Three experiments were conducted to examine cross-language priming in bilinguals. The first was a cross-language primed lexical decision task experiment with Chinese-English bilinguals. Subjects made lexical decisions about primary associate targets in the two languages at the same rate, but priming occurred only when the prime was in their first language (Ll), Chinese, and the target was in their second language (L2), English. Experiment 2 produced the same pattern of asymmetrical priming with two alphabetic languages, French and Dutch. In Experiment 3, the crucial stimuli were translation equivalents. In contrast to the results of Experiments 1 and 2, priming occurred across languages in both the LI-L2 and L2-Ll conditions. However, this priming was also asymmetrical, with more priming occurring in the LI-L2 condition. A tentative separate-interconnected model of bilingual memory is described. It suggests that the representations of words expressed in different languages are stored in separate memory systems, which may be interconnected via one-to-one links between some translation-equivalent representations as well as meaning-integration processes.Bilinguals are able to communicate in either of two languages without experiencing constant intrusions from the inactive language. Yet when a bilingual learns something via one language, there appears to be access to that knowledge via the other language. How can the language systems be kept separate in practice and still share the same information? Are the languages represented as separate, independent modules in memory? Or, are all languages represented in a shared, interdependent semantic module? These questions have important implications for an understanding of bilingual behavior and for more general models of memory and representation.In 1963, Kolers formalized the question and integrated it into psychological issues of representation by proposing a shared-separate dichotomy. He suggested that the representations of words expressed in different languages either are stored in discrete independent memory systems or take totally abstract forms, such as propositions, and are shared by words in the two languages.We would like to thank the staff of Hong Kong University who served as subjects in Experiment 1 and the students and staff of the Center for Life and Learning of Louvain University for their help in carrying out Experiments 2 and 3. We would also like to thank Annette de Groot and David Balota for their comments on earlier versions of the paper, as well as MaryPotter, Judith Kroll, and a third reviewer for their comments on this paper. Correspondence should beaddressed to C. W. Keatley,

No abstract