Divergent effects of win-paired cues on learning from timeout penalties in female and male rats

Abstract

In both males and females, linking rewards with salient audiovisual cues in simulated gambling games increases risky choice in humans and rats. However, the prevalence and severity of gambling problems differs in men and women. In previous work, reinforcement learning (RL) models were applied to data from male rats performing the rat gambling task (rGT) to investigate the computational processes promoting risky choice. In the rGT, the optimal strategy is to favor options paired with smaller per-trial gains but shorter and less frequent time-out penalties. Rewards are either delivered with (cued) or without (uncued) concurrent audiovisual cues. Previous work showed these cues drive risky decision making by causing male rats to under weigh the relative cost of timeout punishments, specifically for one of the highly risky options. Here, we applied the same methodology to a large dataset from female rats performing the cued and uncued rGT to investigate whether the same cognitive mechanism drives risky decision making across sexes. Cues decreased the learning rate from all time-out penalties in female rats, rather than specifically from those paired with a risky option. Although females were less sensitive to the shortest time-outs associated with the one pellet option (P1), this computation failed to promote choice of this comparatively safe option due to the overall lower learning rate from penalties. Differences revealed by computational modeling in the way risky choice develops across sexes may help us understand the divergent trajectory of gambling disorder in men and women.