Computational principles of first impressions: Distinct codes for typicality and attractiveness

Humans spontaneously form first impressions of others, ranging from recognizing who is typical to evaluating who is attractive. We show these judgments arise from correlated but distinct coding principles. Using 3D body perception as a model domain, participants from the United States and China rated bodies sampled from a continuous statistical space. Typicality tracked proximity to the population prototype, consistent with norm-based coding. Attractiveness departed from this norm, favoring bodies with greater curvature or muscularity, consistent with dimension-prioritization coding that amplifies specific features. In a follow-up experiment, we created exaggerated versions of typical and attractive bodies to accentuate these differences. Results showed that bodies were judged as more attractive the farther they moved away from the typical caricature. These findings show that first impressions rely on distinct principles within a shared representational space, explaining how people judge fundamental traits that guide social interactions.