Fig. 4: Context specificity in HPC, lPFC and throughout the extinction network.

a, Context specificity analysis. Representational patterns corresponding to the presentation of individual context videos were compared across same contexts (blue) and different contexts (pink) during acquisition and extinction. Periods of context specificity were defined as those where same-context correlations were significantly higher than different-context correlations. b, Context specificity during video presentation in HPC (top) and lPFC (bottom). During acquisition (left), context-specific representations were observed in the HPC locked to the onset of the video (p_corr = 0.005). During extinction (right), a similar effect was observed in the lPFC (p_corr = 0.025). c, Context specificity during cue presentation. Left: during acquisition, no significant differences were observed between same and different contexts. Middle: during extinction, context-specific representations were observed at the time period from 0.8 to 1.15 s (p_corr = 0.012). Right: context specificity during extinction (brown) was significantly higher than during acquisition (green; p = 0.001). d, Left: trial-specific values of context specificity, defined as average same minus different-context correlations in each individual trial, were computed in each of our ROIs. Right: focusing on the lPFC, we correlated context specificity values between regions and across trials. Significant regions surviving multiple comparison corrections using cluster-based permutation statistics are outlined in black. In b and c, black horizontal lines at the bottom of each panel depict significant time periods after correction for multiple comparisons using cluster-based permutation statistics. Time zero indicates the onset of the CS, and shaded lines depict group average rho values ± s.e.m. Two-sided paired t-tests were applied at every time point (b and c) or every time-by-time point (d).