Abstract
Periodic presentations of variable natural face images in a rapid sequence of variable images of nonface objects elicit sensitive category-selective neural response captured in the human electroencephalogram (EEG) objectively at the periodic face frequency (Rossion et al., 2015, J Vis). Here, we test the resistance to degradation of the face-selective response by systematically varying the proportion of periodic face occurrence in the nonface object stream. High-density EEG was recorded from 16 observers during presentations of 54-s sequences of random object images sinusoidally contrast-modulated at F = 12 Hz (i.e., 12 images/s; 83-ms stimulus-onset asynchrony). Observers performed an orthogonal task by responding to random colour changes of a central fixation cross. There were 9 conditions. In the 100% condition, natural face images were embedded in the sequence always at a fixed interval of F/9 (1.33 Hz; every 9th image). In other conditions, a proportion of periodic face events was omitted, replaced by nonface object images. The percentages of periodic face events tested in the 9 separate conditions were 0% (i.e., no face presented), 12.5%, 25%, 37.5%, 50%, 62.5%, 75%, 87.5%, and 100%. Selective responses to faces recorded at 1.33 Hz and harmonics (2.67 Hz, etc.) mainly over occipito-temporal areas emerged significantly only at 25% of face images and followed a nonlinear power-function relationship (coefficient: 1.54, sum of scalp-averaged responses over significant harmonics) with the percentage of periodic face events. Removing half of the periodic face events reduced the response by 66%. The results reveal that a face-selective neural response results from complex, nonlinear comparisons between face exemplars and nonface objects.
Acknowledgement: This work was supported by Singapore MOE AcRF Tier 1 Grant 2018-T1-001-069 to C.O. & B.R., NTU HASS-SUG to C.O., and ERC Grant facessvep 284025 to B.R.