behavior rather than in experienced feedback probabilities, we
recomputed the correlations between behavior and tract count (see
Section 2.6) while including the experienced probabilities as a
covariate. This did not change the pattern of results in the subsequent
network comparisons.