Multiscale network reorganization during the multidimensional recollection of naturalistic events.

The recollection of complex events is multidimensional. Recent work has gathered evidence for varying degrees of functional specialization for different memory dimensions across the main cortical networks involved in cued recollection. However, how functional brain network topology reorganizes to support the retrieval of different types of information remains largely unexplored. Here, we applied graph theoretical analysis to task-related fMRI functional connectivity data from 22 participants engaged in cued recollection of specific details from naturalistic movie scenes. A beta series correlation approach was used to identify the pattern of whole-brain regional co-activation for cued recollection in general, and separately for each memory dimension of interest (object and character details, spatial layouts, temporal sequences, and verbal dialogues). Subsequently, using graph theory, we investigated the association between indices of brain network topology and cued recollection performance, as well as differences in functional brain network architecture across memory dimensions at the multiscale level. Successful recollection was associated with higher global efficiency and lower modularity at the whole-network level, as well as with lower system segregation within the Frontoparietal control network. Crucially, retrieval conditions placing greater demands on relational integration (i.e., temporal order judgments), compared to those emphasizing single-item details, were associated with more integration in the Default Mode network and increased degree centrality, efficiency, and reduced path length in the right hippocampus. Together, these findings reinforce the idea that successful retrieval requires system integration, especially when accessing complex relational representations like temporal sequences.