Categorical perception of newly learned faces

Abstract

Five experiments investigated identification and discrimination of faces. Stimuli were blends of two faces generated with a morphing algorithm. Two same-gender and two different-gender pairs of faces were tested. Experiment 1 (identification) estimated the point of indifference along the morphing sequence, and the associated differential threshold. Experiment 2 (discrimination, ABX) demonstrated that novel faces are perceived categorically. Identity was a more important factor than gender in generating the perceptual categories. Experiment 3 and 4 (identification) demonstrated that categories are generated progressively in the course of the experiment and depend on the range of morphs tested in any one condition. Confidence ratings (Experiment 5) showed that the multidimensional space where faces are represented can be collapsed onto a single dimension. Response probabilities and response times for Experiments 1–4 were predicted simultaneously by a counting model postulating that quanta of discriminal information are sampled independently from the stimuli.

This study was partly supported by research funds to PV from the University of Geneva and from the UHSR University of Milan. Jerôme Guinot implemented the Loki algorithm. We are grateful to Bruno Rossion and to two anonymous reviewers for their constructive criticisms of a previous version of this paper. We are also grateful to our friends Christelle, Francesca, Maria, Henry, and Philippe for letting us morph away their identity.