The evolution of trilingual code­switching  from infancy to school age: the shaping of  trilingual competence through dynamic  language dominance Hoffmann, C and Stavans, A http://dx.doi.org/10.1177/13670069070110010401 Title The evolution of trilingual code­switching from infancy to school age: the  shaping of trilingual competence through dynamic language dominance Authors Hoffmann, C and Stavans, A Type Article URL This version is available at: http://usir.salford.ac.uk/2217/ Published Date 2007 USIR is a digital collection of the research output of the University of Salford. Where copyright  permits, full text material held in the repository is made freely available online and can be read,  downloaded and copied for non­commercial private study or research purposes. Please check the  manuscript for any further copyright restrictions. For more information, including our policy and submission procedure, please contact the Repository Team at: [email protected]. International Journal of Bilingualism, 11, 1 (2007) pp.55-72. The evolution of trilingual code-switching from infancy to school age: the shaping of trilingual competence through dynamic language dominance Charlotte Hoffmann, University of Salford, UK, and Anat Stavans, Beit Berl College and Hebrew University of Jerusalem, Israel Running head: The evolution of trilingual code-switching Please send correspondence to: Charlotte Hoffmann School of Languages University of Salford Salford M5 4WT Greater Manchester [email protected] 1 Abstract This article reports on a study of the code-switches produced by two children who acquired their three languages in early childhood. We compared formal and functional aspects of their switches recorded at two different stages of their development. Of particular interest was the consideration of sociolinguistic variables that have intervened in the children’s environment. We undertook a quantitative and qualitative analysis of the children’s code-switches to ascertain the frequency of switching, the use of each of the three languages employed for switching and the linguistic complexity of the switches. We assumed that the sociolinguistic conditions that changed the linguistic landscape in which these children operated would be reflected not only in the development of each of their languages, but also in the kind of switches that they produced. We tried to establish whether it is the case that certain forms and functions of code-switches constitute a "core" of trilingual language behaviour while others are prone to change. Ultimately, our aim was to gain an insight into the specific trilingual language production processes over a given period of time that can shed light on the development and nature of trilingual competence. Key words: code-switching, trilingual competence, trilingualism 2 1. Introduction The study of child trilingualism can be frustrating as it is often hampered by practical problems relating to the collection and interpretation of data on the one hand, and the absence of theoretical models that might be used for comparison on the other. The rewarding feature of such endeavours, however, is that they do allow fascinating glimpses into the human capacity of processing language and the linguistic resourcefulness of multilinguals. The study of bilingual code-switching has been undertaken from a variety of linguistic and pragmatic perspectives, with code-switches representing evidence of language processing, formal preferences, or discourse strategies. Not surprisingly, linguistic versatility can become even more enhanced when three languages are in constant contact and use. While trilingualism is becoming a more widespread linguistic faculty in individuals, most of the research on trilingualism concerns individuals who acquire or learn a third language in a school context (Cenoz et. al., 2001; Cenoz & Genesee, 1998) or immigrant minorities who learn a third language in social contexts (Baetens Beardsmore, 1993). Most of these studies and models view the development of a third language consecutively to the development of one or two other languages. Our subjects fall into a particular category of trilinguals, namely children who acquired their three systems almost simultaneously, i.e. children who had contact with three linguistic systems from birth and developed them as first language. Their language acquisition took place in contexts where their wider linguistic environment and particular family language constellation changed, although the languages involved remained the same. It therefore became necessary to bear in mind sociolinguistic factors alongside the developmental aspects as well. The focus of our study is on language production of trilingual children in those instances where their output differs from that of their monolingual peers. As part of their trilingual communication, i.e. with bilinguals and trilinguals who share the same languages, code-switching serves as a highly effective interactional tool. It allows the user to compensate for linguistic gaps 3 in one or several of their languages or to capitalize on the typological efficiency of one linguistic system versus the other so as to achieve successful communication. We examine the children’s non-monolingual language output from a quantitative and a qualitative perspective. Thus, we analyse our data in terms of frequency of switches and mixes and according to their morpho-syntactic shape and the number of languages they contain. A developmental stance has been adopted as we compare the switches produced by the children at different ages and taken at different stages in their development. Ultimately, we hope to contribute some insights into the development and nature of trilingual competence, including its related issues of language choice, language dominance and directionality of switches. 2. Studies in code-switching In line with our two-pronged approach to trilingual competence in children as demonstrated through code-switching behaviour our study is informed by two types of research: i.) Studies that look at the language behaviour of groups in multilingual contexts that have a broad remit as they aim to trace changes in the linguistic systems of languages in contact. Diachronic studies show up particular language strategies of language choice and switching that bilinguals or multilinguals adopt. Those that take a longer-term perspective point towards the emergence of new, fused systems or language shift. In particular, we have been influenced by the conceptual and theoretical proposals put forward by Auer (1998), Clyne (1997, 2003) and Muysken (2000). ii.) Some psycholinguistic investigations that look at developmental aspects of language contact as revealed in code-switching during the process of language acquisition in trilingual children, and that deal with the nature of multilingual competence. Unfortunately only few such studies exist and they are based on small samples (Hoffmann, 2001; Stavans, 1990, 2003, in press). While not exactly comparable to our own research, Clyne’s work with large numbers of adult multilinguals who acquired their third language (English) in immigrant contexts was insightful and thought-provoking, as was Lanza’s work on bilingual code-switching (1997, 2001). 4 2.1 Can children code-switch? We go along with the separate language acquisition hypothesis that has been proposed (Genesee, 1989, Genesee et al., 1995; Comeau et.al., 2003) and widely accepted in the literature on bilingual language acquisition and suggest that it extends to simultaneous trilingual acquisition as well. Similarly, we agree with the view that code-switching (CS) is a feature of adult as well as child language. Like Lanza (1997) and a previous study by Stavans (1992), we are of the opinion that there is no basic qualitative difference between adult CS and infant language mixing. The main difference is that children’s language systems are under development and their competence is still to be achieved, whereas in the adult competence has already been acquired. A typical example of this is the case where trilingual children mix the same proposition in utterances in their different languages. Examples such as ‘Will you play mit me?’, ‘Nina pinta mit lápiz’, Nina Auto fahren con Mami’ (Hoffmann, 1985) show that the child obviously knows that a preposition is required in a particular slot but that its language-specific lexical shape has not yet become fully contextualised, a phenomenon we refer to as permeability of languages. The present study provides evidence for a developing pragmatic competence in trilingual children. It has encouraged us to propose a preliminary model of such an evolving trilingual competence. In this study, we look at language alternation primarily from a formal, structural perspective although functional considerations also come into play. We see code-mixing (CM) as the conflation of various linguistic units (be they morphemes, words, phrases or whole clauses) within the sentence. This is similar to Muysken who in an examination of bilingual speech uses code-mixing “to refer to all cases where lexical items and grammatical features from two languages appear in one sentence” (2000:1). Code-switching, on the other hand, involves mixing linguistic elements (from words to phrases and clauses) across sentences. Thus intra-sentential CM and inter-sentential CS are interpreted as different manifestations of bilingual speech when the concept of the sentence as a grammatical unit is given prominence. 2.2 Studies espousing our approach and perspective For Auer (1998) code-switching is discourse-related. He associates it with a “locally meaningful event by the participants” and situates it at the pragmatics end, whereas language mixing is seen as having a wider remit as it is to be interpreted “in a more global sense", i.e. when it is seen as a 5 recurrent pattern. This he places in the middle of the scale. Fused lects, the new linguistic systems that emerge from (usually prolonged) language contact and are stabilised mixed varieties, are located at the grammar end of his continuum. Auer suggests that his model characterizes not only the typology of language alternation but also highlights the progressive development of bilingualism. “Compared to CS,” he writes, “LM seems to require a higher bilingual competence; in addition, there is some evidence that alternational LM requires more proficient bilinguals than insertional LM. On the continuum from CS to FL [fused lects], the most balanced bilinguals are likely to be found here.” (Auer, 1998:10). One of the pillars of Grosjean’s holistic view of bilingualism (1985, 2001) is his notion of language modes. When bilinguals are in monolingual mode their other language is largely deactivated and their language production is similar to that of monolinguals. Both languages are activated when they are in bilingual mode which is when their language output may display features of bilingual speech such as borrowing and code-switching. Grosjean’s model can be adopted to account for the trilingual’s language production as well. One would then have to posit that trilinguals can operate in monolingual, bilingual and trilingual modes depending on the degree of language activation or deactivation of their three linguistic systems. We have argued before (Hoffmann, 2001; Stavans, forthcoming) that there may well be a difference between bilingual and trilingual competence that goes beyond a quantitative one that results from the fact that there are three instead of two languages involved. Three different linguistic systems, the combinatory choices in their actual or potential use, and the interplay of attitudinal and socio-cultural factors all contribute towards a qualitative difference as well. Bilinguals and multilinguals alike have been observed to be particularly sensitive to the linguistic context of language use and appear to be intuitively responsive to the linguistic needs of their interlocutors. We see this heightened sensitivity as part of the trilingual’s overall metalinguistic awareness and language competence. Multilingual children can be eclectic in their choice of languages and switched linguistic items. For them, code-switching affords economy and efficiency of communication with other multilinguals. A closer look at their CS indicates that certain language structures or properties are 6 more prone to switching than others. We can also see that the patterns of the switches change as children get older and/or other factors such as increasing dominance of one of their languages exert their influence (Hoffmann & Widdicombe, 1999; Reyes & Ervin-Tripp, 2004). The latter study traces such development by looking at school-age consecutive bilinguals’ switching and borrowing practices and assesses their significance from a developmental perspective. They report on their subjects’ change of pattern of switching behaviour and that their switches became increasingly more sophisticated; this we have also observed in our trilingual subjects. Apart from drawing on studies of code-switching, we have looked at research in other, related areas, especially with regard to taking account of sociolinguistic factors that influence language development. As children go through their linguistic formative period they acquire communicative competence through acculturation and socialisation, both processes in which language plays a crucial role. Bryant’s work (2001) highlights the importance of acquiring such competence early on in life: not only do they determine successful social integration and educational development; they also have a bearing on later literary skills. It is therefore to be expected that any changes in the linguistic environment of developing multilinguals is likely to have consequences for their communicative competence. Over the past two decades a number of monolingual and bilingual language processing models with a psycholinguistic focus have been developed (e.g. Green, 1986, 1998; Levelt, 1989; also Walters, 2001). Work in this area has been widened by Clyne’s proposed multilingual language processing model (2003:213). It is ambitious in that it attempts to be all-embracing, aiming to account for linguistic processing as well as the various forces that impinge on it. The result is a construct that in itself incorporates several psycholinguistic and sociolinguistic models of multiple language production, and includes a range of components that all influence and determine a multilingual’s choice of output such as use and context, attitude and identity. 3. The Study1 3.1 The subjects The subjects were two children (siblings) raised trilingually from birth with each parent speaking a different non-community language. The home environment was genuinely trilingual as each 7 parent was able to communicate in all three languages. The first set of data was recorded when the children were nearly 3 and 6 respectively and the family, the father speaking Spanish and the mother Hebrew, was living in the US. English was the language of the children’s wider experience outside the home, while Spanish and Hebrew were used inside the home, Spanish being the language used by the parents to each other. The second set of data was collected three years later when the younger child was almost 6 and the older one was 9. At that time the family had moved to live in Israel and the mother had switched to using English with the children in order to provide the children with continued trilingual input, although Spanish continued to be the language the parents used with each other and the language the father used when addressing the children. 3.2 Data collection and analysis The first set of data (T 1) was collected over a period of 18 months by Stavans (1990) in a naturalistic home environment while the children and family conducted normal family life. The recording sessions included interactions of the children in one of three possible language modes: monolingual – when playing with a friend; bilingual –with relatives or family and friends; trilingual –with parents. The second set of data (T 2) was collected nearly 3 years later (Stavans, 2001). The children were asked to engage in an extensive discourse, a semi-structured experimental task of narration. The children were given the wordless picture-book entitled "Frog Where Are You?" by Mercer Mayer. The children where then asked to retell the story, at 2 week intervals, following the pictures. Recording of the children's narration in each one of the three languages was transcribed and analysed. In this study we work with the terms CS and CM which suit our focus on formal aspects. In our discussion of the developing trilingual competence we draw on Auer’s model when we regard CS and CM as landmarks on a linguistic continuum of multilingual production. We are mindful of the fact that, unlike Auer, we are dealing with trilingual simultaneous language acquisition. Therefore, the emerging linguistic competence is not the result of the same factors as in the case of language contact among adolescents or adults in a multilingual context. We distinguish and 8 define CS along the sociolinguistic lines of alternations that occur in utterance, clause or sentence initial position. CM are alternations that occur at the sentence structure level, at the lexical level, and at the morpho-syntactic level. We examined our data with a view to finding answers to the following three questions: 1) How often did the children code-switch and code-mix when using their three languages? 2) Which language or languages were used for CM and CS? 3) What kind of switches did these trilingual children employ? 4) 4. Results 4.1 Frequency of switching Our first step was to quantitatively assess the overall frequency and distribution of CM and CS by both children over time. We counted the total number of occurrences and compared these to the number of opportunities for switching between the languages that they had. Based on the observed switching behaviour of these subjects we define opportunity to be a morphosyntactic, a sentential or an utterance boundary. Table 1: Percentages of CS and CM by each child at T1 and T2 Subject M E M E Age T1 T1 T2 T2 2;6-4;0 5;5-6;11 6;4 9;3 Proportion of CS 28.70% 35.40% 1.30% 1.70% Proportion of CM 1.20% 3.20% 28.60% 10.20% Total proportion of 17.00% 38.60% 29.90% 11.90% alterations Table 1 shows that E (the older subject) used more switches in the earlier production compared to relatively few switches in the task at a later age. By contrast, child M (the younger subject) 9