Abstract Previous research has found that masked repetition primes, presented immediately prior to the test item in a recognition memory test, increase the likelihood that participants think that the item was present in a previous study phase, even if it was not. This memory illusion is normally associated with a feeling of familiarity, rather than recollection (e.g., as indexed by Remember/Know judgments), and has been explained in terms of an increased fluency of processing the test item, which, in the absence of awareness of the cause of that fluency (i.e., the masked prime), is attributed instead to prior exposure in the study phase. Recently however, we have found that masked conceptual primes (semantically rather than associatively related to the test item) have the opposite effect of increasing Remember but not Know judgments. This result appears difficult to explain in terms of existing theories of recollection and familiarity. Here we report data from a functional magnetic resonance imaging study using the same design, in which we replicate our previous behavioral findings, and find converging evidence for increased activity following conceptual primes in brain regions associated with recollection. This neural evidence supports an account in terms of “true” recollection (for example, conceptual primes reactivating semantically related information that was generated at encoding), rather than an artifact of the mutually-exclusive nature of the Remember/Know procedure.