Goal-directed behaviors allow animals to act to satisfy needs and desires. The outcome devaluation task is an effective method for identifying goal-directed behaviors and distinguishing these from other types of behavior. Rats can be trained to lever-press for one or multiple distinct food rewards. During testing, the previously earned food-or a control food for comparison-is devalued by allowing the animal to freely feed on it until they are sated before testing lever-press performance under extinction conditions (no rewards are delivered). Behavior that adapts to reflect the new value of the outcome is considered goal-directed, whereas behavior that continues as in previous training despite the change in outcome value, is not. As more research groups have used this task, variability in the procedures used has increased. Here, we provide a reliable procedure for conducting the outcome devaluation task with appropriate controls. We describe the most common variants of the task and control conditions and discuss troubleshooting measures such as outcome pre-exposure, habituation to pre-feeding chambers and attention to animals' hunger levels. The method outlined can be executed in ~2 weeks including training (~8 d) and testing (1-4 d) by researchers who are familiar with performing behavioral tasks in laboratory rodents, although longer training may be considered for those who are interested in observing habitual control of behavior. This protocol should facilitate the comparison of results from different studies and laboratories, while allowing flexibility in the application of the outcome devaluation task to different research questions.