A Unified Account of the Effects of Caricaturing Faces

Abstract

Explanations of the caricature advantage (caricatured faces are recognized faster than veridicals) usually involve an encoded prototypical norm face (Rhodes, Brennan, & Carey, 1987). As an alternative to these, the present study describes a mathematical framework based on Valentine's (1991) exemplar-based model, which accounts for the caricature advantage without reference to an explicitly encoded norm face. This framework encodes faces in a multidimensional Voronoi diagram based on normally distributed face-space representations. The properties of this framework are investigated geometrically and computationally. It is demonstrated how a parallel processing system can extract a Voronoi diagram from multidimensional representations. The framework was able to account for many of the empirical findings on face recognition and caricaturization without extracting a norm face.