One of the most relevant phenomena both from a theoretical and clinical perspective is extinction. In particular, several researchers are interested in the response recovery effects from extinction. Reinstatement is an effect that has been proposed as a laboratory model to study relapse from extinction-based therapeutic treatments. We designed two experiments with humans to evaluate the reinstatement effect in a predictive learning task. In both experiments, participants learned a specific relationship between two cues (X and Y) and two outcomes (O1 and O2) during the first phase. Throughout extinction, both cues were presented without outcomes. After an exposure to the original outcomes, reinstatement of the first-learned information was observed during testing in both experiments. However, we found that the reinstatement effect was contextual modulated (Experiment 1; ηp2 = .78, 90% CI [.48, .86], p < .0001). Furthermore, in Experiment 2 we showed a reduction of reinstatement when an extinction reminder was used ηp2 = .45, 90% CI [.07, .65], p = .012. Theoretical implications are discussed, and some potential uses are mentioned.