Investigating the role of social networks in language maintenance and shift

Models predicting language maintenance and shift among immigrant groups have been primarily derived from large-scale group studies. Although these studies have identified a large inventory of factors that influence individuals' decisions about first-language maintenance, they have largely not addressed the issue of the social mechanisms by which individuals negotiate and sustain their choices. This article presents parts of a small-scale study designed to allow such an investigation. The concept of social network is proposed as an organizing mechanism underlying individuals' choices. In particular, this article, after a closer look at a highly elaborated concept of social network, suggests ways that research tools used in large-scale studies might be improved. The subjects are ten immigrant women in the US. By selecting educated, English-proficient, and ethnically isolated subjects, the significance of economic and political pressures was reduced. The subjects were ranked as maintainers or shifters according to frequency of L1 usage across domains. An elaborate social-network questionnaire was designed in order to establish which features of the network would pattern with maintenance ratings. For eight of the ten subjects, there is a strong association between maintenance ratings and the ratio of L1 to L2 speakers in the speaker's secondary network in the US. In addition to this primarily quantitative part of the study, in order to find further factors possibly standing in relationship to observed language maintenance or shift, the ten subjects were studied using qualitative interview data. These revealed factors that are both network-related (for example, pressures from within families toward the bilingual upbringing of children) and non-network-related (pride in the home country). The maintainers also showed a stronger emotional attachment to their L1s than the shifters. As these factors helped explain discrepancies in the first part of the analysis and were not