Address for correspondence:

research into face processing has produced considerable technical and theoretical advances. For example, it is possible to generate photographic-quality colour caricatures of faces; principal component analysis can be used to provide efficient storage of facial images; laser scans can produce a 3D model of an individual's head that can be manipulated by use of software; a model head can be animated to appear to speak, and facial images can be ‘aged ’ by application of a mathematical algorithm. Nevertheless, the question ‘How do we recognise faces? ’ remains extremely difficult to answer with any precision. In contrast, formal models have been used successfully to account for human performance in recognising, categorising and identifying artificial concepts. In this chapter the rôle of formal models in providing a basis for understanding our ability to recognise faces in the real world will be critically evaluated. Figure 1: An example of a set of schematic faces. Similar stimulus sets have been used in many categorisation experiments. The faces differ in the distance between the eyes, the position of the eyes, the length of the nose and the position of the mouth. Can formal models of concept representation tell us anything about face recognition