Acceso usuarios
Recent research on psycholinguistic aspects of bilingual production has been aimed at general modeling issues (De Bot & Schreuder, 1993), control of processing (Green, 1993), and the formulation of output (Azuma, 1993; Myers-Scotton, 1993). In the latter approach, codeswitching, henceforth `CS' has been an important source of evidence, on the concept of a “base” language, and the claim that closed class words do not codeswitch. The present study provides further evidence on these and related issues drawn from a previously untapped source of bilingual performance involving GreekEnglish conversational data. We propose that a useful line of research is opened up by the assumption that CS is revealing of linguistic units that are planned in real time during the opened up by the assumption that CS is revealing of linguistic units that are planned in real time during the production process.
We present a methodology for the systematic observation of CS distribution in our corpus, based on a range of defined units of discourse organization including the turn and different types of units within the turn, down to the level of the word. CS is found not to occur within word unit boundaries.
Findings suggest that the turn and immediate subturn unit boundaries are important loci for CS; that a variety of units beside the clause (Ford, 1982) and the verb phrase are operating in the planning process; that CS occurs preferentially before adjuncts, and before head words in constituents, especially nouns; and that closed class words are involved in CS. We conclude that planning for bilingual language production involves a range of different, and quite small, units, and that both languages contribute to the on-line building of utterance frames.
