Nitzan Guy¹ (), Galit Yovel¹; ¹Tel Aviv University

A well-established model of face identity posits that faces are represented as points in a multidimensional face space. According to a norm-based version of face space, faces are represented relative to a norm, with distinctive faces farther from the norm. Norm-based theories primarily focused on between-identity representations, where distinctiveness of each identity is evaluated relative to a norm, which is an average of different identity faces. These studies therefore overlooked how the rich representation of familiar faces influences the representation of face identity. In particular, it is unknown whether within-identity typicality – how typical an image of a particular familiar person – contributes to the norm-based face space representation of familiar faces. This study examines how within-identity typicality shapes norm-based face space representations and influences distinctiveness judgments of familiar faces. Our experiment included three rating tasks performed by different participants: (1) familiar distinctiveness: between-identity distinctiveness ratings (“how distinct the face is in a crowd of people”?) by participants familiar with the faces; (2) unfamiliar distinctiveness: between-identity distinctiveness ratings by unfamiliar participants; and (3) within-identity typicality: assessing how a/typical an image is of a specific familiar person. All three measures showed high split-half reliability scores. Results revealed that familiar distinctiveness ratings correlated positively with unfamiliar ratings, indicating shared between-identity perceptual representations. Additionally, images rated as more typical of a familiar identity were judged as more distinctive by participants familiar with them accounting for additional variance in between-identity distinctiveness rating beyond unfamiliar distinctiveness. Our study is the first to show the importance of within-identity representation of familiar faces to the norm-based face space. Within-identity variability is not only integral to representing familiar faces but also shapes between-identity distinctions. The findings have implications for face space frameworks of social perception and deep learning models of face recognition.