Figure 3.

Mean rank accuracies of sentence prediction at the group level and individual level. The Group-level indicates the model accuracy on fMRI images aggregated across participants. The Mean over participants indicates the mean model accuracy within individual subjects, i.e., the mean of the next 10 bars. The dashed and dotted lines indicate the critical values of accuracy being significantly different from chance (dashed: p < 0.01; dotted: p < 0.0001). The critical values of accuracy were determined by 100,000-iteration random permutation tests of the corresponding classification paradigms, taking into account the number of cross-validation folds, number of test items per fold, and number of classes. All the other analyses were performed at the individual participant level unless otherwise specified.