Word association responses in first-language (L1) Spanish and second-language (L2) English were investigated by means of response latencies and types of associative response produced. The primary aims were to establish whether (a) some response types are produced more often or faster than others, (b) participants' L2 response time profiles mirror those of their L1, and (c) participants' L2 association responses are mediated by their L1 and modulated by proficiency. Results indicate that responses are faster when a double association link is produced--that is, when the response is associated by form and meaning ("postman" [right arrow] "postbox") or meaning and collocation ("spider" [right arrow] "web"). L2 response time profiles broadly mirror those of the L1, although L2 times are generally slower. A significant priming effect from L1 translation equivalents of cues used in the L2 association task was observed, suggesting L1 mediation in the production of L2 associative responses. Findings are discussed in light of the revised hierarchical model (Kroll & Stewart, 1994). New approaches to modeling and understanding the bilingual lexicon are also suggested.