Figure 2.

BLA inhibition impairs action-based probabilistic reversal learning, whereas vlOFC inhibition only slows early adjustment to reversals. A, B, Trial structure (A) and timeline (B) of the action-based task. Rats were first surgerized with either hM4Di DREADDs on a CaMKII promoter or eGFP null virus, also on the same promoter. Rats were allowed to recover for 1 week before testing on a stimulus- or action-based reversal learning task. C, Initial learning of a rewarded side. D, Learning during subsequent deterministic (100/0) and probabilistic (90/10) reversal. Plots show cumulative P(Correct) for the first 150 trials with a sliding window of 10 trials. Drug order was counterbalanced such that on R2 and R4 animals received VEH if they were administered CNO first on R1 and R3 and vice versa. There was no effect of inhibition on learning in R2 and R4 (not shown). There was no effect of CNO on learning in the eGFP group. *p < 0.05 significant sex by drug order interaction (see Fig. 3 for learning curves plotted by sex). Bonferroni-corrected post hoc comparisons following mixed-effect GLM with sex as a covariate fixed factor wherein a drug × virus interaction was found resulting in ***p < 0.001 effect of drug only in BLA hM4Di, not in eGFP. The impairment in probabilistic reversal learning following BLA inhibition was also evident in animals with no history of stimulus-based learning and was specific to reversal, not initial probabilistic learning (Fig. 4).