Fig. 5: Relationship between behavioral performance and inter-electrode coupling of key DAN, SN, and DMN nodes.

a Locations of peak-responsive dPPC, dAIC, and PMC electrodes (in fsaverage space) within cohorts of subjects with simultaneous dPPC-PMC (n = 6) and dAIC-PMC (n = 4) coverage. b Grand average HFB responses during GradCPT correct omission and commission error trials for the electrodes shown in a) (cf. Fig. 4a). c Illustration of how inter-electrode coupling was calculated from continuous HFB 0.1–1 Hz filtered time series. Using an example 20-sec time series for illustration purposes (top), a lagged cross-correlation was performed (shifting dPPC relative to PMC and vice versa). (Bottom) The zero-lag correlation was taken as the value with no time shift, whereas the lag-minimum correlation was taken as the minimum value across the time shifts. In the main analysis, these metrics were calculated based on whole runs (typically 6 min long). d Coefficients from a linear mixed model with d′ (behavioral performance) as dependent variable and with fixed factors including lag-minimum and zero-lag 0.1–1 Hz HFB correlation between dPPC and PMC (and including subject modeled as random factor). e Coefficients as in d for lag-minimum correlations in separate linear mixed effects models that were each constructed from power amplitudes of distinct frequency bands. Error bars indicate upper and lower 95% confidence intervals. Asterisks indicate significant effects (p <0.05, Satterthwaite’s approximation, F test on linear mixed effects model).