Learned categorical perception for natural faces

Abstract

We present an experiment involving learned categorical perception for natural faces, using four different category learning tasks. In the four tasks, participants learned to classify faces divided on the basis of gender, hair color, a subjective determined level of ‘trust’, or at random. After category learning, participants rated the similarity of each pair of faces, and their judgments were compared to a previously collected ‘base rate’ set of similarities for the same stimuli. Evidence for learned categorical perception was then sought, in the form of increased differences between intra-category and inter-category similarities. Over the four conditions, we observed no learned categorical perception when categories were based on obvious properties (gender and hair color), nor when the category structure was intentionally random. However, when the loosely defined category structure of trust was employed, a learned categorical perception effect emerged.