Abstract
Identification of faces of a non-native race is drastically impaired compared to performance with own-race faces (Hancock & Rhodes, 2008; Meissner & Brigham, 2001). It is known that differential experience brings about this effect yet it is not clear how experience, or the lack thereof, with a particular race impacts neural processing of faces. It has been suggested that unlike own-race faces, other-race faces do not benefit from expert holistic processes but are recognized via general-purpose processes in a piece-meal fashion similar to other, non-face, visual forms (Rossion, 2008; Tanaka, Kiefer, & Bukach, 2004). Alternatively, the difference may stem from quantitative, rather than qualitative, changes in processing. Under this hypothesis, the neural computation that leads to recognition itself is the same in both own- and other-race faces; but in the case of other-race faces the input is noisier due to reduced experience. In order to discriminate between these alternative models we measured contrast recognition thresholds in a 5AFC paradigm for three classes of stimuli: a) own-race faces b) other-race faces, and c) houses. Contrast thresholds were measured in white noise and no-noise conditions. High-noise efficiency and equivalent internal noise for the human observers across the three stimulus conditions were computed relative to an ideal observer that performed the same tasks. We predicted lower efficiencies for houses, which are recognized via general-purpose processes, compared to own-race faces, which are recognized via expert holistic processes. We found both own-race and other-race efficiencies to be significantly higher than that for houses. Efficiencies did not differ between the two face conditions but equivalent input noise was higher for other-race faces. These results do not support the idea of distinct processing styles for own-vs. other-race faces. Instead, they suggest qualitatively similar processing regardless of race.
Meeting abstract presented at VSS 2016