Discrimination of familiar and unfamiliar synthetic faces by North Americans and Koreans

Abstract

A new class of synthetic-face stimuli has been recently introduced for studying visual face processing. Synthetic faces are extracted from digital photographs of individual faces and bandpass-filtered after converting into radial frequencies for the head shape. In face-discrimination experiments Wilson et al (2002, submitted), using synthetic-face cubes, found that discrimination thresholds for highly distinctive face cubes centred on a non-mean face were 1.45 times higher than for faces centred on the mean face. These results might give us a quantitative insight into 'the other-race effect'. To investigate further this problem, we measured the face-discrimination thresholds for non-North-American faces, in this case Korean faces with Korean subjects. We found that the increment thresholds for mean face cubes for three subjects averaged 4.4% in front view and 5.6% in side view. However, the thresholds rose to 6.1% and 7.8% for non-mean front and side face cubes, respectively. In good agreement with the previous study, these results indicate that we do have finer discrimination ability for familiar faces than for unfamiliar faces. We will switch North-American and Korean faces and test each with the other group to get a quantitative measure of the 'other-race effect'.