Abstract
Across many different races, people are better at recognizing own-race faces than faces from other races. Recently, we (Hayward, et al., 2008) reported that, following study of whole faces, performance in an old-new memory task for scrambled (to isolate component processing) and blurred (to isolate configural processing) faces showed an own-race advantage (ORA) in both conditions. In addition, memory for blurred faces was better than for scrambled faces, supporting the importance of configural processing in judgments of facial identity. Here, we present further investigation of these effects. In Experiment 1 we replicated our earlier experiment but inverted the stimuli; participants saw inverted normal faces at study and inverted blurred or scrambled faces at test; this manipulation eliminated the difference between blurred and scrambled faces, consistent with the assumption that face inversion differentially impairs configural processing. However, across both conditions, an own-race advantage was still observed. In Experiment 2, we wanted to test memory performance when participants encoded only one type of information; participants studied upright scrambled or blurred faces, and then were tested with the same type of image, or with normal faces. Study of blurred faces led to a small ORA for blurred and normal test images, but study of scrambled faces led to an own-race disadvantage, where old-new discrimination was better for other-race scrambled and normal faces than for own-race images. Taken together, these results suggest a dissociation in processing of configural and component face information in the ORA; configural information leads to an own-race advantage regardless of study conditions, but the component ORA appears to require encoding of the whole face, upright or inverted. Intriguingly, the lack of whole face information may impair the encoding of face components from own-race faces, as compared with other-race faces.
This work was supported by a grant from the HK Research Grants Council (HKU4653/05H) to WGH, and a grant from the Australian Research Council to GR.