A large number of behavioural models of cognition and emotion in rodents rely on food motivation, which could be affected in models of obesity. Our objective was therefore to evaluate the applicability for long-term interventions in rats of a previously known ethologically relevant test1 , able to pick up subtle cognitive impairments, with test parameters independent of food motivation. Male Sprague-Dawley rats (n=18) were water-deprived (20 hours) prior to training in a lever-based water reward task in a modified Skinner box (Fixed Ratio = 1). Following the Optimal Foraging theory, an element of uncertainty was introduced: 90% probability of a reward being yielded combined with a 20% probability of system shut down. Reactivation required rats to break a distant photobeam. Thus, the rats had to estimate when relocating to reactivate the system was feasible. Following training, the mean number of lever presses in an OFF sequence could be organised in decision curves, stable over time (n=12, no significant difference between test sessions). Animals could be classified into 'fast' (n=15, mean 3, SD 1.5), 'slow' (n=2, mean 10, SD 1) and 'non-learners' (n=1, excluded from study) depending on how many sessions they took to reach criterion. This classification did not influence performance after reaching criterion (n=17, no significant difference in number of rewards obtained between test sessions). After reaching criterion, the animals (n=12) were tested at 3 weeks of distance, without evidencing variations in performance and decision rule (P > 0.05). A subset of the animals (n=6) was then tested 3 additional times, spaced 3 weeks from each other, and no significant difference in decision rules was observed (P > 0.05). This preliminary trial shows the integrity of the Decision Making paradigm for cognitive testing in the context of long-term studies (up to 9 weeks after reaching criterion) involving dietary or pharmacological interventions.