Instructions and experiential learning have similar impacts on pain and pain-related brain responses but produce dissociations in value-based reversal learning

Description: This includes key maps from Atlas et al., 2021 (biorxiv). The experiment involved thermal pain administration preceded by pain-predictive cues, and cue contingencies reversed three times during the experiment. One group (n = 20, Instructed Group) was informed about contingencies before learning & prior to each reversal, whereas a second group (n = 20, Uninstructed Group) learned only through experience. We focused on BOLD responses to medium heat that was crossed with pain-predictive cues and conducted multilevel mediation analyses that measure brain mediators of cue effects on pain. We measured mediation of both current cue effects (i.e. cues that update as contingencies change) as well as original cue effects (i.e. effects of original cues regardless of reversal). We also measured associations with parameters (expected value and prediction error) from a quantitative model based on an adapted Rescorla-Wagner learning model with an additional parameter to measure the effects of instructed reversals. This model was fit to pain ratings on medium heat trials.