Abstract
Humans easily recognize faces, despite the complexity of the task and the visual system which underlies it. Different spatial regions, including both core and extended face networks, and distinct temporal stages of processing have been implicated in face recognition and there is ongoing controversy regarding the extent to which the mechanisms for recognizing a familiar face differ from those of an unfamiliar face. Here, we used electroencephalogram (EEG) and frequency tagging, a high signal-to-noise approach, and whole brain decoding methods to elucidate the processes that subserve familiar and unfamiliar face recognition. Familiar and unfamiliar faces were presented periodically at 15Hz, 6Hz and 3.75Hz either upright or inverted in separate blocks, with the rationale that faster frequencies allow shorter processing times per image. Decoding of 15 Hz trials, which likely evince early visual processing, revealed similar classification for upright and inverted faces for familiar and unfamiliar. At 6 Hz, decoding familiarity was more accurate for upright compared with inverted faces. Decoding face orientation (upright versus inverted) was more accurate for unfamiliar than familiar faces, indicative of view-invariant pattern recognition associated with familiar faces and implicating the core face network. Finally, at 3.75 Hz, there were no main effects of familiarity, but decoding showed significant correlation with individual face familiarity ratings, suggesting that this slow frequency reflected higher-level cognitive aspects of familiarity processing. These results reveal fundamentally distinct temporal stages of processing, only the slowest of which directly correlates with personal familiarity. These findings may reflect a non-linearity in brain activity between early and later stages of processing, as documented by Landi & Freiwald (2017).
Meeting abstract presented at VSS 2018