The relationship between recognition memory and repetition priming remains unclear. Priming is believed to reflect increased processing fluency for previously studied items relative to new items. Manipulations that affect fluency can also affect the likelihood that participants will judge items as studied in recognition tasks. This attribution of fluency to memory has been related to the familiarity process, as distinct from the recollection process, that is assumed by dual-process models of recognition memory. To investigate the time courses and neural sources of fluency, familiarity, and recollection, we conducted an event-related potential (ERP) study of recognition memory using masked priming of test cues and a remember/know paradigm. During the recognition test, studied and unstudied words were preceded by a brief, masked word that was either the same or different. Participants decided quickly whether each item had been studied ("old" or "new"), and for items called old, indicated whether they "remembered" (R) the encoding event, or simply "knew" (K) the item had been studied. Masked priming increased the proportion of K, but not R, judgments. Priming also decreased response times for hits but not correct rejections (CRs). Four distinct ERP effects were found. A medial-frontal FN400 (300-500 msec) was associated with familiarity (R, K Hits > CRs) and a centro-parietal late positivity (500-800 msec) with recollection (R Hits > K Hits, CRs). A long-term repetition effect was found for studied items judged "new" (Misses > CRs) in the same time window as the FN400, but with a posterior distribution. Finally, a centrally distributed masked priming effect was visible between 150 and 250 msec and continued into the 300-500 msec time window, where it was topographically dissociable from the FN400. These results suggest that multiple neural signals are associated with repetition and potentially contribute to recognition memory.