ABSTRACT
This chapter considers what is known about the representations which subserve face recognition in order to think about the possible nature of the dimensions of face space. It describes attempts to correlate the physical deviation of faces with the psychological variable of distinctiveness. In the psychological and the computer science literature there has been a tendency to treat the face as though it were a two-dimensional picture, within which key points or edges can be located and measured. The chapter suggests that an image-based coding of faces might implicitly convey all the other possible ways of measuring face structure, and that the principal components of variation of face images may provide a way to describe 'face space'. It reviews effects of surface detail, negation and lighting on face recognition, which strongly suggest that an image-based approach such as principal components analysis may be more psychologically valid than approaches based on the explicit extraction and measurement of face 'features'.
