Stereotype priming can lead to assimilation or contrast effects on behavior. We argue that self-activation is a moderator of both assimilation and contrast effects. To test this hypothesis, in two studies, we activated independent or interdependent self-knowledge before priming participants with the dumb Blonde stereotype or a control category (Study 1) or no prime (Study 2). Participants then answered a knowledge test. Results support our expectations: Participants presented assimilation under interdependence (i.e., underperformance compared to control group) while they presented no assimilation (i.e., comparable performance with control group in Study 1) and contrast (better performance than control group in Study 2) under independence. We discuss implications of these findings in regards of previous research and recent models such as the Active Self Account [Wheeler, S. C., DeMarree, K. G., & Petty, R. E. (2005). The roles of the self in priming-to-behavior effects. In A. Tesser, J. V. Wood, & D. A. Stapel (Eds.), On building, defending and regulating the self: A psychological perspective (pp. 245–271). New York, NY, USA: Psychology Press].