Abstract
Although the other-race effect (ORE) is a very reliable finding, its underlying mechanisms are still under debate. This study is based on seemingly paradoxical findings in the face learning literature: While other-race faces and caricatures of same-race faces evoke very similar patterns of event-related potentials (smaller P200 and larger N250 components compared to veridical same-race faces), behavioural effects are exactly the opposite (better performance for caricatures, poorer performance for other-race faces). This could suggest qualitatively similar processes for both types of faces at learning, but with different consequences for recognition: When learning any unfamiliar face, deviations from the norm are used for forming a basic mental representation. Such distinctive information is useful in the case of caricatures, because the deviations from the norm are in different directions for each individual face, but misleading for other-race faces, because the most salient deviation from the norm is in the same direction for all members. We tested this idea by using highly distinctive same-race (Caucasian) faces with all noses manipulated in a uniform direction. In a learning/recognition task, we compared performance for these faces to veridical same- and other-race (Asian) faces. Our main aim was to simulate an ORE with the highly distinctive "big-nose" same-race faces. In accuracies and RTs, we found significant costs both for "big-nose" and other-race faces, compared to same-race veridicals. In ERPs, we observed a similar pattern for "big-nose" and other-race faces, with smaller P200, larger N250 and larger LPC compared to veridical same-race faces. Overall, our results support a perceptual account of the ORE. However, they suggest that qualitatively similar processes mediate the learning of unfamiliar same- and other-race faces, but with different consequences due to differences in the usefulness of the respective distinctive information.
Meeting abstract presented at VSS 2018