackay, 1978), it appears that consonants in initial position are different from those in final position. Ini- tials seem to be earlier acquired, more stable, more readily available, and less susceptible to disorders than final consonants. Also in FL literature syllabic position has been mentioned as important. These findings raise questions about whether a sound not occurring in a certain position in the NL, would be produced correctly in that position in a FL, even if the phoneme is the same for each language. MW We: Another factor not regarded by the contrastive analysis approach is the node of performance. Literature on phonological interference fran the contrastive analysis hypothesis' point of view mentions only the production of speech sounds. It seems that the perceptual mode of performance is ignored. In linguistics speech production is considered the spoken realization of the language system, in other words spoken language. Further, in linguistics the terms language perception and speech perception are used to refer to the input of spoken language. There are two components in perceiving spoken language. First, there is the conprehension aspect that refers to the content of the conveyed message. Second, there is the recognition of speech signals. The latter one allows distinction between different forms of a language system, such as words. In this way it contributes to the canprehension aspect. For example, a distinction between /t/ and /9/ allows 15 understanding of the words tree and three as two different meanings. It seems that recognition of speech signals mainly takes place at the phonological level. Canprehension, on the other hand, seems to relate to vocabulary, syntax, and pragmatics. Therefore, the two components often are considered as two levels and labeled as language or speech canprehension and speech perception. Although contrastive analysis seems to disregard perception, literature on language acquisition values the contribution of perception to the learning process, especially fran the view point of language comprehension. Baetens-Beardmore (1982), Garcia (1983), Albert and daler (1978), and Cass and Selinker (1983) are among the authors in the field of second language learning who stress the importance of language input for the development of camnnication skills in a FL. The ignorance of the perceptual mode by contrastive analysis leaves one to assume a synmetrical relationship between perception and production. Literature on both language carprehension and speech perception shows evidence for an asynrnetrical relationship. In literature on bilinguality and learning of a FL, it is stressed that speakers of more than one language have different competencies for the modes of perception and production. It is, for instance, possible for a person to understand a FL well without being able to speak it fluently. Further, this relationship is considered asynmetrical in that perception is thought to be prerequisite to production. Albert and l6 Obler (1978) argue that although a certain connection between the systems exist, they also are to some degree independent of each other (p. 251). They support their statement by referring to the observations that perception precedes production in language learning, and that the production system seems to suffer more than the perception system in those instances where one of the languages has not been used for a while. This independence of the two systems implies that within the general trend individual differences may occur. At the level of speech perception, the asyumetrical relationship has been explored in the acquisition of a FL as well as a NL. Q) the side of speech language pathology it is known that perception and production are related in the acquisition of phonological rules. Edwards (1974) and Ingran (1974) both conclude the existence of the relationship between perception and production in child phonology. Edwards' findings show that "the order of devel0pment in phonemic perception may not be identical with the order in production" (p. 218). But in general it is accepted that phonemic perception precedes correct production. Goto (1971), and Sheldon and Strange (1984) studied the perception and production of /l/ and /r/ by Japanese students of ESL. They concluded that perception and production do not have a synmetric relationship at the phonological level either and that correct product- ion can occur along with incorrect perception. Sleldon and Strange (1984) stressed the important influence of the phonological environment on the target sound, including the phoneme distribution within the 17 word. It seems, thus, that in language comprehension, perception is thought to be prerequisite for production. Speech perception, though, can precede or follow correct production. The evidence pointing to asymetry between perception and production suggests that performance patterns predicted for production need not apply to perception as a contrastive analysis interpretation would inply. Consequently, the question can be raised about whether there is a relationship between errors in the perception and production of the same phonemes. W In reviewing the limitations of contrastive analysis as a means of predicting or explaining phonological interference, it seems that many questions remain without answer. Che of those questions considers the linguistic context of the phonemes that are contrasted especially their distribution within the syllable. Another question regards the mode of performance. The contribution of the perceptual ability and the relationship between perception and production in phonological interference seem not to be taken into account by contrastive analysis. The difference between the modes of performancenperception being a mental process and production being characterized by the motor aspect of realizing the language system--is also ignored. Answering those questions would be of both theoretical and practical inportance. Theoretical knowledge about the existence and extent of the syllable distribution of phonemes as a factor of phonological interference would add to better understanding of the l8 phenomenon. Better and broader understanding of phonological interference would lead to a better and more adequate approach to avoid or reduce its effects. This could be useful for FL teaching and in speech language pathology as well. Statement of the Problem In reviewing the literature it appears that the topic of phonological interference in the acquisition of a FL is of great interest. This interest is shared by not only the fields of linguistics and Second Language Teaching, but also by speech-language pathology. The latter one adds knowledge about the motor aspects of speech, perception of speech, and the phonological develognent in the NL to the theoretical and practical/educational knowledge of the other two fields. It is further noticed that the literature reports little research on the positional effect of segments within the syllable as a factor in phonological interference. Nevertheless descriptive studies consider this factor of value, and in the research on phonological developnent within the NL, the influence of segment distribution in the syllable has been acknowledged. Since the acquisition of a FL, including the phonological acquisition, is regarded to take place similarly to the acquisition of the NL, the question may be raised whether and how far segment position effects play a role in the occurrence of phonology interference. Taking into account that phonology is the actual realization of the language system and thus directly related to cmnmnication, it seems that the effect of segment 19 distribution within the syllable as a factor in phonological interference deserves a systematic investigation. W'EII'IIE In order to study linguistic content as a factor in phonological interference it would seem appropriate to choose contrastive languages with very different segment distributions within the syllable but yet identical phonemic segments. It is hypothesized that the segment distribution within syllables that are accepted according to the phonological rules in the M. will interfere with the different segment distribution within syllables as accepted in the FL. It is further hypothesized that the interference will take the form of adapting the syllable structure of the FL as much as possible to an accepted form in the NL, whether by deletion or addition of segments. Examples of two languages with very different syllable structures are English and Mandarin. They come from the two very different language families: Indo-European and Sino-Tibetan (Gage, 1981). Among these, English represents the Indo-European sub-language Germanic, and Mandarin represents the Chinese sub-language within the Sino-Tibetan language family (Gage, 1981). The sub-languages differ in phonemic inventory and phonemic distribution. In general it can be said that pitch at the syllabic level is phonemic in Chinese, but not in Germanic languages. A main difference in phonemic distribution can be considered the occurrence of consonant clusters in Germanic and the lack of them in Chinese. 20 The differences between the exemplars of the subfamilies can be described in more detail. Within the phonemic inventory category, the feature of aspiration for initial consonants is recognized as phonemic in Mandarin but not in English. Reversed, the distinction voiced- voiceless for consonants has phonemic value in English but not in Mandarin. Within the category of phonemic distribution two main differences are found. In English consonant clusters occur frequently, but Mandarin lacks them. Further are only a few consonants used in Mandarin in the final position of a syllable, whereas most of the consonants in initial position in English do also occur in final position. In the phonetic inventory category similarities also are found. Besides in the close nunber of phonological elements (43 and 42 for respectively English and Mandarin) and in the close distribution of those elements over syllabic and nonsyllabic phonemes (19-24, and 20-22 for respectively English and Mandarin), similarities are found in the occurrence of vowels and consonants. The shared vowels are /i/a/u/, the shared dipthong is /au/, and the shared consonants are /p/t/k/f/s/z/m/n/r)/l/r/h. Of the set two entire natural classes are represented. They are the voiceless stops /p/t/k/ and the nasal stops /m/n/rJ/. The segments of those natural classes represent similarities and differences in the distribution within the syllable for the two languages. All consonants of the plosive class occur word initial in both Mandarin and English. Word final they only occur in English. The class of the nasals shows less uniformity in their segment 21 distribution. The /m/ is identical in position appearance to the plosives, and can occur word initial in Mandarin and in both positions in English. The /n/ occurs word initial and word final in both languages, and the [9/ only appears in word final position in both languages. According to the differences in occurrence of final consonants in the syllable then, the question arises whether native Mandarin speakers would be likely to show more deviations of final consonants than of initial ones. In order to allow for investigation of the distributional effect of segments within the syllable and to rule out the possible influence of unfamiliarity with certain segments, it seemed wise to choose shared consonants and vowels for the test items. These appear to be /p/t/k/m/n/r]/i/a/u/. To avoid any confusion between child language develognent and SIA, it was decided to explore the above mentioned positional influence in the syllable in adult bilingual speakers. . . WWW. El I . I I I As mentioned before, because of its static nature, contrastive analysis appears to have limitations as a means of predicting or explaining phonological interference. The limitation of ignoring the linguistic context of the target sound has been discussed above. Another limitation is the neglected influence of the perceptual mode of performance or the named symetric relationship between perception and production of sounds. The influence of perception on production 22 and the asymetric relationship between the two modes of performance have been acknowledged in child phonological development. There are indications that this knowledge for the NL also holds for the phonological acquisition in SLA. Broadening the knowledge about the role of the mode of performance in phonological interference as occurring in SLA would have inplications for clinical purposes regarding bilingual or lmltilingual students. Taking this into account, it seems of interest to investigate the distributional influences of segments within the syllable as a factor in phonological interference in both modes of performance. Purpose of the Study The purpose of the present study, then, is to investigate the effect of segment distribution within the syllable on the perception and production of those segments as a factor of phonological interference. The study will be particularly considering the initial versus final position of single consonants. The study will attenpt to answer the following questions: 1. (a) Mast are the frequency and pattern of errors made by native Mandarin adult speakers in the production of initial versus final single consonants (oral and nasal staps) in monosyllabic words of CV and VC structures in an English enviroment? (b) Do the obtained error frequency and error pattern by segment position vary in relation to segment class or segment type of the target consonant? 23 (a) “hat are the frequency and pattern of errors made by the same native Mandarin speaking adults in the perception of initial versus final single consonants (the sane oral and nasal stops) in mnosyllabic CV and VC nonsense words in an English context? Do the error frequencies and patterns obtained in the production and perception of the before mentioned nonosyllabic words reveal any relationship? CHAPTER I I PROGDURES Subjects 51.:.. The subjects were recruited so as to achieve as much uniformity as possible in age, field of study, and background in English as a Second Language. A further selection criterion required subjects to have normal hearing acuity . W In order to recruit subjects who met the criteria for the study, a survey of the Chinese population was done, with the help of the Chinese Student Association of Michigan State University. The president of the group agreed to enclose survey forms in the quarterly newsletters at the beginning of Fall term 1984. The questions on the form related to the selection criteria and the willingness of the person to participate in the study. Subjects were selected after the investigator reviewed the forms returned by 20 respondents. A copy of the survey form can be found in Appendix A. The subjects' auditory acuity was screened for pure tones bilaterally at 250 - 500 - 1000 - 2000 Hz, and ZOdB SPL per ASHA recomnended level. All screening was conducted in an audiometric booth with a Tracoustics Audiometer. 24 25 5 I. I . . The subjects were three female and three male students fran Taiwan, Republic of China, who were enrolled at Michigan State Ihiversity at the time of the study. All subjects spoke only Mandarin- Chinese, however, Taiwanese could be understood by soon. The subjects arrived in the USA during Fall 1983, a fact which places the duration of their (5 stay at about 17 months at the time of the study. The age range of the subjects varied from 24.1 to 29.11, with an average of 25.8 years. The estimated total number of hours for studying ESL varied from 1200 to 3200, with an average of 2080 hours. Considering an academic year to include 40 weeks, those hours were calculated by nultiplying the years of study by the hours per week. The subjects were exposed to and had cournunicated in English for at least 17 months at the thme of the test. Regarding judgements of their own communication ability in English, two subjects rated thenmelves as adequate enough to lecture to a group of students. Three rated themselves as having good proficiency, and the remaining one subject rated herself proficient enough to be understood. The professional fields of study differed for all subjects and included Telecommnication, Connxlnication, Advertisement, Mechanical Engineering, Cmputer Science and Marketing. All subjects passed the Audiometric Pure Tone Screening Test at the ZOdB SPL level in both ears. Z6 Table 1 C. . w I S I . , Q . . 1 F 27.0 Canmnication 1920 Sufficient 2 M 25.5 Computer Science 1680 Good 3 M 24.1 Mechanical 1600 Good Engineering 4 M 29.11 Marketing 3200 Lecture 5 F 24.7 Advertising 2880 Good 6 F 24.6 Teleccsnnlnication 1200 Lecture-Good Materials W The subjects were required to read individually a prestructured set of stimulus nonosyllabic words under laboratory-like conditions. Their articulation of segments was judged by native speakers of English. This first task was labeled the production task. In the same sitting, the subjects were required to listen individually to aurally presented structured monosyllables and to select the perceived item from a written list. This task was labeled as the identification task. Each task contained three presentations of the basic set of stimulus items, making a maxinum of 108 items. 27 I . . IS. I I The stimulus items were CV and VC syllables in which the consonants /p/t/k/m/n/q/ were combined with one of the vowels /a/i/u/. Those vowels are present in both languages. The consonants /p/t/k/m/ occur in both English and Mandarin in syllable initial position but in final position in English only. The nasal /n/ occurs in both languages in both positions, and In] occurs in both languages only finally. This would allow then to test for the hypothesis of the syllabic position of phonemes as a factor of influence in the occurrence of phonological interference. Altogether 36 different syllables were created: /pa/ta/ka/ma/na/I]a/ lap/at/ak/an/an/aq/ Ipu/tu/ku/m/nu/nu/ lop/ut/uk/un/un/uo/ lpi/ti/ki/mi/ni/ui/ lip/it/ik/im/in/i'll W In both tasks the stimulus items appeared in the carrierphrase "say . . . again." The purpose of the carrier phrase was to allow for a closer approximation of the stimulus items in connected speech than would be possible when produced singly. The carrier phrase in this study was chosen in such a way that the phonetic enviromlent of the target syllable would be as neutral as possible. E'IES'IJ' The stimulus consonants were chosen for several reasons, one of them being their syllabic position in English and Mandarin. In order 28 to explore the influence of syllabic position of phonemes on the production and perception modes of performance, the study was in need of phonemes that are shared by both languages but differ in their syllabic distributions. The sounds /p/t/k/m/ satisfy this criterion: they occur initially in both languages but finally in English only. The nasal stop /n/ occurs in both positions in both languages. This distribution allows one to test the expectation that the error rate would not differ for the positions. Likewise, the nasal stop [9/ is not expected to present much difference in error rate for the final position. Since Iq/ does not occur initially though, it would allow for testing the hypothesis that an unfamiliar phoneme distribution of the NL interferes with the production and perception of the according phoneme in the FL. Further, these six consonants span two articulatory classes: voiceless oral stops /p/t/k/ and voiced nasal stops /m/n/r)/. Since the places of articulation are related across the two classes /p-m/, /t-n/, /k-']/, conparison at both the consonant class and segment levels was possible. The vowels /a/i/u/ are shared by both languages. Further they are cardinal vowels and contrastively distinctive. The next section will describe the data collection procedures for each task separately, after a brief overview of both tasks is given. 29 Data Collection Procedures W The order of the 108 items was randomized separately for each task. Each subject participated in the study individually; no group testing was done. The order of acconplishing the production and identification within one and the same session was counter balanced. Weak Winn The subjects were required to read a list of stimulus items. The 108 stimulus items in the carrier phrase "say . . . again" were presented in written form to each subject. As shown in Appendix B, only the single stimulus syllables were written down, and the carrier phrase was presented in capital letters at the top of each page as a reminder. The subjects read the stimlus items in an audiometric booth. The two rooms in the booth shared a window. The subject and the study leader were seated at a table in one room with the microphone (Model Ampex 2001) at a distance of six to eight inches from the subject's mouth. The production was taped in the adjacent room to reduce nonrelevant noises as much as possible. For taping, a TEAC A-Z340 4 channel SIMJL-SYNC Stereo reel-to- reel tape recorder was used. One subject per channel was taped on Scotch Magnetic Tapes 212 (45 min.) at a speed of 7 1/2 inches per 30 second. Further, the chronograph function of a digital quartz watch of the brand Meister-Anker nr. 834.710-6 was used as a stopwatch. The subjects were instructed to read the stimlus items in the carrier ‘phrase colunn wise fran the top to the bottom of the page. After each produced item, a three-second interval was held before the reading of the next item took place. To avoid rehearsal preparation for the next stimulus during this interval, the subjects were instructed to cover the item with a sheet of paper until the three seconds transpired. To indicate the end of the interval, a reading signal was given by the investigator. The signal consisted of a forward movement of the right hand. To reduce the nunber of invalid responses due to unfamiliarity of the subjects with the task, a training set preceded the real task. The subjects received a test list with CV and VC syllables containing different consonants. Further, the set was identical to the genuine task. After practicing the reading of the items in the carrier phrase "say . . . again", the three-second-interval was introduced. men from the side of the project leader and the subject's side agreement was reached about the comprehension and the execution of the instructions, the original task was started. The pre-task training was also used to adjust the loudness level of the input. I I. I E I . B The judges were recruited among graduate students in the Department for Audiology and Speech Sciences. Knowledge and experience 31 in phonetic transcription (IPA system) as required for the M.A. Degree for Speech-Language Pathology was the main selection criterion. In addition, hearing ability was required. The three judges were all native Americans. Che judge finished the M.A. Program for Speech-Language Pathology in August 1984 and was employed as a Clinical Fellow at the time of the judging. The two remaining judges were completing their last terms of graduate study for the M.A. degree in Speech-Language Pathology. All three judges had experience in phonetic transcription using the IPA system. Their hearing ability was normal as screened bilaterally with the Program 111 Tracoustics Audiometer at the frequencies of 250 - 500 - 1000 - 2000 Hz, 20 dB SPL in both ears. The judging took place in the same sound treated roan in which the subjects' productions had been taped. The same tape recorder used for taping served for reproducing the taped speech (TEAC A-2340 4 Channel SIMJL-SYNC Stereo reel-to-reel tape recorder). To allow several people to listen to the tape at the same time, a headphone amplifier (HA 100) was connected with the tape recorder and the output fed to headphones binaurally. The intensity of the amplifier output was at a comfortable loudness level as judged by the listeners.. The judges received a list of the stinulus items in the same order as taped by the subjects on the recorder, as can be seen in Appendix C. Behind each word, space was allowed for the transcription and for rating the confidence of their judgments on a scale of one to three. The judges were instructed to do the following: to cover the stinlulus word so any 32 bias would be avoided, to transcribe only the target syllable by using the IPA system, to mark down special occurrences such as breaks and repetitions, and to rate the confidence of their judgment on the three-point rating scale. All this was to be done in the three-second-interval between one heard item and the next one. All three judges listened to the same subject at the same time, which resulted in three judgment lists per subject. The tapes were listened to only once without any break or repetition. A short break was held after the judging of two subjects in succession. The order in which the subjects were judged was randomly chosen. All occurrences of a change in the target consonant, on which at least two of the three judges agreed, were considered as an error. Those items for which no agreement could be reached were considered indeterminant data and were excluded for data analysis. This explains the different n values that may occur among the six subjects. The scoring took place by deriving the percentage of the incorrect responses. The percentage was calculated by dividing the number of errors by the total umber of responses. Maturation The subjects were required to identify on a sheet stimulus items that were presented aurally. Beforehand, a stilmlus tape was prepared. The 108 items of the identification task were taped on a Scotch Magnetic Tape 212, with the help of a reel-to-reel tape recorder (A-2340, 4 channel SIMJL-SYNC 33 Stereo) at a speed of 7 1/2" per second. A microphone of the brand MEX 2001 was used. The speaker of the items was a native American speech-language pathologist who practiced the task individually and under guidance of the investigator before the actual recording took place. The three-second-interval, measured on site by the investigator, was maintained in between the items with the chronograph function of a digital quartz watch (Meister-Anker). The items were randomly ordered. The 108 stinnslus items were presented in the carrier phrase by audiotape. To allow for identification of the presented item, the subjects received a list, on which the target item was written plus two other possible syllables. Those other syllables consisted of the same vowel and two other consonants of the same consonant class. A copy of the identification form is available in Appendix D. The subjects listened to the items in the silent roan of the audiometric booth, in which their own production had been taped. In order to reduce the influence of environmental noises, a headphone of the AVID H brand was used while listening to the tape. The subjects were seated on a chair at a distance of about one meter from the tape recorder. They wrote on the small desk of their chair and were not in touch with the table on which the recorder was placed. The loudness level of the output was at a comfortable loudness level as judged by the listeners. The subjects listened to the tape one at a time. The subjects were instructed to listen to the tape and to circle the item in the carrier phrase that they perceived. Further, the 34 subjects were asked to indicate the confidence in their choice by checking the appropriate number on the rating-scale on the same form. To avoid invalid responses due to misunderstanding or inconplete understanding of the instructions, a trial test set preceded the actual task. The items of this set were the same in syllable structure as the study items but contained different consonants. The investigator read the trial items. The real task was begun only after appropriate responses to the test set. 11' III'E"E The judging determined the correct and incorrect identified itelm by comparing the items circled by the subjects with the list of items that were actually presented. Items that were not identified and those representing two choices were considered indeterminant. The scoring was based on the percentage of incorrect items. This percentage was calculated by dividing the nunher of incorrect responses by the total mater of valid responses. Data Analysis WW3. In both tasks the data that did not meet the judging criteria were considered indeterminant and excluded from the data-analysis. The data were analyzed for the production and identification tasks separately and after that conmared across the tasks in order to find a possible relationship between them. 35 In all cases the data were presented in percentages of errors. The percentage of error was compared for initial versus final position and for each syllable position. The comparisons were extended to nasal and plosive segments considered individually and as a class. The parametric Pearson correlation coefficient statistic was applied to the data where appropriate. Walt The valid responses were analyzed in frequency rates of the errors from two different angles of view. The error rate of consonants in initial versus final position as well as the error rate of consonants per class (oral versus nasal) and segment were regarded. Finally, the analyzed data of the production task were compared with the data of the identification task in order to search for any relationship between the two. '11 II .E. . I I The valid data were analyzed according to two different points of view. At one hand the errors were compared to their occurrence in initial versus final position. At the other hand the errors were carpared as consonant classes and segments oral versus nasal. CHAPTER I I I RESULTS The purpose of the study was to explore the mode of performance and syllable position as factors of phonological interference. In particular, the initial versus final position of single consonants (oral and nasal stops) in the production and perception of CV and W2 nonsense words were considered. Following the order of the three questions posed (see page 22), the results are presented in the order of the production and identification tasks respectively, followed by a section on the relationship between them. Performance on the production and identification tasks is examined in terms of error frequency and error pattern. The Production Task W In analyzing the changes on the target syllable, three major categories of errors were observed. Ole category included errors on just the target consonant in the syllable (on nasal or oral stops). A second category included errors on the target segment plus the surrounding vowel segment. A third category consisted of errors on just the vowel segment while the target consonant was articulated correctly. Since only the first and second error categories were of interest, the third category was not included in the analysis of the results. 36 37 Chuuud1_Error_Ersausaaz Table 2 displays the percentages of error for both categories together and for categories I and II separately. The table reveals three main findings. First, more correct than incorrect responses were yielded. The percentages of correct responses varied from 76.64% for 31 to 87.76% for 56' Second, all subjects showed errors in both categories. Third, the error percentages varied among the subjects, ranging fran 12.26% (56) to 23.36% (SI). The actual number of errors can be derived from the error percentage. For instance, 51 showed 23.36% errors on a total of 107 items. The number of errors would be 25 in this case. Across the subjects there was no systematic tendency for Category I errors to be larger than Category II errors. For example, 81 shows 4.67% errors for Category I and 18.69% for Category 11, whereas 82 shows 10.48% errors for Category I and 3.81% for Category 11. Therefore, the error categories were pooled in subsequent analyses. Table 2 Casra11_2ercaaraaa_of4Err2L_2n_1hegflraducrioa;Iaak Subjects S3 86 83 S4 83 88 Mean Std. .n=1 7 n=1 5 n=103 n§102 n=1 5 n=1 6 CNerall 23.36 14.29 22.33 16.67 20.37 12.26 18.21 4.5 Cat. 1 4.67 10.48 14.56 14.71 6.48 9.43 10.05 5.92 Cat. 11 18.69 3.81 7.77 1.96 13.89 2.83 8.16 6.68 38 The original intention to inmose statistical treatment on the data was not carried out. Statistical treatment typically requires pooling of data, which would obscure the individual variability among the subjects as observed in this study. E E I 5]] 1° E . . The question regarding the frequency and pattern of errors made in the production of syllable initial versus final oral and nasal stops is regarded in Table 3. Table 3 displays the error rate by syllabic position for each of the two segment classes. Inspecting the overall rate, one finds that errors were made in both positions. Half of the subjects (51' S3, 35) show more errors in final than in initial position. For exanple, S1 shows 14.81% errors in initial position and 33.96% errors in final position. The remaining subjects show the reversed picture. The distribution of errors in initial versus final position varied per class. For the plosives all errors occurring for the two subjects (51’ 8,.) that did make errors were in final position. 81, for example, showed no errors in initial position but demonstrated 3.7% errors in final position. Similar observations can be made for S‘. For the nasals a variety in distribution was found. Three subjects (81, 53, SS) show more errors on the final consonants than on the initials. 51' for example, made 33.34% errors on initials and 53.85% of errors on the finals. Conversely, three subjects (82, S4, 85) made more errors on nasals in the initial than final position. 39 :omumoom o~ha->a _a:mu u emu so :omumnoa o—hn_—>m _am.m:m u cm. 1.: hN.m 605+ act—n oo.o~ noenN an.wv mm.mm nv.cm Ne.em numb mo.mN mo.n~ no.oN nm.mn vm.mm MN": RNHC RN": RN“: MNHC ON": NNHC RN": 0NHC RN": ONHC RN": u—aaflz I.” l- 3; i- .3- i- .i- 1. en; --- .1 I- .1 l- A; .{ ANnn AN": n~u= AN": oNNn AN": AN": RNun nNNG AN": AN": AN”: no>mno_m 1N4“ no.“ cu.- atnu anu ~n.v.— hcevN neg: Nnom mmom dog; 3;; NH: mmJL oa.mm ~w.vm :auua NH": fhfl-h fhflfi Th“: 9?": MhUC GVHG *WH: Hm": VhH—u mnH—u an—u uuwumfiu god Gmu cm nmu am emu am nmu am emu cm :mu am am“ am nmu um .15 53: on an em mm ~m .. flm. nuuumnnm 345% n o—nuh 4O Nbre subjects made errors in the nasal class than in the plosive class. In both positions more errors occurred on nasals than on plosives. Tables 4 and 5 show that individual segments within each class did not contribute equally to the error distributions. E E I l I. 'I I S Thble 4 shown the error rate overall and for the segments wdthin the plosives class. It can be seen that /t/ is the only segrnent contributing to the error frequency and that the errors made all occur in final position. Thble 5 shown the error rate by syllabic position for the nasal class and.segments. 'Tbe distribution of errors in initial versus final position differs for fim/ and /n/ in contrast to /q/. For Am/ and [n/ more errors occurred in final than in initial position. For example, the only subject who did make errors on /m/ did so in the final position. SI made no error on /n/ in initial position and 19.23%rin final position. S; made 3.7% errors on /n/ in initial position and 27.27% in final position. 41 momamnom o_nu__mu nomamaom o—An_~>n ~umumnm an: n :— .ueo III III III III III III III III III III III III III III III III J en.“ --- no.” --- --- --- --- --- en.“ --- --- --- --- --- ck.m --- e III III III III III III III III III III III III III III III III E en." --- n~._ --- --- --- --- --- en.“ --- --- --- --- --- an.“ --- _~.nn2u NNWG Nana nNma‘ “MN: emu a. emu am emu ma dNWfi :mu “Nun nuwn “Nu: em nee no mm nuuu_nnw em “N”: CA .wN ANNE e_ m “Nun kumn. .nee .nm em aummuaxq HdflfluuwlddflvdududHlNmluduuflludluuddfluuuqu v 2.3 momumuom 0—ha~_>a _a:_u n :mu .- momumuoa o—Aa-mu ~qmumcm H mm. ac.m on.m ou.NN ov.om oc.o~ mo.n~ mo.n~ mm.mn ec.m~ ~o.vm n~.n~ No.n~ om.n~ ne.o~ No.4n mn.nm r No.cn an.“ no.n— we. I»: ll- ~N.- lal am.n~ ll: n~.n~ n.m no.5 -ll n~.oH ll: : an.n Ill n~.~ 1:. an- ll- ll: as: all ll: vo.m~ la: ll: lll ll: on; E Nu.oa n~.n oo.c¢ no.nn cc.o~ mo.m~ nn.wv mm.nm m¢.¢n No.4m w~.wo mo.mN wo.m~ no.m~ nw.mn mm.mn -auozu wNma. AN": AN": AN": mem. 0N" NNMG “Nu: o~u= NNm: oNua “Nu: no t emu cm nmu mm emu mm emu cm emu cm emu cm :mu cm sommm sum .3"... :32 em an J an Nm Hm Afloumhnm daddudvdqulddddZIHqu gamma—“5% n 63.... 43 In contrast /']/ showed errors for every subject. At least 13% of the responses were in error for each subject in both tasks. The distribution of errors in initial versus final position showed that three subjects (51' $3, 55) made more errors in final position. The remaining subjects showed the reverse. Since /I]/ does not occur in syllabic initial position in Mandarin, the error rate of /r)/ being high in this position is in the expected direction. Since /I)/ is the only segment prone to not occurring in syllable initial position, the high rate in this position may obscure the overall bias toward syllable final errors. Thus, comparison of positional effect without the segment /r)/ may give a better view of bias toward syllable final as opposed to initial error. Table 6 displays the error rate by syllabic position for both classes in such a way that the influence of /I]/ on the overall bias towards more errors syllable final is revealed. It appears that five subjects (81 - 55) made more errors in final position than in initial when /r]/ was excluded. 44 comumuom o_nu__xo _a:mu u :mu comamnoa omna—_an ~amumcq H mm O 2.0 an; 4.: 2; -- -.~ 3.2 -- 26 SJ RAN -.~ I; -- 3.2 -- \S mcmvnuuno VTHG “Tull-G WT": “THC TV": MVHC NV": WVUG VT": MVHC *VH: WV": nun-”a0- :a who no.“ 2.: 3.: en.: 3.: 3.: $.3 2.2 2.3 3.2 3.: ~52 3.: 3.3 2.: \E mambo—ocm an": on": emu: on": ovum an": ovum on”: an": emu: «nu: vnn: nonsens- :- «dug, a: 5 n: E n: E n: 5 n: E a: 5 n: E In: no .3m :32 am nm J an Nm .. am. nauumhnm e o—Afih 45 Emu-11m The error pattern refers to the distribution of errors in termn of the type of error. Table 7 shows the error patterns by syllabic position for both classes. Eight patterns were required to account for the errors across the subjects. There were three simple patterns (substitution, rearrangement, and addition), four combined patterns (substitution -+ addition, rearrangement + addition, deletion + addition, and rearrangement + substitution) and one miscellaneous pattern consisting of other combinations not mentioned above. Table 7 E E l :1 E l E o! . I 1 MW 1 2 3 4 5 6 7 8 Position Subst Rearr Eel Rearr Sound and number and and and and True. mummmmmmm Plosives Initial n=l 100.00 Final n=2 50.00 50.00 Nasals Initial n=50 60.00 10.00 28.00 2.00 Final n=55 45.45 3.64 9.09 18.18 7.27 3.64 1.82 10.91 The variation of patterns by syllable position appeared to be a function of segment class. The segment classes showed shnilarities and 46 differences. Similarity was found in the observation that twice as many patterns were required to describe the errors in final position. The plosives needed two classes for the final position as opposed to one in the initial. The nasals required eight classes to cover final errors as opposed to four for initial errors. The difference was observed in the quantity of error patterns. For both positions, more patterns were required for the nasals than for the plosives. The most prevalent error pattern by position across the segment classes appeared to be substitution. Appendix E provides a table that refers to the master of subjects participating per error pattern. The Identification Task The analysis of the results in this section focuses on the second research question, which considered the identification by syllabic position of consonants in monosyllabic nonsense words. The results are presented by error frequency and error pattern. W251 The nature of this task yielded responses that were judged to be right or wrong. The subjects circled one item that either matched or did not match the spoken item. Criteria for error counting are described on page 34. MW Table 8 displays the overall frequency of errors for each subject. All subjects made errors, though the frequencies vary. The scores ranged from 7.41% (86) to 18.86% (Sl)' 47 Table 8 : 112 IE III 'E' . I] some n: 06 n: 04 n: 0 n=108 .n- 07 n—lQfi Error Leanna-tar Errors 18.86 16.35 14.44 13.89 10.28 7.41 13.53 4.13 E E I 5 II 1° 2 . . Amajor goal of this study was to consider errors as a function of initial versus final position. Thble 9 displays the error frequency by syllabic position overall and for each segment class. .A view on the overall frequency reveals that errors were made in both positions. Ole subject (56) did not show a preference for errors in a certain position. The remaining subjects made more errors in favor of the final position. The distribution of errors in initial versus final position varied per class. For the plosives all errors occurring for the three subjects that did make errors (51' S3, S6) were made in final position. 81, for example, demonstrates no errors in initial position but 4.0% in final position. Similar observations can be made for S3 and 56' For the nasals the distribution of errors by syllabic position varied. Five subjects (51"55) favored the final position. 51' for example, made 25.92% errors in initial position and 44.44 in final position. The rennining subject (56) did the reverse, as can be seen in the 14.81% errors for the initial position conpared to the 11.11% in .comumnom ~ammw on» momnmnoa o_na—~>u ~uamu u mam . as :omumaom o~n-_nn unmamcn n am. no.0u nn.+~ an.nn oo.+~ -.—n um.vu mo.n~ ~n.m— vv.vv Hu.- o~.am nw.—n we.m¢ Na.nN vv.vv Na.nN hNHC NNNC AN": AN": AN": RN”: NNHC NNHC MN”: NNHC RN”: RN": 'nflnflz ma.~ all on." :11 nu.n -- -- can l-l -- Hm.m -- oll all cc.v ll- uNn: NNHC AN": AN": h~uc AN": AN": AN": AN": AN": nNn: «Nu: no>mnoum an.o Nm.n no.5u eo.- "v.5 Hv.n Nn.- o~.o N~.~N on.n cc.c~ -.~H co.c~ oa.- cc.nN om.- _—auo>O on”: on": «an: on”: on": on”: men: we“: can: vnuc ~nu: emu: ddflflflflu dfiflmuww, new an new no n_a no can .ne new .ne nee em new no nan an .nem ease am nm em mm Nm e. um. auuo_n=w o e_nsp 49 in the final position. It seems, thus, that the trend in the distribution of errors by syllabic position is similar for plosives and nasals. More subjects made errors in the nasal class than in the plosive class. In both initial and final position, more errors were observed for the nasals than for the plosives. Tables 10 and 11 show the unequal distribution of the individual segments to the error distribution. 8 E I I I. .1 I 5 Table 10 displays the error frequency by syllabic position overall and for the segments within the plosives class. The three subjects (51' S3, 56) that did make errors did so only in the final position. The segments /t/ (for S1 and S6) and /P/ (for S3) were the contributing segments. Table 11 shows the error frequency by syllabic position for the nasal class and segments. The distribution of errors in initial versus final position showed a difference for /m/ and /n/ in contrast to /r]/. The differences were found in the higher error frequency for /q/ in initial position as compared to /m/ and /n/ and in the larger mate: of subjects who made errors in initial position. For /m/ more errors were observed in final position for the four subjects that did make errors (S1 'S-Q- For example, 51 made no errors on /m/ in initial position but identified 7.4% of the items in final position incorrectly. The same trend of favoring errors in final position is observed for five of the six subjects who made errors on /n/ ($1, 82, 50 comamuom 071:: 12.: n =2: comnmmom 03:1- uumamam N E. III III III III Oll III III III III III III III III All III III u— aa.~ --- o~.~ --- “a.“ --- --- --- --- --- --- --- --- --‘ oc.v --- . ~n.~ --- ~o.o 1-- --- --- --- --- --- -t- an.“ --- --- --- --- --- m ao.~ --- as.“ --- uh.m --- --- --- --- --- Ha.“ --- --- --- oo.+ --- -¢ao2u “Nu: “Nu: -u= NNu: -w= “Nu: \NNua n~ua “Nu: “Nm: \wuualuNuma1 can an am. am emu um emu an an“ an an“ an emu cm ..:mn .cm «auumamm an nun—ah. 51 nomummom u~n¢~umu ~acmu u emu so comumaom 03-3.: namin— u :«s Nm.o 2.: 3.3. 3.3 3.5 3.: -.- ~52 ~52 Sin 8.3 3.3 +93 .3- 3.: 2.: o .2.» on.» 3...: ~m.+ :J .1... an; I..- 3.: .1. ”2.: -.- *5: :4 KJ I... a mofi mm.” cK No. E... E... 2.: l... améa KR 2.: 11.. an.» a-.. 325 I-.. E 3.3 «rif— Zén 3.: 3.: 3.: 3.2 3.: It: 3.: 3.3 3.: 3.3. No.3 12: No.3 232.0 NNnd n "a find “Nu: NNma “Nu: «Nu: Sun: and tuna a; u lawn. a: am :3 am a: cm a: E a: 5 a: 5 a: E can: .5 .3m 5...: cm mm J mm Nm _m Sousa...“ 33% .3 93:. 52 S4, S5, 36)' For example, 54 nude no errors on /n/ syllable initial but misidentified 11.1% of the items with /n/ syllable final. Chly S3 showed 22.22% of errors for /n/ in initial position, and 18.52% in final position. In contrast, lq/ showed higher error frequencies for errors nude in initial position than in final position. Comparison of the error frequency for Aml, /n/, and /q/ reveals for S4 values of respectively 3.71%, 0%, and 7.4%, for exanple. Further, five subjects made errors on /q/'final ($1, 82, 34, SS) as opposed to two on /n/ initial (82, S3) and one on Am/ initial (S4). The distribution of errors in initial versus final position showed that four subjects (51’ 82, S4, SS) made errors that favored the final position, that 56 showed a reversed result, and that S4 showed no preference for either position. .Since /r)/ does not occur syllable initial in Mandarin, the high error rate in this position is according to the expectations. This high error frequency for /I)/ though may obscure the overall bias towards more errors in final position. Thus, comparison by syllabic position without the segment Iq/ may reveal a different view on the distribution of initial versus final. Thble 12 displays the error rate by syllabic position for the two classes ccmbined with and without /q/. All subjects favored errors in final position and that three subjects (51’ $5, $6) nude no errors in initial position when /r]/ was excluded. For the remaining subjects (52, S3, S4) who made errors in initial position, the frequency dropped. 53 nomumuom o~au——>. _unmu u nmw DO :omumnom o~Aa_~>- ~umumnm u am. fig 24 8.2 a; 1;... i- -.~ --- 3.2 -.~ 8.3 and 8.2 24 an.» i- \E names—on» he": mvua awn: nvu: mvuc awn: av": cwnc “cu: nvu: urn: mvn: consume- 2. 86 2.“ 2.2 3.2 3.5 ”fa 2.: 3; 2.2 on.“ 3.8 2.: 8.3 3.28.2 3.2 \S wnmva~unm van: vnna nan: +nu: vnn: emu: mvn: mvnn on": vnnc Nan: vnu: nucsuuon :« as...” a: E a: 5 an 5 a: 5 a: E a: E a: E :2 E .18 5»: on an 3a. mm Nm : Hm. «woowaaw N— odnuh 54 W The error pattern refers to the distribution of errors in ternu of the type of error. Because of the nature of the identification task, which required indication of the perceived itam on a list with choices, the possible error responses allowed for analysis of substitution patterns only. Table 13 shows the substitution patterns by class and segment. The variation of patterns by syllable position appeared to be a function of class. Similarities and differences were observed for both classes. Similarity was found for the distribution of substitution frequencies in initial versus final position. More errors were made in final position in both classes. For the plosives substitutions vmre found only in final position. For the nasals six substitutions occurred in final position, whereas four occurred in initial position. Efifferences were observed in the variability of substitutions and in their distribution by syllabic position. The plosives showed for each involved segment only one other segment substituted. For example, /p/ was only substituted by /k/ but not by /t/. The nasals in contrast showed for each involved segment two other segments as a substitute. For example, fm/ has been substituted by /n/ as well as by /r]/. The difference in distribution of the error patterns by syllabic position showed no substitutions in initial position for the plosives. The nasals showed substitutions in initial position for four of the six possibilities. 55 Table 13 WW 11:. l l. 2 Sound Sound Confusion Julian—— w__'mzr W Initial—Emil.fl Plosive p --- k.l --- 100.00 n=1 t --- k --- 100.00 n=3 Nasal m --- n 20.00 80.00 n=9 m--- 'J --- 100.00 n=5 n --- m 50.00 50.00 n=4 n --- q --- 100.00 n=11 r] --- m 55.56 44.44 n=9 r3 --- n 60.53 39.47 n=38 .p --- 1: should be read as p is substituted by 1:. Inspection of the error pattern frequencies by syllabic position reveals a higher rate on /m/ final and /r)/ initial as could be expected. For /n/ only the even rate on /n/ ---- /m/ would be according to expectatiomAppendix F provides a table displaying the nurrber of subjects participating per error pattern. 56 The Relationship Between Production and Identification The third question for this study considered the relationship between the performance in production and identification of initial versus final oral and nasal stops. Figure 1 displays production and identification errors graphed by syllabic position separately for each subject. Each figure shows the error rate in percentage on the vertical axis, and the phonemic segments on the horizontal axis. For each subject, the error frequencies for the initial position are displayed in the left figure, and those for the final position in the right figure. 3] . 1.] 5“].2.. An overall inspection of the figures reveals that the relationship between production and identification depends on syllabic position. For initial position high and uniform correlations bemeen the two rmdes were observed across all subjects. The correlation coefficients varied from a high of r=1.0 for three subjects (51’ 85' S6) to a low of r=.84 ($3). The uniformity of the curves was observed in terms of the rank order by segment class, and is noticeable in the nasal segments, which were most frequently in error. Note that the segment /9/ had the highest error frequencies on both tasks, followed respectively by /n/ and /m/. The nice uniformly high relationship between production and identification performance for initial segments contrasts with the variable relationship for final segments. Variability was observed for the correlation rates and for the uniformity of the curves. The $1 52 53 Initial position g 75_ r-.-l.0 t Ola-.001 LU ~_ 50 o .4 25 0‘ E E L. w '8 9 o t LU ”5 O! p t k m n Phonetic Segments (e) 0 Identification x Production ‘3 57 100 r Final position r=.82 75' SO 25 0' .l p t k m n g (b) 100- 75b l’:.38 n. S. 100 75 SO 25 Legend k m n g Phonetic Segments (f) p t 0! Significance Level r Correlation Coefficient Figure l. The Relationship By Syllabic Position Between The Error Percentage Of The Production And The Identification Task 58 Final position Initial position 100 x 100' :5“: 75.. l’=.88 75_ l':.53 n. S. .1] 0:.02 Sz. .._ 50 50- O o" 25 25~ 0‘ O 100 100* m g 75- 53 SS ._ 50 0 av“ 25 0‘ 100 100!- U1 § 75_ 75_ r=.80 n. s. m 0:.05 x 56 as— 50 50" O «4' 25 zsr OPTkmng 0apt mng Phonetic Segments Phonetic Segments (k) (l) Legend 0 Identification 0‘ Significance Level X Production r Correlation Coefficient Figure 1. Continued 59 correlation curves varied from a high of r=.97 for S3 to a low of r=.38 for 82. Four of the correlation coefficients did not reach significance at the .05 level, which would require a minimal coefficient of r=.81. Variability in the uniformity of the curves was observed in two ways. First, variability of the correlation curves across the subjects manifested itself in the different participation of the subjects in the both involved segment classes. For instance, 51 in Figure 1b made errors in both tasks on lt/n/ql, whereas 52 in Figure 1d made errors in both tasks on /n/q/. Second, the preference for errors in one task to another varied across the subjects. 51’ for example, showed in Figure 1b no preference for a certain mode in errors on /t/and /v]/, more errors in identification for /m/, and more errors for production on /n/. Orly $5 in Figure lj showed consistently more errors on production than on identification for all involved segments. Conzparison of the correlation curves for both tasks by syllabic position for each subject revealed that the height of correlation in one position does not necessarily imply a corresponding height in the other position. Subjects 2, 4, 5, and 6 showed high correlation rates in initial position but nonsignificant coefficients in final position. Contrastively, S3 who showed a coefficient of r=.84 at 0r=.05 in initial position, reached a r=.97 at Or=.001 in final position. Further, a tendency towards more errors in one mode for initial position is not necessarily followed by the same tendency in final 60 position. Only S, (Figure li-j) favored errors on production in both positions. The findings as described above suggest a dependency of the relationship between the two modes of performance on syllabic position. 8 1 . l. l 5 :1 II ”T . I Slll'E°' E 51° In the initial position, nnch uniformity was observed for the plosives in such a way that no errors were made in both tasks. The correlation for the nasals showed uniformity across the subjects for /m/ and /r)/. 01 /m/, five subjects (51' 52, S3, 85' 56) made no errors in both tasks. 0) /q/ all subjects made errors on both tasks. Five of the subjects ($1, 82, S3, S5, 56) showed a higher error percentage for the production task than for the identification. Less uniformity on error correlation for the modes of performance across the subjects was demonstrated for /n/. Four subjects (51' S4, 55’ 56) showed no errors on /n/ for both tasks, whereas S2 and 83 showed errors for at least one task. In the final position less uniformity was noticed for the relationship between the two modes of perfornnnce across the subjects. Errors were made in both segment classes by four subjects (81, 83, $4, S6). Three of these subjects (51' 53, S6) showed a preference for more errors in perception than in production, whereas S4, showed more errors in production. These results suggest a dependency of the correlation between the modes of performance on segment class. 61 ‘1': . l 5 l] I. E . . E? l. 5 I. In the initial position, unifornfity within the subjects could only be inspected for the subjects who had more than one segment involved (82, S3, S4). Chly S; (Fig. 1e) showed consistent preference for more errors in one mode as opposed to $2 and S4, who favored more errors in a certain mode depending on segment (Figures 1c and 1g). In the final position only one of all subjects (85) showed consistently more errors for production than for identification (Figure lj). The remaining five subjects showed inconsistent preference for one mode to another. These findings suggest a dependence of the relationship between the two modes of performance on individual segments and on individual subjects. This dependency tends in initial position towards more errors on the production task, whereas no consistent preference for any task is found in final position. CHAPTER IV DISCUSSICN AND mucanms Purpose of the Study The purpose of the study was to explore syllabic context of phonemes and mode of performance as factors in phonological interference. Interference was defined as a result of language contact, manifesting itself in the occurrence of elements, characteristics, and rules in one language that belong to another language. The questions posed in the study regarded the difference in error rate for phonemes in initial versus final syllabic position. The influence of syllabic position was tested in both the production and identification mode of performance in order to learn about their relationship. Further, a possible relationship between error rate at one side and segment class and type of the phonemes at the other side was regarded. The studied phonemes were oral and nasal stop consonants combined with the three cardinal vowels. Summary of the Results The results showed for both production and identification more correct than incorrect responses. More errors were observed for the nasals than for the plosives. For most of the subjects a positive effect on the initial versus final bias was observed, with slightly higher evidence for the identification task. 62 63 Correlation was observed between error rate at one side and segment class and type at the other hand. The class of the plosives appeared to be less in error rate involvement, participating subjects, and involved segments. The nasals showed a high error rate overall and per segment, more participating subjects, and involvement of all segments. Occurrence of errors appeared to be highly related in the production and identification tasks in such a way that errors occurring on items in one task were very likely to occur in the other task. For errors made in only one mode of performance, no bias towards a certain mode could be observed. High and uniform correlation between the two modes was shown in initial position, as opposed to the variable correlation for performance in final position. The relationship between the production and perception mode of performance seemed to depend on syllabic position of the phonemes, segment class and type, and individual subjects. The discussion examines those findings in terms of the inconsistency of the bias towards more errors in final position and the higher involvement of nasals than of plosives. The questions will be discussed in this order. Discussion of the Results II . {11"11! E'lEEE Three main factors may have played a role in the observed inconsistency of the initial versus final effect across and within the subjects. The factors relate to the linguistic theory about 64 phonological interference, to the task, and to the subjects. They will be regarded in this order. E B l l l I. . . I] The way contrastive analysis approaches phonological interference inplies a static nature of the phenomenon. This would, mean that a FL learner's pronunciation as measured at a certain point in time would not be subject to change. Much of the literature describes phonological interference in this way (Sumuknr, 1957; Saunders, 1962, 1969; Asia, 1973; Roy, 1975; Jackson, 1981; Nfirhassani, 1983). In the last two decades, though, transfer is more and more considered as a dynamic process. It is explained that mastering of new structures in the FL will start with performances close to structures in the NL. Gradually, the performances will get closer to the structure of the FL. Authors like Van Teslaar (1965), Dickerson (1976), and Flege (1980) apply this version of transfer to the phonological level. Dickerson (1976) studied sound changes within the same language group, concluding that the phonological acquisition of a FL learner resembles the sound acquisition process of a NL learner within his NL. Van Teslaar (1965) and Flege (1980) apply the theory directly to the acquisition of new phonological structures in FL learning. In this vision on transfer, interference reflects a temporary stage in the acquisition of a new phonological systan. Application of this view to the subjects and the results of this study would explain the differences wdthin and across the subjects as different stages within the process of transfer. 65 W The nature of the production task very likely had an influence on the responses. Wilsonand Msllergird (1981) and Chen (1976) mention the differences between single word reading, sentence reading, prose reading, and spontaneous speech for the production of speech. From single word reading to spontaneous speech there is a gradual shift from focusing on form to focusing on content. “hen more content comes in, less energy, time, and concentration can be given to the pronuncia- tion. Conversely, where less content exists, more attention can be given to the form. The latter condition,then,may explain the results of the study. Recall that the production task in this study was a highly structured sentence reading task with many norxneaningful stimulus items. Consequently, it is possible that because of the reduced meaning of task, the subject's awareness of the motor act was increased. The nature of the stinuli may also have influenced the performance in both tasks. The CV and VC syllable structures are very basic in all languages (Jakobson, 1968) and might have been too simple to evoke a more consistent initial versus final effect than they did. The presentation of the stinmli may have had impact on the results as well. The three-second-interval in the production task may have allowed for increased awareness of the motor act. The sinple and ever returning carrier phrase might have provided a context so ideal that it reduced the error rate. A last factor of influence related to the task could be orthography. Although many languages make use of the same Latin 66 alphabet, the pronunciation of the same letters may vary among the languages. An exanple would be the letter 2, which is used for the sound combination /ts/ in German but pronounced /z/ in English. The N1. of the participating subjects in this study does not make use of single letters to form written words but uses characters for writing. (he of the implications is that sounds occurring in the NL are not always recognized as such in a letter orthography system as the Latin alphabet. In this way the English spelling may have influenced the results on both tasks. Sunuktu (1957), Wilson and Mollerglrd (1981), and Mirhassani (1983) reported orthographic influence in their studies and descriptions on the production of ESL for students with several language backgrounds (resp.Sundanese and Javanese; Norwegian; Iranian). WW Although the main selection criterion considered uniformity of the subjects' backgrounds, there still remain factors that could not be controlled in the framework of a study like this. (he of the factors is the language background of the subjects in the country of origin. The teacher's pronunciation, for exanple, may have influenced the subjects' present performances in the FL. The emphasis on ESL for reading or speaking purposes may have a strong inpact on fluency and pronunciation in the FL. If emphasis is placed on speaking skills, the amount of practice in speaking ESL could have been a factor. Although all subjects reported to speak only Mandarin, it should be taken into consideration that in Taiwan several dialects are frequently used for connxtnication. These dialects have their own phonological structures, 67 that differ in many cases from Mandarin. In this way they may have influenced the production of Mandarin from other speakers and so the perception of the subjects. Exanples would be the presence of the final consonants Ip/t/k/ in inplosive form in Hokkien and the /m/ (Tay, 1969), or the interchangeability of final /n/ and /r]/ in the Shanghai dialect (Shen, 1953). Another possibility is that Nhndarin was used as a second language by both parents, since their own dialects differed too mach to allow for conmunication. Traits of their dialects might have occurred in their Mandarin and have passed to their offspring, thus influencing perception and/or production of their offspring's Mandarin. Another factor might be the different language experience in English related to the field of study in the LEA. The requirement for a certain level of English varies across professional fields and consequently during training. For example, nujors such as conmunication, teleconrmnication, advertisement and marketing might emphasize speaking skills more than the majors of computer science and unchanical engineering do. Related to the major areas of study nfight be the personal tendency towards speaking. Nhyhe students in the first four mentioned fields of study tend to speak more than students in the remaining areas. For this study this explanation does not seem to hold though. Subjects from the more mechanical areas of study did not consistently show'higher error percentages than the subjects fronlother fields of study. 68 A third factor, related to the subjects, concerns the individual differences in a broad way. Strevens (1973) and Flege (1980) in their studies on pronunciation in SLA mention the physical aspect of pronunciation. The individual difference in physical-motor skills nfight have influenced the subjects to respond to the task differently. For some of them, physical abilities might have been too sophisticated for the task or not sophisticated enough. The same would hold for the different perceptual skills of the subjects. Further the motor-cognitive make-up is very likely to differ per subject and thus to influence difference in results. Suter (1976) and Thhta, Loewenthal and “bod (1981) found that the language spoken at home influences the pronunciation in the FI... This would inply that students who spoke Nhndarin nnst of the tint would be likely to score higher error rates than those who spoke nore English. A.factor like this went beyond the scope of this study. The same holds for the mental and social influences on the pronunciation as nentioned by Sapir (1927) and Strevens (1973). Also hard to control within the scope of this study was the awareness-level of the phonemic inportance of final consonants which is indicated as a factor by Shen (1959). According to Suter (1976) in his study on pronunciation accuracy, two more factors are of importance: the speaker's concern about his pronunciation as having a great impact and the ability for oral nfinficry. Those factors were not regarded in this study. 69 If I H E E II I I] E El . According to the contrastive analysis approach, more errors would be expected to occur in the FL on phonemes in a certain position that do not occur in that position in the NL. For the phonemes and positions as used in this study it would mean that more errors would be expected on /p/t/k/m/ in final position, since they do not occur finally in Mandarin. Further, it would be expected that /n/ would show no difference in error rate since it occurs in both positions in Nhndarin. Finally, lq/ would be expected to show'a higher error rate in initial position since it does not exist initially in.Nhndarin. The results, though, showed fewer errors and participating segnents for the plosives in final position than for the nasals. The /k/ showed no errors at all for both tasks, the segments /t/ and /p/ showed a low rate of errors in the perception task, and on /t/ only errors in the production task occurred. Contrastively, all nasal segments were affected on both tasks in one or both positions. Although errors were biased toward final position as expected, nore errors were found on nasals than on plosives. This differential error rate by segment class contrasts with the prediction of contrastive analysis (See pages 19-21). Reasons for the higher error rate on nasals mdght be found in the different acoustic properties of the two phoneme classes and in the multilingual columnication situation on Taiwan. The different acoustic properties may only account for the errors on perception. 70 Liberman et al. (1963) studied the role of consonant-vowel transitions in the perception of plosive and nasal stop consonants. They described the importance of the transition between consonant and vowel, especially in the second formant, as a cue to the perception of place within the classes of oral and nasal stops. The second formant cue appeared to be less effective for the nasals though. This might explain the difference in error rate between the nasal and oral stops as observed in this study. Malécot (1963) studied acoustic cues for nasal consonants. A nasal resonance functioned prinarily as a class marker, and only secondarily as placemarker, mainly in final position. He noted that the /m/ resonance differed from /n/ and /q/ for which the resonances appeared similar. This might explain the unexpected high error rate on final /n/ and /'3/ in the perception task. The high error rate on /n/ final is also observed by Riekert and Swennen (1984) in their study on speech problems in Vietnamese and Chinese speaking Dutch. The nultilingual situation in Taiwan, might also help to explain the results. A possible explanation for the lower than expected error rate for the plosives and /m/ in final position might be found in the syllable structure of the widely spoken Taiwanese dialect (also referred to as Hokkien, Fukienese, Annoy or Southern Min-dialect) which exhibits those phonemes in word final position, /p/t/k/ in unreleased form though (Tiee, 1969; Tay, 1970). As discussed before the subjects might have developed an unconscious awareness of phonemic values of final consonants in hearing Taiwanese although not speaking 71 it themselves. Further, the Mandarin spoken in Taiwan might hear influences from Taiwanese also in regard of phonological structure. The error rates on /n/ and /q/ in final position were higher than expected on both tasks. It was observed that those phonemes were confused with each other in this position. According to Shen (1976), /n/ and /I)/ are interchangeable in the Shanghai dialect. This study did not include data about the subjects' parents' language background and therefore can relate this information only as a possibility to the obtained results. Another possibility might be the existence of some assimilation rules in spoken Mandarin which reduce the phonemic distinction between final /n/ and /q/ in certain contexts. TlEl'l'B 21. 3111.. Whereas the constrastive analysis hypothesis regards the production mode of performance in phonological interference extensively, the perception mode of performance is not mentioned at all. One is led to assume a synmetrical relationship between the two modes. The results of this study show indeed a strong relationship between production and identification errors though the relationship is stronger for initial than final position. Especially for the initial position, a strong uniformity is observed in class and segments that are involved in the error rate, thus favoring the assmzption of synmetry as predicted by contrastive analysis. However, the picture seems more conplex for phonemes in syllable final position. 72 More nonsignificant correlation coefficients occur, and those that are significant look less uniform than their counterparts in initial position. No consistent preference for more errors in production or perception can be observed within and across the subjects. Further, there is not nuch uniformity in class and segments that participate in the error rate. Some subjects show involvement of all segments, some of only a few segments. It would follow, then, that the relationship between production and perception seems to depend on syllabic position, segment class and type, and individual subjects. The discrepancy between the modes provides clues for the learning process. For the initial position, the error bias towards production would suggest that correct perception does not necessarily imply correct production. This supports the generally accepted vision of perception preceding production. This vision is used in SIA for perception defined as language comprehension or language input as opposed to language production. Perception at the phonological level has been less considered in SLA. Those studies that have been done on phonological acquisition are not always in agreement with the vision on the relationship between production and perception as used in SLA. Goto (1971) and Sheldon and Strange (1982) studied identification and production of /l/ and /r/ in English words for Japanese adult students. Their findings showed an overall better production than identification performance of the target phonemes. Catford and Pisoni (1970) conducted an experiment to describe the efficacy of articulatory 73 versus auditory training of foreign sounds. Their findings showed a better performance of the students in the articulatory group for production as well as for identification of the target sounds than for the auditorily trained subjects. In this study, the observation of the relationship between perception and production of phonemes in final position seem to be consistent with the findings of the above cited studies. The results would advocate that besides unidirectional support from perception to production, nutual support between the two modes of performance should be regarded as a clue on FL phonological acquisition. The results further suggest that intensity and form of the relationship between production and perception are related to segment class and type. Limitations of the Study The main limitations of this study that reduce the generalization possibility of the results are related to the selection of the subjecrs and to the nature of the tasks. Since the subjects were not randomly chosen but rather chosen according to certain selection criteria (language background and experience, age, etc.), they do not represent the overall population of English-speaking Chinese with Mandarin as NL. It follows that the results in this study can only be generalized to subjects with similar characteristics. Further, the number of subjects is very small, which means that even to a population with similar characteristics the results should be generalized with caution. 74 The nature of the task also can be considered as a limitation. Both tasks were highly structured and did not have a direct relationship with daily-live performance. Examples are the use of pure monosyllabic stimulus items, a three-second-interval between the readings, the very controlled phonological context in the fonm of only one carrier phrase, and the orthographic presentation of both tasks. For the identification task , it can be added that only one native American-English speaker had to be identified, whereas identification of more speakers and of the subjects' own productions had not been taken care of. Further, no conparison is made with the performance of native speakers. The literature showa that even for NL speakers errors tend to occur nore often in initial than final position. Therefore, in this study it cannot be separated out how'much the error bias towards final position is due to phonological interference and how nuch may be due to the inherent perceptual-motor differences between syllable initial and final position. Finally, analysis of the adjacent vowels could supply further information on the segmental differences in error rate. Implications for Practical Purposes Information about syllabic position as a factor in phonological interference and about the relationship between production and identification of phonemes in FL acquisition has direct practical inplications for both teachers of ESL and speech-language pathologists. 75 Knowing that not only unfamiliarity with a phoneme but also the unfamiliar position of a known phoneme may cause differences in perception and production shifts the emphasis from training of only new phonemes to training of phonemes in new contexts. Knowledge about phonological acquisition that considers mastering of phonemes in context would lead to stressing the teaching of sounds in context instead of as single phonemes. Being informed about the asynxnetrical relationship between production and identification would lead one to consider both perception and production aspects in teaching pronunciation. It would be realized that development of one of the modes could stinulate development of the other mode of performance. For speech-language pathology, knowledge as described before is of great inportance in order to distinguish correctly between errors as a result of phonological interference and those resulting from speech disorders. Imlications for Future Research Closely related to the learning of a FL is the phenomenon of interference. Phonological interference is just one part of the phenomenon, but it has direct impact on the commnication skills. Since societies nowadays require more and more mltilinguality, it will be worthwhile to continue research on phonological interference. The research should also include aspects such as the role of syllabic distribution of sounds in perception and production on segment class 76 and type. Inplications of this study for future research would be related to the selection of subjects and the nature of the task. In regard to the selection of subjects, subsequent studies on phonological interference could consider to select a larger population and/or a population exhibiting more variety in background (lingual, educational, social, vocal) and age in order to collect data that could be applied to the whole population of native Mandarin speakers who speak.ESL. The nature of the task could be adapted in several ways, depending on the focus of the further research projects. More natural envirorsnents, such as reading of prose or spontaneous speech, would provide information about the manifestation of phonological interference in daily life situations. The use of multisyllabic words and/or consonant-clusters would make the task more natural and thus increase the possibility of generalizing the results to the actual speech in daily life. Further, the study of different segments could be considered. Adding to the perception task the identification of not only more native speakers' speech but also of the subjects' own speech would provide anorher dinension of the perception mode. Participation of native subjects would allow for conparison between perfornance of native and foreign speakers on the tasks. This would allow to investigate how much of the error bias towards final position is due to phonological interference and how'much is due to the position. Finally, the adjacent vowel as a variable in error frequency 77 on the segments would be worth an investigation. It would provide better insight in the process of perception. Conclusions It is concluded from the findings of this study that syllabic position and mode of performance have an effect on the error frequency and error pattern of phonemes in a FL. This effect interacts with sement class and type of the phonemes and with the subjects. The positional effect of phonemes manifests itself in an error bias towards final position. The interaction manifests itself in the higher error frequency for nasals as opposed to plosives for both positions, the higher frequency of participating nasal segments as opposed to plosives for both positions, and the variability in participation of the subjects particularly in final position Further, the existence of a relationship between the two nodes of performance is concluded. The relationship interacts with the syllable position of the phonemes, with segment class and type, and with the subjects. The interaction with the syllabic position is demonstrated by the high correlation between production and perception on initial segments as opposed to the variability on final segments. The interaction with segment type and class is noticed in the low error frequencies for the plosives as opposed to the nasals in both positions and the different error frequency for the segments within each class. Variability in participation of the subjects, particularly in final position, shows interaction with the subjects. 78 It is concluded from the findings about syllabic position and mode of performance of phonemes as influencing factors in error frequency and pattern of segment class and type that phonological interference is a complex phenomenon, that cannot be predicted or explained by merely conparing the sound inventories of two languages. APPENDIX A SURVEY FORM FOR RECRUITMENT AND SELECTION OF SUBJECTS 79 This week I was approached by Erna Pluut who is a graduate student at the Department of Audiology and Speech Sciences. She has a great interest in Chinese languages as her husband is a Chinese himself. For her thesis she wants to do research on the linguistic features of Gainese speakers for the English language. In order to do so she needs some basic informtion and later on she will need volunteers. They will read an English text for her, which will not require nuch time. Filling out the form and returning it soon (by October 9) to the address that will be most convenient to you, will be of great help. I hope you will support her and encourage you to do so. Chang, Ta - (hing Before starting the research, I would like to conduct an anonymous survey which will not bind you in any way to continued participation in the program. If you would please fill out and return form PART I, it would be greatly appreciated. I will also need some further volunteers for continued participation in my research program. For those people, if you would fill out and return (either together or separately) both form PART I and PART II, I would appreciate it very nuch. You will be contacted later. Thank you very nuch for your help. Erna Pluut PARTI PIEASE,T'OBEFILI.EDCIJI'BYEVER‘ICNE a) Date of birth: month , year / Major _/ Male Female b) “hen did you cane to the (BA? month year c) mat dialect did you speak at home? Mandarin . Taiwanese , Other d) Vihat Chinese dialects do you speak? e) “hat Chinese dialect do you use mainly? f) At what age did you start to learn English? 3) How many years have you studied English? I-bw many hours per week? h) How many years have you been speaking English? i) What education did you receive in your native country? high school , vocational school , college university 80 j) mat statement would be applicable to you? Please, mark with an X My English is such that a) I could lecture a group of students b) I can comrnicate well c) I can make myself understood d) I prefer not to speak English unless I have to. ----------------------- separate here if you wish---------------------- PART II FILL CIJI' IF “11 ARE WILLING TO PARTICIPATE Name Telephone number Age a) first Chinese dialects do you speak? b) How long have you been in the LEA? c) Are you presently a student? Major A P P E N'D I X B SUBJECTS’ READING LIST FOR THE PRODUCTION TASK ook poo 0°C paa aang maa oom koo poo ngaa een 000 SAS' 81 3811 poo pee taa eek oop kaa een kee IKiAIhl paa aang aak oong eep aang eep oong aat kaa taa 82 SAY . . . . AGAIN mee too oot aam koo kee eeng ngaa eeng maa ngoo aam eem eet oop aak pee aat tee ngaa ngoo eeng 00? pee mee kee too aak aap ngee naa kaa nee aap naa eem noo 000 koo oop eep paa oong am 000 ngoo een 000 SAH' 83 ngee tee 00!) 33.0 t8! tee oom aan 0GB eem too 83.! Afihhlhl “CC ook naa eet ngee eet eek 00¢ ook nee eek A.P P E NiD I XI C JUDGES' LIST FOR THE PRODUCTION TASK 84 l - 2 - 3 moo ook nno poo oop 00! kaa paa een kee paa aang koo aak poo oong moo een eep BOO aan eep oong poo aat pee kaa taa taa eek Z££1_£21£ nee _____ 1 - Z - 3 sun ._____ 1 - 2 - 3 eeng ______ 1 - 2 - 3 xnaa ______ 1 - 2 - 3 ean ._____, 1 - 2 - 3 ask ______ l - 2 - 3 tee _____, 1 - 2 - 3 eeng ______ 1 - 2 - 3 nee ______ 1 - 2 - 3 aak _____, 1 - 2 - 3 naa _____ 1 - 2 - 3 aap _____, 1 - Z - 3 too ______ 1 - 2 - 3 koo _____ 1 — 2 - 3 ngaa _____ 1 - 2 - 3 ngoo _____, 1 - 2 - 3 eet ______ 1 - 2 - 3 pee _____, 1 - 2 - 3 85 l - 2 - 3 Zill.£fli£££ ngaa oop kee aap kaa naa oot kee eeng amn oop aat ngoo pee too ngee nee eem Remarks: 86 1-2-3 0CD noo I“! I100 0CD koo eem oop t00 eep Ctt paa nee oong ook naa 0011 eet ngoo ngee een eet 00D eek ngee 00! tee ook 0011 nee aan eek taa 11106 tee APPENDIX D SUBJECTS' WRITTEN FORM FOR THE IDENTIFICATION TASK 17. 18. 19. 20. 21. 22. oop oong ngoo paa koo emu aan ngaa ngee too ook oop paa pee nee aan ook poo maa tee nee ngaa 00t 0GB moo €88 too eeng aam naa nee poo oop 00t taa tee ngee aan oop too ngaa 87 kee 1 ..... z ..... 3 Dace 1 ----- z ----- 3 on 1 ----- z ----- 3 ook 1 ----- 2 ----- 3 oon 1 ----- 2 ----- 3 noo 1 ..... z ..... 3 kn 1 ----- z ----- 3 P00 1 ----- 2 ----- 3 een 1 ----- 2 ----- 3 sang 1 ----- 2 ----- 3 man 1 ----- 2 ----- 3 mee 1 ----- 2 ----- 3 koo 1 ----- z ----- 3 oot 1 ----- 2 ----- 3 ook 1 ----- 2 ----- 3 kaa 1 ----- 2 ----- 3 kee 1 ----- 2 ----- 3 mee 1 ----- 2 ----- 3 ans 1 ----- 2 ----- 3 oot 1 ----- 2 ----- 3 koo 1 ----- 2 ----- 3 ma 1 ----- 2 ----- 3 kee eeng eeng nee poo taa aat aan noo mas eek een kaa kee tee ook eep ngee ngoo ngoo 001‘! tee een een ngee too paa aak aan ngoo naa eep can paa tee pee oop eet nee nno nno oong pee can emu nee koo kaa aap aang ngoo noo ngaa eet eeng taa pee kee oot eek nee noo noo oom 88 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- z 1 ----- z 1 ----- 2 1 ----- z 1 ----- 2 1 ----- 2 1 ----- 2 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. aan pee taa nno eep eek nna koo aak oot aap naa eet can oong koo ngaa aan aap oon mee aan tee paa ngoo eet eep naa too aap ook aat ngoo naa eep eeng aan too naa oong aat oong [10¢ 89 tans 1 ----- 2 ----- 3 kee 1 ----- 2 ----- 3 kaa 1 ----- 2 ----- 3 noo 1 ----- 2 ----- 3 eek 1 ----- 2 ----- 3 eet 1 ..... z ..... 3 nets 1 ----- z ----- 3 poo 1 ----- 2 ----- 3 aat 1 ----- 2 ----- 3 009 1 ----- 2 ----- 3 ask 1 ----- 2 ----- 3 noo 1 ..... z ..... 3 ngaa 1 ----- z ----- 3 eek 1 ..... z ..... 3 een 1 ----- 2 ----- 3 con 1 ----- Z ----- 3 poo 1 ----- z ----- 3 naa. 1 ----- 2 ----- 3 oon 1 ..... 2 ..... 3 aak 1 ----- 2 ----- 3 can: 1 ----- 2 ----- 3 ngee 1 ----- 2 ----- 3 tee aap mun aak tee paa oon aak oot too ngee eek aat sang taa pee oong eet eet naa kaa eep pee aat oong aap pee taa oong aap ook poo nee eep aak aan paa tee omn eep eep ngaa paa eet kee aak oon aat kee kaa aan aat oop koo eet aap aan kaa kee oon eek eek naa taa eek 90 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- z 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- 2 1 ----- z 1 ----- 2 1 ----- 2 1 ----- 2 l6. 17. 18. 19. 20. aang aat nee poo een eeng emu kaa noo aan oop oom too ngaa oot een noo nee aang aan aak ngee too ean een eeng paa fine aan 00! oong poo naa ook eem nee aam aan 91 aan aap nee koo eeng ean een taa ngoo aang ook oon koo naa oop eeng ngoo ngee aan aang APPENDIX E NLMBER OF SUBJECTS PARTICIPATING PER ERROR PATTERN IN THE PRODUCTION TASK 92 Table 14 We . l E l . I l 2 3 4 5 6 7 8 Subst Rearr Del Rearr Sound _ and and and and PIOSIVCS Initi‘l C-“ --- CD-- --- 0-- --- u-- Final -"- --9 1 CO- --- --- O-- Nasals Initial --- 2 3 4 --- --- 1 Final 1 l 4 2 1 1 2 APPENDIX E NLMBER OF SUBJECTS PARTICIPATING PER ERROR PATTERN IN THE IDENTIFICAle TASK 93 Table 15 Sound Sound __£9.ai.1.ian__.. Tyne Centuries Initial.__£inal__ Plosive --- k‘I "" 1 n=1 ..-- k --- 2 n=3 Nasal --- n 1 3 n=5 --- r) --- 3 n=5 --- m 2 1 n=4 --.. I) --- 5 n=11 --.. m 4 1 n=9 --.. n s 5 n=38 .p --— k should be read as p is substituted by k. BIBLICXIRAPHY Akmajian, A.; Demers, R.; and Harnish, R. (1979) W W Wseschueetts: MIT Press. Albert. M and Cbler. L. (1978) W W Academic Press- American Speech-Language-Hearing Association. (lune 1985) Clinical management of conmunicatively handicapped minority language populations . 63:15. Aziz, Y. (1973) Some problems of English consonant sounds for the Iraqi le‘tne‘s may a. 166-1680 Baetens-Beardmore, H. (1982) WWW. Tieto Ltd. Clevedon Avon. Bower. T. (1977) WWW. Cmbzidse: Harvard University Press. Blumstein, S. (1978) Segment structure and the syllable structure in ephuie. In: W1. A. Bell and 1.3. Hooper. Eds., North-Pblland Publishing Company. Broselow, E. (1983) Nonobvious transfer: on predicting epenthesis errors. In: W. Selinker. M. Ed. Massachusetts: Newbury I-buse Publishers. Bunte, P., and Kendall, M. (1981) Vlaen is an error not an error? Notes on language contact and the question of interference. W laxicaLLinuiatin. 2.3.. 1. 1- 7 Clyne, M. (1975) Forschungsbericht Sprachkontakt, Kronberg. Catford, J., and Pisoni D. (1970) Auditory versus articulatory training in exotic sounds. W. 11. 477'431- Chen, CeY. (1976) Pronunciation of English by students from the Chinese stream in Singapore: some salient features. REE: 191111111. 2.. 54-60. Dickerson, W. (1976) The psycholinguistic unity of language learning and language chance. Wain. 2.9. 215- 231 Edwards, M. (1971) Perception and production in child phonology: the teetins of four hypotheses. W. 1.. 205- 219. 94 95 Flege, J. (1980) Phonetic approximation in second language acquisition. War. 2.0. 117-133. Gage, W. (1981) A list of Non-Indo-European languages grouped by families. In: V. Clark; P. Eschholz; andA. Rosa (Eds. ) Language, Wings. New York: St. Martin' s Press, Inc., 597-6020 Garcia. 5. (1983) W. University of New Mexico Press, Albuquerque. Gen. 8.. Selinker. L. (Eds.) (1983) W W. Newhury I-buse Publishers, Inc., Massachusetts. Goto, H. (1971) Auditory perception by normal Japanese adults of the sounds "1"and " ". 82912211111219.1151. 9.. 317-322 Hirschfeld, U. (1983) Zur Interferenz im Bereich der Phonologie und Phonetik. W. 2.0.. 51-55. Hodson, B.W., and Paden, E.P. (1981) Phonological processes which characterize unintelligible and intelligible speech in early childhood. 1m, 1s, 359-73. Ihenacho, A. (1980) Foreign language teaching and pronunciation - aspects of the Nigerian situation. Sum, 8.. 245-255. Ingram, D. (1976) Phonological disability in children. Elsevier New York. Ingram, D. (1974) Phonological rules in young children. mil—91 Mama. 1. 49-64 Jackson, H. (1981) Pronunciation of English consonants by Indian learners. E111, 31, 418-420. Jakobson. R. (1968) W. Laroche, I. (1981) A view of transfer and interference in contrastive analysis. 81m, 9__, 29-34. Liberman, A.; Delattre, P.; Cooper, P.; and Gerstman, .(1963) The role of consonant- -vowel transitions in the perception of stop and nasal consonants. WW.V01WI. 67-79 Mackay, D. (1978) Speech errors inside the syllable. In: A. Bell and J. Hooper (Eds.) Wm, North-Pblland Publishing Company, 201-211. 96 Malécot, A. (1963) Acoustic cues for nasal consonants. -W SW. 1. 93-103. Mirhassani, A. (1983) Pronunciation problems of Iranian students learning English. 1381... 21, 320-330. Olmsted. D. (1971) What. The Hague: Mouton. Owens, Jr. R. (1984) W. Charles E. Merrill Publishing Company, . Reed, D.; Lado, R.; and Shen, Y. (1949) The inportance of the native language in foreign language learning, W, 17-23. Riekett. 0.. and firemen. M. (1984) MW. Rivers. W. (1968) W. University of Chicago Press . Roy, M. (1975) Bengali difficulties with the sounds of English. ELIl. 10, 66-72. Sapir, E. (1927) Sound patterns in language. W, 1,, 37-51. Saunders, W. (1963) Six vowel English. Wing, 11.. 177-87. Saunders, W. (1962) The teaching,of English pronunciation to speakers of Hokkien. Was. 12.. 151-157. Schumann, I. and Stenson, N. (Eds.) (1974) W W. Newbury Pbuse Publishers, Inc. Sheldon, A. and Strange, W. (1982) The acquisition of /r/ and /l/ by Japanese learners of English: Evidence that speech production can precede Speech perception. WW. 3.. 243- 261 Shen, Y. (1962) Linguistic experience and linguistic habit. Luggage W. 12.. 133-150- Shen, Y. (1961) Sound-arrangements and sound-sequences, Language, Lamina. 11. 17-32. Shen, Y. (1959) Some allophones can be important. W, 19, 7-18. Shen, Y. (1954) Phonemic charts alone are not enough. W tannin. z. 122-29. Strevens, P. (1973) A rationale for teaching pronunciation: the rival virtues of innocence and sophistication. 171.11,, 18, 182-189. 97 Sumuktu, R. (1957) Some examples of Sundanese and Javanese phonic interference in relation to learning English. W, 8, 37-48. Suter, W. (1976) Predictors of pronunciation accuracy in second language learning. mm. 2.6. 233-53. Tahta, 8.; Wood, M.; and Loewenthal, K. (1981) Foreign accents: factors relating to transfer of accent from the first languge to a second language. M.A. 265-272. Tay, M. (1969) I-bkkien phonological structure. Wu... 1, 81‘880 Templin. M. (1957) WWW. wivereitr of Minnesota Press. Teslaar van, A. (1965) Learning new sound systems: problems and prospects. 1861.. L 79-95. Tiee, H. (1969) Contrastive analysis of the monosyllabic structure of American English and Mandarin Chinese. W. 12. 1-15. Wallace, B. (1949) Pronunciation as a two-fold process. W Learnin. 2.. 44-46. Wardhaugh, R. (1974) The contrastive analysis hypothesis. In: Schumann. I. and Stenson. N. (eds. ) W W. Newbury House Publishing, Inc. Weinreich, U. (1953) W. The Hague: Mouton Wilson, D., and WllergArd, E. (1981) Errors in the production of vowel no. 10 /A/ by Norwegian learners of English. 1361.. 12, 69-76.