A corpus-based analysis of codeswitching patterns in bilingual communities

The first aim of the current study is to compare codeswitching (CS) patterns in the bilingual speech of three communities: Miami, Patagonia, and Wales. In order to classify the CS patterns found in our data, we apply the Matrix Language Frame (MLF) model (Myers-Scotton, 1993, 2002) to all bilingual clauses contained in nine recordings of natural conversation. We selected the MLF as a means of classifying our data because the model has been tested successfully on Welsh, English and Spanish data in previous studies (Deuchar, 2006; Deuchar & Davies, 2009; Davies & Deuchar, forthcoming). Our second objective is to investigate the relationship between these CS patterns and sociodemographic community characteristics. We propose that variation found in these patterns may be linked to community-wide extralinguistic factors. We predict that the choice of matrix language (ML) will be affected by relative language proficiency levels, the language used in education, the language of their social networks and the social identity of the participants. We expect the ML to be language in which most speakers are proficient (cf. Myers-Scotton 2002: 27), and for the ML to also match the language used in education, the main language of social networks, and the language most associated with their perceptions of their own identity. For example, for the Welsh-Spanish bilinguals in Patagonia, if the language of education and their social network is Spanish, and they have a higher proficiency in Spanish than in Welsh, then we would predict that Spanish would be the preferred ML. If there is variability with respect to these factors, then we expect a mixed ML distribution in the data