Code-switching among proficient adult bilinguals has been extensively studied, and it is by now universally concluded that intra-sentential code alternations are rule-governed and systematic, displaying dependency relations that reflect the operation of underlying syntactic principles. The central, guiding question to be addressed herein is whether and, if so, how second language learners acquire the knowledge that defines structural coherence and allows them to render well-formedness judgments for code-switched forms. This exploration takes on particular significance given that learners receive no evidence which could guide them in rendering such judgments, and therefore results consistent with those observed among competent bilinguals could be imputed to unconscious, abstract linguistic knowledge. The investigation thus proves doubly fruitful, in the discovery of developmental patterns and in the evaluation of linguistic-theoretical methodologies and constructs.

The competitive nature of the lexical selection process during spoken word production is well established in monolinguals. In this paper we explore the implication of this process for spoken word production in bilinguals. A cross-language semantic competitor priming effect was demonstrated, which shows that cross-language lexical competition is a feature of the word production system of the unbalanced English–French bilinguals who participated in the experiment. Experimental evidence was also found which suggests that a selected language bias effected through inhibition of the unwanted language plays an important role in resolving the cross-language lexical competition during bilingual word production in a selected language. The data further suggest that the dominance of the unwanted language relative to the selected language determines the presence/absence or “strength” of inhibition through which the selected language bias is effected. These findings are also interpreted in terms of a recent language-specific lexical selection account of bilingual lexical access.

We investigated 563 German–English nouns for the purposes of identifying cognates, false cognates and non-cognates. Two techniques for identifying cognates were used and compared: (i) De Groot and Nas's (1991) similarity-rating technique and (ii) a translation-elicitation task similar to that of Kroll and Stewart (1994). The results obtained with English-speaking participants produced 112 cognates, 94 false cognates, and 357 non cognates and indicated that the two techniques yielded similar findings. Rated similarity of German–English translation pairs and translation accuracy were positively correlated. We also investigated whether the presence of German-specific characters and the availability of German pronunciation information influenced similarity ratings and translation accuracy. Ratings for translation pairs in which the German word contained a language-specific character were lower and the word was translated less accurately. Participants provided with pronunciation information rated German–English translation pairs as being more similar and translated German words correctly more often than participants who did not receive pronunciation information. We also report the relationships among word frequency, rated imageability and the performance measures. The resulting database of information is intended to be a resource for researchers interested in cognitive processing in German–English bilinguals.

An analysis of 519 gender errors (out of 9,378 modifiers) in the advanced French interlanguage of 27 Dutch L1 speakers confirms earlier findings that gender assignment and/or agreement remain problematic for learners at all levels. A hypothesis derived from Pienemann's Processability Theory (1998a) that accuracy rates would be higher for gender agreement in structures involving no exchange of grammatical information between constituents was not confirmed. The analysis of interindividual and intra-individual variation in gender accuracy rates revealed effects from avoidance and generalisation strategies, from linguistic variables, sociobiographical variables and psycholinguistic variables. We argue that gender errors can originate at the lemma level, at the gender node level, or at the lexeme level. Different psycholinguistic scenarios are presented to account for intra-individual variation in gender assignment and agreement.