Abstract
Variations in faces are often judged and classified in terms of discrete categories (e.g. identity, gender, race, and emotional expression). We examined how the boundaries defining these categories can be influenced by prior adaptation to images within a given category, in order to test whether adaptation effects can be manifest for the types of natural variations and judgments typically involved in face perception. Stimuli were frontal-view images of faces taken from the Ekman and Matsumoto image set. Pairs of faces were morphed (Gryphon Morph v. 1.5) to form a finely graded series of images that varied between two genders, expressions, identities, or ethnicities (Caucasian and Japanese). Observers initially adapted for 1 to 3 minutes to one of the two original faces, and then made a forced-choice classification of a composite morphed image (e.g. responding male vs. female, or happy vs. angry). The proportion of the two faces making up the morph was varied over trials in a staircase to define the category boundary at which the two judgments were equally likely. Adaptation produced strong shifts in category boundaries toward the adapting image, so that the original neutral point appeared biased away from the adapting image and thus toward the second, more novel image. For example, after adapting to a male face or an expression of surprise, a face that was previously judged to be neutral was rated as female or happy, respectively. These adaptation effects are consistent with a renormalization of facial categories according to the observers' recent visual experience. The fact that large aftereffects can rapidly occur for natural and common variations in faces suggests that adaptation may routinely influence face perception in everyday viewing.