Intersubject correlation

We take the ISC to be the mean of the lower triangle of the pairwise correlation matrix of signals between each participant [see (115) for a tutorial]

where

-transformation and

S is a matrix of subject time series data from a total of N participants, and i and j indicate the subject indices. We performed hypothesis tests using a subject-wise bootstrap procedure as recommended by (116), which randomly draws participants with replacement to create a new sample by sample similarity matrix. Correlations between the same participants are dropped before computing ISC. We computed ISC for each participant’s mean vmPFC time series, their vmPFC spatial pattern at each moment in time, and their unique spatiotemporal pattern, defined as the vectorized lower triangle of the participant’s spatial recurrence matrix across all voxels in their vmPFC. We also carried out an analogous series of calculations with each participant’s V1 activity patterns. The similarity of spatiotemporal patterns has an interesting connection to computing distance correlation (117, 118), a measure of general (i.e., possibly nonlinear) statistical dependence between two signals of arbitrary dimensionality.