We investigated the consequences of bilingualism for verbal fluency by comparing bilinguals to monolinguals, and dominant versus non-dominant-language fluency. In Experiment 1, bilinguals produced fewer correct responses, slower first response times and proportionally delayed retrieval, relative to monolinguals. In Experiment 2, similar results were obtained comparing the dominant to the non-dominant languages within bilinguals. Additionally, bilinguals produced significantly lower-frequency words and a greater proportion of cognate responses than monolinguals, and bilinguals produced more cross-language intrusion errors when speaking the non-dominant language, but almost no such intrusions when speaking the dominant language. These results support an analogy between bilingualism and dual-task effects (Rohrer et al., 1995), implying a role for between-language interference in explaining the bilingual fluency disadvantage, and suggest that bilingual fluency will be maximized under testing conditions that minimize such interference. More generally, the findings suggest a role for selection by competition in language production, and that such competition is more influential in relatively unconstrained production tasks.