ELSEVIER Journal of Pragmatics 23 (1995) 281-299 Conversational code-switching in a Chinese community in Britain" A sequential analysis * Li Wei, Lesley Milroy Department of Speech, University of Newcastle upon Tyne, Newcastle upon Tyne NE1 7RU, UK Received January 1993; revised version February 1994 Abstract The informal conversational data presented in this paper were collected in the course of a larger participant observation study of a set of ten Chinese/English bilingual families in the Tyneside area of northeastern England. A sequential analysis is offered in an attempt to illuminate patterns underlying code-switching both within and between speaker turns. The procedures of Conversational Analysis (CA) are adopted, with particular attention to the way in which code-switching might be said to contextualise particular conversational procedures. Evidence is presented to suggest that these bilingual conversationalists deploy code-switching as a resource to help them contextualise preference organisation and repairs. It is argued that code-switching functions in a manner similar to other contextualisation cues which have been discussed in the literature, and that it is available to bilingual conversationalists as an additional conversation management resource. It is suggested that some of the code-switching patterns discussed in the paper are specific to generational subgroups in the community, and an attempt is made to associate the strategic use of conversational code-switching by these subgroups with wider intergenerational patterns of language choice and language competence in the community as a whole. 1. Introduction The purpose of this paper is to examine the use of code-switching as a conversational resource by Chinese/English bilingual speakers in Tyneside in the North East * The study reported here is partly supported by two grants from the Economic and Social Research Council of Great Britain (numbers R000 221074 and R000 232956), and partly by a Ridley Fellowship at the University of Newcastle upon Tyne awarded to the first author. Versions of this paper were presented in the European Science Foundation Summer School on Code-Switching in Bilingual Studies in Pavia, Italy and the 4th International Pragmatics Conference in Kobe, Japan and we thank participants for their helpful comments. We particularly thank Peter Auer for his comments on earlier analyses. Most of all of course our gratitude is due to the Tyneside Chinese families for their co-operation with this work. 0378-2166/95/$09.50 © 1995 Elsevier Science B.V. All rights reserved SSDI 0 3 7 8 - 2 1 6 6 ( 9 4 ) 0 0 0 2 6 - B 282 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 of England. The conceptual apparatus upon which we build our analysis is Gumperz's notion of contextualisation. In general terms, this notion refers to the strategic activities of speakers in varying their communicative behaviour within a socially agreed matrix of conventions, which are used to alert participants in the course of the on-going interaction to the social and situational context of the conversation (Gumperz, 1982: 132-135; 1992: 42-43). Conversation participants appear to exploit variable spoken language elements at all linguistic levels (prosodic, phonological, morphological, syntactic; see, for example, Local, 1986; Local et al., 1984, 1986) and at the non-verbal level (gestural, kinesic and proxemic; see for example Duncan, 1969, 1972; Kendon, 1977) as procedures for signalling contextual presuppositions. In Gumperz's terms, these are contextualisation conventions or contextualisation cues, their chief function being to signal participants' orientation to each other. Sometimes they are used primarily to contextualise imminent completion of a turn at talk or topic shifts, but at other times they have the capacity to signal meanings such as irony or seriousness, and social identities and attitudes of the participants. Auer (e.g. 1984, 1991) argues that bilingual code-switching should be analysed as a contextualisation cue, because it works in many ways like other contextualisation cues. However, code-switching has some characteristics of its own in addition to those it shares with such elements as gestures, prosodies and phonological variables. In particular, the sequential organisation of alternative choices of language provides a frame of reference for the interpretation of functions or meanings of conversational code-switching. In this paper, we shall adapt the analytic procedures of Conversation Analysis (CA) in our examination of Chinese/English code-switching. The principles of CA are clearly explained by, for example, Heritage (1989); Atkinson and Heritage (1984:1-11), and Levinson (1983). In the present context, a CA framework implies a sequential examination of bilingual language choice, with attention to the location of the two languages in the turn-by-turn organisation of interaction. It also entails a minimum imposition of previously given theoretical frameworks, preferring attention to recurrent patterns which can be seen to emerge after detailed analysis of numerous short conversational sequences. The sequential analysis offered here contrasts sharply with various classificatory models which have sought to categorise the grammatical structures or discourse functions of code-switching. From a structural point of view, it is possible to distinguish code-switching at three different discourse levels (see Table 1). In a given piece of conversation, we may find two speakers using different languages in consecutive turns (Level A). Within a turn, a single speaker may switch code at sentence-utterance boundaries (Level B). This is what Poplack (1980) referred to as 'inter-sentential code-switching'. Level A and Level B are closely linked in that the end of a sentence-utterance is potentially a turn transition point. The third level of code-switching refers to different constituents within a sentence-utterance being coded in different languages. The existing analytic models such as the one proposed by Poplack (e.g. 1980; Sankoff and Poplack, 1981) (and indeed the more recent models proposed by DiSci- 283 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 Table 1 Levels of code-switching Code-Switching SPEAKER 1 SPEAKER 2 SPEAKER 1 SPEAKER 2 SPEAKER 1 LanguageA LanguageA LanguageA LanguageB Sentence1 Sentence 2 Sentence 3 + Language B Language A Constituent 1 Constituent 2 LanguageA LanguageB + Level A Level B + Level C ullo et al., 1986 and Myers Scotton, 1990) focus on intra-sentential code-switching and do no take into account the sequential organisation of language choices by different conversation participants. Thus they cannot account for code-switching at turn and sentence boundaries (Levels A and B). It is clear from the literature that while frequent intra-sentential code-switching (Level C) is found most prominently in stable bilingual communities with an established history of language contact, contrastive choices of language by two different speakers at turn boundaries~ i.e. Level A code-switching, are often found in conversational interactions involving participants of differing language abilities and attitudes (e.g. Poplack, 1980, 1988; Scotton, 1976, 1983; Li Wei, in press). We shall see shortly that the Tyneside Chinese community, from which the present study draws its data, is undergoing a language shift across three generations; speakers from different generations have very different preference for and ability in using Chinese and English. As a result, code-switching occurs most often in inter-generational conversation and overwhelmingly at turn and sentence boundaries. The sequential approach adopted here can grant insight into such code-switching practices. Several researchers, most notably John Gumperz (1982), have attempted to assign discourse functions to code-switching behavior (such as addressee specification and rhetorical emphasis). Although we shall refer later to procedures for revealing such functions as they become manifest in discourse, we note here that the taxonomic analyses by researchers such as Gumperz are in practice difficult to replicate, since the identification, definition and discrimination of functions is problematic (see further Moffatt and Milroy, 1992). Sequential analysis avoids such an imposition of analyst-oriented classificatory frameworks, attempting rather to reveal the underlying procedural apparatus by which conversation participants themselves arrive at local interpretations of language choice. Auer has argued that participants do not interpret code-switching by subsuming a given instance under one of a pre-established set of functional (or indeed structural) types; rather, they make use of specifiable procedures in coming to a situated interpretation, and the meaning or function of a particular pattern of language choice in a given utterance or sequence of utterances derives both from contextual information and from these more general proce- 284 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 dures (1984:11). Thus, merely enumerating the functions of code-switching cannot account for the (in principle) infinite number of ways in which language choice may become meaningful. The application of a CA-style sequential approach to bilingual code-switching data is relatively new and not yet well-developed, and Peter Auer's analysis of German/Italian code-switching data represents the best effort to date (see especially Auer, 1984). His analysis begins, however, with a distinction between discourserelated and participant-related code-switching. The former contributes mainly to the structural organisation of the on-going conversation, by establishing a contrast in language choice between two continuous stretches of talk, while the latter invites assessment by participants of the speaker's preference for and competence in one language or the other. Yet this distinction is not always clear-cut, as Auer himself recognizes, and is sometime difficult to maintain in practice. Auer's own analysis demonstrates, for example, that turn-internal switches which often have such discourse functions as reiteration or emphasis may at the same time be participantrelated, in the sense that they reflect the speaker's competence in the language and his/her interpretation of the reactions from the audience. Instead of making such a binary distinction, we shall in the present analysis adopt a two-step approach, beginning with detailed description of the sequential organisation of language choice in conversation, which will then be interpreted with reference to the more general patterns of language preference and language ability of the speakers involved. Before we present this detailed analysis of Chinese/English bilingual conversational data, we describe briefly the Tyneside Chinese community, as it provides the social context of the study. 2. The social context of Chinese/English code-switching The Tyneside Chinese number somewhere between 5,000 and 7,000 persons. They are the second largest non-indigenous ethnic community (after those from the Indian subcontinent) in the North East of England, and they speak as their first language one of several Chinese languages (but mainly Cantonese), for which we shall use the generic label 'Chinese'. Like other researchers who have worked with immigrant communities, we are conscious of the need for a model of on-going social and linguistic change, since code-switching and language choice patterns need to be modelled very differently from those in well-established bilingual communities (Boeschoten, 1990). The material presented in this section is derived chiefly from participant observation carried out by the first author over a period of 18 months, although the contact between the fieldworker and the community has been established long before the actual fieldwork began and is still being maintained (see further Li Wei, 1994). On the basis of extensive participant observation within the community, three groups have been identified which are not always exactly isomorphic with the three generation cohorts of grandparents, parents and children: Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 285 (i) first-generation immigrants; (ii) sponsored immigrants, who are either immediate kin of the first-generation migrants or have personal connections with people already established in this country; (iii) the British-born. Subsequent analysis revealed that these groups contract quite different kinds of interpersonal network ties, which need to be interpreted within the framework of a social organisation which gives primacy to the family and an economic dependence on the catering trade. Over the years, most first-generation immigrants and those sponsored immigrants who are actively involved in the food trade have contracted network ties with mainly Chinese non-kin who are associated with their business and professional activities. The rest of the sponsored immigrants, mostly women and the elderly, more or less confine themselves to the household and family. The Britishborn generation differs from both these groups in having developed extensive network ties outside the family and often outside the Chinese community also. The educational level of this group is much higher than that of the other, and most British-born Chinese seem to want to go into occupations other than the catering trade. Thus, the exchange networks of the economically active group, both men and women who belong mainly to the 'parent' generation cohort, are strongly Chineseoriented but not restricted to kin; those of the economically less active adults are also Chinese-oriented but largely restricted to kin; and those of the British-born generation are less kin-oriented and less ethnically oriented than either of these groups (see further Li Wei et al., 1992; Milroy and Li Wei, 1994). Parallel to this change in social network ties of different generations of speakers is an on-going language shift from Chinese monolingualism in the 'sponsored immigrants' generation (especially elderly women), through various proportions of Chinese/English bilingualism amongst the first-generation immigrants and sections of the sponsored immigrants (those who are economically active), to English-dominant bilingualism characteristic of the British-born Chinese. Table 2 shows the observed patterns of language choice with a range of interlocutors by 58 speakers from three generations of ten Tyneside Chinese families (see further Li Wei, 1994). In Table 2, Pattern 1 indicates the use of Chinese only in all situations. Pattern 2 is a clearly Chinese-dominant pattern. Patterns 3 and 4 can be described as slightly differently balanced bilingual patterns. Pattern 5 suggests the use of both Chinese and English with all interlocutor types. Patterns 6 and 7 indicate the use of English only with the child generation and either Chinese only or both Chinese and English with female interlocutors of the grandparent generation who are non-family members. These last two patterns can be described as English-dominant bilingual patterns. As we can see from the number of speakers listed in the table, language choice preferences of members of the three generations are very different: Chinese tends to be used by grandparents and to grandparents (most of whom are sponsored immigrants), while English tends to be used by children and to children (most of whom are British-born). Both Chinese and English may be used by parents and to parents (those who are either first-generation migrants or younger, economically active 286 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 Table 2 Generalised patterns of language choice (Total number of speakers: 30 males + 28 females = 58) Interlocutors Pattern Pattern Pattern Pattern Pattern Pattern Pattern 1 2 3 4 5 6 7 No. of speakers 1 2 3 4 5 6 GP P Ch C C C C CE C CE C C C CE CE CE CE C C CE CE CE CE CE C C CE CE CE CE CE C CE CE CE CE E E C CE CE CE CE E E 9 2 - 14 2 4 - 14 3 5 5 Interlocutors: 1 = Female grandparent generation; 2 = Male grandparent generation; 3 = Male parent generation; 4 = Female parent generation; 5 = Male child generation; 6 = Female child generation Language choice: C = Chinese; E = English; CE = Chinese and English Speakers: GP = Grandparents; P = Parents; Ch = Children s p o n s o r e d immigrants). Gal (1979) suggests that it is through this k i n d o f association b e t w e e n language choice and interlocutor types that languages acquire their social s y m b o l i s m . F o r e x a m p l e , since in the T y n e s i d e c o m m u n i t y the e x c l u s i v e use o f Chinese is associated p r i m a r i l y with the grandparents, it m a y be d e s c r i b e d as the ' w e c o d e ' for that generation and for older speakers generally; English, on the other hand, which is associated chiefly with British-born children, m a y be r e g a r d e d as their ' w e c o d e ' . Details o f the generational changes in language choice, l a n g u a g e ability and social networks o f the T y n e s i d e C h i n e s e c o m m u n i t y are reported in Li W e i (1994). It is clear even from the fairly general description presented so far that codeswitching cannot be a c o m m u n i t y - w i d e p h e n o m e n o n a m o n g s t the T y n e s i d e Chinese. The nine Chinese m o n o l i n g u a l grandparents (seven females and two males) in our s a m p l e (see T a b l e 2), for e x a m p l e , are not c a p a b l e o f speaking English at all. E v e n a m o n g s t those who have the ability to use both l a n g u a g e s (though at very different levels), there are a d d r e s s e e - r e l a t e d variations in language preferences: none o f the speakers speak English to the m o n o l i n g u a l grandparents; ten British-born children use only English with their peers, and the m a j o r i t y o f parents speak only Chinese with m e m b e r s o f their o w n generation. Note also that the C E labels in Table 2 m e r e l y indicate that both l a n g u a g e s have been o b s e r v e d in conversational interaction. T h e y do not give any information o f the m a n n e r in which the two l a n g u a g e s are used ( e . g . whether they are used in separate turns or m i x e d in the same utterance or turn). Such information can only be obtained through a detailed turn-by-turn analysis o f conversational data. Since the general preference is therefore for m o n o l i n g u a l i s m , c o d e - s w i t c h i n g in conversation is a l w a y s socially m a r k e d and structurally noticeable (Scotton, 1988; M y e r s Scotton, 1993). In other words, it is r e a s o n a b l e to a s s u m e that speakers generally have an awareness o f the preferred l a n g u a g e for an o n - g o i n g interaction and that they c h o o s e a different l a n g u a g e in order to draw the attention o f the addressees Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281--299 287 to some specific discourse structure, so that that particular portion of the conversation could be interpreted differently from the rest of the interaction. Myers Scotton (1993: 79-80) argues that all bilingual speakers possess a 'markedness metric', an innate, internalised model which enables them to recognise which language in the community repertoire is marked and which is unmarked in a specific interactional context and that choice of one language rather than another carries social import. In the following section, we examine in some detail how speakers of different generations and social background, and with apparently different language preference and language attitudes, make use of code-switching as a linguistic resource for specific communicative purposes. The data base for the study is a corpus of approximately 23 hours of spontaneous conversational exchange recorded in ten Tyneside Chinese families (30 male + 28 female = 58 speakers). The actual code-switching utterances, however, amount to around 50 minutes only, which may in itself be seen as a consequence of the community-wide preference for monolingualism (Chinese for grandparents and parents and English for the British-born children respectively). We shall concentrate on two structures taken from the complex architecture of conversational organisation, namely, preference organisation and repair, and present evidence which suggests that Chinese/English bilinguals use code-switching to contextualise (dis)preference and to repair trouble spots. 3. Contextualising preference organisation Preference organisation in simple terms refers to the ranking of alternative second parts of the so-called adjacency pairs, such as acceptance or refusal of an offer, or agreement or disagreement with an assessment (Levinson, 1983; Pomerantz, 1984; Atkinson and Drew, 1979). It has been argued that such alternatives are not generally of equal status; rather, some second parts are 'preferred' while others are 'dispreferred' (for example, acceptance of offers as opposed to rejections; agreements with assessments rather than disagreements). Critical to the notion of preference is the observation that preferred second parts tend to be structurally simpler and to latch smoothly on to the first parts, while dispreferred second parts are structurally more complex, and are typically delivered after a filled or unfilled pause, with some preface marking their dispreferred status (often the particle well), and sometimes with some account of why the preferred second cannot be performed (see further Levinson, 1983: 307; Lesser and Milroy, 1993: 193). In short, speakers evidently manage their disagreements, refusals and rejections by marking them in a predictable way. Let us now look at some examples of the way(s) in which Chinese/English bilinguals in Tyneside contextualise preference in conversational interaction. Consider the following sequence: (1) (Dinner table talk between mother A and daughter B.) A: Oy-m-oy faan a? Ah Ying a? (Want some rice?) 288 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 B: A: B: A: B: (no response) Chaaufaan a. Oy-m-oy? (Fried rice. Want or not?) (2.0) I'll have some shrimps. Mut-ye? (.) Chaaufaan a. (What?) (Fried rice.) Hai a. (OK.) In (1), mother A, speaking Chinese throughout, offers her daughter rice. The initial offer, or in CA terms the first pair part, sets up a conditional relevance and expectation which needs to be fulfilled by a second part. Failure to fulfil conditionally relevant expectations by producing appropriate second pair parts results in a 'noticeable silence', which is exactly the case in this example when B, the daughter, delays her response to the mother's offer. We can see the speakers' orientation to the paired structure in the mother's response to child's silence - the offer is repeated in an emphatic form to elicit a response after the first attempt failed. The child's response to the mother's repeated offer is a request for an alternative to rice, which the mother apparently understands as a refusal, as she requests a repair in the following turn which is followed with another offer of rice. Notice that B's indirect refusal is marked in two steps - first a two-second delay before delivery, a commonly occurring signal of an imminent dispreferred response; then the choice of English which contrasts the code choice in the immediately preceding turn by the mother. The child's final acceptance of the mother's offer of rice is in Chinese, which corresponds to the language choice of the mother, but differs from the one she has used to mark her indirect refusal. Similar patterns of preference marking are evident also in the following two sequences: (2) (B, A: B: A: a twelve-year-old boy, is playing with a home computer. A is his mother.) Finished homework? (2.0) Steven, yiu mo wan sue? (want to review (your) lessons) B: (1.5) I've finished. (3) (A, a man in his late twenties; B, a woman in her early forties; C, B's teenage daughter) A: Sik gai a. (Eat chicken.) B: mm. (5.0) A: Haven't seen Robert Ng for a long time. (2.0) A: Have you seen him recently? B: No. Li Wei, L. Mihoy / Journal of Pragmatics 23 (1995) 281-299 A: B: 289 Have you seen Ah Cbing? ... (2.0) (To C) Ning ngaw doei haai lai. (Bring my shoes.) (To A) Koei hoei bindou a? (Where was she going?) As in (1), second pair part responses in (2) and (3) are delivered after a period of silence and other delay components including insertion sequences in (3). But what is remarkable here is that all these dispreferred second parts are accompanied also by contrastive choices where the next turn speakers use a language different from that of the preceding turn speakers. This kind of code non-alignment is reported also by Auer (1984) and Sebba and Wootton (1984) in German/Italian and London English/Jamaican creole bilingual communities respectively. It should be pointed out that the analysis as instances of code-switching of such contrastive choices of language by different speakers in consecutive turns can only be done within a CA framework which focuses on the sequential organisation of interaction, but the recognition of such bilingual strategy is important because it often reflects the language ability and language preferences of participants. It has been argued that preference marking is an important procedure for the preservation of social solidarity during delicate and potentially face-threatening encounters (Silverman, Ms.). An examination of ways in which code-switching is being used as a conversational resource in preference marking may therefore throw light upon how bilingual speakers maintain their social relationships. Let us take a look at one further example of different language choices which mark preferred versus dispreferred second parts. (4) (A is the mother, B her nine-year-old daughter, and C, her twelve-year-old son.) A: Who want some? [Crispy a. B: [Yes. A: Yiu me? (Want some?) B: Hai a. (Yes.) (A handing over some spring-rolls to B.) A: (To C) Want some, John? C: Ngaw m yiu. (I don't want.) A: M yiu a? Crispy la. (Don't want?) C: (Shaking head) mm In this sequence, B twice accepts A's offer of spring-rolls, twice using the same language as A (Chinese and English respectively) for this preferred response. In contrast, when C declines A's offer, we find a pattern similar to the one which is evident in (1), (2) and (3). After a short pause, C selects a language different from the lan- 290 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 guage used for the first pair part; A uses English for the offer, whereas C uses Chinese for the refusal. The general pattern seems to be that while preferred second pair parts are accompanied by language alignment, dispreferred responses are marked by contrasting language choices. We have already noted earlier that dispreferred responses in monolingual English conversation are generally marked by various structural complexities including pauses before delivery, prefaces such as but and well, token agreements, appreciations and apologies. Also common are accounts or explanations of why the preferred second part is not forthcoming. On the basis of the examples we have seen so far, it seems reasonable to suggest that contrasting choices of language occurring in the same sequential context can be used to mark dispreference in bilingual conversation in much the same way as those markedness features in monolingual conversation. In fact, Auer (1991) argues that code-switching is the most significant discourse marker in bilingual conversations in the sense that apparently deviant (or marked) choices of language are more noticeable than other linguistic features. Our data indeed suggest that while code-switching co-occurs with some dispreference markers such as pauses (see (1) and (2) above), it seems sometimes to substitute for particular language-specific markers. For example, English dispreference markers such as well and but do not occur in this function in our corpus. This raises the question of how far dispreference markers are universal and how far they are language specific, a question worth exploring in future research in bilingual communities or in cross-linguistic research. We conclude this section with a brief look at two further examples of codeswitching which marks different kinds of dispreferred seconds - respectively a disagreement with an assessment and a refusal of an offer. (5) (Two young women are looking at new dresses.) A: Nau, ni goh. (This one.) B: Ho leng a. (Very prettty.) A: Leng me? (1.5) Very expensive. (Pretty?) B: Guai m gaui a? (Expensive or not?) A: Hao guai. (Very expensive.) (6) (A, a man in his early thirties is talking with B, who is a woman of a similar age.) A: Manhing drive you home. B: (1.0) ngaw daap basi hoei. (I'll take a bus.) A: Yiga m ho hoei. Yan daw (.) ho naan daap basi a. (Don't go now. So many people. Very difficult to get on a bus.) B: (Waits for A to call Manhing.) Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 291 In (5), B offers her assessment of A's new dress - Ho leng a (Very pretty). A's response to this consists first of a 'reflective' question in Chinese - Leng m e ? (Pretty?). This type of question is formed by partial repetition plus the question marker 'me' and has discourse functions similar to English tags such as 'isn't it?' or 'really?'. Here, the subsequent interaction suggests that it functions as a 'hedge' heralding a further dispreferred assessment, and itself indicates only a qualified agreement with B's original assessment (see Pomerantz, 1984, for comments on responses to assessment). Notice that A switches to English for her second assessment. When B asks for confirmation in the following turn Guai m guai a ? (Expensive or not?), A's 'preferred' response is in Chinese, the same language as B's question. In (6), A is offering to ask Manhing, his son, to drive B home. B declines the offer by saying that she can take the bus. She marks her refusal first with a one-second pause - a common dispreference marker, and then by choosing Chinese, which contrasts with A's choice of language. But A repeats his offer and gives a reason for not taking the bus at that particular time of the day. Notice that A's reformulation of his offer is accompanied by switching from English, the language he used for his original offer, to Chinese, the language B has used for her dispreferred response. Preference organization affects not only the second part speaker's contributions but also operates across a given speaker's turns, giving rise to 'repairs' of first parts of the pair in subsequent turns. In (6) A's reformulation of his original offer after a dispreferred response is one example of such a repair. In the following section, we shall look more closely at the language choices deployed in initiating and making repairs in bilingual conversation. 4. Contextualising repair Consider (7) below: (7) (A and B are both female speakers in their early forties.) A: ... koei hai yisaang. (He's a doctor.) B: Is he? A: Yichin (.) hai Hong Kong. (Before) (In Hong Kong.) In this extract, A and B are talking about the occupation (as a doctor) of a Chinese man who has recently settled in Tyneside. However, A here has not specified his period of employment as a doctor as past or present time; in Chinese, time reference is expressed not by alternative verb forms as in English but by adverbials (e.g. yesterday, next year). The question 'Is he?' (in conversation-analytic terms, a next turn repair initiator) offers A an opportunity to confirm or to reformulate her original assertion. In the event, A subsequently repairs her original statement by specifying both the time and place of the man's period of employment as a doctor. Note that B 292 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 realises her repair initiator in a language which contrasts with the language of A's turns. A comparable example is (8) below where B's repair initiator is also marked out with a contrasting choice of language. (8) (A (in her forties) and B (in her mid-twenties) are both women.) A: D a m do. Koeige telephone gonggan. Koei dang yatjan joi da. (Can't get through. Her telephone is engaged. She'll ring again in a short while.) B: She ring? A: Hai a, ngaw da. (Yes, I'll ring.) Here, A is trying to telephone a friend, but the line is engaged. Initiating a repair on A's first turn Koei dang yatjan joi da (She'll ring again in a short while), B switches to English to ask 'She ring?'. This is followed by A's repair Hai a, ngaw da (Yes, I'll ring). Again, we find that the language of the repair initiator token contrasts with the language of the preceding and following turns. It should be pointed out that the fact that the two second speakers in both examples ask the first turn speakers to confirm or to reformulate their statements suggests that to them as participants there is a problem in the on-going interaction, although to non-participants what needs repair may not be immediately transparent. Only through a sequential analysis which focuses on each move of the conversationalists themselves can we, as analysts, detect any repairable spot and infer the social meaning of code-switching. Now compare (7) and (8) with (9) below. (9) (A is woman in her late thirties and B is a man in his late twenties.) A: He's a [ku:] ... (.) I don't know how to say (.) send message (.) Nay ji-m-ji a ? (Do you know?) B: Oh, courier. A: Yes, courier. In (9), A uses English to comment on a relative of hers who travels frequently between Britain and Hong Kong. As the subsequent interaction reveals, she is trying to locate and pronounce the lexical item 'courier'. The subsequent repair by B differs from those exemplified in (7) and (8) (other initiated self-repairs), in that it is a self-initiated other repair invited by A herself. Nevertheless, (9) is comparable to the two preceding sequences in that the language of the repair initiator contrasts with the language of the preceding and following turns. Further examples show that code-switching can also contextualise self-initiated self-repairs, that is, repairs that are done by the speaker within the same speaking turn without prompting from others. Consider, for example, (10) below. Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 293 (10) A: His sister (.) koei-ge mooi is my good friend. (his younger sister) Here, the English kin term sister is problematic; Chinese has two different kin terms which differentiate younger sister from elder sister. To accomplish a repair which clarifies the referent of 'sister', the speaker code-switches to Chinese, switching back to English to continue with her talk. Sometimes speakers deploy other resources to mark a particular item within an utterance as problematic. Instead of replacing it with a contrasting-code item as in (10), an 'attention catcher' (or in Sacks and Schegloff's (1979) terms 'try marker') may be inserted in a different language before the speaker repeats or clarifies the problematic item. In (11) below, for example, A inserts a code-switched tag, 'you know', to flag her subsequent repetition of the word daji (typist). (11) A: Koei hai gongsi jo daji you know daji yuen. (She works in a company as a typist (you know) typist.) The code-switches in (10) and (11) have one feature in common - a predictable end at which point the speakers switch back to the base language of the turn. Patterns similar to those displayed here have been reported elsewhere in the codeswitching literature. Researchers have frequently observed that code-switching can serve the functions of word-finding, self-editing (with or without discernible errors), repetition, emphasis, clarification, confirmation, and so forth. However, all these functions can be described within a more principled, integrated framework as phenomena arising from repair, a very general conversation organisational procedure. A more general, methodological point to be made here is that the conversationanalytic approach to code-switching (or the sequential analysis of language choice), as presented here, is intended to incorporate the existing models into an 'interpretative' framework which focuses on the participants' methods of using code-switching as a communicative strategy in conversational interaction. This point is worth emphasising, because very often new analyses of code-switching are presented as if they contradict rather than complement each other, and without suggesting ways in which they might be integrated into a more coherent model. The data presented above suggest three major ways in which code-switching can contextualise repair. First (as in (7), (8), (9) above) a repair initiator may be issued in a contrasting language, the speaker of the repairable item then doing the repair him- or herself. This is a type of self-initiated self-repair. Second the repairable item(s) may be replaced with an equivalent in a different language; this can also be done by the speaker without prompting from others (self-initiated self-repair, as in (10)) or by different speakers (other-initiated other-repairs, as in (9)). The third (as in 11)) is to insert an item in a different language to draw the listener's attention to the repairable. Certain English language expressions such as you know, right, see seem often to be used for this purpose. Schiffrin (1987) examines the functions of such 'discourse markers' in monolingual conversation, but the existing conversation 294 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 analysis literature has little to say about the marking of repairs in bilingual discourse. Further analysis along the lines suggested here may therefore contribute not only to our understanding of code-switching as a contextualisation cue but to conversational management procedures more generally. Conversation analysts have frequently pointed out that repair procedures reflect to a high degree the collaboration and cooperation among participants which is an essential aspect of orderly conversation, and which itself embodies a higher level of inter-subjective social consensus (for a recent and comprehensive discussion, see Schegloff, 1992). Failure to respond to repair initiators or to do required repairs can have undesirable communicative consequences, as is illustrated by the following sequence. (12) (A, eighteen, son of B, female, early forties, is looking for the car keys to go out.) A: Where's the keys? B: Mut-ye? (What?) (2.5) Gaha lokyu a. (It's raining.) A: I w o n ' t be long. B: No. In this extract, which requires a rather complex exposition, A's question about the location of the keys can be understood as a pre-request to use the car to go out. However, B does not reply directly but follows A's utterance with another question Mutye? (What?). This may be seen as a candidate 'next turn repair initiator', which gives the first part speaker A an opportunity to re-formulate the prior turn in the next turn in order to avoid an imminent dispreferred response. Notice that although the question is contextualised by being realised in a language contrasting with the language of A's pre-request, A fails to make the apparently required repair. After a twoand-a-half second gap, B's comment on the weather may be seen (in light of the interactional outcome) to constitute a hedged refusal of A's pre-request, which she has evidently interpreted as a request to use the car. Again, B has chosen to use Chinese for her turn, but A continues to indicate his wish to take the car and fails furthermore to align his choice with B's. His mother's ultimate non-compliant response follows without further delay. While it is not possible to predict on an 'if only' basis possible alternative outcomes of requests, we can note that A's failure to achieve the desired compliant response from B is associated in (12) with his failure to read the contextualisation cue provided by A's language switch and to use appropriate procedures. Gumperz (1982: 133) discusses a comparable case, where a failure to read such a cue gives rise to a similarly unsatisfactory interactional outcome. A clear example of failure to carry out repair leading to communication breakdown is provided in (13) below. Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 295 (13) (A, an eight-year-old girl, and C, a boy of about fifteen, are children of B, mother of mid-forties.) A: Cut it out for me (.) please. B: (2.5) A: Cut it out for me, mum. C: [Give us a look. B: [Mut-ye ? (What?) A: Cut this out. B: Mut-ye? (What?) C: Give us a look. ... (2.0) B: Nay m ying wa lei? (You don't answer me?) A: (To C) Get me a pen. Following a null response from B to A's request for help, she repeats it, using a vocative on this occasion to specify her mother as the next turn speaker. B's subsequent next turn repair initiator Mut-ye ? (What?) overlaps with C's turn - he selfselects as next speaker. A then issues her request for the third time, but B repeats the same next turn repair initiator token. Again A fails to do the repair which B apparently expects, and the lack of alignment between the language choices of A and B is again noticeable, as in (12). A's three repeated requests are in English and show no sign of the change of form which would indicate a repair, while B's repair initiators are in Chinese. It is instructive to compare this example with the alignment evident in the last two turns of (1), where the same repair initiator is followed both by a compliant response and a shift of language. At the end of (13), however, we find something close to a communicative breakdown, in the sense that B offers no response at all to A's repeated requests. After a two-and-a-half-second silence, B asks A why she does not respond to her. A then turns to C, abandoning the exchange between herself and B. Compare this with (12), where a response is ultimately provided, albeit a non-compliant one. These extracts serve to highlight the role of code-switching as a central device for successful communication. Essentially, code-switching is a discourse strategy whereby bilingual speakers accommodate and collaborate with each other. By changing from one language to another, speakers indicate their awareness of potential trouble spots in the interactional process and repair those which have already arisen. Failure to do so in accordance with the matrix of contextualisation cues and conventions apparently accepted by co-interlocutors can, as we have seen, lead to the breakdown of an on-going conversation. The wider issue of how such discoursal meanings are related to the broader social significance of code-switching has been elaborated by various researchers. For example, Scotton (1988) argues that while the tendency is for speakers to use the language with which they feel more comfortable, they are generally aware of the set 296 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 of rights and obligations involved in the on-going exchange and would choose the form of their conversational contribution appropriate to that set, even though it may sometimes mean that speakers have to use a language which they know less well. Any move that is inappropriate to this matrix of mutual rights and obligations may be interpreted by the participants as deliberate and poses a potential threat to social interaction (see also Heller, 1982). 5. Conversational code-switching and community norms of language choice We have tried to demonstrate in this paper that Chinese/English bilingual speakers switch languages to contextualise preference organisation and repairs. Elsewhere we have also discussed code-switching being deployed to contextualise turn taking (including turn allocation and competition) and various kinds of inserted sequences (see Li Wei, 1994; Li Wei et al., 1992). By building a contrast in language choice for two stretches of conversation, bilingual speakers are able to draw attention to details of the projected course of conversation and to check each other's understandings. It is important to emphasise that code-switching is only one of many linguistic resources available to bilingual speakers, and they may (and indeed do) select alternative cues in the sequential contexts which we have studied (recall that codeswitching makes up only 50 minutes of a total 23 hours of tape-recording). It is equally important to bear in mind that in actual conversation a given utterance may simultaneously perform a number of discourse functions. For example, an utterance serving as a turn-allocation component may also function as a repair initiator, and a dispreferred second part may also be an insertion sequence. Accordingly, codeswitching as a contextualisation cue is multi-functional, and this is one reason why traditional classificatory approach which attempts to enumerate ad hoc functional categories to which instances of code-switching may be subsumed are unsatisfactory (see Auer, 1984, 1991). At various points in the analysis, we have alluded to the link between conversation structures and language choice (see particularly Table 2 above). This relationship, however, is variable in so far as certain (groups of) speakers tend to use one language for a particular conversation structure, while others use a different language for the same interactional purpose. We have seen, for example, that parents and grandparents sometimes use English to contextualise offers or requests when they talk to the British-born children (e.g. extracts (2) and (4)). In both cases they provide (in CA terms) first parts of adjacency pair which have the function of turn allocation. This particular interactional task may, however, be contextualised in Chinese in peer-group interaction among the British-born Chinese speakers. This is because the base language for the two types of interaction (inter-generational and intra-generational respectively) is very different - parents and grandparents normally speak Chinese and only switch to English for certain communicative effect in intergenerational talk; in contrast, the British-born generation normally use English for peer-group interaction and switch to Chinese for special purposes (see further Li Wei, in press). Clearly, generational variations in code-switching strategies cannot Li Wei, L. Milroy /Journal of Pragmatics 23 (1995) 281-299 297 be fully explained without reference to the more general patterns of language choice and language shift which we have described in the first part of the paper (see also Li Wei, 1994). In her 'rnarkedness' model of code choice, Scotton (1983) argues that on the one hand bilingual speakers are creative, rational social actors and they exploit the available choices (linguistic or otherwise) for specific communicative purposes; on the other, they act within a normative framework - a set of rights and obligations which is determined by the community to which they belong. With regard to the Tyneside Chinese community, the generation- and network-specific language choice preferences provide such a frame and the various code-switching practices discussed in this paper must be interpreted within this frame. For example, it is clear from both long-term participant observation and detailed analysis of tape-recorded conversational data that English is the preferred language of the British-born children. The implication of this general language choice pattern is that the base language of a conversational exchange initiated by these speakers is very likely to be English. Thus, in order to build a contrast in code choice in an ongoing conversation, they switch to Chinese. In contrast, the preferred language of parents and grandparents is Chinese. They do not normally switch language when talking with members of their own generations, except perhaps for a few cases of self-repair in English (most often taking the form of temporary lexical borrowing). Speakers' awareness of the community norm and of the interactional, as well as social, effect of language choice is most evident in the marking of preference organisation. While there is a strong likelihood that the British-born children will use English to contextualise dispreferred seconds in inter-generational conversation, they normally switch to Chinese after the adults involved issue them repair initiators (e.g. as in (1) above). The adults (parents or grandparents), for their part, normally insist on using Chinese when their first pair parts have been responded to by the children either in English or in silence, giving rise to contrastive language choices. An interesting example of the speakers' ability to exploit the generational language choice norm is (2), in which the mother asks her twelve-year-old son in English, rather than Chinese, whether he has finished his homework. If Scotton's view of bilingual speakers' 'markedness metric' is correct, it seems plausible to argue that in some senses the mother knows that by choosing the preferred language of the child instead of her own preferred language, she is turning a simple question into an indirect request for the child to do his homework before playing with the computer, an implicature evidently understood by the child. Had the child not interpreted the meaning of this particular choice correctly, one assumes that he could have been ordered by the mother to pack up his computer games and do his homework. This example further illustrates that there is no simple, one-to-one relationship between code-switching structure and community-level language preference. The latter does not constrain individual instances of code-switching; rather it offers a frame of reference for participants (and analysts too for that matter) to interpret the meaning of language choice in conversation. While some instances of code-switching seem to serve primarily discourse functions (e.g. either Chinese or English may be used to contextualise turn competition where contrasting code choice is more sig- 298 Li Wei, L. Milroy / Journal of Pragmatics 23 (1995) 281-299 nificant than the use o f a particular language) and are not easily related to language preference o f the speaker, the v e r y fact that speakers change from one l a n g u a g e to another in conversation calls for special interpretation. In sum, the sequential a p p r o a c h to conversational c o d e - s w i t c h i n g offered in this p a p e r helps p r o v i d e e v i d e n c e at the m i c r o - i n t e r a c t i o n a l level o f the generation- and n e t w o r k - s p e c i f i c l a n g u a g e choice preferences o f the T y n e s i d e Chinese/English bilingual speakers. It enables us to understand better the strategies w h i c h bilingual speakers with differing l a n g u a g e preference and l a n g u a g e ability use to m a n a g e conversational interaction and the p r o c e d u r e s to arrive at local interpretations o f code-switching. This i n turn p r o v i d e s the foundation for understanding the interactional basis o f the processes o f social and linguistic change. References Atkinson, J.M. and P. Drew, 1979. Order in court: The organisation of verbal interaction in judicial settings. London: Macmillan. Atkinson, J.M. and J. Heritage, eds., 1984. Structures of social action: Studies in conversation analysis. Cambridge: Cambridge University Press. Auer, P., 1984. Bilingual conversation Amsterdam: Benjamins. Auer, P., 1991. Bilingualism in/as social action: A sequential approach to code-switching. In: Papers for the symposium on code-switching in bilingual studies, 319-352. Strasbourg: European Science Foundation Network on Code-Switching and Language Contract. Auer, P. and A. di Luzio, eds., 1992. The contextualisation of language. Amsterdam: Benjamins. Boeschoten, H., 1990. Asymmetrical code-switching in immigrant communities. In: Papers for the workshop constraints, conditions and models, 85-100. Strasbourg: European Science Foundation Network on Code-Switching and Language Contact. DiSciullo, A.-M., P. Muysken and R. Singh, 1986. Code-mixing and government. Journal of Linguistics 22: 1-24. Duncan, S., 1969. Nonverbal communication. Psychological Bulletin 72: 118-137. Duncan, S ;, 1972. Some signals and rules for taking speaking turns in conversations. Journal of Personality and Social Psychology 23: 283-292. Gal, S., 1979. Language shift: Social determinants of linguistic change in bilingual Austria. New York: Academic Press. Gumperz, J.J., 1982. Discourse strategies. Cambridge: Cambridge University Press. Gumperz, J.J., 1992. Contextualization revisited. In: P. Auer and A. di Luzio, eds., 39-53. Heller, M., 1982. Negotiations of language choice in Montreal. In: J.J. Gumperz, ed., Language and social identity, 108-118. Cambridge University Press. Heller, M., ed., 1988. Codeswitching: Anthropological and sociolinguistic erspectives. Berlin: Mouton de Gruyter. Heritage, J., 1989. Current developments in conversation analysis. In: D. Roger and P. Bull, eds., Conversation: An interdisciplinary perspective, 2147. Clevedon: Multilingual Matters. Kendon, A., 1977. Studies in the behaviour of social interaction. Lisse: Peter de Ridder. Lesser, R. and L. Milroy, 1993. Linguistics and aphasia: Psycholinguistic and pragmatic aspects of intervention. London: Longman. Levinson, S.C., 1983. Pragmatics. Cambridge: Cambridge University Press. Li Wei, 1994. Three generations two languages one family: Language choice and language shift in a Chinese community in Britain. Clevedon: Multilingual Matters. Li Wei, in press. Variations in language choice and code-switching patterns of three Chinese groups in Newcastle upon Tyne. To appear in Multilingua. Li Wei, L. Mihoy / Journal of Pragmatics 23 (1995) 281-299 299 Li Wei, L. Milroy and S.C. Pong, 1992. A two-step sociolinguistic analysis of code-switching and language choice. International Journal of Applied Linguistics 2(1): 63-86. Local, J., 1986. Patterns and problems in a study of Tyneside intonation. In: C. Johns-Lewis, ed., Intonation in discourse, 181-198. London: Croom Helm. Local, J.K., J. Kelly and W.H.G. Wells, 1986. Towards a phonology of conversation: Turn-taking in Tyneside English. Journal of Linguistics 22:411-437. Local, J.K., W.H.G. Wells and M. Sebba, 1984. Phonology for conversation: Phonetic aspects of turn delimitation in London Jamaican. Journal of Pragmatics 9: 309-330. Milroy, L. and Li Wei, 1994. A social network perspective on code-switching and language choice: The example of the Tyneside Chinese community. In: L. Milroy and P. Muysken, eds., One speaker, two Languages: Cross-disciplinary perspectives on code-switching. Cambridge University Press. Moffatt, S. and L. Milroy, 1992. Panjabi/English language alternation in the early school years. Multilingua 11(4): 355-385. Myers-Scotton, C., 1990. Intersections between social motivations and structural processing in codeswitching. In: Papers for the workshop constraints, conditions and models, 57-81. Strasbourg: European Science Foundation Network on Code-Switching and Language Contact. Myers-Scotton, C., 1993. Social motivations for codeswitching: Evidence from Africa. Oxford: Clarendon Press. Pomerantz, A., 1984. Agreeing and disagreeing with assessments: Some features of preferred/dispreferred turn shapes. In: J.M. Atkinson and J. Heritage, eds., 57-101. Poplack, S., 1980. Sometimes I'll start a sentence in Spanish Y TERMINO EN ESPANOL: Towards a typology of code-switching. Linguistics 18: 581-618. Poplack, S., 1988. Contrasting patterns of code-switching in two communities. In: M. Heller, ed., 42-70. Sacks, H. and E.A. Schegloff, 1979. Two preferences in the organisation of reference to persons in conversation and their interaction. In: G. Psathas, ed., Everyday language: Studies in ethnomethodology, 15-21. New York: Irvington. Sankoff, D. and S. Poplack, 1981. A formal grammar of code-switching. Papers in Linguistics 14(1): 3-45. Schegloff, E.A., 1992. In another context. In: A. Duranti and C. Goodwin, eds., Rethinking context, 191-228. Cambridge: Cambridge University Press. Schiffrin, D., 1987. Discourse markers. Cambridge: Cambridge University Press. Scotton, C.M., 1976. Strategies of neutrality: Language choice in uncertain situations. Language 52(4): 919-941. Scotton, C.M., 1983. The negotiation of identities in conversation: A theory of markedness and code choice. International Journal of the Sociology of Language 44:115-136. Scotton, C.M., 1988. Code-switching as indexical of social negotiations. In: M. Heller, ed., 151-186. Sebba, M. and A.J. Wootton, 1984. Conversational code-switching in London Jamaican. Paper presented at Sociolinguistic Symposium 5, Liverpool. Silverman, D., Ms. Preserving social solidarity: Preference organisation and advice in Aids counselling.