Although research has revealed that the activation of a social category can influence subsequent behavior, a number of unresolved issues remain. In Experiment 1, we investigated whether people's ability to learn new information would be influenced by accessible knowledge representations. Extending previous research, we predicted that behavior would differ as a function of the type and direction of primed information. The results supported the hypothesis: Test performance on a previously unfamiliar topic was enhanced when individuals had previously thought about either a category associated with intelligence or an exemplar from a category associated with unintelligence. In Experiment 2, we investigated whether a common set of priming stimuli can produce both assimilation and contrast effects. Based on Mussweiler's (in press) Selective Accessibility Model, we predicted that having participants focus on the similarities among a set of supermodels would elicit behavioral assimilation, whereas a dissimilarity focus would elicit behavioral contrast. The results supported the hypothesis: Participants behaved unintelligently under a similarity focus and intelligently under a dissimilarity focus. The implications of the findings for perception-behavior effects and the Selective Accessibility Model are discussed.