Next, we measured which voxels correlated with the number of cognitive components engaged at each time point. We calculated this correlation for each subject in each task, and then took the mean for each task. In every task, we saw positive correlation between the number of cognitive components engaged and activity at voxels in the fronto-parietal network and dorsal attention network. We observed negative correlations between the number of cognitive components engaged and activity at voxels in the default mode network. These results are shown in Figure 5 for every task. Moreover, results were practically identical when ignoring voxels where more than 3 cognitive components had high probabilities during the joint probabilistic reverse inference.