Abstract
Adapting to a distorted face induces an aftereffect in the appearance of a subsequently viewed test face. Previous studies have found clear transfer of the adaptation across stimulus changes including size (Zhao and Chubb, Vision Research, 2001; Leopold et al. Nature Neuroscience, 2001) and color (Yamashita et al., IOVS, 2000). We compared the degree of selectivity for a range of image properties to characterize which stimulus dimensions the adaptation might be contingent upon. The properties examined included spatial variations (changes in size or spatial frequency content) and contrast variations (changes in color, contrast, or contrast polarity). Observers adapted to a single image (expanded or contracted) or to opposite distortions in two different images (e.g. large expanded face vs. small contracted face) that alternated at a rate of 1 image per second. They then judged the appearance of a test face (presented for 1 sec interleaved with 6-sec adaptation topups) using a 2AFC staircase task that varied the expansion or contraction of the image in fine steps. Selectivity was assessed by comparing the aftereffects induced by the same (e.g. both large) or different (e.g. large adapt, small test) images. The degree of selectivity varied widely for different image manipulations. In general the strongest selectivity was found for differences in frequency or contrast polarity, while the weakest for color or high vs. low contrast. These nominal differences are roughly related to the degree to which the adapt and test images appear drawn from a similar original face, and consistent with this, stronger selectivity was also found between faces of different individuals that were rated as more dissimilar. This suggests that a component of the adaptation is specific for higher-level properties of the images that are related to facial identity.