Abstract: The training outcome plays a critical role not only in the initial acquisition of behavior, but also in the selection and initiation of previously acquired responses. Although models of outcome encoding differ substantially in their assumptions about the content of learning and the specific function of the outcome representation in response selection, they tend to make the common prediction that the noncontingent presentation of an outcome should selectively reinstate the performance of an extinguished response with which it had been paired. The current experiments were designed to delineate further the psychological and neural processes that underlie outcome-mediated response selection. We found evidence that the capacity of an outcome to reinstate performance was mediated through an outcome-response association. This contrasts with the finding that the influence of outcome value on instrumental performance is primarily mediated by a response-outcome association. Additional studies found that instrumental reinstatement was insensitive to a reduction in the incentive value of the mediating outcome, providing a further dissociation between these two forms of outcome encoding. Although the medial prefrontal cortex (mPFC) has been implicated in instrumental outcome encoding, its specific involvement remains unknown. We found that, although pretraining mPFC lesions disrupted the sensitivity of instrumental performance to outcome devaluation, posttraining lesions failed to have this effect, suggesting the mPFC may be selectively involved in the acquisition, but not the storage or expression, of response-outcome learning. Neither lesion had an effect on the outcome's capacity to reinstatement instrumental performance. Finally, we found evidence that the orbitofrontal cortex (OFC) is selectively involved in Pavlovian learning; posttraining lesions disrupted Pavlovian-instrumental transfer, but left intact the sensitivity of instrumental performance to outcome value. OFC lesions also impaired Pavlovian contingency learning, suggesting that this structure may play a role in using Pavlovian outcome expectancies to both guide behavior and inform subsequent learning.