The reinforcer devaluation task is often used to model flexible goal-directed action, the ability to adaptively modify behavior when the value of a reinforcer changes. Deficits in goal-directed action are reported in multiple neuropsychiatric conditions, including schizophrenia. However, dysfunction is not always apparent in early stages of schizophrenia, possibly due to neural compensation. We designed a novel devaluation task in which goal-directed action could be guided by stimulus-outcome (S-O) [presumably orbitofrontal cortex (OFC)-mediated] or response-outcome (R-O) associations [presumably prelimbic cortex (PL)-mediated] to maintain. Therefore, if our task is able to model neural compensation, damage to either PL or OFC should not impair devaluation because the non-damaged region can compensate for the loss using the alternate strategy. In Experiment 1, rats received bilateral OFC, PL, combined OFC+PL, or sham lesions and then completed our devaluation task. Sham, OFC, and PL lesioned rats showed intact devaluation, whereas the OFC+PL lesion group exhibited impaired devaluation. In Experiment 2, rats received bilateral PL or sham lesions and then completed behavioral training. During devaluation testing, rats were divided into Cue Normal or Cue Switch test conditions. In Cue Switch test conditions, the cuelight and spatial lever location predicted conflicting outcomes so we can determine whether rats are devaluing using a S-O or R-O strategy. Sham rats exhibited normal devaluation performance (based on the lever-location) in both Cue Normal and Cue Switch conditions. PL lesioned rats showed normal devaluation performance in the Cue Normal condition but a reverse devaluation effect in the Cue Switch condition suggesting the PL-lesioned rats were devaluing based on the outcome predicted by the cuelight. In Experiment 3, rats received unilateral infusions of cholera-toxin-b (CTb), a retrograde tracer, into OFC and either bilateral PL or sham lesions and then completed behavioral training. Rats were sacrificed on the last day of training to double-label for Arc and CTb in the mediodorsal thalamus (MD) and Arc in the OFC. We found increased Arc+CTb in MD when PL is lesioned and increased Arc+ neurons in OFC when PL is lesioned. There were no effect of PL lesion in the basolateral amygdala. Our results suggest that our devaluation task can successfully model neural compensation between OFC and PL and this compensation may involve MD. This research demonstrates a method to study how functional neural circuitry is subtly altered like in early stages of schizophrenia.