Four face overgeneralization hypotheses are offered to explain consensual first impressions of faces. These hypotheses hold that the psychological qualities that are accurately revealed by the functionally significant facial qualities that mark babies, emotion, a familiar identity, or unfitness are overgeneralized to people whose facial structure resembles that of babies, a particular emotion, a particular identity, or a particular level of fitness. Research supporting the first three hypotheses is briefly reviewed and recent studies supporting the fourth, anomalous face overgeneralization hypothesis, are discussed in more detail. The results supported the hypothesis that accurate impressions of faces that signal low fitness are overgeneralized. Traits indicative of fitness were accurately perceived to vary from low to moderate levels of attractiveness or averageness in a representative sample, while they were erroneously perceived to vary from moderate to high levels, and impressions of the fitness of normal faces were predicted by their resemblance to anomalous ones. The present findings reveal the value of evolutionary theory and ecological theory for generating hypotheses that elucidate the origin of consensual first impressions, a socially significant topic that is less readily addressed within a traditional social cognition framework.