Sociolinguistics Soziolinguistik An International Handbook of the Science of Language and Society Ein internationales Handbuch zur Wissenschaft von Sprache und Gesellschaft 2nd completely revised and extended edition 2., vollstandig neu bearbeitete und erweiterte Auflage Edited by / Herausgegeben von Ulrich Ammon . Norbert Dittmar Klaus 1. Mattheier . Peter Trudgill Volume 1 / 1. Teilband Offprint I Sonderdruck Walter de Gruyter . Berlin· New York 70. Code-Switching 589 70. Code-Switching/Sprachwechsel 1. Linguistic manifestations of language conjunctions and their conjuncts), but these contact were soon met with a host of counter- 2. Theories of CS examples. 3. Fitting theory to data The first general account of the distribu- 4. The data of cs tion of stemmed from the observation 5. Community strategies for CS es 6. Summary that es is favored at the kinds of syntactic 7. Literature (selected) boundaries which occur in both languages. The Equivalence Constraint (Poplack 1980) states that switched sentences are made up 1. Linguistic manifestations of of concatenated fragments of alternating language contact languages, each of which is grammatical in the language of its provenance (see also Code-switching (eS) is but one of a number Lipski 1978; Muysken 2000; Pfaff 1979). of the linguistic manifestations of language The boundary between adjacent fragments contact and mixing, which variously include occurs between two constituents that are borrowing on the lexical and syntactic levels, ordered in the same way in both languages, language transfer, linguistic convergence, ensuring the linear coherence of sentence interference, language attrition, language structure without omitting or duplicating death, pidginization and creolization, among lexical content. others. There is little consensus in the litera- That general principles, rather than atom- ture over which aspects should be subsumed istic constraints, govern es is now widely under the label code-switching. In this ar- accepted, though there is little consensus as ticle, es refers to the utterance-internal jux- to what they are or how they should be rep- taposition, in unintegrated form, of overt resented. Many theories assume that the linguistic elements from two or more lan- mechanisms for language switching are di- guages, with no necessary change of inter- rectly predictable from general principles of locutor or topic. (monolingual) grammar. As extensions of Mixing may take place at any level of lin- the formal linguistic theories successively in guistic structure, and a long research tradi- vogue, these tend to appeal to such abstract tion has grown up around questions of lan- grammatical properties as inter-constituent guage choice and language negotiation relationships (e.g., government, case assign- among interlocutors in bilingual contexts ment) andlor language-specific features of (Gumperz 1976/1982; Heller 1982). But the lexical categories (i.e., subcategorization of combination of languages within the con- grammatical arguments, inherent morpho- fines of a single sentence, constituent or logical features). even word, has proved most intriguing to Di Sciullo et al. (1986), for example, ident- linguists. This article surveys the treatment ified the relevant relations as C-command in the literature, linguistic and social, of and governme;nt: es cannot occur where a such intra-sentential es. government relation holds. Replacement of the function of government in standard the- 2. Theories of CS ory by the notion of feature agreement led to a parallel focus on feature matching in es First dismissed as random and deviant (e.g., studies. The Functional Head Constraint Weinreich 1953/1968) intra-sentential es (Belazi et al. 1994) adds language choice t6 is now known to be grammatically con- the features instantiated in functional and strained. The basis for this is the empirical lexical categories, prohibiting es where a observation that bilinguals tend to switch mismatch occurs. MacSwan's (1999) C:ldap- intra-sententially at certain (morpho )syn- tation of the Minimalist proposal restricts tactic boundaries and not at others. Early CS at structural sites showing cross-lan- efforts to describe these tendencies (e.g., guage differences in monolingual features. Gumperz 197611982; Timm 1975) offered The distinction between lexical and func- taxonomies of sites in the sentence where es tional categories is a hallmark oftheories in- could and could not occur (e.g., between voking the complement structure of individ- pronominal subjects and verbs or between ual1exica1 items to characterize permissible 590 IV. The Social Implications of Levels of Linguistic Analysis CS sites (e.g., Joshi 1985 and its sequel, the predictions of the theory with the data of the Null Theory of CS (SantorinilMahootian actual bilingual behavior. 1995); see also BentahilalDavies' Subcat- egorisation Constraint (1983)). Perhaps the 3.1. CS vs. borrowing most detailed mo-del involving the contrast It is uncontroversial that CS differs from the between lexical properties and functional other major manifestation of language con- (or "system") morphemes is the Matrix tact: lexical borrowing. Despite etymological Language Frame model (Azuma 1993; identity with the donor language, estab- Myers-Scotton 1993). Here, structural con- lished loanwords assume the morphological, straints on CS result from a complex inter- syntactic, and often, phonological, identity action between a dominant matrix language of the recipient language. They tend to be -and the prohibition against embedding "sys- recurrent in the speech of the individual and tem" morphemes from the "embedded" lan- widespread across the community. ,The stock guage in matrix language structure. of established loanwords is available to The assumption that bilingual syntax can monolingual speakers of the recipient lan- be explained by general principles of mono- guage, along with the remainder of the re- lingual grammar has not been substantiated. cipient-language lexicon. Loanwords further While such formal theories at grammar may differ from CS in that there is no involve- account well for monolingual language ment of the morphology, syntax or phonol- structure (including that of the monolingual ogy of the donor language. fragments in CS discourse), there is no evi- dence that the juxtaposition of two lan- 3.2. Borrowing vs. nonce borrowing guages can be explained in the same way. As Recent research on borrowing as a syn- described in ensuing sections, bilingual chronic process (e.g., the papers in Poplackl communities exhibit widely different pat- Meechan 1998a; Poplack et al. 1988) has terns of adapting monolingual resources in shown it to be far more productive than its their code-mixing strategies, and these are result (established loanwords) would imply. nQt predictable through purely linguistic Crucially, the social characteristics of recur- considerations. The equivaltmce constraint, rence and diffusion need not be satisfied, re- as formalized by Sankoff (1998a; 1998b; sulting in what has been called, after Wein- Sankoff/Mainville 1986; Sankoff/Poplack reich (195311968), nonce borrowing (Sankoff 1981), is a production-based explanation of et al. 1990). Like its established counterpart, the facts of CS, which incorporates the no- the nonce borrowing tends to involve lone tions of structural hierarchy and linear lexical items, generally major-class content order, and accounts for a number of empiri- words, and to assume the morphological, cal observations in addition to the equival- syntactic, and optionally, phonological iden- ent word order characterizing most actual tity of the recipient language. Like CS, on switch sites. These include the well-formed- the other hand, particular nonce borrowings ness of the monolingual fragments, the con- are neither recurrent nor widespread, and servation of constituent structure, and the nonce borrowing necessarily requires a cer- essential unpredictability of CS at any po- tain level of bilingual competence. Distin- tential CS site. guishing a nonce borrowing from CS of a lone lexical item is conceptually easy but 3. Fitting theory to data methodologically difficult, especially when this item surfaces bare (i.e., morphologically Which of these competing (and often con- uninflected, or in a syntactic slot shared by flicting) models offers the best account of both languages), giving no apparent indi- bilingual CS? Testing the fit of theory with cation of language membership. the data of CS should be a straightforward The classification of such lone other-lan- matter; however, disparate assumptions, guage items is at the heart of a fundamental goals and domains of application have thus disagreement among CS researchers over 1) far hindered such efforts. Assessment of the whether the distinction between CS and bor- descriptive adequacy of a theory of CS rowing should be formally recognized in a requires that at least two methodological is- theory of CS, 2) whether these and other sues be resolved. One Involves identification manifestations of language contact can be and principled classification of language identified in bilingual discourse, and 3) mixing phenomena, the other, confronting criteria for determining whether a given 70. Code-Switching 591 item was switched or borrowed. Researchers parallels to those of their counterparts in the who classify lone other-language items as recipient language, while at the same time CS tend to posit an asymmetrical relation- differing from relevant patterns in the