Fig. 3: Information measures computed from randomly sampled 10-node subsets.

a Fraction (left) and number (right) of subsets with negative O-information, obtained by randomly sampling subsets from the HCP (blue) and MICA (red) FC matrix. Fraction and number estimated from samples of 5000 (3–12 nodes) or 200 (13–15 nodes) subsets with negative O-information. As the size of the subset grows, the fraction expressing an overall synergy-dominated structure (negative O-information) drops precipitously, while their absolute number continues to climb due to combinatorics. b The relation between O-information and TSE complexity in 100,000 randomly sampled 10-node subsystems (HCP data). While very few randomly sampled sets have negative O-information (see panel a), TSE complexity generally increases with the strength of the dependencies visible to the O-information (R = 0.642, p = 0). c The participation quantifies, for each node pair, how often they are encountered as part of a subset with negative o-information (10 nodes, 5000 random samples, HCP data), The plot shows the relation of the participation against the FC, with each data point representing one of the 19,900 unique node pairs. Node pairs with strong mutual FC (positive or negative) are rarely encountered as part of the same synergistic subset, while node pairs that are more frequently encountered tend to show weak FC. Spearman’s rho between absolute FC and participation: ρ = − 0.504, p = 0.