The "Perruchet effect" refers to a dissociation between the conscious expectancy of an outcome and the strength or speed of responding in anticipation of that outcome. This dissociation is considered by some to be the best evidence for multiple learning processes with expectancy governed by participants' explicit beliefs and responding driven by the associative history of the cues that partially predict the outcome. However, an alternative nonassociative explanation is that the trends in responding are the result of recent experience with the same outcome (i.e., repetition priming based on event recency). This explanation casts doubt on the theoretical import of the dissociation because it suggests that associative learning may not be involved in generating the observed trends in response strength. Associative accounts of the Perruchet effect predict a weakening of the response strength trends when the cues perfectly predict the occurrence of the outcome. In two experiments, we compared a condition with two cues that were each perfect predictors of an outcome with control conditions in which the cues did not perfectly predict the outcome. In both experiments, the typical downward trend in response time (RT) observed in Perruchet effect experiments was substantially weaker (and indeed absent) for the predictive group, suggesting an associative contribution to the effect.