Face Familiarity and Similarity: Within- and Between-Identity Representations Are Altered by Learning

Face familiarity is thought to alter distances between representations in psychological “face space,” resulting in substantial improvements in recognition. However, the underlying changes are not well understood. In Experiment 1 (n = 192), we investigated the effect of familiarity based on everyday exposure to celebrities. Participants judged the similarity of pairs of face photographs, and we found that greater familiarity increased perceived similarity for two images of the same person, while decreasing similarity for two images depicting different people. In Experiment 2 (n =157), familiarity was manipulated through the learning of new identities by watching 5-min video clips. Again, when judging the similarity of image pairs, familiarity increased the perceived similarity of images of the same person, while having the opposite effect on images depicting different people. In Experiment 3, we trained a computational model with images of 333 different identities (totaling 3,949 photographs) and manipulated its familiarity with two new identities. The changes in distances between novel images of these identities (a proxy for similarity) replicated our behavioral findings. Overall, we build upon recent evidence by demonstrating two transformations through which familiarity alters representational space to likely benefit face perception.