Is All Face Processing Holistic? The View from UCSD

There has been a great deal of progress, as well as controversy, in understanding how complex objects, in particular, human faces, are processed by the cortex. At the same time, sophisticated neural network models have been developed that do many of the same tasks required by these cortical areas. Such simplifying models allow us to explore hypotheses concerning relatively complex domains such as face processing. In this chapter, we give a somewhat idiosyncratic history of the development of neural network models of face processing, concentrating on work at UCSD, and show how these models have led to a novel hypothesis concerning processing of facial expression. While our models have suggested a role for holistic representations of faces in identification, general local features appear to be important in recognition of expression. Is all face processing holistic? 3