hide
Free keywords:
-
Abstract:
We are very good at classifying familiar and unfamiliar faces in terms of their race or sex, but compared to the robust identification of familiar faces, discriminating unfamiliar faces, especially other-race faces, is more difficult (other-race effect). In this talk, I will present three studies investigating what is important in a face for race classification and person identification. First we investigated what gives a face its perceived ethnicity. To this end, mixed-race faces were created by embedding one
facial feature (e.g. Caucasian mouth) into the face of the other ethnicity (e.g. Asian face). The perceived ethnicity of these mixed-race faces was assessed in a classification task. The eyes and the texture (skin) proved to be major determinants of ethnicity for Asian and Caucasian participants. Second, we examined what is at the base of the other-race effect. We dissociated ethnicity from identity information by creating Asian and Caucasian faces that shared the same identity (e.g. making a Caucasian face look more Asian), and tested the other-race effect while controlling identity-related facial information. Participants showed equal race discrimination performance for same- and other-race faces. Thus no other-race effect
appeared when ethnicity was the only varying factor between the test faces, suggesting that the other-race effect cannot be attributed to face race per se. Finally, we tested what type of facial information is most relevant for the identification of familiar faces. We created both
sex-morphs and identity-morphs of very familiar faces, and asked participants to pick the original familiar face among its sex- or identity-morphs. We found a better performance for identity- than sex-manipulated faces, indicating that sex-related facial information is represented less accurately than identity-related information. The implications of these results for models of face
representation will be discussed.