Psycholinguistic, metalinguistic and socio-psychological accounts of code-switching: a comparative analysis of their incidence in a large Croatian-English sample/Psiholingvisticke, metalingvisticke i sociopsiholoske motivacije za prebacivanje kodova: komparativna analiza njihove ucestalosti u govoru Australaca hrvatskog podrijetla.
Uriel Weinreich's contention that "[t]he 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" remained unchallenged for a short time only (1953, p. 73. Round brackets his). Since this assertion, models based on specific data sets and particular perspectives have been developed to account for why code-switching can and does occur. These are briefly discussed in section 2.0 of this paper. The third section introduces the data sample of 100 recorded interviews with Croatian-English bilinguals whose speech is predominantly Croatian. In the following sections, approaches which examine psycholinguistic, metalinguistic and socio-psychological features are presented consecutively. This paper examines the theoretical basis and explanatory power of each approach and applies them to this large, quantitative sample. These approaches have customarily been applied to smaller, qualitative samples: this paper seeks to test their amenability to a large corpus. This paper seeks to demonstrate which models that account for the incidence of code-switching have explanatory power in relation to a large number of code-switches. Based on a large sample, this paper seeks to provide an answer to the question: which of the following features--psycholinguistic, metalinguistic and socio-psychological --are most frequently located in speech containing code-switching?
2. Code--switching and accounts for its incidence
While the bulk of studies that examine code-switching focus on its grammatical and structural features, research in contact linguistics has periodically also focussed on speakers' apparent motivations for language alternation. Not long after Weinreich's (1953) contention, studies by Clyne (1967) and Gumperz (1976, 1982) document mid-sentence code-switching and attempt to offer explanations for its incidence. In Clyne's German-English and Dutch-English corpora, code-switching frequently occurs due to the (momentary) psycholinguistic state of speakers' (in-) ability to distinguish or select words from their own discrete or combined mental lexica when accessing or retrieving homophonic items, or those words with ambivalent membership. Psycho-linguistically-focussed accounts of code-switching are of relevance in bilingual situations involving typologically closely-related languages and/or where storage of items is likely to be shared rather than separated in speakers' lexica. 'Triggered' code-switching of this type is located and discussed in further studies on code-switching (eg. Zentella, 1997, Gregor, 2003).
Clyne's concept of 'triggering' as an overt catalyst for code-switching is based on his notion of how 'available' each language is for bilinguals. The notion of 'availability' has been taken up by others in discussions of how 'activated' and/or 'selected' speakers' language varieties are (Treffers-Daller, 1997, Grosjean, 2000) and the 'mode' that speakers find themselves in, due to environment. Sociolinguistic features such as interlocutor or setting are obvious and uncontroversial when bilinguals code-switch purely on the basis of specific addressee and the unmarked code usually employed with him or her. In fact, this type of code-switching may be re-termed a uni-directional shift in the language of interaction with specified interlocutor/s (cf. Fishman, 1967, Blom & Gumperz ,1972, Hlavac, 2010).
Overt activation, a psycholinguistic feature, is present in examples of code-switching which contain metalinguistic talk. In spontaneous conversation, comment from either language about choice of language or a contrastive and covert use of discourse markers from both languages (cf. Halmari, 1997) can occur which may draw attention to the discourse or "languaging" of text itself (Maschler, 1994:326). Metalinguistic talk as an overt indicator of code selection is distinguished here from examples which indicate that an informant has difficulty in the production of target speech forms. Diagnoses about proficiency are generally a peripheral or irrelevant factor in most accounts of bilingual data sets which presume and demonstrate speakers' ability to control and produce text in both languages (cf. Myers-Scotton's (2002:105) term 'classic codeswitching' practised by speakers who require a high level of proficiency in both languages to produce bilingual text). However, in recent years, the role of proficiency level and its influences on type and frequency of bilingual discourse has been taken up by Bullock and Torribio (2004) and Montrul (2008).
Building on Blom and Gumperz's (1972) notion of ' situational' code-switching, two models have developed which focus on the setting of interactions and socio-psychological features that pertain to it: Giles's Speech Accommodation Theory (Giles, Bourhis & Taylor 1977), later known as Communication Accommodation Theory (Giles, Coupland & Coupland 1991) and Myers-Scotton's Markedness Model (1993), later extended as the Rational Choice Model (2001). Both have similar approaches, but contrasting points of departure: Giles's theory is 'listener-centred' and employs the term 'convergence' to refer to processes by which fellow interlocutors adapt their speech so that there are more similarities and fewer differences between them. This is related to general socio-psychological processes which Giles and his colleagues see as a universal trait in human interactions: speakers adapt their speech patterns to others where they seek to win others' social approval or to increase the 'communicative efficiency' of the interaction. Myers-Scotton's model is 'speaker-centred' and top-down; that is, macro-societal linguistic conventions and universalist assumptions about rationally-based choices are considered determinant features in linguistic behaviour, and speakers' speech style is based on how they believe a particular style will best suit their purposes in communicative events.
This study does not go beyond socio-psychological features and does not explore the 'metaphorical' functions of code-switching (Blom and Gumperz, 1972) or discourse-generated phenomena and input from both two codes in performing certain speech acts or discourse-pragmatic functions, eg. Auer (1995), Li (2005). This study focuses on three psycholinguistic/socio-psychological models which are chosen as they locate common congruent psychological processes as a cognitive basis for individuals' code-switching behaviour. This study is novel in that it seeks to quantitatively apply these models to a large sample.
3. Informants and corpus
The corpus, on which the data of this paper are based, was collected from recorded interviews that the author conducted in Croatian with 100 young adult Croatian-Australians. Informants were aged between 16 and 32 (average age: 21), 50 were female and 50 male and 88 were born in Australia and the remaining 12 arrived in Australia younger than five years of age. Forty-two informants were university students; fifteen were professionals, twelve high school students; eleven were employed in administrative or clerical occupations and seven worked as labourers or in unskilled occupations. The author is a Croatian-English bilingual and made contact with respondents through relatives, common social networks, schools and cultural organisations. The we or unmarked codes for interactions amongst second-generation Croatian-Australians include English and English with insertional or alternational code-switching to Croatian with or without an emblematic function. Monolingual Croatian or 'Croatian-dominant' speech is a more highly marked code for intra-second-generational speech but one of the unmarked codes for inter-generational speech with first-generation speakers. Informants' knowledge that their interviewer was bilingual meant that there was no barrier for them to employ English-origin forms in their discourse.
No attempt was made to test proficiency in either language and English was presumed to be the dominant language and Croatian (although chronologically their first-learnt language) the non-dominant language of informants (Hlavac, 2003:338-347). The recordings with informants consisted of a loosely-structured interview which contained twelve questions relating to place of residence, family circumstances, schooling experiences, occupational experiences/ambitions, leisure activities, recounting the content of a film recently seen, travel experiences and the telling of a joke/anecdote. In addition, informants were shown two pictures: a 'typically' Croatian one of villagers dressed in national dress dancing kolo; a 'typically' Australian one of holidaymakers on Bondi Beach. The same questions asked to all informants allows some uniformity and comparability across the sample. The shortest recordings went for 20 minutes, the longest for 2 hours--many informants conversed freely about topics beyond the 12 questions put to them. All interviews consist of dyads only and most informants were interviewed in their homes. For most, the initial nervousness of the interview situation passed quickly and most appeared to be able to converse freely. Fourteen informants were previously known to me while the remaining 86 were unknown. This relationship differential is further explored only in section 7.0 below where differences in means of contacting informants are examined. Elsewhere in this paper, no distinction is made between informants according to familiarity with the researcher/author.
A 15 minute segment was chosen from each interview and transcribed according to Croatian orthography. English-origin forms, where they appear in otherwise Croatian text, are in capitals. Those English items which are the centre of focus in particular examples are also underlined. English glosses are non-literal but the ordering of items in English translations generally also resembles Croatian SVO word order. Names that appear in examples are pseudonyms. Data on each informant are given in round brackets after each utterance. The first number refers to informant number, 'M' or 'F' refers to gender while the last number indicates the informant's age, eg. '73,M,21' signifies: informant number 73, male, 21 years old.
Table 1 below presents the overall sample which shows that 40% of the 5677 turns contain code-switching. Code--switches are defined here as the presence of lexemes contributed by two languages in one utterance. Code--switching or bilingual speech is a common phenomenon in this speech community as I expected. This indicates the following: code-switching is so frequent that its occurrence is relatively unmarked.
The vast majority of the 4,223 English-origin items--3,615 or 86%--are one-word code-switches. Single--item code-switches are included here as their occurrence can be closely related with that of multiple-item switches. To show which segments of the data sample I am referring to and to show how frequent multiple-items are, Table 2 below sets out the incidence of English-origin items in relation to their length and clause boundaries.
Three categories of code-switching are distinguished: extra-clausal, inter-clausal and intra-clausal. The first group is identified by its exclusively discourse-specific function:
(1) ..idemo vise puta na tjednu, YOU KNOW, SO.. svaku godinu nase, am.. YOU KNOW.. uveca se.. znas.. (5,F,17)
.. we go many times a week, YOU KNOW, SO.. every year our, um.. YOU KNOW.. it increases.. you know..
Inter-clausal code-switches are defined as code-switches between clauses regardless of syntactic (non-)equivalence of the two or more clauses, ie. co-ordinate, subordinate or relative clauses.
(2) .. ali kad imam zadacu ja cu to raditi HOWEVER LONG IT TAKES.. ja kao test radim.. (95,M,16)
.. but when I have an assignment I will do it HOWEVER LONG IT TAKES.. I do like a test..
Intra-clausal code-switches are identifiable according to clause-internal position of switch site, regardless of hesitation phenomena, pause-fillers or other elements which may accompany the code-switch:
(3) .. sam morao raditi kao, am.. ah.. WORK EXPERIENCE, trecu godinu i tamo sam.. (49,M,23)
.. I had to do like, um.. ah.. WORK EXPERIENCE, third year and here I was..
4. Psycholinguistically conditioned code-switching and metalinguistic code-switching
Psycholinguistically conditioned code-switching and metalinguistic code-switching are grouped together due to the feature of code activation or selectedness as an 'unconscious' feature for the former and as a 'conscious' process for the latter. Psycholinguistically conditioned code-switching refers to production of forms from a speaker's 'other' language through that language's activation and (momentary) selection due to cross-linguistic similarity of forms (eg. bilingual homophones) or through selection of language-specific referents from one language (typically proper nouns) that precipitate an alternation. Metalinguistic code-switching refers to overt features, (eg. explicit warnings, apologies) that a change in language has just happened or will happen.
Among the four types of triggering identified by Clyne (1967:84-89) are 'consequential', where a trigger precedes a code-switch, and 'anticipational' triggering in which the trigger word is anticipated and the code-switch precedes it. Figure 1 below demonstrates the former:
[FIGURE 1 OMITTED]
Example (4) below contains an instance of consequential triggering. The English-origin proper noun Ford is the trigger word which precedes a code-switch into English.
(4) Um.. po svjetlima, nije.. ja nikad nisam vidio ovakve svjetle u Australija, ja sam bio u Sydney i u Brisbane, bio sam po cijeloj Australi, um.. kako ljudi su se.. imaju robe na sebi, to vise nije kao australsko, umm.. auti, um.. oh.. cekaj to je u Australi, to ima kao FORD, YEAH GOOD OLD FORD STATIONWAGON, COMMODORE, to je u Australi, mislim.. (53,M,32)
Um.. because of the lights, it is not.. I have never seen these sorts of lights in Australia, I have been to Sydney and Brisbane, I have been all around Australia, um.. how the people have.. they are wearing clothes, that is more not like Australian, um.. cars, um.. oh.. wait that is in Australia, there is like FORD, YEAH GOOD OLD FORD STATIONWAGON, COMMODORE, that is in Australia, I think..
There are four English proper nouns in the example (4) above, Sydney, Brisbane, Ford and Commodore. Neither the occurrence of Sydney nor Brisbane precipitates switching, however that of Ford does. Recognition of a Ford car and the verbalisation of its name, is accompanied by two things. Firstly, the code-switch appears as a clear consequence of the form Ford. Secondly, cars belong to the semantic/thematic field of transport and technology, a field that is more likely to be spoken about in English in the vernacular of the informants.
Example (5) has a bilingual homophone trigger, "tennis".
(5) Imam puno zadaca i sutra mi igramo TENNIS.. THAT'S ABOUT ALL.. (38,M,18)
I have a lot of assignments and tommorow we are playing TENNIS.. THAT'S ABOUT ALL..
The item tennis is produced according to its English phonetic representation. The succeeding switch contains a topic-comment shift which terminates the turn and the code-switch could be motivated by the change in discourse style just as much as by homophonous tennis.
Figure 2 below schematically shows how anticipated triggering occurs.
[FIGURE 2 OMITTED]
The following example, (6), has two instances of anticipated triggering:
(6) Oh, OKAY.. sad.. ce mi bit jako tesko na hrvatski.. mogu radit, um.. I COULD BE LIKE A FOOD STYLIST, mogla to, i.. sto se jos moze.. jako, posao je tako VERY BROAD.. moze.. (24,F,18)
Oh, OKAY.. now.. it will be very hard for me in Croatian.. I can work, um.. I COULD BE LIKE A FOOD STYLIST, I could (do) that, and.. what else could I do.. very, the job is therefore VERY BROAD.. it's possible..
The first anticipated switch, 'food stylist' in (6) above is preceded by metalinguistic comment about limitations in Croatian control: .. ce mi bit jako tesko na hrvatski ('.. it will be very hard for me in Croatian'). Later, a second code-switch very broad occurs, triggered by broad. The Croatian equivalent jako ('very') is produced just before the switch which indicates that non-control of the referrent very does not account for the switch.
Example (7) below contains an example of a metalinguistic talk in English, preceding an integrated code-switch:
(7) YEAH.. zaboravila sam koji zadnji film sam gledala, um.. knjiga sto sada citam je.. um.. pisala Jane Austen, citam 'Emma', to zato sto ja cu to citat NEXT SEMESTER za engleski, pa hocu da ga sad procitam. To je od, kao jedna djevojka i ona voli da svakoga, DON'T KNOW HOW TO SAY IT, SETAPOVATI.. znas.. (3,F,19)
YEAH, I've forgotten the last film I saw, um.. the book that I'm reading is.. um.. was written by Jane Austen, I'm reading 'Emma', that's because I'll read it NEXT SEMESTER for English, so I want to read it now. It's about, how one girl and she likes to, DON'T KNOW HOW TO SAY IT, SET UP EVERYONE.. you know.. (3,F,19)
Table 3 below shows the number of code-switches that were motivated by production of English-origin items. Although these are usually single words, the code-switch that accompanies their election brings about a multiple word chunk in English.
Table 3 above shows that consequential is more frequent than anticipational triggering which is congruent to Clyne's (1967: 84-95) original findings for German-English data. Phonotactic dissimilarities between Croatian and English, and therefore, a comparatively lower number of potential bilingual homophones between the two languages mean that homophones are not a productive category responsible for triggering. Triggering is generally an infrequent phenomenon that may occur after single key words are elected. Those forms which are most likely to precipitate triggering, unintegrated English-origin nouns, rarely do so. Of the 962 unintegrated, intra-clausal, single-item code-switches, 686 (71%) are nouns, whether common or proper. Only 42 or 6% of the nouns co-occur with triggering. This indicates that triggering is a relatively uncommon phenomenon in this sample and that a small number of items such as proper nouns are the ones most likely to precipitate it.
As Table 3 above suggests, metalinguistic talk is a more frequent occurrence that accompanies code-switches than triggering. Examples (8) and (9) contain explicit markers of the code or the forms that informants are using:
(8) Ja volim uvijek bit van nesto radeci, znas.. ja sam KAKO ONI KAZU FIDGETY.. znas.. (83,M,24)
I always like to be outside doing something, you know.. I am WHAT THEY CALL FIDGETY.. you know..
(9) .. oni priznaju kao oni imaju ACCREDITED TRANSFER ILI ARTICULATION kako oni to zovu.. (74,M,22)
.. they recognise like they have ACCREDITED TRANSFER OR ARTICULATION how they call it..
Examples of 'introduced' or 'justified' metalinguistic talk as in examples (8) and (9) above occur 144 times (cf. Table 3 above). Although relatively infrequent, occurring in 3% of code-switching examples, metalinguistic talk is the third most common of the overt or 'flag' features that can surround code-switching, after lexicalised filled pauses and unfilled pauses (cf. Hlavac, 2011: 3779). The relative prominence of metalinguistic talk is congruent to Maschler's findings which record elements that draw attention to the discourse or "languaging" of text itself (1994:326).
5. Overt limitations in speaking a non-dominant language
Lower proficiency is rarely documented as a cause for code-switching: most studies examine data in which speakers have the ability to perform particular functions in each language without apparent difficulty. Levels of proficiency, where relevant, commonly reflect the functionally-specific settings of acquisition and use. Some structurally based studies on code-switching examine dominance and its effects on type and structure of code-switching, Poplack (1980), Sridhar and Sridhar (1980), Bentahila and Davies (1992), Lanza (1997) and Montrul (2008). In these studies which focus also on language communities experiencing a language shift speaker's relative dominance and non-dominance in the respective languages is treated as a sub-determining variable associated usually with generational membership.
Non-dominant proficiency in Croatian is explicitly expressed as the motivating factor for code-switching in a small number of cases only. Example (10) below, also containing metalinguistic talk, is an example of this:
(10) ... kako su zrtve, ovaj, ostavljene na milost i nemilost sudskog procesa, ili citavog sistema, ili kako.. ovaj.. (J.H.) ... how victims are, um, left to the mercy or lack of mercy of the legal process, or of the whole system, or how.. um..
YEAH, WELL, bas o tom govorimo, YOU KNOW, YOU KNOW.. koji je, YOU KNOW, koji je.. oh.. IT'S LIKE.. I CAN'T EXPLAIN IT, IT'S LIKE.. I CAN'T EXPLAIN IT, IT'S REALLY DIFFICULT, I JUST HAVEN'T GOT THE WORDS.. (smije se). (9,F,20)
YEAH, WELL, we're just talking about that, YOU KNOW, YOU KNOW.. which is, YOU KNOW, which is.. oh.. IT'S LIKE.. I CAN'T EXPLAIN IT, IT'S LIKE.. I CAN'T EXPLAIN IT, IT'S REALLY DIFFICULT, I JUST HAVEN'T GOT THE WORDS.. (laughter)
Examples such as that in (10) above are found in only fifteen of the 608 multiple item code-switches. Amongst these few examples are those which do not contain admissions of lack of proficiency but which contain overt discourse and morphosyntactic features indicating difficulty in text production.
(11) Ja sada radim to.. TERTIARY ENTRANCE.. u svi UNIVERSITY, um.. koji pruzaju drugi, cetiri.. osam.. ja isto imam.. LIKE.. pismo sto ja pisam.. i I SENT IT TO THEM THEN.. (88,F,17)
I'm now doing that.. TERTIARY ENTRANCE.. into all [pl.] UNIVERSITY, um.. which offer second [ord. no.], four.. eight [card. nos.] .... I've also got.. LIKE.. a letter that I'm writing to them.. AND I SENT IT TO THEM THEN..
Discourse and morphosyntactic features present in example (11) are comparable to examples of syncretism and morphological innovation present in the speech of Swedish-born Croatian-speakers (Durovic, 1983) and American-born speakers of Russian (Polinsky, 2008) and are in line with Montrul's (2008) description of default forms found amongst young speakers of immigrant languages in America. Although these features often co-occur with code-switching they are not a necessary pre-requisite for code-switching. Overall, there are 15 examples of expressed limitations in Croatian and 102 passages or examples that bear features similar to those in example (11) above that indicate limitations to informants' ability to express themselves freely. A larger number of code-switches can be attributed to this feature than psycholinguistically conditioned or sociolinguistic code-switching. The total number of code-switches that appear to be motivated by difficulties with proficiency is 117 or less than 3%. This sample shows that code-switching is a phenomenon that is rarely motivated by limitations to proficiency, whether overt or covert.
6. Convergence and divergence
Convergence and divergence are terms commonly used in Communication Accommodation Theory which refer to participant-related or listener-oriented code choices. As a process or postulate, Communication Accommodation Theory seeks to explain how and why modifications in interlocutors' linguistic behaviour occur to the effect that linguistic features become more or less similar to each other, ie. ".. the processes whereby individuals shift their speech styles to become more like that of those with whom they are interacting." (Giles & Smith, 1979:46).
Accommodation may be conscious or unconscious (cf. Backus, 1996:15) and is usually associated with intergroup dynamics (cf. Genesee & Bourhis, 1988). But accommodation can also be examined as a phenomenon within in-group settings as an inter-individual phenomenon. It may be that in almost all examples of spontaneous speech, accommodation as a linguistic representation is thought to be unconsciously or automatically determined. For this study based on semi-spontaneous speech I posit that accommodation, at least during the initial part of the interview, is conscious. This has the consequence on the linguistic form of speech to the effect that convergent linguistic forms, eg. mainly monolingual Croatian text, predominate. As Giles & Smith (1979) suggest, accommodation is very much the rule in most speech situations and is, in Backus's (1996) terms, 'unconscious'. This is not the case for non-accommodation.
Convergence and divergence are measured here primarily according to language choice. The author recognises that language choice in itself need not be a key determinant in informants' verbal behaviour and the things that they wish to convey and enact through speech (cf. Gafaranga's (2005) 'language-blind' conversationalist approach). Language choice is examined across the sample inasmuch as it is congruent or non-congruent to the requested language of the interview, Croatian or 'Croatian-dominant' speech. For many speakers, monolingual Croatian may not be an available option as a language variety because of their proficiency level in Croatian. Nonetheless, all 100 of the informants who are included in this data corpus are those who claimed to and who were able to 'communicate in a conversation in Croatian'. The following examples show excerpts from two interviews which are typical for the type of exchanges contained in the sample:
(12) Oh, jedan dan, ja sam bio na poslu i jedan dan, imamo jedan LUNCH-ROOM UPSTAIRS gdje su ti, Y'KNOW, mozete da jedete, i jedan dan ja sam pao DOWN THE STAIRS i jedan je bio.. glavni BOSS je bio tamo i, um, samo stanem i smijeham, jer ja sam pao na pod.. (56,M,21)
Oh, one day, I was at work and one day, we have a LUNCH-ROOM UPSTAIRS where there are those, Y'KNOW, you can eat, and one day I fell DOWN THE STAIRS and one was.. the main BOSS was there and um, I just stood there and laughed because I fell on the ground..
Je li ti pomogao ili nista.. samo se smijao? (J.H.) Did he help you or nothing.. he just laughed?
Oh, imam jedan drugi.. drugi BOSS, on je.. on je LIKE, te gleda, Y'KNOW, on samo te gleda gde mi radimo i ima, i on ima jedan sef ... (56,M,21)
Oh, I have another.. another BOSS, he's.. he's LIKE, he looks after you, Y'KNOW, he just looks after you, where we work, and there's, and he has a boss as well ...
Example (12) above contains a number of English discourse markers, you know, like and lexemes such as lunch room and boss whose appearance is predictable and unremarkable. The extra-clausal discourse marker code-switches 'punctuate' the informant's speech and their occurrence is congruent to the findings of other studies that focus on discourse pragmatic markers that report common frequencies. eg. Salmons (1990), Blankenhorn (2003), Hlavac (2006). Further, although representing 'other-language' forms, the 'convergent-like' function of them is apparent. You know is a marker of metaknowledge, either about what interlocutors share or about what is generally known (Schiffrin, 1987). Polyfunctional like can perform functions such as a dramatisation of internal feelings (Romaine and Lange 1991) or be a quotative verb (Tagliamonte and Hudson, 1999) both of which are conversational devices that accommodate to the interview setting. Example (13) below contains similar English-origin items:
(13) YEAH.. dobijem, oni ne placaju puno, YOU KNOW, zato ja trebam puno raditi, pa ja puno kupim i onda trebam puno raditi.. YOU KNOW, platiju, WHAT, SIX DOLLARS AN HOUR.. (100,M,17) YEAH, I get, they don't pay a lot, YOU KNOW, that's why I have to work a lot, well I go shopping a lot, and then I have to work a lot.. YOU KNOW, they pay, WHAT, SIX DOLLARS AN HOUR ...
To nije puno. (J.H.)
That's not a lot.
Tesko, tesko je radit.. ali, YOU KNOW.. IT'S GOOD FUN, SO.. ljudi.. (100,M,17)
It's hard, it's hard to work.. but, YOU KNOW.. IT'S GOOD FUN, SO ... the people..
I tamo radis kao kuhar ili na kasi ili... ? (J.H.)
And you work there as a cook or at the cash register or..?
YEAH, YEAH.. ima, ima, LIKE.. pet.. mozda.. YEAH, pet LIKE, DIFFERENT, UM, JOBS, sto mozes kuhat, ovaj.. meso ili mozes stavit sve ove ONIONS i DRESS salat, i mozes kuhat ovaj, BREAD, stavit tu BUNS.. i onda, svima puno, YOU KNOW, ali ja, ja sam.. puno na, ovaj, COOKING MEAT i.. u RESTAURANT, onako.. YEAH, IT'S GOOD.. (100,M,17)
YEAH, YEAH.. there are, there are, LIKE.. five.. maybe.. YEAH, five LIKE, DIFFERENT, UM, JOBS that you can.. cook, um.. meat or put on all the ONIONS and DRESS sa lat (dressing salad), and you can cook, um, BREAD, put the BUNS on.. and then to everyone lots of, YOU KNOW, but I, I am.. often on, um, COOKING MEAT and.. in the RESTAURANT, so like.. YEAH, IT'S GOOD..
In (13) above, discourse markers such as you know recur as well as ambient yeah which is barely distinguishable from its Croatian homophonous synonym je (3.SG. 'to be'). Occurrence of English-origin discourse markers reflects adoption of Australian English pragmatic norms and is, in itself, not an example of conscious code-switching where motivations for shifts in role or footing are ascertainable (cf. Hlavac, 2006). In examples (11) and (12) above, neither informant assumes a role other than that of interviewee. Further, in example (13) the code-switches you know, it's good fun, so and yeah, it's good do not so much reflect role realignments but summative evaluations of previously narrated events. These are, in Chan's (2004) terms, "textual" rather than contextual features. They appear at the end of turns and evaluate the content of the turn.
Generally, informants, as shown above in examples (11) and (12) accommodated to the desired and unmarked language of the interview situation, predominantly Croatian speech. The examples of text in which (largely monolingual) Croatian is the language choice of informants are countless and informants' accommodation was taken as axiomatic during the interviewing process. Therefore I will not discuss the many examples in which Croatian turn follows Croatian turn between interviewer and interviewee. However, where English-origin items--extra--or intra-clausal code-switches, or entire clauses in English--precede a code-switch into Croatian (whether or not Croatian items or text preceded English-origin items in the turn), this type of code-switch into Croatian can be interpreted as an example of convergence to the macro-discourse dynamics of the interview setting. This is because the presence of English-origin items, particularly single items, is unmarked. Within a turn, English-origin items are divergent if they precipitate longer stretches of English speech. What is significant about English-origin items is that they are succeeded by switches back to Croatian. These switches back to Croatian are overt and converge to the unmarked language of the interview setting.
Below are a selection of examples of code-switching to Croatian that show convergence. Examples including extra-clausal code-switches such as those in (14) below are included as they 'punctuate' informants' speech and they are relevant to a discussion of convergence inasmuch as informants code-switch back afterwards. The point at which the informant converges back to a more marked variety of the interview is marked with double vertical lines: '||' Examples (14) and (15) below contain intra-clausal and inter-clausal code-switches respectively.
(14) YEAH || volim ali nisam vec dugo bio, zbog skole i nogometa, YOU KNOW || nemam vremena jednostavno, ali inace volim ic.. (63,M,17)
YEAH || I like to but I haven't been for a while, because of school and soccer, YOU KNOW || I simply don't have the time, but otherwise I like to go..
(15) Za sada nemam zapravo nista.. stavili su me, [am].. gdje radim sada, imam, um.. PART-TIME JOB || u Big W i stavili su me tu za SUPERVISOR || ali.. ne znam, sto ucim sada.. tol'ko ne volim sto ucim.. ja sam uvik htila ici u TEACHING || ili nesto tako i nisam, ili ne znam, mozes izabrat sada.. (4,F,20)
At the moment I don't really have anything.. they put me as, um.. where I'm working now, I have a, um.. PART-TIME JOB || at Big W and they put there as SUPERVISOR || but.. I don't know, what I'm studying at the moment.. I don't really like what I'm studying.. I always wanted to go into TEACHING || or something like that and I didn't, but I don't know, now you can choose..
(16) Ne znam. I'M TRYING TO LOOK, UM.. IS IT A WEDDING? || Ne znam. (2,F,27)
I don't know. I'M TRYING TO LOOK, UM.. IS IT A WEDDING? || I don't know.
Extra-clausal English items are followed by a code-switch to Croatian in 67% of instances (1803 of 2688) while for intra-clausal switches the percentage of those which are followed by a switch back to Croatian is 73% (913 of 1258). As far as inter-clausal switches are concerned, a majority of full-clause English switches (59%) are followed by Croatian full clauses (163 of 277) so that Croatian is the language choice for the final clause of the turn. While the number of switches (4223) does not correspond to the number of turns containing switches, 2276, (giving an average of 1.9 switches per turn in those turns that contain code-switches), the high percentage (66%) of code-switches which are followed by Croatian text (2763 out of 4233 switches) indicates that in most turns containing code-switches convergence to or back to the dominant language of discourse occurs. In other words, the high percentage of English code-switches which are followed by code-switches to Croatian suggests that a similarly high percentage ([approximately equal to] 66%) of turns containing switches also contain Croatian turn-final clauses, ie. 66% of these 2276 turns gives a figure of ca. 1550. This number combined with the number of Croatian monolingual turns (3043) gives a figure of 4593, or over 80% of all 5677 turns.
Accommodation to macro-discourse language-choice norms appears to be a strategy employed by informants in most turns, regardless of whether the turn contains code-switching or not. Table 4 below contains sets out the categories of code-switching found in the sample. Those examples, after which a code-switch to Croatian follows, are classified as convergent.
Not every English-origin item can be considered an example of divergence. For this reason, Table 4 above does not contain in its third column divergent but non-convergent examples. The reservoir of possible examples of divergence is enormous and also needs to be relativised. My application of divergence does not typically refer to linguistic forms such as single word or lexically simplex intra-clausal transfers and single-word extra-clausal or tag transfers (usually discourse markers) because these 'shorter' transfers are perceived to be less overtly divergent from the imposed language variety. Divergence here is restricted to clause-length transfers into English and multiple-word transfers into English which terminate a turn.
Below are three examples of divergence. Example (17) contains a turn-final, extra-clausal code-switch, while examples (18) and (19) also contain longer chunks of English speech:
(17) Ne znam, meni se vise svida tude ali.. zavisi.. oh, dobro je, ali.. || I DON'T KNOW.. (45,F,20)
I don't know, I prefer foreign (food) but.. it depends.. oh, it's good, but.. || I DON'T KNOW..
(18) .. i bi htio ic sa tatom zasto bi vise naucio sa tatom, a bi isao sam, bi naucio i znam da sam vec vidio prijatelje i sto su bili, i nisu znali pricat hrvatski.. puno slabije nego ja pricam.. isto dosli su natrag i.. mogu pricat.. onda, ja znam da bi.. ja bi to || PICK IT UP VERY QUICKLY.. (47,M,27)
.. I'd like to go with dad why (because) I'd learn more with dad, but I'd (also) go alone, I'd learn and I know I have already seen friends who were, and they couldn't speak Croatian.. a lot worse than how I speak.. they also came back and.. they can speak.. then, I know that I would.. I would it || PICK IT UP VERY QUICKLY..
(19) NAH.. bilo je kad su bili || YOUNG AND WHEN THEY WERE GROWING UP, i sto su radili, || IT WAS ABOUT GIRLS, TEENAGE GIRLS.. (36,F,17)
NAH.. it was when they were || YOUNG AND WHEN THEY WERE GROWING UP, and what they did, || IT WAS ABOUT GIRLS, TEENAGE GIRLS..
Examples of divergence are predictably less frequent than those of convergence. The number of extra-clausal transfers which contain examples of potential divergence is 884 (out of 2688) or 33% of extra-clausal transfers. The number of divergent intra-clausal switches is 257 (out of 1258) or 20% of all intra-clausal switches while the number of multiple word inter-clausal switches which are potentially divergent is 140 out of 277 or 51%. The combined total of turns with divergent-like characteristics is 1281 or ca. 20% of all turns.
The in-group nature of the interview suggests that 'interindividual' rather than 'intergroup' dynamics determine the nature of accommodation. Shared ethnicity and linguistic proficiencies between the interviewer and the interviewee mean that their speech is less likely to contain features that show convergence or divergence on the basis of inter-group dynamics. Instead, interindividual dynamics, relating to interlocutors' roles to each other, topic and content of conversation and conversational-internal features determine the ways that interlocutors converge to or diverge from the unmarked codes of the interview situation. Examples (10) to (17) vindicate this. Topic (the employment-related items in (12), (13) and (15)), habitualised discourse markers (extra-clausal you know, yeah, like and nah in (12), (13) and (14)), textual features (self-directed speech in (16) turn-terminator in (17), idiomatic phrases in (18) and a change from narrative to summative text in (19)), accompany code-switches in the interviewees' turns that are largely convergent-like. Thus, communication accommodation theory can be applied to in-group situations in the same way that it is applied to out-group situations such as those described in Thakerar, Giles & Cheshire (1982), Gallois & Giles (1998).
7. Markedness Model and Rational Choice Model
As stated above, the language of conversation is Croatian or Croatian-dominant speech. This is the unmarked choice of the interview conversation. Such an allocation of language choice is the basis of the Markedness Model: changes to this choice go outside expected allocations and reasons for such change are sought by examining what a speaker seeks to accomplish through a change of code. The focus of the Markedness Model is not the interaction itself, but the types of participants in an interaction and how they present themselves. This means that choices of code, directly or indirectly, are attributable to the interpersonal relationships that speakers seek to strategically index. Speakers "need situational factors as input..." (Myers-Scotton, 1993: 110) but it is speakers' decisions themselves which determine code choices, not situational factors arbitrarily. Further, "[w]hen speakers engage in CS [code-switching], it means they perceive the interaction--either initially or as it progresses--as one in which they can best maximize their rewards [...] by using two or more linguistic varieties." (Myers-Scotton, 1993: 110. Square brackets mine).
As shown above in tables 1 and 2, code-switching is a frequent occurrence across the whole sample. The high number of code-switches indicate, in the first place, that code-switching is an unmarked variety. Some of Myers-Scotton's (1993: 114, 125) criteria for unmarked code-switching are met: ie. "situational factors remain[-ing] more or less the same during the course of the exchange"; a "good deal of intrasentential switching" in addition to alternating switching. But other criteria are not so clearly met. For example, although both interlocutors are bilingual peers, unmarked code-switching is not expected amongst those who are strangers to each other (Myers-Scotton, 1993:119) and the motivation to register dual membership through frequent Croatian-English code-switching is perhaps tenuous as monolingual English or English with select Croatian imports are the usual speech varieties marking membership as (second-generation) Croatian-Australians.
In an application of this model to this data, I focus on participant roles. Non-naturalistic, semi-formal interviews are a common source of data for studies on spoken language and while most linguists state their role relationship to those from whom data are gathered, few studies systematically examine possible differences in speech depending on whether speakers are speaking to an 'in-group peer', 'in-group researcher', 'out-group researcher' etc. Beebe (1981) and Clement and Noels (1992) examine the role of different interviewers and effects on informants' speech. My role to the informants was that of 'in-group researcher/peer' and their role to me was 'in-group informant'. The interview situation, with its comparable, open-ended questions and identical picture descriptions, represents a stable and unchanging, if not static interaction. Informants have consented to be in a situation about which they have pre-formulated ideas wherein the adoption of the role of interviewee is the unmarked role. They have the choice to adopt the role of interviewee or to step outside of this role. Two informants not included in the sample who interrupted and broke off the interview code-switched to English when doing this. The change in role coincided with selection of the marked code for the interview.
In addition to my role as 'in-group researcher/peer', the relationship that I had with some informants was different to that with others. Fourteen informants were previously known to me and therefore familiar: eight-six informants were unfamiliar and their participation in this research study was secured through relatives, friends or the co-operation of ethnically affiliated associations/schools. A premise of the different role-relationships in addition to that of 'researcher/peer' and 'informant' is that across the sample of informants, those who were already known to me would be most likely to conform to the intentions of the setting, and therefore provide longer turns, and to conform to the desired code of the interview, and therefore code-switch less. In Table 5 below, relationship to me, the interviewer, is differentiated and matched with length of turns and average number of switches per turn.
Table 5 above shows variation of no more than 10% between informants' total number of words and turns compared to the average word-length (2110) and number of turns (57) for each informant segment. Fourteen informants were known to me at the time of the interview--relatives, friends/acquaintances, work colleagues or former students. Of the remaining 86 informants unknown to me before the interview, 9 were introduced to me via relatives, 42 were introduced to me through friends or acquaintances, sometimes mutual friends. Thirty-five informants were included through ethnically-specific associations or Saturday morning school.
The four groups are not equally distributed across the sample. This affects the comparability of findings. There is remarkably little difference in the number of words per turn between the groups: 35--41 words. There is little numerical variation in the frequency of code-switching according to the role relationship with the interviewer. In the turns of informants known to the interviewer there are on average 0.74 code-switches per turn, in the turns of those contacted through friends and relatives 0.67 and 0.66 code-switches respectively, while those contacted through organisations recorded the highest number of code-switches per turn: 0.85 code-switches per turn. Thus, previously unfamiliar informants gained through organisations with a more removed relationship to the interviewer record slightly shorter turns and slightly more frequent code-switching. Those known to the interviewer or introduced through friends or relatives recorded slightly longer turns and a slightly lower level of code-switching. As stated, role-relationship and code-switching are examined here as a static features, independent of the conversational features of the interactions.
The relevance that Myers-Scotton's markedness model has to such a corpus is its understanding of the role of the interviewee and how s/he adopts the role of commentator of his/her own text. This text can often appear self-directed. Thus, the examples of turn-terminal switching in example (13) above "you know, it's good fun, so" and "yeah, it's good" appear as role-relationship changes. Example (20) below contains example of 'lexically-motivated' switches such as 'pest' and 'Renaissance type of era', but there are changes of footings internal to turns which also reflect changes in role-relationships, from that of question answerer to that of an interlocutor making appraisals and seeking my agreement.
(20) Oh, dosta sarana, sarana, krapa.. moze se nac, tamo koliko god hoces.. oh, sve je puno sarana tamo.. nije lose, nije lose tamo, samo sarane kao sada.. kad se umetne te sarane u te jezero, ne znam ni ja.. osamdeset godina prije, onda to nema, IT'S OUT OF CONTROL, to je sad, LIKE.. to je kao PEST, i to sve sada kao pojede drugu ribu, i nema bas tol'ko ribe, SO.. YEAH.. (95,M,17)
Oh, a lot of carp, carp, black umber.. can be found, as much as you like there.. oh, it's all full of carp there.. it's not bad, it's not bad there, it's just carp like now.. when they put those carp in the lake, I don't know myself.. eighty years ago, then there's no, IT'S OUT OF CONTROL, now that's, LIKE.. it's like a PEST, and now they're like eating the other fish, and there's not that much other fish, SO.. YEAH..
To je to, i kazi mi, kad jos imas slobodnog vremena, cime se bavis? Trebas li dosta citat za skolu? sto najvise volis? (J.H.) That's that. Tell me, when you otherwise have free time, what do you do? Do you have to do a lot of reading for uni? What do you like the most?
Oh, volim vise, LIKE, kao, staro englesko.. kao 'Macbeth'.. ili te, LIKE THE RENAISSANCE TYPE OF ERA.. to volim, INTERESTING nesto, staro englesko onda citas, nekad.. knjige citam kao o sportu.. i uzmem i to, YEAH.. OH, WHAT'S IT CALLED?.. to je, to ucimo u mojoj skoli.. 'Macbeth' sada.. bas smo pola, pola knjiga LIKE FINISHED.. za engleski, nije bas losa stvar.. (95,M,17)
Oh, I prefer, LIKE, like old English.. like 'Macbeth'.. or those, LIKE THE RENAISSANCE TYPE OF ERA.. I like that, something INTERESTING, old English then you read, sometimes.. I read books like about sport.. I borrow that as well, YEAH.. OH, WHAT'S IT CALLED?.. that's, we're learning about that in my school.. 'Macbeth' now.. actually we've just LIKE FINISHED half, half the book.. for Engl ish, it's not such a bad thing..
To je bas interesantno, posto se radi o.. ljubomoru, o zlocinima, o ubojicama i.. (J.H.)
And that's interesting, because it's about.. jealousy, about crimes, about murderers and..
Sve nase sada, sve sto vidis na televizor, sve filmovi su BASED UPON THAT, UPON THAT, sto je Macbeth uradio i, YOU'VE GOT TO GIVE HIM CREDIT FOR HIS WORK.. YEAH.. INTERESTING ... (95,M,17)
Everything that we have now, everything that you see on the television, all films are BASED UPON THAT, UPON THAT, what Macbeth did and, YOU'VE GOT TO GIVE HIM CREDIT FOR HIS WORK.. YEAH.. INTERESTING..
At the end of the final turn the male, 17-year-old informant appears to want to reconfigure role-relationships between us, to be more 'peer-like' and appeals to a shared sentiment of admiration towards Shakespeare. The extract contains evaluative and affective passages, many of which are in English: it's out of control, you've got to give him credit for his work. There is a hedge which may suggest retrieval difficulty but which also appears self-directed, oh, what's it called? Elsewhere, there are yeah and like, which co-occurs with its Croatian counterpart kao.
Myers-Scotton's markedness model provides a framework for examining how speakers approach situations on the basis of assumed and projected role-relationships. These are axiomatic considerations for human interactions in a global sense and where role-relationships ascertainably shift in a narrow or inter-individual sense (cf. Myers-Scotton & Bolonyai, 2001). Such shifts can coincide with incidences of code-switching, such as the longer English stretches in example (17) above.
This paper is based on a large sample, analysed quantitatively, but nonetheless containing sufficient detail on the data's attributes to allow in-depth examination. Triggered code-switching on the basis of proper nouns or some English-origin common nouns is relatively infrequent, co-occuring with 6% of the unintegrated, singly-occuring English nouns in the sample. Differences in the phonotactic structures of words between the languages results in a smaller number of homophones and therefore a lower number of possible trigger words.
Code--switching is rarely accompanied by expressed declarations of proficiency shortcomings. In less than 3% of the instances of code-switching are there verbal comments or admissions and/or discourse-pragmatic or morphosyntactic features which indicate difficulties in production of Croatian speech. Metalinguistic features accompanying code-switching are much more common:
there are 144 examples of 'introduced' or 'justified' code-switches. Metalinguistic text is but one of the 'flags' that can commonly surround code-switching. The relatively small number of code-switches that can be attributed to triggering (42), that contain features of metalinguistic talk (144) or that occur on the basis of limitations in Croatian proficiency (117) in a sample containing over 4,000 code-switches indicates that these features do not substantially account for why bilinguals code-switch, at least not in semi-spontaneous speech in a communicative interaction in which code-switching is fairly unremarkable and therefore relatively unmarked. The cognitive/psychological processes associated with voluntary or involuntary activation of particular lexical items and awareness of linguistic output and control (ie. proficiency in the language) are typical features of bilinguals' self-conceptualisations of their language skills. But these are features that enable code-switching to happen without predicting or describing why code-switching actually occurs in given circumstances. In order to account for the incidence of code-switching in a sample in which it is highly frequent and unmarked, other models have been applied to examine their explanatory power to this large sample.
Giles's Communication Accommodation Theory and Myers-Scotton's Markedness Model are commonly applied to situations in which interlocutors have different first languages, and in which the languages or interlocutors occupy asymmetrical power relations. Their application to the data of this paper shows that these models can be employed in samples in which interlocutors have equivalent commands and practices of use for their languages and in which code-switching is a lowly marked, if not unmarked variety. The Communication Accommodation Theory notion of convergence is tested and applied in this paper to refer to where informants follow the desired language choice of the interview situation: Croatian. Convergence as overt behaviour is less easily ascertainable where informants comply with the desired choice of the interview situation and simply speak monolingual Croatian. Convergence is ascertainable where informants succeed English-origin items or text with code-switching to Croatian--this is behaviour that more overtly indicates the informant's readiness to 'switch back' to the desired code. There are many examples where code-switching back to Croatian occurs, and in particular there are a very large number of examples where a turn is commenced with an extra-clausal, English-origin item such as yeah or well and the informant then code-switches straight into Croatian. This large number of convergent-like code-switches, that number 2,763 or two-thirds of the sample, suggests it is possible to conceptualise code choices as indicative of speakers' moves to be similar to those with whom they are interacting. A great number of the code--switches found in the sample can be accounted for by an analysis of bilingual speech that explains them as examples of 'convergent moves', to the code of the interviewer and the desired code of the interview. At the same time, it is not apparent that every English-origin is an example of divergence. In instances where informants seek to change footing, to 'round off' or even terminate their turns, this can be done with multi-word English code-switches which appear to have a distancing function where informants wish to disengage from parts of the interview.
The last model which is applied to this sample is that of Myers-Scotton (1993). Firstly, Myers-Scotton defines features of interactions in which code-switching is unmarked and/or frequent--largely unchanging situational factors that are also likely to bear a good deal of intra-clausal switching. There are some differences in the role-relationships beyond the interviewer and interviewee that pertain to the informants: some were already familiar to the interviewer; some were introduced via friends or relatives; for others there was a more removed means of contact between the interviewer and interviewee. At least in the initial stages of the interview, such different conceptualisations of role relationship may lead to differences in length of turns and levels of code-switching as features which mark interviewees' different roles to the interviewer and situation.
Although there are only small differences in the lengths of turns and in the average number of code-switches in turns between informants on the basis of role relationship, the most 'distant' informants do record shorter turns and higher levels of code-switching which supports this contention. Beyond role-relationships projected at a macro-societal level, Myers-Scotton's Markedness Model can be applied to instances of code-switching that co-occur with discourse-internal shifts in role, position, addressee or textual reference.
Accounts for the incidence of code-switching in this large sample can be found in these two latter models which are based on how speakers position themselves towards others. Bilingual speech is found to be a consequence not of involuntary co-activation of equivalent or similar forms or of a lack of ability to speak monolingually. Instead, bilingual speech, in the context of semi-spontaneous discourse, is determined by speakers' employment of 'other--language' items which may signal re-positioning of roles and/or signal discourse-internal features. Code--switching into English and back to Croatian reflects the speaker's and the listener's desired linguistic choices with the roles and discourse contexts that they are able to enact within them.
Izvorni znanstveni clanak
Prihvaceno za tisak 05.03.2012.
AUER, P. (1995) The pragmatics of code-switching. In L. Milroy & P. Muysken (Eds.) One Speaker, Two Languages: Cross-Disciplinary Perspectives on Code--switching. (pp. 115-135). Cambridge, UK: Cambridge University Press.
BACKUS, A. (1996) Two in one. Bilingual Speech of Turkish Immigrants in The Netherlands. Tilburg: Tilburg University Press.
BEEBE, L. (1981) Social and situational factors affecting the strategy of dialect code-switching. International Journal of the Sociology of Language. 32, 139-149.
BLANKENHORN, R. (2003) Pragmatische Spezifika der Kommunikation von Russlanddeutschen in Sibirien. Entlehnung von Diskursmarkern und Modifikatoren sowie Code--switching. Frankfurt am Main: Peter Lang.
BENTAHILA, A. & DAVIES, E. (1992) Code--switching and Language Dominance. In R. Harris (Ed.) Cognitive Processing in Bilinguals. (pp. 443-458). Amsterdam: North Holland.
BLOM, J-P. & GUMPERZ, J. (1972) Social meaning in linguistic structures: code-switching in Norway. In J. Gumperz & D. Hymes (Eds.) Directions in Sociolinguistics. (pp. 407-434). New York: Holt, Rinehart and Winston.
BULLOCK, B. & TORRIBIO, A. (2004) Introduction: convergence as an emergent property in bilingual speech. Bilingualism: Language and Cognition 7(2), 91-93.
CHAN, B. (2004) Beyond "contextualization". Code--switching as a "textualization cue". Journal of Language and Social Psychology. 23(1), 7-27.
CLEMENT, R. & NOELS, K. (1992) Towards a situated approach to ethnolinguistic identity: The effects of status on individuals and groups. Journal of Language and Social Psychology. 11, 203-232.
CLYNE, M. (1967) Transference and Triggering. The Hague: Martinus Nijhoff.
DUROVTC, L. (1983) The case systems in the language of diaspora children. Slavica Lundensia 9, 21-94.
FISHMAN, J. (1967) Bilingualism with and without Diglossia; Diglossia with and without Bilingualism. Journal of Social Issues 23(2), 29-38.
GAFARANGA, J. (2005) Demythologizing language alternation studies: conversational structure vs. social structure in bilingual interaction. Journal of Pragmatics. 37, 281-300.
GALLOIS, C. & GILES, H. (1998) Accommodating mutual influence. In M. Palmer (Ed.) Mutual Influence in Interpersonal Communication. Theory and Research in Cognition, Affect and Behavior. (pp. 135-162). New York: Ablex.
GENESEE, F. & BOURHIS, R. (1988) Evaluative reactions to language choice strategies: the role of sociostructural factors. Language and Communication. 8, 229-250.
GILES, H., BOURHIS, R. & TAYLOR, D. (1997) Towards a theory of language in ethnic group relations. In Giles, H. (Ed.) Language, Ethnicity and Intergroup Relations. (pp. 307-348). London: Academic Press.
GILES, H. & SMITH, P. (1979) Accomodation theory: optimal levels of convergence. In H. Giles & R. St Clair (eds.) Language and Social Psychology (pp. 45-65). Oxford: Oxford University Press.
GILES, H., COUPLAND, J. & COUPLAND, N. (1991) Contexts of Accommodation. Developments in Applied Sociolinguistics. Cambridge, UK: Cambridge University Press.
GREGOR, E. (2003) Russian-English Code--switching in New York City. Frankfurt am Main: Peter Lang
GUMPERZ, J. (1982) Discourse Strategies. Cambridge, UK: Cambridge University Press.
HALMARI, H. (1997) Government and Codeswitching. Explaining American Finnish. Amsterdam: John Benjamins.
HLAVAC, J. (2003) Second generation speech: lexicon, code-switching and morpho-syntax of Croatian-English bilinguals. Bern/New York: Peter Lang.
HLAVAC, J. (2006) Bilingual discourse markers: Evidence from Croatian-English code-switching. Journal of Pragmatics. 38, 1870-1900.
HLAVAC, J. (2010) Shifts in the language of interpreting with bi--or multi-lingual clients: circumstances and implications for interpreters. Interpreting 12(2), 186-213.
HLAVAC, J. (2011) Hesitation and monitoring phenomena in bilingual speech: A consequence of code-switching or a strategy to facilitate its incorporation? Journal of Pragmatics. 43, 3793-3806.
LANZA, E. (1997) Language Mixing in Infant Bilingualism. A Sociolinguistic Perspective. Oxford: Clarendon Press.
LI W. (2005) "How can you tell?" Towards a common sense explanation of conversational code-switching. Journal of Pragmatics. 37(3), 375-389.
MASCHLER, Y. (1994) Metalanguaging and discourse markers in bilingual conversation. Language in Society. 23(3), 325-366.
MONTRUL, S. (2008) Incomplete Acquisition in Bilingualism. Re-examining the Age Factor. Amsterdam: John Benjamins
MYERS-SCOTTON, C. (1993) Social Motivations for Codeswitching. Evidence from Africa. Oxford: Clarendon.
MYERS-SCOTTON, C. & BOLONYAI, A. (2001) Calculating speakers: codeswitching in a rational choice model. Language in Society. 30 (1), 1-28.
MYERS-SCOTTON, C. (2002) Contact linguistics. Bilingual encounters and grammatical outcomes. Oxford: Oxford University Press.
MYERS-SCOTTON, C. & BOLONYA, A. (2001) Calculating speakers: Codeswitching in a rational choice model. Language in Society. 30, 1-28.
POLINSKY, M. (2008). Heritage language narratives. In Donna Brinton, Olga Kagan & Susan Bauckus, (eds.) Heritage language education. A new field emerging. (pp. 149-164). New York/ London: Routledge.
POPLACK, S. (1980) Sometimes I'll start a sentence in Spanish y termino en espanol: toward a typology of code-switching. Linguistics 18, 581-618.
ROMAINE, S. & LANGE, D. (1991) The use of like as a marker of reported speech and thought: a case of grammaticalization in progress. American Speech. 66(3), 227-276.
SALMONS, J. (1990) Bilingual discourse marking: code switching, borrowing and convergence in some German-American dialects. Linguistics 28, 453-480.
SCHIFFRIN, D. (1987) Discourse Markers. Cambridge, UK.: Cambridge University Press.
SRIDHAR, S.N. & SRIDHAR, K. (1980) The syntax and psycholinguistics of bilingual code-switching. Canadian Journal of Psychology 34, 407-16.
TAGLIAMONTE, S. & HUDSON, R. (1999) Be like et al. beyond America: The quotative system in British and Canadian youth. Journal of Sociolinguistics 3(2), 147-172.
TREFFERS-DALLER, J. (1997) Variability in code-switching styles: Turkish-German code switching patterns. In R. Jacobson (Ed.) Codeswitching worldwide. (pp. 177-200). Berlin: Mouton de Gruyter.
WEINREICH, U. (1953) Languages in Contact. The Hague: Mouton.
ZENTELLA, A. (1997) Growing up bilingual. Malden/Oxford: Blackwell.
Table 1: Number of turns and examples of code-switching across sample Number Percentage Turns 5677 Monolingual Croatian turns 3043 53% Monolingual English turns 311 6% Non-lexicalized turns (ie. uh-huh or mm.) 47 1% Turns containing code-switch/es 2276 40% English-origin items/code-switches 4223 Lexical tokens (Croatian + English) overall 211000 Table 2: Categories and numbers of code--switches Single items Multiple items Total Extra-clausal switching 2516 172 2688 Inter-clausal switching 137 140 277 Intra-clausal switching 962 296 1258 Total 3615 608 4223 Table 3: Categories of triggers and language of metalinguistic talk Consequential triggering Proper nouns 7 Single-word common nouns 24 Bilingual homophones 1 Total 32 Anticipational triggering Proper nouns 2 Single-word common nouns 8 Total 10 Anticipational metalinguistic talk In Croatian 78 (introduced) In English 5 Total 83 Consequential metalinguistic talk In Croatian 58 (justified) In English 3 Total 61 Table 4: Number of convergent and non-convergent code--switches Convergent Non-convergent Total Extra-clausal switching 1804 884 2688 Inter-clausal switching 161 116 277 Intra-clausal switching 798 460 1258 Total 2763 1460 4223 Table 5: Role relationship to interviewer and number of turns and code--switches. Previously Known No. of No. of Turns Ave. words Switches Ave. switches informants words per turn per turn 14 27279 671 514 Ave. 1948 48 41 30 0.77 Previously Unknown Contact through Relative No. of No. of Turns Ave. words Switches Ave. switches informants words per turn per turn 9 20698 518 342 Ave. 2300 58 40 31 0.66 Contact through Friend No. of No. of Turns Ave. words Switches Ave. switches informants words per turn per turn 42 91426 2373 1594 Ave. 2177 57 38 32 0.67 Contact through Ethnically Affiliated Association/School No. of No. of Turns Ave. words Switches Ave. switches informants words per turn per turn 35 71656 2115 1797 Ave. 2047 60 35 46 0.85 Overall No. of No. of Turns Ave. words Switches Ave. switches informants words per turn per turn 100 211008 5677 4223 Ave. 2110 57 37 42 0.74