Extended Data Fig. 5: Lack of diversity in OFC and LPFC.

We additionally ran analyses in all reward selective neurons (as opposed to only RPE-selective neurons) in OFC and LPFC as an exploratory analysis to assess whether consistent diversity was present when more neurons entered the analysis, since these regions have a smaller proportion of RPE-selective neurons compared to ACC. Same analysis as in Fig. 1, but for OFC and LPFC on all reward-selective neurons or RPE selective neurons. We applied exactly the same criteria and analyses to these brain regions as we did in ACC. As in Fig. 1, we computed the Pearson correlation for each of 1000 independent data partitions, and calculate the mean and geometric mean of the R and p-values, respectively. The coloured (left) histograms are the distributions of the reversal points, and the grey (right) histograms are the log(p-values) from the correlations. a) OFC reward-selective neurons. b) OFC RPE-selective neurons. c) LPFC reward-selective neurons. d) LPFC RPE-selective neurons. With the exception of the reward-selective neurons in OFC (a), none of these analyses were significant. Moreover, when we compared the diversity of these reward-selective neurons in OFC (a) across stimulus set (that is, Fig. 1e analysis), the correlation between stimulus set 1 and 2 was not significant (R = 0.15; P = 0.35). This may suggest the diversity in these OFC neurons is due to, for example, stimulus-selectivity, whereby some neurons are selective for stimuli coding the edges of the reward distribution, which could appear as optimism/pessimism in a given stimulus set, but does not generalise across stimulus set as would be expected from diversity related to value. The RPE-selective neurons had no consistent diversity, and as RPE selectivity is a requirement to test further predictions of distributional RL, we did not look for further distributional RL signatures in these brain regions.