From Unfamiliar Face Processing to Familiar Face Recognition: Behavioural and Neural Effects of Real World Face Learning

There are large differences between familiar and unfamiliar face recognition, however, progress in the field has been slow (Burton, 2013). Familiar face recognition is faster (Clutterbuck & Johnston, 2002), more accurate (Burton, Wilson, Cowan & Bruce, 1999), robust to changes in viewpoint (Hill & Bruce, 1996) and expression (Bruce, 1982). To date, these studies have used traditional methods of face familiarisation in which participants are exposed to different pictures of the target face. The central aim of my proposed PhD project is to establish a new method of face familiarisation based on real-world social interactions in which participants interact with actors whom they have never met. This is important for three reasons; firstly, face recognition happens every day and in every social interaction, my research adds real world validity to lab-based experiments. Secondly, these real-world interactions are likely to drive stronger neural responses that will increase the power of my proposed fMRI studies. Finally, it is important to improve the methods used to tease apart the differences between familiar and unfamiliar face recognition as they have important implications in the use of eye-witness testimony (Brewer & Wells, 2011) and in the use of photo ID in airports and other high-security environments (White, Kemp, Jenkins, Matheson & Burton, 2014).

Although the behavioural effects of familiar face recognition are well studied, the neural correlates are less well established. The occipital face area, fusiform face area, and superior temporal sulcus are face-selective regions, they respond more to faces than objects or scenes (Kanwisher, McDermott & Chun, 1997; Pitcher, Dilks, Saxe, Triantafyllou & Kanwisher, 2011). However, neuroimaging and neurostimulation studies examining face familiarity in the brain have produced mixed results. Some fMRI studies find increased FFA activation to familiar faces compared to unfamiliar (Rotshtein, Henson, Treves, Driver & Dolan, 2005), and others find no such modulation (Dubois et al., 1999). Some TMS studies demonstrate the OFA represents face parts in an early processing stage prior to identity recognition (Pitcher, Walsh, Yovel & Duchaine, 2007; Pitcher, Walsh & Duchaine, 2011), whereas other TMS experiments find the OFA plays a role in face recognition (Solomon-Harris, Mullin & Steeves, 2013).

Crucially, all these prior studies have used conventional methods of face familiarisation. This is an issue as such studies use 1 photo to represent the face to be learnt. However, faces display within-person variation which cannot be captured by 1 photograph (Jenkins, White, Van Montfort & Burton, 2011). Familiarisation techniques using ambient images instead, natural photographs of an individual under different lighting conditions and angles etc., could be used as they are not blind to within-person variation. Moreover, live interaction with a target face also captures within-person variation but in an ecologically valid way.
Furthermore, face learning is not yet fully understood. It is known that longer durations of time observing a face leads to better face recognition (Downes et al., 1997), however, more recent studies have shown that exposure to variation during face learning, e.g. presenting faces from different viewpoints, also leads to better face recognition (Dwyer, Mundy, Vladeanu & Honey, 2009). The precise weightings of variation and length of time have not yet been systematically tested. Therefore, further study is required to tease apart the different factors of face learning.

In my master's project (Sliwinska et al., in prep), participants interacted with target identities in person, a new highly ecologically valid familiarisation technique, which lead to 15% increases in recognition accuracy. In my PhD, I aim to refine this new and promising method to clarify the conflicting literature on the OFA and FFA's roles using fMRI and TMS.