Abstract
Effective human interactions depend on our ability to rapidly recognize a familiar face in a dynamic visual environment. We investigated how much visual information (spatial frequencies, or SF) is needed to achieve this fundamental brain function. Human participants viewed 63-s sequences of unfamiliar face images presented at a fast rate of 6 Hz, with different images of different familiar faces embedded every 6th image (1 Hz). Each sequence comprised 9 SF steps at two stimulation orders: (1) from coarse and indistinguishable images to fine and clearly recognizable images; or (2) the reverse order. We captured an objective and sensitive neural index of automatic familiar face recognition in participants’ Electroencephalogram (EEG) at 1 Hz and harmonics. In the coarse to fine sequences, the neural recognition response emerged over the occipito-temporal cortex at around 8 cycles/image. Response increased significantly and plateaued at 11 cycles/image (by a factor of 7). In the reverse sequence order (fine to coarse), significant recognition responses were observed for faces filtered around 40 – 15 cycles/image without inter-SF step differences, which were equivalent with the response amplitudes observed from the other stimulation order at the same SF steps. However, the familiarity recognition responses reduced strongly (by a factor of 8) at around 11 cycles/image. Our results provide neural thresholds and tuning functions consistent with observations in behavioral studies that a middle SF range (8–12 cycles/image) is important for individual face recognition, showing here that this is valid for automatic rapid recognition of ecological images of familiar faces embedded in a stream of unfamiliar faces. While the lack of hysteresis effectmay be due to the high variability of identity and views appearing in the stimulation sequences, we attribute the delayed threshold for reversed sequences to unnatural stimulation conditions (i.e., from fine to coarse).