Table of Contents
2. Code-switching as a Field of Research
2.1. The Object of Research
2.2. The Origins of Code-switching Research
2.3. The Research History
3. Major Explanatory Frameworks
3.1. Gumperz: Situation – Metaphor – We and They
3.2. Myers-Scotton and the Markedness Model
3.3. Auer and the Conversationalist Approach
4. Language and the Internet
4.1. Computer-mediated Communication and Language Use
5. CS Research in Computer-mediated Settings
5.1. Paolillo and the Indian Diaspora
5.2. Georgakopoulou and Greek-English E-mail Discourse
5.3. Androutsopoulos and German-Based Web Forums
5.4. Hinrichs: English and Jamaican Creole in E-mail Communication
6. Theory in Practice
6.1. The Corpus
6.2. The Users
6.4.1. Prior expectations versus actual language use
6.4.2. Conversational loci and linguistic structures susceptible to CS
6.4.3. Referential code-switching
6.4.4. Directive code-switching
6.4.5. Metaphoric code-switching
6.4.6. 'Poetic' code-switching
6.4.7. Expressive code-switching
6.5. Summary of findings
9. Appendix – The Data
10. Appendix – German Summary
It has now been about a decade and a half since the Internet and the World Wide Web have come to represent a major realm of research in various fields of linguistics. This is of course largely due to the fact that they offer easy access to a massive and unlimited amount of language data, which do not have to be transcribed in arduous ways as is the case with speech recordings. However, alongside this major cause of attraction, and despite the overall dominance of English, it is also the multilingual nature of the Internet which has naturally sparked the interest of bilingualism research as well as language contact research. It is the choice of and the switching between the available codes of the users' repertoires which mark a major topic of interest, and which shall be explored in the thesis at hand. The final focus of investigation will be the communicative functions and meanings of the phenomenon called code-switching (CS) as it naturally occurs in a Canadian-Croatian discussion forum.
In order to prepare the theoretical ground for the analysis of my own corpus of computer-mediated communication (CMC) data, the first part of this research thesis will define the subject-matter and the origins of research into it, and a general overview of the classic linguistic treatment of the phenomenon will be given. Following, this paper will illustrate the major concepts and approaches relevant to the final purpose of the thesis, and point out potential critique of their assumptions. In order to establish a link between code-switching phenomena in general and their communicative setting in the conducted study, the subsequent chapter will address several issues in computer-mediated communication research. Thereafter a review of renowned studies of code-switching phenomena in computer-mediated communication will be provided.
Finally, the Croworld corpus, including its compilation and structure, and the backgrounds of the users involved, will be presented. My subsequent analysis of selected forum posts will aim to relate the tenets of the classic approaches, as well as the insights of the CMC studies regarding the significance of code-switching practices.
2. Code-switching as a Field of Research
The following chapters serve to prepare the ground for the subsequent theoretical treatise. They present an advance clarification of several relevant terms and concepts.
2.1. The Object of Research
The title of this paper features the term code-switching for the phenomenon under scrutiny. Everyone with a little linguistic knowledge is roughly familiar with the basic concept behind the term. However, once you delve into the relevant academic literature, you soon find that there is no such thing as a generally accepted definition for code-switching. Rather, we are dealing with a highly ambiguous label that is generously applied by some, and deliberately shunned by others where the same linguistic occurrences are concerned. Celso Alvarez-Cáccamo depicts the term's career as follows:
From its origins in the physical sciences [...] until its current circulation in political anthropology [...], the notion of 'code-switching' has experienced the characteristic multiplication, fragmentation and metamorphosis that a conceptually rich term is prone to experience [...]. The increasing lexicalisation of the expression (from 'switching code' to 'code-switching') indexes its central place in academic fields dealing with so-called bilingual behaviour. (Alvarez-Cáccamo 1998: 29)
An initial sign of the term's ambiguous nature is that not even the spelling is agreed upon and consistent: besides code-switching, which will be used for this paper, there exist the forms codeswitching as well as code switching, with no distinction of meaning intended. According to Lesley Milroy and Pieter Muysken (1995: 12) there have been deliberate attempts within a Research Network of the European Science Foundation (ESF) to standardise the terminology that has evolved around the concept of code-switching, resulting in the insight that it is an impossible task. On this account, the reader is left with a confusing range of terms describing various aspects of the phenomenon. What causes the confusion is the fact that with some terms there is a partial overlap in meaning, while other particular terms are used in different ways by different authors. In what follows, an overview of the most salient or prevalent terms and their referential scopes will be given.
The term code originally comes from the fields of communication theory and semiotics (cf. Encyclopedia of Language and Linguistics, Vol. 2, 1994). According to a dictionary definition, code is “a system of words, letters, numbers or symbols that represent a message or record information secretly or in a shorter form” (OALD). Within the field of sociolinguistics, particularly during the late 1960s and 1970s, the term code became popular by Basil Bernstein's deficit hypothesis and the associated elaborated and restricted codes. However, with respect to the concept of code-switching, in sociolinguistics it has generally been used as a neutral label for any communication system involving language, thereby subsuming the more specific notions of dialect, language and variety. Along these lines, Carol Myers-Scotton mentions code as referring to “linguistic systems at any level, from separate languages to dialects of a single language to styles or substyles within a single dialect” (Myers-Scotton 1998c: 3). However, there have equally been voices which caution against such a simple equation and stress the need to keep the notions of code and speech variety apart (cf. Alvarez-Cáccamo 1998: 37f).
One of the earliest definitions of code-switching was rendered by Uriel Weinreich in his description of bilingualism as “the practice of alternately using two languages” (Weinreich 1953). This is what one could call the broad definition, which, so to speak, leaves it at indicating the juxtaposition of elements from two or more languages or dialects. In the course of the development of CS theories, each researcher added some details to this definition, thereby specifying it for their own purposes. Consequently, disagreement on criteria such as the following arose: the length of the juxtaposed utterance (whether a whole discourse, a single word, maybe even containing morphemes from two languages, or something other in between the poles of that spectrum), the density of switches in a given spoken or written text, whether the switch is an instance of an individual type or rather of a common type, whether or not it bears social significance, and if so, of what nature (cf. McCormick 2001). Disagreement as to the length of a code-switched utterance yielded the subcategories termed intersentential code-switching and intrasentential code-switching. Intersentential CS occurs at sentence or clause boundaries, while intrasentential switches appear within a clause and may take the form of a phrase, a single word, or may also cross morpheme boundaries (cf. Mahootian 2006). There have been researchers relating intrasentential CS to 'coordinate bilinguals' (who learned their languages in different times or contexts) and intersentential CS to 'compound bilinguals' (who learned both languages at the same time/ in the same context). However, there are also other views, which hold that most bilinguals show features of both types of switching (cf. Montes-Alcalà 2001: 197). Some researchers add tag switching as a separate third category, which they consider independent from intersentential switching.
While code-switching is the more well-established term, language alternation is another term to be found in use. It is defined as an umbrella term for those instances where two languages are used functionally and locally in a given interaction. It is more or less bound to a particular explanatory framework, as will become clear later.
According to Rajend Mesthrie (2001: 442), code-mixing is a fluid term which overlaps with code-switching and mixed code, but can still be distinguished from them in certain respects. His description is somewhat vague when he states that “prototypically code-switching is relatively 'clean' while mixing is relatively 'ragged'”. I take it from his elaborations on 'clean' and 'ragged', though, that he refers to the rather usual matching up of code-switching with intersentential switching and of code-mixing with intrasentential switching. In the scientific community, however, there is no consensus on whether or not the terms switching and mixing are distinguishable, which entails their being used interchangeably by many authors, and as referring both to intersentential and intrasentential CS. Some researchers have launched another distinction recently, namely that between code-mixing as referring to the use of two or more languages at the discourse level and mixed code as referring to switches within clauses or words (cf. Mahootian 2006). As a link-up with the subsequent definition of borrowing it may be noted here that code-mixing can involve borrowing, but “is theoretically (if not always in practice) distinguishable from it” (Mesthrie 2001: 443).
A form that has spread from one language or variety into another is called a borrowing. It may take the form of a stem just as well as that of a whole phrase. The term is rather misleading, as it implies a temporary nature, which is the very opposite of what the term is supposed to convey (cf. Heath 2001: 433). A borrowing differs from a code-switched element in that it is 'nativised' into the target language to some degree or other, and that it is used by both bilinguals and monolinguals. The 'perfect' instance of a borrowing, since fully nativised and yet distinguishable at the same time, is that of which the adoption dates back to past centuries. For the English language we are here dealing with words of Latin, Scandinavian and French origin, such as 'expectation', 'skirt' and 'money'. They are instances of borrowing which fall into the loanword subcategory. However, the cases where borrowings and code-switches are not so clearly distinct quite possibly outweigh those instances of explicit borrowing. This is due to the fact that “borrowings may resemble code-switches in retaining a foreign status and/or a discernible internal structure, while code-switches often resemble borrowings in brevity (words, short phrases) and in being fitted into another language's syntax” (Heath 2001: 433). Furthermore, there have lately been scholars attempting to show that borrowing and code-switching are expressions of the same process, in the form of divergent points on a continuum (cf. Jacobson 1998).
The description of the above terms and their scope of meaning has made it clear that terminology in the domain of code-switching research is a delicate matter to discuss and that it renders comparison across studies a rather difficult undertaking. In summary, what should be emphasised at this stage is the universal fact that in the mode of speaking or writing labelled with the umbrella term code-switching, speakers who are proficient in two or even more languages opt for combining their bilingual or multilingual resources in their discourse, such that [both languages] share the responsibility for the speaker to get a given message across... (Jacobson 1998: 52)
What we should also keep in mind is that it is not the labelling that creates the confusion, but “the intrinsically gradient and fuzzy nature of the continuum” (Heath 2001: 433) we encounter when dealing with code-switching.
2.2. The Origins of Code-switching Research
In search of the very origins of the research branch of code-switching one can go back to the 19th century, when early linguists such as the German Hugo Schuchardt engaged in the study of language contact. In his treatise on Slavo-German and Slavo-Italian from 1884 he goes so far as to credit the study of language mixing priority over all other language phenomena of his time. He in turn refers back to yet earlier work composed by a scholar named Lucio in 1666 on the mixture of Croatian and Romance in Dalmatia on the basis of 14th century records (cf. Clyne 2004: 799). Schuchardt is considered a precursor of the major aspects of the field of language contact studies. Due to the historical orientation of 19th century linguistics, he and his colleagues looked into matters such as the influence of neighbouring languages on each other, language islands and the impact of a national language on long-standing ethnic minorities and immigrant languages (cf. Clyne 2004: 800). Within the realm of these subjects much attention was paid to lexical transference (borrowing), integration, and phonological and syntactic transference and their role in language change. Lexical transfer mostly appeared under such headings as 'curiosity' or 'evil threat to the purity of language'. The phenomenon of code-switching in the speech of bilinguals was first actually noted by Max Braun in 1937, who noticed that borrowed lexemes triggered switches between Russian and German. However, code-switching was to be an underrated phenomenon for still some time (cf. Clyne 2004: 801).
Eventually, in the early 1950s the famous work by Uriel Weinreich, Languages in Contact (1953), brought about a turning point in various respects. Weinreich was the first to aim at a systematic study of language contact, for one. Also, despite still standing in the structural linguistic tradition, he pointed out that one should not lose sight of the fact that it is the bilingual individual which is the ultimate locus of contact (Romaine 2004: 49), thereby identifying the need to include social and psychological aspects into the study of language contact and bilingualism (cf. Clyne 2004: 799). For this reason, Weinreich and the language contact research from the early 1950s are considered the early beginnings of sociolinguistics (cf. Clyne 2004: 803). What is more, and most relevant to the topic of the thesis at hand, Weinreich's work kindled interest in code-switching phenomena by such famous and oft-cited statements as the following:
The ideal bilingual switches from one language to the other according to appropriate changes in the speech situation (interlocutors, topics, etc.), but not in an unchanged speech situation, and certainly not within a single sentence. (Weinreich 1953: 73)
Besides language contact and bilingualism research, a further research trend contributed to the emergence of CS research, namely information theory with its notion of code as a mechanism for the unambiguous transduction of signals between systems. In the 1950s, first R.M. Fano looked at “[t]he information theory point of view in speech communication” (Fano 1950), and subsequently Roman Jakobson adapted the notion of 'switching code' to the change a speaker must accomplish in order to successfully interpret, or 'decode', another person's system, or to produce such a change themselves (cf. Alvarez-Cáccamo 1998: 30).
2.3. The Research History
As pointed out above, code-switching was not consciously observed in any field of research until the mid-twentieth century. When it finally was, in most cases the terminology had clearly negative connotations, even on the part of the (socio-)linguists most responsible for stimulating research in the field. The prevailing term for code-switching phenomena at that time was interference, a label which paints the picture of a deficient bilingual who switches between languages out of inability to finish a conversation in the language he or she entered. The aforementioned, oft-cited comment by Uriel Weinreich (1953: 73) reflects the contemporary commonly held view, and dismisses a great deal of common code-switching practice by stating that the ideal bilingual switches due to situational changes, but never in an unchanged situation and under no circumstances within a sentence. Along similar lines, Einar Haugen stated a couple of years later, that “[a]ny item that occurs in speech must be a part of some language if it is to convey any meaning to the hearer...The real question is whether a given stretch of speech is to be assigned to one language or the other” (Haugen 1956 as cited in Alvarez-Caccàmo 1998).
Some early treatments of code-switching phenomena stood in the tradition of the so-called allocation paradigm. A prevailing model within this framework is Joshua A. Fishman's domain model. This deterministic perspective considers linguistic choices to be predictable on the basis of the domain in which they occur, with domain being defined in terms of “institutional contexts and their congruent behavioral co-occurrences” (Myers-Scotton 1993: 49). Although Fishman's model does not explicitly require that exactly one code is to be associated with one domain, this is clearly the model's unmarked choice: “'Proper' usage dictates that only one of the theoretically coavailable languages or varieties will be chosen by particular classes of interlocutors on particular kinds of occasions to discuss particular kinds of topics” (Fishman 1972: 437, author's emphasis). For this and other early lines of research “it was (and is) crucial to understand which 'language' a bilingual was using at a given moment” (Alvarez-Caccàmo 1998: 32, author's emphasis).
What Alvarez-Cáccamo (1998) terms the interactional turn was a change in the direction of research towards a functional, interactional view of language behaviour in general. It emerged within the wider frame of the likewise evolving cross-discipline of discourse analysis. “Interactional sociolinguistics focus on how meanings emerge, are managed, and are negotiated through a variety of devices, processes and strategies that pervade interaction” (Schiffrin 2004: 95). For the code-switching phenomenon two essential questions rose: why does the bilingual code-switch and what exactly does he/she mean by this kind of linguistic behaviour? (cf. Jacobson 1998). The launching of this new type of (code-switching) research is generally attributed to the linguistic anthropologist John J. Gumperz, who emphasised the need to consider the social function of language use. While he conducted research in this direction all through the late 1950s and the 1960s, as for instance with 'Hindi-Punjabi code-switching in Delhi' (1964), it is his 1972 exploration of Social Meaning in Linguistic Structure: Code-Switching in Norway (together with Jan-Petter Blom) which was and is considered to be seminal. As will be seen in due course, the basic ideas Gumperz developed concerning CS are to a considerable extent still valid and do still figure in contemporary research.
In general, “[t]he two functions most discussed in the literature until the early 1980s were the referential and the expressive functions of code switching” (Mahootian 2006: 515). The referential function refers to code-switches which are mainly due to lexical gaps, lack of fluency about a particular topic in one language, or failure of lexical retrieval. The listed motivations mark challenges to the notion of balanced bilingualism (cf. Mahootian). The expressive function of code-switching is supposed to act out meta-communication, where the form of mixed speech discourse itself “is a comment about the speaker rather than the speech” (Mahootian 2006: 515).
In 1980, Shana Poplack initiated the grammatical line of CS research through the pioneering article 'Sometimes I'll start a sentence in English Y TERMINO EN ESPANOL'. The underlying theory of her hypothesis is generative syntactic theory. Poplack and various researchers along with her claim that there are universal rules for the generation of sentences which are not even violated within a code-switched environment. Their aim is to show that CS is not to be considered aberrant behaviour, but rather proof of linguistic competence in bilingual subjects.
Carol Myers-Scotton, another well-known figure in code-switching research, belongs to this grammatical research line on the one hand, under the heading of which she developed the Matrix Language Frame Model (MFL). The model divides the two languages of a code-switched stretch of speech into the matrix language, which renders the morphosyntactic frame for the embedded language. Myers-Scotton claims predictability of CS and the existence of syntactic restrictions for code-switching, two assumptions which have given rise to sociolinguistic protest. On the other hand, however, Myers-Scotton has also contributed to the sociolinguistic and interactional approaches herself by another model, namely the Markedness Model (MM). This particular model she understands as “an explanation accounting for speakers' socio-psychological motivations when they engage in CS” (Myers-Scotton 1993: 75). She elaborated the concept of marked versus unmarked language choices, which attributes purposeful - albeit unconscious - behaviour to speakers' decisions, as will be seen in more detail below.
Another line of research within the interactional paradigm, which was largely established by Peter Auer, applies the methodology of conversation analysis to filter potential meaning and functions out of interactants' subtle moves in conversation.
3. Major Explanatory Frameworks
The following chapters provide an overview of three major contributors to classic CS research, of which the object of analysis is face-to-face interaction. The chapters are titled after the most famous advocates of the respective approaches, but are still broad enough to include references to other fellow researchers. Each chapter features a positioning of the approach within its frame of contemporary scientific reference. In order to warrant readability of the table of contents, there are no further sub-sections within the following chapters. Yet for the sake of clarity, the text is structured by way of major keywords being indicated by headings in between paragraphs.
3.1. Gumperz: Situation – Metaphor – We and They
As mentioned above, the earliest of the approaches to be treated here was launched by the linguistic anthropologist John J. Gumperz between the late 1960s and the early 1970s. The alignment of his work has various labels, but it is of course sociolinguistic in nature. Within sociolinguistics, Gumperz' work is assigned to the interactional research direction, of which Gumperz is also founding father, and which was shortly touched on above. Within interactional sociolinguistics in turn it is part of the approach called Ethnography of Speaking, and later Ethnography of Communication, of which Gumperz was one of the foremost supporters and which is ascribed primarily to Dell Hymes, a sociolinguist and anthropologist. Hymes therein called for a radical reorientation of research at the interface of anthropology and linguistics: he stressed the primacy of function over form and the recognition of the speech event as the basic unit of analysis, among other things. A major concern lies with how language use reflects and, moreover, to a considerable extent even accomplishes social order (cf. Saville-Troike 2004: 109). Hymes favoured inductive research methods to address these questions.
The most prominent output of Gumperz' early work on code-switching phenomena he authored together with Jan-Petter Blom, a Norwegian social anthropologist. Blom and Gumperz studied naturally occurring verbal interaction in the small town of Hemnesberget in northern Norway. Contrary to the vast majority of code-switching research which followed this seminal work, their study is not about bilingual switching in the (today) conventional sense, but about standard-vernacular alternation between the varieties of Ranamål (a northern Norwegian dialect) and Bokmål (the standard; language of official transactions, religion, and the mass media). Among the findings was that in spite of the linguistic similarity of Ranamål and Bokmål, the local people thought of them as distinct systems, which was mainly due to the differing social functions that the varieties fulfilled (local cultural identity versus formal education/official transactions, religion, mass media). The aim of their contribution in Hymes and Gumperz' Directions in Sociolinguistics from 1972 is to clarify the social and linguistic factors involved in the communication process and to test Bernstein's hypothesis by showing that speakers' selection among semantically, grammatically, and phonologically permissible alternates [...] is both patterned and predictable on the basis of certain features of the local social system (Blom and Gumperz 1972: 409).
One major claim of Blom and Gumperz' is that “linguistic alternates within the repertoire serve to symbolise the differing social identities which members may assume” (Blom and Gumperz 1972: 421). At the same time they caution against the notion of a “one-to-one relationship” between speech variety and social identity and they express the need to take into account additional contextual information. We will turn to identity again later, when Blom and Gumperz' we code/they code contras is depicted, but let me first introduce their famous dichotomy of situational and metaphorical code-switching.
In Blom and Gumperz' view, a switch of codes can be called situational when “within the same setting the participants' definition of the social event changes”, that is when participant constellation, topic, or mode of interaction alter, and when this change is signaled through linguistic clues, among others. In their Norwegian dialect/standard research situation, situational code-switches were marked by a shift from Ranamål to Bokmål and vice versa, and simultaneously by a change in channel cues such as speed, rhythm, amount of pauses and so forth. Blom and Gumperz stress the direct relationship between language and the social situation to the point that any violation of commonly accepted norms may entail the termination of the conversation or other social sanctions. Furthermore, they attribute to situational code-switching a clear change in the participants' definition of each other's rights and obligations (all Blom and Gumperz 1972: 424). Gumperz later compared this type of clear code alternation to Fishman's above mentioned domain model with its compartmentalisation and habitual kind of language use (Gumperz 1982: 61). He made this comparison in his Discourse Strategies, which were published a decade after his seminal work with Blom. Therein he also dedicated a considerable part to code-switching, more precisely, to metaphorical code-switching. In this 1982 publication, Gumperz offers some structural constraints and extends his earlier ideas. He does not elaborate greatly on situational CS, but he does add the afterwards controversially absorbed statement that “[t]here is a simple, almost one to one, relationship between language usage and social context” (Gumperz 1982: 61), which means that language choice can be predicted from situational parameters.
In contrast to situational code-switching, metaphorical code-switching does not involve a fundamental change of the speech event, which also means that there is no shift in the definition of the rights and obligations. The code-switch here “allows for the enactment of two or more different relationships among the same set of individuals” (Blom and Gumperz 1972: 425). Those different relationships are alluded to through the temporary use of a different code which serves as a metaphor for another social relationship that is generally associated with it. Blom and Gumperz call it semantic effect:
The semantic effect of metaphorical switching depends on the existence of regular relationships between variables and social situations [...]. The context in which one of a set of alternates is regularly used becomes part of its meaning, so that when this form is then employed in a context where it is not normal, it brings in some of the flavor of this original setting.” (Blom and Gumperz 1972: 425)
What the quote basically comprises is that the 'almost one-to-one-relationship' between code choice and situational factors in situational CS serves as a point of reference for the interpretation of metaphorical CS; or phrased differently, that it violates the one-to-one-relationships and thereby induces conversational implicatures. In his 1982 publication, Gumperz renamed the phenomenon conversational code-switching, a term that was soon to be broadly adopted across the research community. What he elaborates on in more detail at that point is the notion of automaticity in the selection of linguistic alternants. It is his assumption that the participants in an interaction are usually unaware which code is used and are not able to consciously recall their choice. What they unconsciously do according to Gumperz, however, is that they convey metaphoric information about how an utterance is to be understood by building on general knowledge about situational norms (Gumperz 1982: 61). And it is also in the Discourse Strategies that with regard to metaphorical CS, Gumperz refers to a bilinguals' awareness of alternative communicative conventions as “a communicative resource”, “which can be built on to lend subtlety to what is said” (Gumperz 1982: 65, 69).
We code versus They code
Under the heading of conversational code-switching, Gumperz then draws up a second dichotomous concept: that of two given codes or languages as we code and they code, respectively. His proposition holds that bilinguals' two codes or languages at a general level “directly reflect or signal the contrasting cultural styles and standards of evaluation which they encounter in daily interaction” (Gumperz 1982: 66). An ethnically specific minority language according to his model tends to be regarded as the we code and in people's minds is linked with in-group and informal activities, while a majority language is looked upon as the they code, with which the common associations are: more formal, stiffer, less personal and out-group-related. Yet, it seems to have been a matter of considerable concern to Gumperz to point out at the same time that the connection between communicative code and group identity, as drawn in the we code/they code contrast, is to be understood as a symbolic one, and that it is by no means supposed to directly predict actual usage, but for a minute number of interaction types, such as talk to older monolinguals, small children, or in extremely ritualised interaction. Outside of those restricted circumstances, options are diverse, and Gumperz as for the interpretation of utterances refers to discourse context, social presuppositions and speakers' background knowledge (Gumperz 1982: 66). With reference to Susan Gal's study of Hungarian-Austrian code-switching (1979), Kathryn Woolard depicts as a classic metaphorical switch a scenario where speakers of a minority language switch temporarily to the majority language to win an argument. She argues that in such cases, what happens is that “the dominant language metaphorically bestows greater authority on the speaker” (Woolard 2004: 77, my emphasis). In a similar manner, in Gumperz' model anger, distancing, softening, a joking tone and intimacy are all considered to be achieved through the contrast of we and they codes.
Having described the broader meanings of code-switching phenomena by setting up a dichotomy of types first, and another dichotomy concerning the two cultures involved, Gumperz moves on to fine-tune his approach by listing six roughly divided conversational functions of CS. Since the principles of interpretation in speakers seem to be of a tacit understanding by their nature, Gumperz relies on discourse analysis to isolate the functions, which are the following:
- Quotations – either as direct quotations or as reported speech
- Addressee specification – the switch serves to direct the message to one of several possible addressees
- Interjections – the switch serves to mark an interjection or sentence filler
- Reiteration – a message is repeated in the other code either literally or in somewhat modified form; may serve clarification and elaboration, amplification or emphasis
- Message qualification – qualifying constructions such as sentence/verb complements or predicates following a copula
- Personalisation versus objectivisation
The last group is somewhat blurred in definition. Gumperz here relates the code contrast to the following opposition pairs: talk about action versus talk as action, degree of speaker involvement in versus distance from a message, reflection of personal opinion versus objective fact, reference to specific instances versus general truth (entire account as depicted in Gumperz 1982: 75ff). As to the listing of functions, Gumperz does not claim exhaustive coverage, but he points out the usefulness of such a listing for discussing the data with the subjects themselves. Once he obtained participants' agreement on an interpretation of a switched passage, he felt to be able to deduce that this identical judgement must generally be due to similar linguistic perceptions. This then led him to explore CS in terms of his contextualisation cues, which give rise to those perceptions (Gumperz 1982: 82). It constitutes an attempt of Gumperz' to incorporate the phenomenon of code-switching into his more general theory of conversational inference.
CS as a Contextualisation Cue
In his elaborations of contextualization conventions in the above mentioned 1982 collection of essays, Gumperz' starting point is the hypothesis that any utterance offers various ways in which it can be understood and that people arrive at decisions about how to interpret utterances by their definition of what is happening at the time of interaction: “[T]hey define the interaction in terms of a frame or schema which is identifiable and familiar ”. The frame alone does not create meaning, but it “constrains interpretations by channelling inferences so as to foreground [...] certain aspects of background knowledge and to underplay others” (Gumperz 1982: 130, author's emphasis). This channelling of interpretation is effected by so-called contextualisation cues, the use and perception of which occur mostly habitually. Gumperz defines them as “any feature of linguistic form that contributes to the signalling of contextual presuppositions” (Gumperz 1982: 131). As potential linguistic realisations he gives prosody, choice among lexical and syntactic options, formulaic expressions, conversational openings, closings and sequencing strategies and, not least, code-, dialect- and style-switching phenomena. Gumperz thus puts the signalling of contextual information by means of a code-switch on a level with the signalling conveyed through prosody or syntactic and lexical processes in monolingual utterances, being the first to point out the important link between CS and monolingual stylistic choices. With regard to his metaphorical CS then, the signalling value consists in a shift in contextualisation cues which does not co-occur in any depending way with a shift in topic.
Gumperz' Contribution to CS Research
Blom and Gumperz' as well as Gumperz' elaborations rather form compilations of numerous ideas than one solid argument, which the above account certainly reveals. However, if it had not been for the two anthropologists, most of the work that the CS research branch can look back on today would presumably not have been done at all, as it was they who first spread the insight that code-switching is nothing less than skilled communicative behaviour which can convey social meaning and can help accomplish interactional goals. Also, with their interactional model, they were the first to offer an alternative approach to the aforementioned allocation paradigm, doing pioneer work for microlinguistic analysis. At the time of Blom and Gumperz' project in Norway, sociolinguists such as Joshua Fishman and William Labov were studying order in language use solely at the macro-level of societies, suggesting that linguistic choices are idiosyncratic at the individual level, and need to be quantified across groups for any patterns to emerge. Blom and Gumperz' ambitions were geared towards proving that variation found in individual conversation is not arbitrary either (cf. Myers-Scotton 1993: 51ff).
Critique of Gumperz' notions
The concept of situational code-switching has marked one focal point of criticism expressed among CS researchers. It is the explicit absence of any volitional action on the part of the speaker which many researchers have not agreed with and which has rendered it an outdated model. It has been argued that “even code selections which are made in reaction to a change in setting, topic, or addressee involve some element of speaker volition” (Hinrichs 2006: 45), as it is eventually any speaker's choice whether or not they meet the conversational expectations. Resulting from this is the rejection of a conceptual division between situational and metaphorical CS, as there is no need for a dichotomous model when all instances of code-switching are assigned intentionality, albeit of differing degrees. This is also where Peter Auer's critique sets in: he points out that in the course of any given situation there is plenty of opportunity for a renegotiation of parameters and that therefore Gumperz' model “is based on an odd, too confining conception of 'the situation' – something like a cage, chosen by the co-participants themselves, but inescapable as soon as the cage door is locked (i.e., the situation defined)” (Auer 1984b: 90). Regarding the we code/they code contrast, there is doubt among researchers whether CS “always signals such a macrosocially informed contrast in identities” (Woolard 2004: 77). These scholars also argue that Gumperz' 'minority-within-majority-setting' is only one of many possible versions of bilingual communities. And where there are researchers in approval of the notion of distinct in-group versus out-group values, they have sounded a note of caution on a priori presumptions about which code is the we code and which the they code respectively, since this can easily change across generations, lifespans as well as contexts, and since a code may even be 'refunctionalised' (cf. Woolard 2004: 78).
3.2. Myers-Scotton and the Markedness Model
Carol Myers-Scotton's model for code-switching is the most comprehensively stated contribution to the study of social motivations for code-switching phenomena. Although the following account will mainly draw on her 1993 version of Social Motivations For Codeswitching – Evidence from Africa, I do think that as for chronology as well as alignment, this chapter best fits in between the ones on Gumperz and Auer, since she already engaged in research in the matter as early as the 1970s. Just as Gumperz' approach, Myers-Scotton's is termed an interactional one, and it is furthermore considered as complementing more general social psychological approaches (cf. Sachdev and Bourhis 2001). This is also obvious from the build-up of the above mentioned publication, which features as motivations, or background respectively, an extensive account of prominent themes from the disciplines of the sociology of language, pragmatics, social anthropology and linguistic anthropology, on which Myers-Scotton based the subsequent depiction of her code-switching model. I will here proceed similarly, first pointing out those themes of the study of language in use highlighted by Myers-Scotton, which I find particularly salient with regard to the paper at hand, then giving an account of the Markedness Model (MM).
In consideration of the above discussed terminological diversity in CS research, it may be worth mentioning that Myers-Scotton uses the terms codeswitching and code selection or choice in turns, and as for her own model does not give distinctions of any kind.
To begin with, Myers-Scotton reflects upon precursors to her model, here being of interest her views on Gumperz' approach. She perfectly well agrees with Gumperz' proclamation that motivations for language use are to be sought not only in social identity and situational factors, that is in what she calls the stable factors, but just as well, or partly even rather, for that matter, in the dynamic factors, namely in the possibility of “conveying (intentional) meaning of a socio-pragmatic nature through code choice” (Myers-Scotton 1993: 57, my parentheses, since Gumperz himself does not explicitly use the term 'intentional'). In her critique of Gumperz, Myers-Scotton calls his work “primarily descriptive” and disapproves of his and many fellow researchers' tendency to produce what she calls “open-ended taxonomies”, that is listings of (stylistic) functions of the CS phenomenon (Myers-Scotton 1993: 59). The taxonomists she also relates to the insistence on looking at each interaction individually and on its own terms, while she herself thinks it vital to “believe in the possibility of generalizing across interactions” (1993: 59), in order to allow for explanatory models of a more general kind to emerge.
Prerequisites to and Premises of the Markedness Model
Before stating the actual model, Myers-Scotton establishes several basic conditions. For a start, participants in any given conversation are understood to share an (unconscious) awareness that their interaction partners' expectations regarding various aspects of the communicative event more or less meet their own (Myers-Scotton 1993: 75).
A second assumption, which encompasses various concepts to be discussed in succession, is that of interactants acting purposefully when engaging in conversation. Regarding the question of the degree of consciousness involved, 'goal-oriented' may be a less ambiguous description, since Myers-Scotton's participants in conversation act unconsciously, just like Gumperz' do. The goal-oriented interactant is called rational actor by a number of researchers in speaker motivations. This social psychological framework considers social interaction as a process of exchange and uses economic metaphors to describe the matter: rationality, the linchpin of the entire communicational event, directs interactants to perform a cost-benefit analysis of given alternatives and to finally make choices which optimise their rewards (cf. Myers-Scotton 2000: 1261), that is choices which yield the best possible returns. Myers-Scotton incorporated this mindset into her CS model in the early 1970s, and then kept developing its consequences in the following decades. From the beginning, her rationale has been that “a major motivation for using one variety rather than another as a medium of an interaction is the extent to which this choice minimizes costs and maximizes rewards for the speaker” (Myers-Scotton 1993: 100). Myers-Scotton's speakers thus engage in code-switching practice as the result of their calculation showing that the aimed for rewards will be greater with an utterance involving a switch than with a monolingual one.
Howard Giles' speech accommodation theory (or, more recently, communication accommodation theory) similarly ties in with both the speaker as rational actor and the cost-benefit analysis. Myers-Scotton hence includes the concept in her markedness model in the sense that interactants achieve accommodation or convergence with and divergence from their conversation partners by their code-switching practice.
Another premise of Myers-Scotton's model, which takes us one step further towards the actual markedness model, is the picture of several sets of rights and obligations (RO), which are thought of as abstract constructs representing the attitudes and expectations that interactants hold towards each other. She shares that aspect with Gumperz, though in comparison he did not elaborate greatly on the concept. In Myers-Scotton's version, any code choice is indexical of a corresponding RO set between interactants in a given type of interaction (Myers-Scotton 1993: 84). The switching between codes then is according to her a clear signal of two possible scenarios: (1) the speaker is either unwilling to, or uncertain whether or not to, commit to any single RO set between interactants, or (2) the speaker seeks to change the established RO set, which makes CS a negotiation for a new set of rights and obligations (Myers-Scotton 1993: 6f). When in the complex course of language acquisition speakers come to know the community's available RO sets, they simultaneously form a sense of indexicality of code choices for these sets (Myers-Scotton 1993: 88). As for the source of the RO sets, Myers-Scotton describes them to derive from situational features (relative statuses of the speakers, topic, setting etc.). This statement has entailed that a couple of critics put her model on one level with Gumperz' situational switching, and only with his situational switching. In a later publication she thus felt the need to clarify that for the RO sets to be derived from situational factors does not at all mean that the situational factors per se predefine code choices. She allows for speakers' awareness that situation, RO set and code choice are somehow linked, yet at the same time insists that this awareness is only one of several ingredients that make up the optimal choice (Myers-Scotton 2000: 1263).
A final premise now concerns the main constituent of the markedness model: markedness is a concept that originated from the Prague School of linguistic thought in the first half of the 20th century and entered structural linguistics as a phonological concept. For its – meanwhile – diverse applications to various kinds of linguistic as well as cultural phenomena the general meaning of marked “is characterized by the conveyance of more precise, specific, and additional information than the unmarked term provides” (Roman Jakobson as cited in Woolard 2004: 80). However, Myers-Scotton's understanding of markedness does not match that of the Prague School. She equates what she terms the unmarked choice with the expected medium, and accordingly the marked choice with the unusual medium in conventionalised interaction (Myers-Scotton 1993: 89). Myers-Scotton's premise is that all interactants are equipped with a so-called markedness metric, which she adds to a general concept of linguistic competence (1993: 79) and which allows for interactants to assess each and every code choice as more or less marked or unmarked. Since Myers-Scotton (1993: 109) argues that in spite of an intrinsic markedness measurement, it is through practical experience of language use in a specific community that readings of markedness occur, she sees a normative framework as inextricably linked with the notion of marked and unmarked choices. Remove it, and an interpretation of choices will be rendered unfeasible, as interactants would be lacking any kind of script or schema to orient to, such goes her argumentation.
 Although admittedly only to the (historical) linguist's eye.
 Fisman's statement rules out intrasentential CS as well as conversational CS, which will be explained in chapter 3.
 Gumperz shares this role with Erving Goffman.
 Basil Bernstein's hypothesis holds that social relationships act as intervening variables between linguistic structures and their realisation in speech (cf. Blom and Gumperz 1972).
 Gumperz had already finished the Discourse Strategies in 1976, though.
 A notion taken from language philosopher Herbert Paul Grice's Theory of Implicature.
 Gumperz later preferred to depict situational and metaphorical CS as two points on a continuum, however, rather than as two contrasting types (cf. Woolard 2004: 76).
 A notion which Gumperz took from the sociologist Erving Goffman.
 Among them Jon Elster since the late 1970s and Lawrence Lessig in the 1990s.