Abstract
The inability of the visual system to process detailed information from multiple, simultaneously-viewed faces (i.e., an ensemble) is circumvented by encoding the various identities into a single summary representation. The attributes important in forming representations of face ensembles have yet to be elucidated. In single-face perception, attributes of shape and surface properties are both used to process identity, with differing accounts of which attribute is more influential. In this study, we investigated the importance of shape and surface properties in ensemble face processing. In each trial, participants saw an ensemble of six faces, after which they were shown a single probe face and asked to report whether it was a member of the preceding ensemble. The probe could be a member of the set, the average identity of the set, a member of a different set, or the average identity of a different set. Importantly, performance was assessed on trials where only shape, only surface, or both attributes varied. Consistent with the dominance of global processing reported in the ensemble literature, we found that participants were biased to report the average identity as a member, despite it never being explicitly seen in any ensemble. Moreover, participants were above chance at correctly identifying single faces as set members and rejecting faces from different sets. Interestingly, when shape changed, participants were more likely to mistake faces from different ensembles as members of the target set, relative to when surface properties or both attributes changed. This suggests a dominant role of surface properties in ensemble face processing, which is consistent with the known link between texture and ensemble processing (Cant & Xu, 2012). These findings have important implications for bridging models of single-face and ensemble-face processing, potentially revealing how different facial attributes contribute to the representation of identity in both domains.
Acknowledgement: NSERC Discovery Grant to Jonathan S. Cant