Expertise and configural coding in face recognition

Abstract

Diamond & Carey (1986) have argued that expertise in face recognition depends on the ability to code configural properties in addition to isolated features. We tested this hypothesis in two experiments by comparing the effect of inversion on recognition of ‘own race’ (high expertise) and ‘other race’ (low expertise) faces. Use of configural information should be associated with a larger inversion effect than use of isolated features, and therefore inversion should produce a larger recognition decrement for own race than for other race faces. In Expt 1 there was a larger inversion effect in reaction times for recognition of own race faces than other race faces, for both European and Chinese subjects (ceiling effects made interpretation of accuracy difficult). In Expt 2 a larger own race inversion effect was found for recognition accuracy, when test face pairs were randomly selected, but not when they were matched on isolated features. Our results are largely consistent with the hypothesis that expertise is associated with greater use of configural information in faces.