There has been a discrepancy among past studies with regard to the relation between negative priming and familiarity of stimuli. That is, Malley and Strayer (1995; Strayer & Grison, 1999) reported that the more familiar the stimuli were, the larger negative priming became (i.e., a positive correlation), whereas DeSchepper and Treisman (1996; Treisman & DeSchepper, 1996) reported that the less familiar the stimuli were, the larger negative priming became (i.e., a negative correlation). These studies differ not only in their experimental tasks (identification vs. matching) but also in their respective manners of arranging unfamiliar stimuli (pure vs. mixed). In the present study, using an identical set of stimuli, we examined whether these factors caused the opposite results. An identification task with a pure arrangement produced a positive correlation, and a matching task with a mixed arrangement produced a negative correlation. These results suggest that the past opposing results are both replicable and that they have reflected the different causal mechanisms of negative priming.