In what follows, we refer to such results as cross-network interactions. A major feature evident in Figure 2 concerns the dependence of BLP interactions on frequency band. Within-network correlations were weak and statistically

nonsignificant in the δ and θ bands. Significant within-network correlations were observed in the α (p < 0.01, FDR corrected), and even more Selleckchem SCH772984 so in the β band (p < 0.005, FDR corrected). No results are shown for the γ band because there were no identified MCWs in that frequency range. Among all RSNs, the DMN showed the strongest interaction with other networks, and this effect was especially clear in the β band (all contrasts p < 0.005, FDR corrected). Significant interactions (all contrasts p < 0.01, FDR corrected) were also observed in the α band. Other networks with significant cross-network interactions include the DAN (α, all contrasts,

p < 0.01 except versus language; β, all contrasts: p < 0.01) and the somatomotor network (α, all contrasts, p < 0.01 except versus visual; β, all contrasts, p < 0.01). VAN, language, and visual networks appeared relatively segregated. The analysis of cross-network interaction was extended to the level of single nodes, confining the analysis to the β band. Figure 3A shows the pairwise interaction matrix for all presently considered nodes (Table S1). Because the spatial resolution of source-space MEG is limited, correlation between closely spaced nodes is high. To minimize the impact of this effect, pairs of nodes closer than 35 mm were excluded (white cells in Figure 3). The 35 mm figure was derived below Duvelisib clinical trial from previous results (de Pasquale et al., 2010). However, the principal features evident in Figures 2 and 3 are insensitive to varying the node-pair proximity limit within a range of 0–100 mm (Figure S3). Five out of seven nodes of the DMN showed significant interactions with nodes of other networks in the β band (bar plots Figure 3A, all contrasts, p < 0.05). Among these nodes, the posterior cingulate cortex (PCC) showed the highest mean interaction with all other nodes (p < 0.001 Bonferroni corrected). Other significant nodes included

the left and right angular gyrus, and medial prefrontal cortex. In contrast, only two nodes of other networks reached a significant level of interaction: left posterior intraparietal sulcus, part of the DAN (p < 0.05, Bonferroni corrected), and the left central sulcus, part of the somatomotor network (p < 0.05, Bonferroni corrected). These findings, at the single node level, are consistent with results obtained by averaging over nodes within networks (Figure 2). Control analyses presented in Figure S4 show that both within or across-network interaction results are insensitive to choice of external node used to compute MCWs. A key finding of this study was that the results shown in Figures 2 and 3A reflect nonstationary phenomena.