Learning faces from variability

There are large behavioural differences in the perception of familiar and unfamiliar faces. However, little is known about face learning - how faces make the transition from unfamiliar to familiar. Most experimental work on this topic examines the effects of study time, and systematic variation of exposure (changes in pose etc). We have argued that a critical component of face learning is within-person variability - i.e. exposure to the range of naturally-occurring variations which are idiosyncratic for a particular face. Here, we present two experiments which manipulate the degree of variability to which viewers are exposed during face learning. Participants learned name and face associations for twenty unfamiliar identities presented with high and low within-person variability. Stimuli were ambient images, showing faces 'in the wild'. We show more accurate performance on a speeded name verification task for identities learned in high compared to low variability. We go on to show that exposure to high variability across just two images improves performance on a face matching task. The results demonstrate the critical role of variability in face learning. Meeting abstract presented at VSS 2015.