The multi-modal cortical parcellation used information related to the four areal properties of architecture, function, connectivity, and topography2 (link). Architecture was measured using T1w/T2w myelin content maps plus cortical thickness maps with surface curvature regressed out5 (link),9 (link),10 (link) (Supplementary Methods 1.5). Function was measured using task-fMRI responses to 7 tasks in 86 task contrasts (47 unique; 39 were sign-reversed contrasts). Effect size maps (beta maps) after correction for the receive field were used instead of Z statistic maps because we are interested in regional differences in the magnitude of the BOLD (blood oxygen level dependent) signal change induced by the tasks, rather than differences in the significance of the BOLD signal change. Functional connectivity was measured using pairwise Pearson correlation of the denoised resting state time series of each pair of grayordinates. Topographic organization was explored using resting state time series in visual cortex, with spatial regressors representing polar angle and eccentricity patterns in area V1 combined with a modified ‘dual-regression-like’ approach that weights each surface vertex according to the cortical surface area that it represents (see Supplementary Methods 4.4). The semi-automated multi-modal parcellation was generated using group average data for all of these modalities from the 210P group of subjects (see Supplementary Methods 3.1–3.3 for details on how the group averages were created for each modality). The reproducibility of these group average maps was assessed by correlating the spatial maps for the 210P and 210V groups (see Supplementary Results and Discussion 1.1).