Based on attribution theory and the logic of conversational norms, we predicted that image-based health communications can alter prevalence estimates for health behaviors. In two studies, participants were exposed either to a positively-framed or negatively-framed communication advocating for specific health behaviors. As predicted, participants who read a health communication rated healthy behaviors as less common when positive attributes were associated with healthy choices than when negative attributes were associated with unhealthy choices. The second study revealed that this pattern was most pronounced among participants who reported initial uncertainty about behavioral norms. These findings suggest that positively-framed influence attempts can promote prevalence assumptions that work against the influence attempt. Copyright © 2002 John Wiley & Sons, Ltd.