Fig. 2: Somatotopic and shared representation across subjects in a common space.

a, Each subject’s anatomy was warped to the left hemisphere of the MNI152 atlas, and somatotopic delineation is plotted in aggregate across subjects (only channels exceeding 50% of the within-subject maximum are included). b, Shared activity is plotted in aggregate with scaled asterisks symbols (again, only channels exceeding 50% of the within-subject maximum are included). c, A clustering analysis was performed to see what representations emerge from the data naïvely. K-means clustering was applied to a 3D feature space of foot, hand and tongue movement r² values, and analyses found that five clusters best captured the tradeoff between error and overfitting. The emergent clusters are shown color-coded in the 3D r² feature space and clearly correspond to sites representing hand, tongue or foot movement, sites of shared representation (RMA) and sites unrelated to movement (Supplementary Fig. 10). d, Channels within pericentral cortex are plotted with colors from c, omitting the cluster for sites unrelated to movement. Inset histograms show the density of clustered sites in x, y and z coordinates in MNI space. Note the position of the RMA cluster in the sulcal depths at the mid-lateral aspect of the PCG. e, A histogram of depths shows that RMA cluster sites are deeper in the brain than somatotopic sites. Note that while it appears that tongue-selective somatotopic sites are posterior to the most superficial aspect of the central sulcus on the MNI rendering, examination of individual axial MRI slices shows them to be mostly anterior to the sulcus (Extended Data Fig. 4).