Fig. 2: Autobiographical mental image features were reconstructed from fMRI data using a general semantic decoding model.

A Reconstruction accuracy of autobiographical mental image features, decoded from fMRI activation across the entire cortex. Accuracy was computed as Spearman correlation against participants’ feature ratings across the 20 autobiographical scenarios (Wedding, Exercising, etc). Bars indicate mean accuracy across participants. To estimate the statistical significance with which each feature could be reconstructed, one sample signed ranks tests were run to compare correlation coefficients to zero (n = 50 participants, one-tailed). P-values were adjusted according to the false discovery rate (FDR)³⁵. 11/20 features had accuracies significantly greater than zero [FDR(p) < 0.05]. B Reconstructed mental image feature ratings reflect interparticipant differences: to evaluate whether reconstructed feature ratings reflected interparticipant differences, a two-alternative forced choice decoding analysis was run, testing how often participant pairs could be told apart by comparing the reconstructed feature ratings to each other’s idiosyncratic feature ratings. This was evaluated on the entire vectorized matrix of 20 features for 20 scenarios, as well as on individual features (see plot bars). Red bars indicate significant decoding accuracies [FDR(p) < 0.05, permutation test, one-tailed]. Orange bars indicate decoding accuracies where permutation tests returned p < 0.05, but values did not survive FDR correction. The analyses were computed on all fifty participants as well as on each age group, as labeled beside the plots. Some icons were adapted from public domain resources on ClipArtMax (https://www.clipartmax.com/). Source data are provided as a Source Data file.