It is often assumed that social models influence people's eating behavior by providing a norm of appropriate food intake, but this hypothesis has not been directly tested. In three experiments, female participants were exposed to a low-intake model, a high-intake model, or no model (control condition). Experiments 1 and 2 used a remote-confederate manipulation and were conducted in the context of a cookie taste test. Experiment 3 used a live confederate and was conducted in the context of a task during which participants were given incidental access to food. Participants also rated the extent to which their food intake was influenced by a variety of factors (e.g., hunger, taste, how much others ate). In all three experiments, participants in the low-intake conditions ate less than did participants in the high-intake conditions, and also reported a lower perceived norm of appropriate intake. Furthermore, perceived norms of appropriate intake mediated the effects of the social model on participants' food intake. Despite the observed effects of the social models, participants were much more likely to indicate that their food intake was influenced by taste and hunger than by the behavior of the social models. Thus, social models appear to influence food intake by providing a norm of appropriate eating behavior, but people may be unaware of the influence of a social model on their behavior.