Fig. 4: Multivariate analyses.

a Average accuracies (diamonds) and individual accuracies (dots) extracted from the FFA, PPA, FEF, SPL (left panel), early visual areas (V1–V4, right panel) for the pairwise classifications in Experiment 1 (F: face, H: house, iF: inverted face). The shaded areas indicate accuracies below the 95th percentile of the null distribution obtained from the permutations. Individual data under different pairwise classifications are shown as color scales (blue dots for face vs. house, green dots for house vs. inverted face, red dots for face vs. inverted face). b The results of the same analysis as (A) in Experiment 2.c Average accuracies (diamonds) and individual accuracies (dots) extracted from the FFA, PPA, FEF, SPL (left panel), early visual areas (V1–V4, right panel) for the pairwise classifications in Experiment 3 (SF: self-face, SH: self-house, OF: other-face, OH: other-house). Individual data under different pairwise classifications are shown as color scales (blue dots for self-face vs. self-house, green dots for other-face vs. other-house, red dots for self-face vs. other-face, purple dots for self-house vs. other-house). d Scatter plots (with best-fitting regression lines) illustrate the individual sensitivity (area under the curve, see methods and Supplementary Fig. 2) for the classification ‘self-face vs. self-house’ in FFA as a function of the cross-correlation between the gaze-track position and the actual eye position (left panel), and as a function of the variation of the first fixation locations for self-face (right panel) in Experiment 3. The variation was obtained by calculating the mean distance (in pixel) among the first fixation locations across all trials.