Extended Data Fig. 9: Comparing VIDS, PINES and VIFS.

a, River plots displaying spatial similarity (calculated as cosine similarity) between the stable decoding maps and the subcortical as well as consciousness network. Ribbons are normalized by the max cosine similarity across networks. Stable decoding models were thresholded at FDR q < 0.05 and positive voxels were retained only for similarity calculation and interpretation. Ribbon locations in relation to the boxes are arbitrary. Pie charts show relative contributions of each model to each network (that is, percentage of voxels with highest cosine similarity for each map). b, The multilevel mediation analytic results showing that VIDS response partially mediates the association between VIFS response and subjective disgust rating in both discovery and validation cohorts. c, The VIFS response plays a partial mediation role in the effect of VIDS response on the disgust rating. In b and c, the mediation analysis examines whether the observed covariance between the independent variable (X) and the dependent variable (Y) can be explained by the third variable (M, also mediator), details see Methods section. Two-sided P values are based on bootstrap tests with 10,000 samples, uncorrected.