Multiple interacting brain areas underlie successful spatiotemporal memory retrieval in humans.

Description of Experimental Design.(A) Participants (n = 16) navigated a virtual city where they picked up a passenger (depicted in the left panel) and delivered the person to a series of stores; they encoded the spatial layout of the city and the temporal order of deliveries. The middle panel provides an example of a store the participant visited. The right panel shows an aerial view of the virtual environment with the stores distributed unevenly around a circle. (B) After encoding, participants underwent functional imaging while performing the spatiotemporal retrieval task shown here. Participants were first presented with a single reference store and asked to indicate whether they saw this store before. After responding, they were subsequently presented with the spatiotemporal contextual retrieval question. Participants were shown two additional store pictures and asked to make a distance judgment regarding the spatial or temporal proximity to the reference store.

Increases in Network Connectivity during Successful Retrieval of Spatiotemporal Details.(A and B) Each ROI (colored circle) is plotted with significant connections (pink lines) between the nodes; network during correct context responses shown in A and during incorrect context responses in B. Both size and color indicate node degree; a larger radius and warmer colors represent a higher node degree, while a smaller radius and cooler colors represent a lower node degree. The ROIs are spatially distributed according to their x and y coordinates in standardized brain space but have been collapsed in the z coordinate direction. The right panel displays node hierarchy where anterior nodes (red) and posterior nodes (blue) are plotted in ascending order of node degree, with the highest node degree level at the top; nodes belonging to the same horizontal level have the same node degree. The abbreviations for the ROIs are as follows: SFG (Superior Frontal Gyrus); MFG (Middle Frontal Gyrus); IFOr (Inferior Frontal Gyrus, Orbitalis); IFTr (Inferior Frontal Gyrus, Triangularis); IFOp (Inferior Frontal Gyrus, Opercularis); PrCG (Precentral Gyrus); aHPC (Anterior Hippocampus); pHPC (Posterior Hippocampus); pPHG (Posterior Parahippocampal Gyrus); IPL (Inferior Parietal Lobule); SPL (Superior Parietal Lobule); PCN (Precuneus); Calc (Calcarine sulcus). (C) Left: The adjacency matrix (right of diagonal: correct – incorrect, left of diagonal: incorrect – correct) depicts the data to optimize viewing the intra-lobular connections. The colored boxes indicate intra-lobular connections in the frontal lobe and temporal lobe (blue, green, respectively). For successful retrieval, numerous connections exist within the MTL but many are found throughout the neocortex, compared to low levels of functional connectivity throughout the brain during unsuccessful retrieval. Right: Comparison of the total number of edges (significant functional connections) between conditions, with the correct – incorrect network containing significantly more edges. *p < 0.0001. (D) Bar graphs showing betweenness centrality for each ROI (collapsed across hemispheres). Asterisks indicate pcorrected < 0.05 significant differences using a chi-squared test on node connectivity between conditions (i.e. correct – incorrect and incorrect – correct).

Individual Patterns of Network Connectivity Predict Memory Performance.The scatter plot shows a significant linear relationship between total connectivity and individual participant behavioral performance (p < 0.05). To determine which ROIs were driving this correlation, each node's contribution to overall network density was regressed against individual performance using a multiple linear regression. The initial model in the stepwise regression consisted of the nodes labeled with asterisks, and a final more parsimonious model resulted in the nodes highlighted in red (MFG, HPC, PCN, and IFG). The beta coefficients from the linear regression and stepwise regression are presented on a line next to the independent variables. For example, the network density resulted in a coefficient of 0.20 for the linear regression.

Differential Spatial and Temporal Network Connectivity.Panels A, B, and C are arranged similarly to Fig. 2 but show the spatial – temporal and temporal – spatial retrieval networks. (A, B) Connectivity plots, using standardized brain space (left plot) or hierarchical arrangement based on node degree (right plot), for spatial and temporal networks. (C) Left: Adjacency matrix where right of diagonal are significant nodes during spatial retrieval and left of diagonal are significant nodes during temporal order retrieval. Middle: Bar graphs comparing the number of edges for anterior versus posterior brain regions for the two conditions. Anterior brain regions showed higher connectivity during temporal retrieval while posterior brain regions showed higher connectivity during spatial retrieval. Right: Bar graph showing no difference in the total number of edges present in each network. (D and E) The MTL (aHPC, pHPC, pPHG) regions displayed high betweenness centrality values but no differences between condition. The SPL and calcarine ROIs showed significantly greater betweenness centrality for the spatial – temporal network, while IFG and MFG showed greater betweenness centrality in the temporal – spatial network. *pcorrected < 0.05.