We spend our lives looking at other people’s faces. Although we seem to perceive them instantaneously as wholes, the face is made up of multiple different sources of overlapping information – about different socially-relevant characteristics of the face such as sex, age, race, emotional expression and eye gaze direction. The ease with which we perceive all of this information belies the immense computational complexity involved in processing it.

Each dimension relates to a different source of information about the person. Questions remain as to how information from these sources interact or interfere with one another in processing. The field of social vision has emerged to tackle such problems, integrating research in vision science and in social psychology around the idea that perceiving information about other people relies on the complex, dynamic interplay of different types of information. This information is either sensory or semantic; relating to that which is encoded through the senses, or to meaningful information about a given object that we have already encountered and represented in memory respectively.

In this thesis, I aim to extend the social vision approach to the problem of perceptual integration in face processing by exploring the under-investigated perceptual relationship between sex and age. I utilise five well-established experimental paradigms from the cognitive psychology literature to test participants' capacity to selectively attend to sex and age cues across several sets of faces. Namely, I carried out a Garner speeded classification task, an Eriksen flanker task, a face priming study, a word priming study, and a task-switching study.

Taken together, the studies suggest that sex and age are asymmetrically integrated in human face processing at sensory and semantic levels of representation, such that sex information interferes with age judgements more so than vice versa. I propose that a speed-of-processing account (SoPA) of dimension discriminability (Melara & Mounts, 1993) - by which it is assumed that the less discriminable (harder to process) dimension interferes more with the more discriminable (easier to process) dimension than vice versa - cannot fully account for the current results. I suggest an alternative to this account which incorporates a perceptual mechanism to account for differences in dimensional discriminability and asymmetric integration between sex and age: that each involves a different “depth-of-processing”, such that age perception can take place using relatively superficial or “shallow” featural processing, whilst sex perception typically requires relatively complex or “deep” configural processing. I call this the depth-of-processing account (DoPA)