Abstract
Numerous experiments have shown that the recognition of a facial feature is dependent on the other features of the face stimulus, a phenomenon taken as evidence that faces are represented holistically (e.g. Tanaka & Farah, 1993). Even though this effect has been found in matching tasks with unfamiliar stimuli (e.g. Pellicano & Rhodes, 2003), supporting the view that it occurs at a perceptual stage, this remains controversial (Wenger & Ingvalson, 2002). Here, we used fast event-related fMRI-adaptation to test whether the regions responding preferentially to faces in the human visual cortex represent facial features independently or not. On each trial, subjects (N=8) were shown a whole face stimulus followed either by a whole face or a single feature (the eyes). The eyes feature in the second stimulus could either be identical to the first stimulus, or different, leading to 4 conditions. Subjects were better at matching or discriminating the eyes when two whole faces were presented, replicating behavioral observations. In face-sensitive areas of the inferior occipital cortex, there was an adaptation when the same feature was presented in the two stimuli of a pair, whether the feature was presented in the same format (whole-to-whole) or not (whole-to-isolated). In contrast, and as predicted, there was a much larger adaptation in the whole-to-whole than in the whole-to-part condition in the middle fusiform gyrus (MFG). These results supports the view that facial features are not represented independently but rather integrated into a global representation in the middle fusiform gyrus.