ABSTRACT
The ability to distinguish between one face and another forms a focus of most common-sense ideas about face recognition, but, as we describe in this chapter and in the next, it is also important to understand how we ‘tell faces together’ when different images are actually variants of the same person’s face. Since Galton’s speculations about the process, the sources of information used in face recognition have been carefully explored, offering useful insights into how we achieve this feat. As we will see, Galton was probably wrong to suggest rapid but serial inspection of different aspects of a single face. Holistic processing of faces, where all elements are perceived together, is the explanation most favoured by today’s evidence. In this chapter we review this evidence. We describe what is known about the holistic visual representations which allow us to recognise faces, and how representations and processes may change as faces become familiar. We describe how inversion and photographic negation affect face recognition and the impact of mismatch in ethnicity or age between observers and faces. We revisit adaptation effects and consider the widely used ‘face-space’ framework and its limitations. Most of the effects we describe are consistent with faces being represented holistically via low-level, image-based analysis. In the final section, we show how recent work on face recognition ‘in the wild’ has emphasised the importance of variations in facial appearance as faces become familiar.
