Abstract
The visual system encodes summary face information (e.g., an average identity) from groups of simultaneously presented faces (i.e., an ensemble), which can be decoded and visualized from EEG data (Roberts et al., VSS 2018). However, the interplay between the processing of an ensemble and specific target faces within the set remains to be clarified. Moreover, the relative contribution of different facial attributes, such as shape and surface properties, to ensemble encoding remains to be explored. To investigate these issues, we conducted an EEG study in which face ensembles could vary in shape properties, surface properties, or both. Each ensemble contained six faces surrounding a seventh central face which was either the average identity of that ensemble (consistent face; CF), or the average identity of a different ensemble (inconsistent face; IF). Importantly, facial ensemble variability in shape and surface properties was matched across the two attributes. Pattern analysis of EEG data revealed, first, that the central face within an ensemble could be decoded regardless of whether it was embedded in CF or IF ensembles. Second, CF and IF ensembles could be discriminated above chance, which was driven primarily by surface information in the proximity of the N170 ERP component. Third, overall decoding was more successful for ensembles varying in surface properties or in both properties relative to shape properties only. These results resonate with the finding that texture and ensemble perception rely on shared neuroanatomical substrates (Cant & Xu, 2012). Further, they are consistent with the dominant role of surface information in single-face representations as revealed by neural and behavioral data (Nemrodov et al., 2019). More generally, these findings further our understanding of the factors mediating face ensemble processing, and suggest that shape and surface properties both contribute, in differing degrees, to the representation of identity in crowds of faces.