The direction and magnitude of transfer from stimulus familiarization (SF) training to a subsequent paired-associate (PA) task were assessed under 18 variations of 3 factors, namely, presence or absence of irrelevant items in the SF task, frequency of presentation of items in SF, and mode of responding (articulation or nonarticulation) to the words both during the SF task and when they appeared as stimulus terms in the subsequent PA task. When only relevant words were familiarized, both positive and negative transfer were obtained and systematically related to response mode and frequency. Introduction of irrelevant items to control for total amount of SF time abolished the consistency and predictability of SF effects. Positive transfer was discussed as the expected outcome of SF and negative transfer as an artifact of certain experimental procedures.