Recognizing familiar faces rapidly seems crucial in everyday life. The actual speed at which a familiar face can be recognized remains however unknown. The current thesis aimed at tracking down the minimal behavioral and neural processing time necessary to recognize known faces. To address this issue, we used different go/no-go paradigms and a new task relying on highly time-constraining task (the Speed and Accuracy Boosting procedure, "SAB"). Relying on minimum reaction times analyses, we report that 360-390 ms are needed to recognize famous faces among unknown ones when bottom-up recognition task is required (subjects did not know the identity of the celebrities that they had to recognize before the test; this situation can be compare to the ecological situation of unexpectedly bumping into someone in the street) (Article 1). This latency could not be decreased even after extensive training (Article 1), or using the SAB (Article 2). Overall, this is 100 ms more than when subjects have to detect human faces in natural scene or process gender (Article 1). Bottom-up recognition is much slower than top-down recognition (recognizing somebody whom you know you are going to meet, corresponding to the ecological situation of looking for someone in particular in a crowd), which takes about 300 ms (Article 3). Additionally, MVPA (Multivariate pattern analysis) was applied on EEG data recorded from the scalp surface to determine at which latency familiarity could be read-out. We report that famous faces could be robustly distinguished from unknown faces as soon as 230 ms after stimulus onset. This familiarity-selective signal was directly linked to the subject's recognition speed (Article 5). Such latency was agin much longer than the latencies observed in face categorisation task, in which case category could be read out starting around 80 ms post-stimulus (Article 4). These latencies are with respect to the different models of visual ventral stream and models of face recognition. Three main models are identified and one is favored in particular.
