Parallel distributed networks dissociate episodic and social functions within the individual

Abstract

Association cortex is organized into large-scale distributed networks. One such network, the default network (DN), is linked to diverse forms of internal mentation, opening debate about whether shared or distinct anatomy supports multiple forms of cognition. Using within-individual analysis procedures that preserve idiosyncratic anatomical details, we probed whether multiple tasks from two domains, episodic projection and theory of mind (ToM), rely on the same or distinct networks. In an initial experiment (6 subjects, each scanned 4 times), we found evidence that episodic projection and ToM tasks activate separate regions distributed throughout the cortex, with adjacent regions in parietal, temporal, prefrontal, and midline zones. These distinctions were predicted by the hypothesis that the DN comprises two parallel, interdigitated networks. One network, linked to parahippocampal cortex (PHC), is preferentially recruited during episodic projection, including both remembering and imagining the future. A second juxtaposed network, which includes the temporoparietal junction (TPJ), is differentially engaged during multiple forms of ToM. In two prospectively acquired independent experiments, we replicated and triplicated the dissociation (each with 6 subjects scanned 4 times). Furthermore, the dissociation was found in all zones when analyzed independently, including robustly in midline regions previously described as hubs. The TPJ-linked network is interwoven with the PHC-linked network across the cortex, making clear why it is difficult to fully resolve the two networks in group-averaged or lower-resolution data. These results refine our understanding of the functional-anatomical organization of association cortex and raise fundamental questions about how specialization might arise in parallel, juxtaposed association networks.NEW & NOTEWORTHY Two distributed, interdigitated networks exist within the bounds of the canonical default network. Here we used repeated scanning of individuals, across three independent samples, to provide evidence that tasks requiring episodic projection or theory of mind differentially recruit the two networks across multiple cortical zones. The two distributed networks thus appear to preferentially subserve distinct functions.

Keywords: association cortex; default network; prospection; remembering; theory of mind.

Conflict of interest statement

No conflicts of interest, financial or otherwise, are declared by the authors.

Figures

Fig. 1.

Procedure for testing functional dissociation within individuals. A: within each subject, networks A and B were identified using k-means clustering. Each network’s border was defined (B) and overlaid on the unthresholded contrast maps for each task domain (C). D: the distribution of z-weighted values within each network’s boundaries was extracted and plotted. E: for each network, plots from both task domains were then visualized on a single graph [theory of mind (ToM), red; episodic projection (EP), yellow], and potential differences between the domain distribution means were quantified using effect sizes. While this figure illustrates distributions for network A only, the procedure was performed identically for both networks A and B.

Fig. 2.

Networks A and B identified within individuals [subjects 1–6 (S1–S6)], using both seed-based and k-means parcellation strategies. Estimates of networks A and B from experiment 1 identified using seed-based (left) and k-means (right) methods exhibit comparable maps. Although exact network boundaries differ by method, both techniques reveal interdigitated network patterns along the medial and lateral frontal cortex, as well as similarly juxtaposed network boundaries along the posteromedial cortex and inferior parietal lobule. k = 17 for all k-means parcellations shown here; network A appears in navy, and network B in light blue. Seed-based maps are thresholded at r = 0.40.

Fig. 3.

Networks A and B show functional dissociation within individuals [subjects 1–6 (S1–S6)] in experiment 1. Left: network A (navy) and network B (light blue) for each subject, as defined by k-means clustering. Right: the distributions plot the functional responses within each network for the two task domains [theory of mind (ToM), red; episodic projection (EP), yellow; overlap, orange]. See Fig. 1 for method. For network A, all 6 individuals reveal a functional response increase for EP over ToM, and most strongly in 4 subjects (with Cohen’s d range = 0.88–2.03 for 4 subjects; d = 0.19 for S1 and d = 0.20 for S4). For network B, all 6 individuals reveal the opposite pattern: the ToM response is increased over EP (Cohen’s d range = 0.51–1.76). The consistent opposing patterns between networks are evidence for functional double dissociation.

Fig. 4.

Parahippocampal (PHC) and ventral posterior cingulate (vPCC)/retrosplenial cortex (RSC) regions of network A exhibit robust functional dissociation in experiment 1. Left: PHC and RSC/vPCC regions, as defined in each subject [subjects 1–6 (S1–S6)], outlined in white. The distributions plot the functional responses within each region of network A for the two task domains [theory of mind (ToM), red; episodic projection (EP), yellow; overlap, orange]. In all subjects for both regions, a clear functional response increase is evident for EP over ToM (Cohen’s d ranges: RSC/vPCC, 0.54–4.43; PHC, 1.67–3.24).

Fig. 5.

Procedure for visualizing task response patterns within individuals. Within each individual, the separate task contrasts were first estimated. Past: Past Self versus Present Self; Future: Future Self versus Present Self, within the episodic projection domain. False Belief: False Belief versus False Photo; Other Pain: emotional (Emo) Pain versus physical (Phys) Pain within the theory of mind (ToM) domain. Color bars indicate z-values. Within each domain, the contrasts were averaged to yield a single best estimate. The two domain maps were then thresholded and plotted on the same brain (center image) to reveal overlap (ToM, red; episodic projection, yellow; overlap, orange).

Fig. 6.

Networks A and B exhibit differential recruitment by episodic projection (EP) and theory of mind (ToM) tasks across multiple cortical zones within individuals [subjects 1–6 (S1–S6)]. Within each column, the lateral (left) and medial (right) surfaces of the left hemisphere are shown. The colors represent the task responses (ToM, red; EP, yellow; overlap, orange; see Fig. 5). For each subject, the left column displays the functional response patterns in relation to the network A boundaries. The right column shows the same response patterns in relation to the network B boundaries. The network boundaries are illustrated by black outlines. EP and ToM are either partially or fully dissociable across multiple cortical regions in all subjects. The most striking double dissociations are evident in S2 and S6, where small idiosyncratic features of the differential task response patterns are predicted by the network boundaries.

Fig. 7.

Networks A and B exhibit differential recruitment by episodic projection (EP) and theory of mind (ToM) tasks in the right hemisphere. Using the same procedures as in Fig. 6, the right hemisphere is displayed for each of the 6 subjects (S1–S6) from experiment 1. Each row shows the lateral (right) and medial (left) views of the right hemisphere for all subjects, with network A boundaries (left) or network B boundaries (right).

Fig. 8.

Functional dissociation of networks A and B observed across task contrasts in experiment 1. The bar graphs plot the functional responses (mean z-values and standard deviations across runs) within each network for each task-specific contrast. Tasks within a domain exhibit comparable patterns of network recruitment. Within network B, for most individuals [subjects 1–6 (S1–S6)], the Other Pain contrast (ToM PAIN) exhibits stronger recruitment than the False Belief contrast (ToM BELIEF). For network A, functional response increases for both episodic projection (EP) contrasts (EP Past and EP Future) over both theory of mind (ToM) contrasts are evident in 5 out of 6 subjects (Cohen’s d range = 0.08–2.25 for pairwise comparisons, with most d > 0.80). For network B, increases for both ToM contrasts over both EP contrasts are also evident in 5 out of 6 subjects (d range = 0.20–2.01, with most d > 1.20).

Fig. 9.

Networks A and B identified within individuals [subjects 7–12 (S7–S12)] in experiment 2 using both seed-based and k-means parcellation strategies. Estimates of networks A (navy) and B (light blue) identified using seed-based (left) and k-means (right) methods exhibit comparable maps within individuals in experiment 2. In experiment 2, k was set to the lowest number of clusters featuring differentiation of networks A and B, labeled for each individual. Similar to experiment 1 results, estimated network boundaries differ by method, but the two networks could be identified in all individuals with both methods. Seed-based maps are thresholded at r = 0.40.

Fig. 10.

Networks A and B replicate functional dissociation within individuals [subjects 7–12 (S7–S12)] in experiment 2. Left: the spatial distributions of network A (navy) and network B (light blue) for each subject as defined by k-means clustering. Right: the distributions plot the functional responses within each network for the two task domains [theory of mind (ToM), red; episodic projection (EP), yellow; overlap, orange]. For network A, 5 out of 6 subjects demonstrate a functional response increase for EP over ToM contrasts (Cohen’s d range = 0.75–1.52). For network B, all subjects demonstrate the opposite pattern: an increase for ToM over EP (d range = 0.09–3.01; d > 1.20 for all except S11). These findings replicate and support a functional double dissociation.

Fig. 11.

Parahippocampal (PHC) and ventral posterior cingulate (vPCC)/retrosplenial cortex (RSC) regions of network A replicate functional dissociation in experiment 2. Left: PHC and RSC/vPCC regions as defined in each subject [subjects 7–12 (S7–S12)]. These regions are outlined in white and are contained within network A, as in Fig. 4. The distributions plot the functional responses within each region for the two task domains [theory of mind (ToM), red; episodic projection (EP), yellow; overlap, orange]. For each region, in all subjects, a significant functional response increase for EP over ToM is evident (Cohen’s d ranges: RSC/vPCC, 1.60–3.42; PHC, 2.35–3.39).

Fig. 12.

Networks A and B replicate differential recruitment by episodic projection (EP) and theory of mind (ToM) tasks across multiple cortical zones within individuals [subjects 7–12 (S7–S12)] in experiment 2. Within each column, the lateral (left) and medial (right) surfaces of the left hemisphere are shown. The colors represent the task responses (ToM, red; EP, yellow; overlap, orange; see Fig. 5). For each subject, the left column displays the functional response patterns in relation to network A boundaries, and the right column in relation to network B boundaries, with boundaries shown as black outlines. EP and ToM are either partially or fully dissociable across multiple cortical regions in all subjects in experiment 2, replicating results from experiment 1. The most striking double dissociations are evident in S7 and S12, where unique features of the differential task response patterns are predicted by the network boundaries in all cortical zones.

Fig. 13.

Networks A and B replicate differential recruitment by episodic projection (EP) and theory of mind (ToM) tasks in the right hemisphere in experiment 2. Using the same procedures as displayed in Fig. 12, the right hemisphere is displayed for each of the 6 subjects [subjects 7–12 (S7–S12)] from experiment 2. Each row shows the lateral (right) and medial (left) views of the right hemisphere for all subjects, with network A boundaries (left) or network B boundaries (right).

Fig. 14.

Functional dissociation of networks A and B replicates across task contrasts in experiment 2. Functional responses (mean z-values and standard deviations across runs) within each network for each specific task contrast are shown. Replicating findings from experiment 1, tasks within a domain exhibit comparable differences in network recruitment, and the Other Pain contrast (ToM PAIN) shows the strongest recruitment of network B in multiple individuals [subjects 7–12 (S7–S12)], rather than the False Belief contrast (ToM BELIEF). For network A, functional response increases for both episodic projection (EP) contrasts (EP Past and EP Future) over both theory of mind (ToM) contrasts are evident in 5 out of 6 subjects (Cohen’s d range = 0.37–1.68 for pairwise comparisons, most d > 0.90). For network B, increases for both ToM contrasts over both EP contrasts are evident in 5 out of 6 subjects (d range = 0.52–2.94, most d > 1.50).

Fig. 15.

Exploratory analysis of network recruitment across temporal orientations. Results from the time-matched contrasts in the episodic projection task are displayed. The distributions plot the functional responses within each network for these contrasts (Past Self versus Past Non-Self in yellow, Present Self versus Present Non-Self in blue, and Future Self versus Future Non-Self in red). For network A, all subjects [subjects 7–12 (S7–S12)] show greater responses in both the Past and Future conditions relative to the Present (Cohen’s d range = 0.20–1.47; most d > 0.60). For network B, the Present condition shows a shifted response relative to the Past and Future conditions in almost all instances (with the one exception of the Future condition in S9; d range = 0.10–1.67 for all but S9, most d > 0.60).

Fig. 16.

Trial-level analyses support functional dissociation across multiple cortical zones. A–C: region-specific, trial-level analysis procedures are shown. A: five cortical zones of interest were selected approximating LTC (lateral temporal cortex), PFC (prefrontal cortex), PPC (posterior parietal cortex), PMC (posteromedial cortex), and MPFC (medial prefrontal cortex). B: region masks were overlaid on each individual’s network map to identify region-specific network boundaries. Beta values were extracted, for every trial of each task, from within these boundaries and averaged. Contrast estimates were then calculated by subtracting control trial values from target trial values. Within each region, factorial ANOVA tested for network × domain interaction effects on the contrast values. C: betas were then standardized to z-weighted values and plotted for ease of visualization. Significant or trend interaction effect: ++P < 0.05, +P < 0.1). This procedure was performed for each network in its entirety (D) and for each region (E), using both task and null data. D: full-network results reveal significant interactions for 5 out of 6 individuals’ [subjects 7–12 (S7–S12)] task data in experiment 2 (all but S10; S7 not shown), with network A exhibiting preferential recruitment for episodic projection (EP; yellow) and network B for theory of mind (ToM; red). No significant interaction effects were seen in null data; one null dataset showed a trend (S8). E: region-specific results are shown for a single subject (S7). Across regions, the task data show significant interaction effects. No significant interaction effects appear for the null data in S7 (Tables 6 and 7 show results from all individuals in experiments 2 and 3).

Fig. 17.

Networks A and B identified within individuals [subjects 13–18 (S13–S18)] in experiment 3 using seed-based and k-means parcellation strategies. Estimates of networks A (navy) and B (light blue) were comparable across seed-based (left) and k-means (right) methods for individuals in experiment 3. As in experiment 2, k varied by individual. A clustering solution was selected, for each individual, that featured differentiation between networks A and B, as well as between these and other large-scale networks. Seed-based maps are thresholded at r = 0.40.

Fig. 18.

Networks A and B triplicate functional dissociation within individuals [subjects 13–18 (S13–S18)] in experiment 3. Left: spatial distributions of networks A (navy) and B (light blue). Right: distributions plot the functional responses within each network for theory of mind (ToM; red) and episodic projection (EP; yellow), with overlap in orange. For network A, all 6 subjects demonstrate higher EP than ToM values (Cohen’s d range = 0.99–1.94 for 5 subjects, d = 0.07 for S15). For network B, all subjects demonstrate higher values for ToM (d range = 1.10–2.67). These findings further replicate and support a functional double dissociation.

Fig. 19.

Parahippocampal (PHC) and ventral posterior cingulate (vPCC)/retrosplenial cortex (RSC) regions of network A triplicate functional dissociation in experiment 3. Left: PHC and RSC/vPCC regions (outlined in white, within network A) as defined in each subject [subjects 13–18 (S13–S18)] in experiment 3. The distributions plot the functional responses within each region for theory of mind (ToM; red) and episodic projection (EP; yellow). All subjects exhibit higher values for EP over ToM in both regions (Cohen’s d ranges: RSC/vPCC, 1.65–3.37; PHC, 1.42–3.45).

Fig. 20.

Networks A and B triplicate differential recruitment by episodic projection (EP) and theory of mind (ToM) tasks across multiple cortical zones within individuals in experiment 3. Within each column, the lateral (left) and medial (right) surfaces of the left hemisphere are shown. Functional response patterns are shown in relation to the network A (left) and network B (right) boundaries. For all subjects [subjects 13–18 (S13–S18)], EP (yellow) and ToM (red) maps are largely dissociable across distributed cortical regions. Preferential overlap between network A and the EP contrasts and between network B and the ToM contrasts are strongly apparent in most subjects, in multiple cortical zones.

Fig. 21.

Networks A and B triplicate differential recruitment by episodic projection (EP) and theory of mind (ToM) tasks in the right hemisphere in experiment 3. The right hemisphere is displayed for each of the 6 subjects [subjects 13–18 (S13–S18)] from experiment 3, using the same procedures as those described for Fig. 21. Lateral (right) and medial (left) views of the right hemisphere are shown in each row, with network A (left) or network B (right) boundaries outlined.

Fig. 22.

Functional dissociation of networks A and B triplicate across task contrasts in experiment 3. For each individual [subjects 13–18 (S13–S18)] in experiment 3, bar graphs show the functional within-network responses (mean z-values and standard deviations across runs) for each task contrast. Replicating findings from experiments 1 and 2, for network A, higher responses for both episodic projection (EP) contrasts over both theory of mind (ToM) contrasts are evident in 5 out of 6 subjects (all but S15; Cohen’s d range = 0.55–1.97 for pairwise comparisons). For network B, the Other Pain contrast (ToM PAIN) shows the strongest recruitment in most subjects, over the False Belief contrast (ToM BELIEF). Increases for both ToM contrasts over both EP contrasts (EP Past and EP Future) are evident in all 6 individuals (d range = 0.58–2.91).

Fig. 23.

Differential recruitment of networks A and B across temporal orientations replicates in additional individuals [subjects 13–18 (S13–S18)]. Within each network, distributions plot the responses for the time-matched contrasts in the expanded episodic projection task (Past Self versus Past Non-Self in yellow, Present Self versus Present Non-Self in blue, and Future Self versus Future Non-Self in red). Replicating experiment 2 results, all subjects show greater recruitment of network A by both the Past and Future conditions relative to the Present (Cohen’s d range = 0.18–2.03; most d > 0.80), and greater recruitment of network B by the Present condition relative to the Past and Future (d range = 0.22–1.19).