Abstract
How fast are visual stimuli categorized as faces in the human brain? Because of their high temporal resolution and the possibility to record simultaneously from the entire system, electromagnetic (EEG, MEG) scalp measurements seem to be the ideal method to clarify this issue. However, despite years of research, this question remains debated, with studies reporting face-sensitive responses on the scalp varying from 50 ms to 200 ms following stimulus onset. Here we disentangle the contribution of low-level properties and high-level visual representations in accounting for early face-sensitivity in the human brain, by means of a factorial paradigm including faces, familiar objects (cars) and their respective phase-scrambled counterparts. In fifteen human participants, we replicated an early face-sensitivity - larger amplitude to pictures of faces than cars – on the positive event-related potential (ERP) P1 (80–100 ms). However, this larger response to faces was accounted for completely by low-level parameters, the P1 being also larger in response to scrambled pictures of faces than scrambled cars. In contrast, the following visual responses (N1, or N70) reflected the brain response to meaningful shapes, being almost non-existent for scrambled stimuli. As usually observed, this N170 was larger in amplitude for pictures of faces than cars, especially in the right hemisphere. Contrary to the earlier P1 effect, this N170 face effect could not be explained at all by differential low-level parameters, the N170 amplitude to scrambled faces and cars being equal. Altogether, these observations indicate that the early visual potential P1 is determined by properties of the stimulus while the following N1 or N170 reflects what the observer perceives/interprets. Moreover, preferential responses to faces may arise at several points in time during visual stimulation, but the earliest access to high-level face representations does not precede the occipito-temporal N170 onset in the human brain.
Belgian National Fund for Scientific Research (FNRS).