The default network and the combination of cognitive processes that mediate self-generated thought

Abstract

Self-generated cognitions, such as recalling personal memories or empathizing with others, are ubiquitous and essential for our lives. Such internal mental processing is ascribed to the Default Mode Network, a large network of the human brain, though the underlying neural and cognitive mechanisms remain poorly understood. Here, we tested the hypothesis that our mental experience is mediated by a combination of activities of multiple cognitive processes. Our study included four functional MRI experiments with the same participants and a wide range of cognitive tasks, as well as an analytical approach that afforded the identification of cognitive processes during self-generated cognition. We showed that several cognitive processes functioned simultaneously during self-generated mental activity. The processes had specific and localized neural representations, suggesting that they support different aspects of internal processing. Overall, we demonstrate that internally directed experience may be achieved by pooling over multiple cognitive processes.

Keywords: default mode network; fMRI; internal processing; internally directed cognition; mind wandering; self-generated cognition.

Figure 1

Experiment 1: Schematic flow of the experimental trial, results of the group-level analysis of four internal tasks and location of Regions of interest (ROIs). (a) Schematic flow of the trial (from left to right). After seeing a picture, the participants received the task instruction, and started to execute the task with their eyes closed. The task execution ended with a beep sound, followed by vividness rating of the experience. There were five tasks (conditions): imagine what happened before ("past imagery") or after ("future imagery") the depicted scene, recall a personal episodic memory related to the depicted scene ("episodic memory"), imagine yourself in the place of the person in the image ("empathizing"), and generate words that rhyme with a provided word (baseline condition; the target word was provided at the stage of task instruction). Each of the pictures was repeated once for each of the five experimental conditions. Notably, the image was presented in all conditions including baseline, to preserve identical visual stimulation. Accordingly, the contrast between internal task and baseline does not include the activity elicited by visual scene. (b) Results of the group-level random effect analysis of Experiment 1 (n=36): Four tasks of interest (episodic memory, future imagery, past imagery, and empathizing) contrasted separately against the rhyme-generation baseline task (voxel-wise primary threshold p-value < 0.001 and a cluster-level threshold p-value < 0.05, corrected. Note the typical DMN activations for all four conditions. Statistical maps were overlaid on T1 SPM template brain. (c) Average location of the ROIs. The ROIs used in the analysis were defined as individual clusters and were not spherical (see Methods). Locations here represent average location across participants (see also Supplementary Table 1). Abbreviations: mPFC is medial prefrontal cortex, PCC is posterior cingulate cortex, AG is angular gyrus, LTC is lateral temporal cortex, and PHC is parahippocampal cortex.

Figure 2

Experiment 1 (n=36): Percent signal change time-courses for five experimental conditions ("future imagery," "past imagery," "episodic memory," "empathizing," and "rhyme generation") in the DMN. The units of the X axis are TRs (2.5 sec). The first bin of the X axis corresponds to the onset of task instruction (see Fig. 1a). Error bars represent the standard error of the mean. Note: a) large differences between four internal tasks and rhymes generation baseline condition; b) differences in shape of time-courses across internal tasks: a clear positive and inverted U-shape response in the PCC, angular gyrus, and mPFC in "episodic memory" condition, and a clear negative and U-shape response in the PHC; c) a higher response to "episodic memory" compared to other conditions in the mPFC and PCC.

Figure 3

Experiment 2 (n=34): Self-referential processing. (a) Group-level random effect analysis of self-referential processing of Experiment 2 (contrast: "self-referential" > "non-self-referential"). Statistical threshold: primary voxel-wised threshold p < 0.001, p<0.05, cluster size corrected. The blue contour line denotes the DMN identified using the first (independent) session of the experiment ("four internal conditions > baseline"). Note, that significant clusters within the DMN were found in the mPFC, PCC, and left angular gyrus, but not in the PHC, LTC, and right angular gyrus. (b) ROI representational similarity analysis between internal processing (Experiment 1, four internal tasks combined) and self-referential processing (Experiment 2). The values reflect the average across participants within-ROI Spearman correlation between the internal processing contrast of Experiment 1 (four internal tasks > baseline) and the self-referential processing contrast of Experiment 2 ("self-referential" > "non-self-referential"). Similarity values denote Fischer z-transformed correlation results. Note, high similarity values in the mPFC, PCC, and to a lesser extent in the left angular gyrus. Significance above zero was established using one sample, two-tail t-test (multiple comparison Bonferroni correction for number of tested regions, N=8, alpha=0.05/8=0.00625). For regional specificity and task specificity analyses, see the Results section. Error bars represent standard error of the mean. For individual data, see Supplementary Figure 1a. (c) Searchlight ROI representational similarity analysis between internal processing (Experiment 1, four internal tasks combined) and self-referential processing (Experiment 2). Top: unthresholded statistical results map; Bottom: thresholded significant clusters (primary voxel-wised threshold p < 0.001, p<0.05, cluster size corrected). Note the high similarity in the mPFC, PCC, and left posterior parietal cortex, but not in other regions of the cortex. Also note the strong left lateralization in the posterior parietal cortex.

Figure 4

Experiment 3 (n=33): Scene construction. (a) Group-level random effect analysis, "scenes" > "objects" contrast. Statistical threshold: primary voxel-wised threshold p < 0.001, p<0.05, cluster size corrected. Note, that the largest and most significant clusters within the DMN were found in the PHC. (b) ROI representational similarity analysis between internal processing (Experiment 1, four internal tasks combined) and "scenes" > "objects" contrast (Experiment 3). Note, much higher than zero similarity values in the bilateral PHC. Significance above zero was established using one sample, two-tail t-test (multiple comparison Bonferroni correction for number of tested regions, N=8, alpha=0.05/8=0.00625). Error bars represent standard error of the mean. For individual data, see Supplementary Figure 1b. (c) Searchlight ROI representational similarity analysis between internal processing (Experiment 1, four internal tasks combined) and "scenes" > "objects" contrast (Experiment 3). Top: unthresholded statistical results map; Bottom: thresholded significant clusters (primary voxel-wised threshold p < 0.001, p<0.05, cluster size corrected). Note the high similarity in the PHC, and to a lesser extent in the retrosplenial cortex and middle occipital gyrus, but not in other regions of the cortex.

Figure 5

Experiment 4 (n=34): Language-related processing. (a) Group-level random effect analysis of language-related processing of Experiment 4 (contrast: meaningful sentences > meaningless non-words). Statistical threshold: primary voxel-wised threshold p < 0.001, p<0.05, cluster size corrected. Note, that the largest and most significant clusters within the DMN were found in the LTC. (b) ROI representational similarity analysis between internal processing (Experiment 1, four internal tasks combined) and language-related processing (meaningful sentences > meaningless non-words, Experiment 4). Note, highest similarity values in the bilateral LTC. Significance above zero was established using one sample, two-tail t-test (multiple comparison Bonferroni correction for number of tested regions, N=8, alpha=0.05/8=0.00625). Error bars represent standard error of the mean. For individual data, see Supplementary Figure 1c. (c) Searchlight ROI representational similarity analysis between internal processing (Experiment 1, four internal tasks combined) and language-related processing (meaningful sentences > meaningless non-words, Experiment 4). Top: unthresholded statistical results map; Bottom: thresholded significant clusters (primary voxel-wised threshold p < 0.001, p<0.05, cluster size corrected). Note the highest similarity in the bilateral LTC.

Figure 6

Summary results: neural loci of three cognitive processes. Results reflect thresholded and binarized maps of corresponding searchlight representational analyses (Figs. 3c, 4c and 5c). The three cognitive processes are self-referential processing (yellow color), mental scene construction (magenta color), and language-related processing (green color). Note that cognitive systems had specific loci (i.e., no spatial overlap between cognitive systems).