Accessing Cross Language Categories in Learning a Third Language

It is well established that early bilinguals can accurately produce and perceive sounds in both of their languages (e.g. Flege and Eefting, 1987, Sundara, Polk, and Baum, 2006; but see also below). This applies both to sounds that occur in only one of the two languages and to sounds that are common to both but differ at the phonetic level. For example, the trilled /r/ in Spanish and the approximate /ɹ/ in English are appropriately produced from a young age by early Spanish-English bilinguals (Goldstein and Washington, 2001; Goldstein, Fabiano, and Washington, 2005). Similarly, English-Spanish bilinguals produce short-lag voice onset time (VOT) with initial /p/ in Spanish but long-lag VOT with initial /p/ in English (Flege and Eefting, 1987). This suggests that bilinguals do indeed have two separate phonetic and phonological systems for each of their languages. However, much early research on bilinguals’ phonological inventories focused on investigating if bilinguals behaved as two monolinguals, or if instead the languages in some way interact in speech production and perception. Generally, results have shown that bilinguals do not perfectly match monolinguals, however. This can result in convergence between the two languages (Sundara et al., 2006; Flege and Eefting, 1987 for English VOT of English-Spanish bilinguals) or even divergence, as a means to further differentiate the two languages (cf. Flege and Eefting, 1987, for Spanish VOT of English-Spanish bilinguals). Similar results have been found in perception, with language mode affecting phonological boundaries (Elman, Diehl, and Buchwald, 1977; Flege and Eefting, 1987). This is in contrast to theories positing that only one language is available at a time (Macnamar and Kushnir, 1971), or that bilinguals have access to both languages in production but only to their dominant language in perception (Cutler, Mehler, Norris, and Seguí, 1992).