Factual and Counterfactual Action-Outcome Mappings Control Choice between Goal-Directed Actions in Rats

The capacity to extract causal knowledge from the environment allows us to predict future events and to use those predictions to decide on a course of action. Although evidence of such causal reasoning has long been described, recent evidence suggests that using predictive knowledge to guide decision-making in this way is predicated on reasoning about causes in two quite distinct ways: choosing an action can be based on the interaction between predictive information and the consequences of that action, or, alternatively, actions can be selected based on the consequences that they do not produce. The latter counterfactual reasoning is highly adaptive because it allows us to use information about both present and absent events to guide decision-making. Nevertheless, although there is now evidence to suggest that animals other than humans, including rats and birds, can engage in causal reasoning of one kind or another, there is currently no evidence that they use counterfactual reasoning to guide choice. To assess this question, we gave rats the opportunity to learn new action-outcome relationships, after which we probed the structure of this learning by presenting excitatory and inhibitory cues predicting that the specific outcomes of their actions would either occur or would not occur. Whereas the excitors biased choice toward the action delivering the predicted outcome, the inhibitory cues selectively elevated actions predicting the absence of the inhibited outcome, suggesting that rats encoded the counterfactual action-outcome mappings and were able to use them to guide choice.