Using data from four national surveys, we examined the role of affective-cognitive ambivalence in moderating the relative impact of affect and cognition on overall attitudes and behavior. Each survey assessed the affective and cognitive components of attitudes toward presidential candidates, as well as overall candidate attitudes and reported voting behavior. We found support for a primacy of affect (vs. cognition) effect among respondents withambivalentaffective-cognitive structures: For respondents with oppositely valenced affect and cognition, affect generally exerted a stronger influence on candidate attitudes and voting behavior than did cognition. However, for respondents withunivalentaffective-cognitive structures (i.e., similarly valenced affect and cognition), affect and cognition exerted a roughly equal influence on overall attitudes and voting behavior. Results are discussed in terms of the processes through which the ambivalence-moderated primacy of affect effect occurs, and its potential consequences.