Figure 3.

Surface searchlight based modeling of representational geometry. Neural representation of faces in each searchlight was modeled with 3 model similarity structures as regressors using ridge regression. Representational geometry in the rIFFA correlated with the identity model (top), whereas representational geometry in EVC correlated with the head view model (bottom). Maps are thresholded at P < 0.05 after correcting for multiple comparisons using permutation testing. Correlation with the mirror symmetry model did not reveal any significant clusters.