Impact of Potential ConfoundsFirst,

Impact of Potential ConfoundsFirst, we re-calculated the functional inter-individual variabilitywithout removing the global signal in the preprocessing.Variability maps, estimated with (Figure 1) and without(Supplementary Figures 1, 2) GSR, demonstrated a highlysimilar pattern (Pearson correlation r = 0.92, p < 0.0001).The cerebellum retained the largest mean inter-individualvariability compared to that of cerebral regions (two-samplet-test, p < 0.0001). The network-level variability also consistentlydemonstrated significant statistical difference for the functionalvariability among the networks (p < 0.001), with graduallyincreased variability occurring in the subcortical network, thenthe primary networks (i.e., visual, sensorimotor, and auditorynetworks), to the association networks (i.e., default and frontoparietal networks), and to the cerebellar network (Supplementary Figure 2B). However, as expected, the GSR largely affectedthe connectivity–cognition correlations, such that the GSRpreprocessing introduced more negative correlations (Figure 2)than the nGSR preprocessing (Supplementary Figure 3). Asan overall trend, this was consistent with the GSR resultsregarding the inter-network connectivity, especially for theconnections from the superior and orbital prefrontal cortex,hippocampus, and the cerebellum predominating individualcognitive ability. It is important to note that the retention ofthe global signal diminished the correlation between the longrange connections and cognition, with a larger proportion ofthe long-range connections only found in the global measure