Fig. 3: Functional connectivity-based predictive modeling of trial-wise novelty and appropriateness fluctuations.
a A schematic of the analysis pipeline. Within training data, novelty score (orange arrows), appropriateness score (blue arrows), and functional connectivity matrices (based on a 300-node whole-brain atlas) were extracted for each participant. Red indicates positive functional connectivity of novelty/appropriateness. Blue indicates negative functional connectivity of novelty/appropriateness. Significant edges were identified at a threshold of p < 0.01 (uncorrected). b Predicted novelty ratings and predicted appropriateness ratings of the held-out subset were calculated based on the training result. c Edges strongly contributing positively (red) and negatively (blue) to the predictive model of novelty (top) and the predictive model of appropriateness (bottom). A degree threshold of 1 was applied; i.e., nodes involved in at least one contributing edge are displayed. d The correlation between the predicted and observed novelty ratings based on brain connectivity for the significant predictions. Source data are provided as a Supplementary Data file. e The correlation between the predicted and observed appropriateness ratings based on brain connectivity for the significant predictions. Source data are provided as a Supplementary Data file. f Correlation value between predicted and observed novelty ratings was compared with a null distribution of r values derived from 1000 permutations of shuffled functional connectivity matrix. The gray lines represented the results (r values) of 1000 permutations. The red line represented the correlation value between predicted and observed novelty ratings and it (r = 0.099) was greater than the r values of all 1000 permutations, which showed that the CPM of novelty was significant (ppt < 0.001). Source data are provided as a Supplementary Data file. g Same as (f), except for the predicted and observed appropriateness ratings. The blue line represented the correlation value between predicted and observed appropriateness ratings and it (r = 0.121) was greater than the r values of all 1000 permutations, which showed that CPM of appropriateness was significant. Source data are provided as a Supplementary Data file.
