Fig. 4: Neural Transitions Coincide with Subjectively Experienced Belief Updates.

a Using Hidden Markov Models, we identified neural transition points within an ROI during the movie (top row). We then compared the timing of these transitions with the belief updates by counting how many updates occurred within different time windows (TRs/scans) following these neural transitions. This varied from 1 to 8 scans. As the window length increases (light grey to black), more belief updates are naturally included. However, the critical prediction is that each region’s transition windows will contain more updates from its preferred domain than from the other domains. b Proportion of belief updates that fall within neural transition windows in each PFC region, for various window sizes. (Left) dmPFC neural transitions are most closely aligned with Action model updates, while amPFC (middle) and vmPFC (right) transitions capture more Agent and State updates, respectively. Statistical significance was assessed via a permutation test on phase- randomised neural time-series. (Asterisks denote window sizes whose observed statistic exceeded the 95th percentile of its permutation-null distribution (p < 0.05, one-tailed test)).