Across experiments, all images were attributed a mean (i.e., the global luminance) and RMS contrast of 0.45±0.1. Both the main experiment and the functional localiser had a block design.
Stimuli for the localiser experiment consisted of three image categories (faces, hands, instruments) each containing 20 greyscale images (See supplementary; Stigliani, Weiner, & Grill-Spector, 2015) along with their phase-scrambled counterparts. The stimuli were superimposed onto the Fourier phase-scrambled backgrounds (procedure as described above) and consisted of unfamiliar faces (10 males) of various viewpoints (including hair), isolated hands in various poses, and stringed instruments (e.g. guitar, cello, lute, etc) positioned in different orientations.
The functional localiser run had a block design, of 10-second blocks, alternated by 10 seconds of fixation with a 12-second fixation period at the beginning and end of the run. To ensure subjects paid attention during functional runs, we instructed them to detect a rare and brief colour change of the stimulus by pressing a button with the right index finger. In each block, there were two targets: HSV profile 1.0, 1.0, 0.8. One colour change occurred per block half, but never during the first stimulus of a block. A fixation cross, made of two thin black lines that connected opposite corners of the square stimuli, was visible throughout all runs.
The functional localiser was used to find voxels responding more to faces than non-face objects. Subjects viewed grayscale images of either intact or scrambled faces, hands or instruments. Images from all categories were presented in 36 blocks in total (one image category per block), with 6 blocks per condition. A condition was never repeated twice in a row and the order of conditions was counterbalanced across subjects. A block consisted of 10 images in a random order, all images were repeated three times during the whole localiser run. Each image appeared for 500ms, followed by a 500ms interstimulus interval.
The FFA, responding preferentially to faces, was identified independently for each subject, based on the localiser scan. First, to select the brain regions more responsive to faces than non-face objects we computed the conjunction between [intact faces - intact hands] and [intact faces - intact instruments] contrasts. Next, to exclude activity to low-level image properties we selected the voxels that showed a significantly larger response to intact than scrambled faces [intact face - scrambled face]. Significant voxel clusters on t maps were selected at a q[false discovery rate, FDR] < 0.01. After visual inspection, the threshold for some subjects was set increased to delineate between face preferring clusters. Overall, there were more face-selective voxels in the right FFA (X ± X) compared to the left (X ± X).