Cross-language naming and translation in highly skilled bilinguals: How task requirements activate lexical and conceptual information

Current models of bilingual lexical access emphasize separate lexical and conceptual routes to processing. Fluent bilinguals may utilize lexical information differently depending on the retrieval requirements of a particular task. Is the cognate advantage during bilingual lexical access merely an artifact of task demands, or does it reflect semantic structure across languages? This hypothesis was tested using two natural bilingual tasks, naming and translation.;Fluent French and English bilinguals were recruited from the Montreal area. Subjects were classified based on how frequently they use their second language (L2) in order to test the hypothesis that high L2 users would show similar response times in L2-L1 and L1-L2 translation while low L2 users would produce a translation asymmetry (L2-L1 faster than L1-L2).;In a masked priming paradigm subjects saw between-language translation primes (prime: eglise - target: church), semantic primes (prime: Dieu - target: church), and unrelated primes (prime: girafe - target: church) prior to the target word which they were required either to name or to translate. Cognateness was also manipulated, so that subjects saw both cognate (pilote-pilot) and noncognate (eglise-church) word pairs.;In the translation task, cognates (pilote-pilot) were responded to faster than noncognates (eglise-church) while in the naming task, noncognates were faster than cognates. A large lexical effect in the naming task suggested that naming occurs primarily through a lexical route to processing. Performance in the translation task activated both lexical and conceptual routes to processing with a significant conceptual interference effect when the prime and target words were in different languages. Bilinguals with low use of their L2 produced a translation asymmetry (L2-L1 faster than L1-L2) while high L2 users showed similar response times in both directions of translation, but only for the native English speakers.;Multiple task demands during translation caused interference, especially when activating the correct response required both a language and a conceptual switch. Subjects performed better when they were aware which language was activated, suggesting that semantic information is most closely organized around a single language, even in fluent bilinguals. Frequency of second language use was an important individual difference variable affecting processing when bilinguals were balanced in their two languages.