Purchase this article with an account.
Daniel Fiset, Caroline Blais, Frédéric Gosselin, Daniel Bub, Jim Tanaka; Potent features for the categorization of Caucasian, African American, and Asian faces in Caucasian observers. Journal of Vision 2008;8(6):258. doi: https://doi.org/10.1167/8.6.258.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
What is the information mediating race categorization? Here, we applied the Bubbles technique (Gosselin & Schyns, 2001) to reveal which areas of faces at five different spatial scales are efficient for race categorization in Caucasian participants. We asked 30 participants to categorize 700 “bubblized” faces selected randomly from sets of 100 male Caucasian faces, 100 male African American faces, and 100 male Asian faces. All face photos were normalized using SHINE, a new Matlab algorithm for luminance and power spectrum equalization (Willenbockel et al., in preparation). Separate multiple linear regressions between information samples and accuracy were performed for each race. The resulting classification images reveal the potent features for the categorization of Caucasian, African American, and Asian faces in Caucasian observers. For African American faces, the participants used mainly the nose and the mouth in the spatial frequency (SF) bands ranging from 10 to 42 cycles per face width. For Asian faces, they used the eyes in the SF bands ranging from 10 to 84 cycles per face width and the mouth in the SF band ranging from 5 to 10 cycles per face width. For Caucasian faces, they efficiently employed the eyes in the SF bands ranging from 5 to 21 cycles per face width as well as the mouth and the region between the eyes in the second highest SF band ranging from 21 to 42 cycles per face width. Interestingly, and congruently with the results of Smith et al. (2005), we observed almost no overlap between the information used for each stimulus category.
This PDF is available to Subscribers Only