Fig. 2: Neural representational dissimilarity matrices (RDM) derived from fMRI were entered into a semipartial correlation analysis.

fMRI RDMs were created using 1—Pearson correlation between the activations of voxels in that region for each pair of images. Five representational dissimilarity matrices, model RDMs, used as predictors in the semipartial correlation analysis. These matrices represent the logarithmic distance between pairs of stimuli in terms of number, average item area, total field area, total surface area, and density.