The mental representation of ingroup and outgroup faces.

Ethnicity critically impacts perceivers' ability to individuate and recognize faces. Valentine (1991) proposed a face space model in part to account for these effects, and although it has received significant attention, basic questions derived from that model have yet to be satisfactorily tested. Across three large-scale studies, over 10,000 human participants provided similarity judgments of pairs of faces. All studies used a full ingroup-outgroup design, such that participants rated both ingroup and outgroup faces. From these ratings, we estimate the configuration of an empirical multidimensional face space. Based on these configurations, we test five questions central to the face space model and to underlying mechanisms like perceptual expertise. We find that eight dimensions capture most of the variation in ratings and that these dimensions correspond to existing models of face perception. Contrary to widespread hypotheses, we find no evidence that perception is "tuned" based on ethnicity. Rather, perceivers of different ethnic groups demonstrate very similar configurations of face space. Also surprisingly, we observe evidence of outgroup homogeneity in only one study, when intergroup contact is particularly low. Finally, we find that sensitivity to face ethnicity is negatively correlated with sensitivity to variation on other dimensions and that greater intergroup contact is associated with reduced sensitivity to face ethnicity and enhanced sensitivity to other dimensions. These findings seem broadly inconsistent with exemplar-based coding accounts and offer relative support for norm-based coding. We discuss the implications of these findings for Valentine's face space framework and for face perception in intergroup contexts. (PsycInfo Database Record (c) 2025 APA, all rights reserved).