To create our retrieval networks, we used a beta time series analysis17, which assumes two regions are coupled during a specific condition of a task if the activity of both is significantly correlated across trials. In other words, two regions are functionally connected if their responses fluctuate similarly under the same condition. To characterize differential connectivity between conditions (correct versus incorrect contextual retrieval), we took the set difference between two networks, i.e., connections that were exclusive to either correct or incorrect contextual retrieval (labeled here as correct – incorrect and incorrect – correct, see Methods for analytic and statistical details). When participants accurately judged the spatial or temporal distance of probe questions compared to incorrect trials, we found a significant overall increase in large-scale network connectivity, indexed by the number of connections (also termed edges) between our regions of interest (also termed nodes) (χ2 (1) = 63.75, p < 0.0001, Fig. 2A,B,C). This result suggests that the increase in coordinated activity of multiple distributed regions underlies successful retrieval of spatiotemporal contextual details.
