Disfluency and language membership? Interpreting flagging behavior in bilingual speech
RACHEL VARRA
The College of William and Mary

This paper investigates the interpretation of flagging phenomena (false starts, filled (eh, uh) and unfilled pauses and metalinguistic commentary (como dicen aquí)) in the vicinity of lexical contact phenomena (codeswitching and lexical borrowing) using a sociovariationist analysis. Decades of research have shown that codeswitching and borrowing are rule-governed processes (cf. Di Sciullo Muysken, & Sing 1986; Myers-Scotton 1997; Poplack 1980) and indicative of complex linguistic knowledge and behavior on the part of the bilingual (Zentella 1997). In arriving at this conclusion, researchers have sought to ensure that the data examined represent ‘natural’ and fluent speech from the communities studied. In doing so, they not infrequently exclude other-language strings that co-occur with flagging phenomena (see Pfaff 1979; Poplack 1987; Teschner 1972: 1140-1142), considering such behavior to be indicative of disfluency on the part of the speaker (e.g. Poplack 1983: 123; Poplack, Sankoff & Miller 1988: 54) or else as an indication that the speech being produced is not an ‘authentic’ part of the linguistic repertoire for that speaker/community. It is the purpose of this paper to examine these assumptions – that flagging behavior is indicative of disfluency and that it is an indication of an individual’s awareness of the foreign origin of an utterance – with respect to a particular community of Spanish speakers in New York City.

Analysis revealed that, while the overall incidence of flagging is small (between 10 and 20% of English-origin material), flagging indeed seems to be an indication of an individual’s awareness of the foreignness of English-origin material and that it might not ‘belong’ to Spanish. Flagging was more frequent among speakers that had more knowledge of, exposure to or use of the ‘other language’ in this study, i.e. English. (e.g. Results showed more frequent flagging among those with greater English proficiency, that that grew up in English-speaking contexts and those that use only English in their day-to-day lives). Flagging was also more frequent in the vicinity of words hypothesized to be new (i.e. ‘nonce’ items) than English words that were ‘widespread’ (Poplack et al 1988) in the corpus). On the other hand, flagging in this corpus was not an indication of linguistic disfluencies. (Results showed that flagging was not related to Spanish proficiency.) These results offer evidence that subverts the assumption that flagging in bilingual speech is associated with poor language skills. Overall, then results of this investigation suggests that the use of flagging a diagnostic for data collection cannot be assumed a priori but must be empirically validated for a particular corpus or group of speakers.