Abstract
When viewing groups of faces, people with typical face recognition are able to recognise individuals, as well as extract the average identity of the group, known as an 'ensemble' representation. It is currently uncertain whether dissociable processes underlie these two kinds of representations. People with congenital prosopagnosia (CP) have specific difficulties recognising individual's faces but it is currently unclear whether they also have difficulty extracting ensemble representations. Leib and colleagues (2012, Neuropsychologia) argued that ensemble coding of face identity was unimpaired in CPs (N = 4), consistent with the view that individual and ensemble processes may be dissociable. However, CPs in this study could have used non-identity, image-based strategies. We re-examined ensemble coding of identity in CP using a task that measured both individual recognition and ensemble coding. Participants were shown groups of faces and subsequently asked to decide whether target faces had been present in this group. Targets were either single faces (exemplars) or morphs of a group of faces (averages). Importantly, we could rule-out use of image-based strategies because both image- and identity-based versions were included. Target faces were either the same image (identical photograph) or same person (different photographs) respectively. Ensemble coding was measured by the extent to which group averages were endorsed as group members despite never being seen. CPs (N = 11) showed ensemble coding for both same-image and identity-based conditions, indicating that CPs do show identity-based ensemble coding. However, ensemble coding in both conditions was significantly weaker than for typical recognisers (N = 20). Recognition of individual faces was also weaker for the CPs. Overall our results indicate that ensemble coding of facial identity is weaker in CPs, consistent with their poor individual face recognition skills. Therefore, we found no evidence of a dissociation between individual recognition and identity ensemble coding.
Meeting abstract presented at VSS 2017