Two experiments with highly fluent Spanish-English bilinguals examined repetition priming of picture identification and word retrieval in picture naming. In Experiment 1, between-language priming of picture naming was symmetric, but within-language priming was stronger in the non-dominant language. In Experiment 2, priming between picture naming and translation was symmetric within the dominant language and within the non-dominant language, but priming was stronger in the non-dominant language. A mathematical model required only three process parameters to explain the pattern of priming across eight conditions. These results indicate that shared processes are the basis of priming, that difficulty influences priming only at the process level, and that translation in both directions is concept mediated in fluent bilinguals.