The response strategy and the place strategy in a plus-maze have different sensitivities to devaluation of expected outcome

Previous studies have suggested that spatial navigation can be achieved with at least two distinct learning processes, involving either cognitive map-like representations of the local environment, referred to as the “place strategy”, or simple stimulus-response (S-R) associations, the “response strategy”. A similar distinction between cognitive/behavioral processes has been made in the context of non-spatial, instrumental conditioning, with the definition of two processes concerning the sensitivity of a given behavior to the expected value of its outcome as well as to the response-outcome contingency (“goal-directed action” and “S-R habit”). Here we investigated whether these two versions of dichotomist definitions of learned behavior, one spatial and the other non-spatial, correspond to each other in a formal way. Specifically, we assessed the goal-directed nature of two navigational strategies, using a combination of an outcome devaluation procedure and a spatial probe trial frequently used to dissociate the two navigational strategies. In Experiment 1, rats trained in a dual-solution T-maze task were subjected to an extinction probe trial from the opposite start arm, with or without prefeeding-induced devaluation of the expected outcome. We found that a non-significant preference for the place strategy in the non-devalued condition was completely reversed after devaluation, such that significantly more animals displayed the use of the response strategy. The result suggests that the place strategy is sensitive to the expected value of the outcome, while the response strategy is not. In Experiment 2, rats with hippocampal lesions showed significant reliance on the response strategy, regardless of whether the expected outcome was devalued or not. The result thus offers further evidence that the response strategy conforms to the definition of an outcome-insensitive, habitual form of instrumental behavior. These results together attest a formal correspondence between two types of dual-process accounts of animal learning and behavior.