Abstract
Background: It is hypothesized that upright faces are represented holistically in the human visual system, with only local part-based feature representations for inverted faces. Objective: To test this, we used a novel inter-feature adaptation transfer technique to determine if face aftereffects resulted from adaptation in whole-face representations. Method: We divided face images into upper and lower halves, and examined first if each half created aftereffects for its own perception (within-feature adaptation), and second if each half created aftereffects for perception of the other half (inter-feature transfer). This was done for upright face identity, inverted face identity, and expression judgments for upright happy versus sad faces. Results: Inter-feature transfer of adaptation was significant for upright face identity, and equivalent in magnitude to within-feature adaptation. No significant inter-feature transfer was found for inverted identity aftereffects. For upright facial expression, we found robust within-feature adaptation but no inter-feature transfer. Conclusions: Most, if not all, identity aftereffects of upright faces may be generated in integrated whole-face representations, but not for inverted identities, consistent with hypotheses of an orientation-dependent expert holistic mechanism. However, our findings for expression suggest that happy/sad features in upright faces may be processed at a featural level.
MD was supported by a grant from Fight for Sight and a SIGN fellowship from the American Academy of Neurology. This work was supported by NSERC Discovery Grant RGPIN 355879-08.