Searchlight RSA (outcome presentation; whole-brain).

The searchlight analysis was implemented in the same way as described above for the searchlight RSA at time of door onset. Here, the onset events read into the searchlight script were the outcome onset events generated in GLM2b. Again, the model RDMs specified that the on-diagonal (identical choice–outcome combinations for the two contexts within a condition) was more similar than the off-diagonal (see Fig. 5a) and that the analysis was conducted separately for the two conditions.

a, RSA at trial outcome. We examined the BOLD similarity at the time of outcome between matched choice (door)–outcome state combinations and mismatched combinations between contexts in the two conditions. b, In our MTL ROI, the difference between representational similarity of matched and mismatched combinations was significantly greater in dependent than independent blocks. Error bars show SEM. *significant at p<0.05.