Substance addiction has been a worldwide problem characterized by the lack of control over the use of substances in spite of increasingly negative consequences. It not only damages the addicts’ physical and mental health, but also threatens the social stability. There is evidence that reversal learning plays an important role in the occurrence and development of addiction. Reversal learning is the ability to update the stimulus-outcome associations according to the changing environment. It has to do with people’s social and emotional behavior. In a classical reversal learning task, subjects are presented with two visual stimuli simultaneously. One is accompanied by a reward and the other accompanied by a punishment. Subjects are required to choose the stimulus that could bring the reward to them through trial–error learning. After a period of time, the stimulus previously associated with reward will lead to punishment, whereas the stimulus previously associated with punishment will lead to reward. Then the subjects need to establish the new stimulus–outcome associations. However, the research on reversal learning in substance addiction has showed that addicts usually have problem with adapting their choice to reward contingencies. Once they learned that a specific stimulus has a certain outcome, they may struggle to learn that the same stimulus is later associated with a different outcome. Specifically, substance users may continue to attribute positive outcomes to some stimuli even in conditions when it has changed. These findings have also been confirmed by studies?on?laboratory?animals. The present paper reviews the research on addicts’ reversal learning: First of all, although a few studies reported that there was no reversal learning deficits in certain substance addicts (e.g. heroin users), most studies showed impaired reversal learning among substance addicts. In addition, there are two forms of reversal learning deficits: perserveration and excessive switching, respectively reflecting impulsivity and compulsivity. In the next place, anomalous activities in reversal–specific brain areas, such as ventrolateral prefrontal cortex (VLPFC) and dorsolateral prefrontal cortex (DLPFC), have been found among the substance addicts compared to the controls. Such results provided evidence for understanding behavioral impairment from a physiological perspective. What’s more, the findings of reversal learning in the context of substance–related stimuli is mixed, which may derive from different types of reversal learning task. In the end, the reversal learning ability could predict addictive behavior. Individuals with poor reversal learning are more likely to use addictive substances. However, there are still some unclear issues need to be further explored in the future: (1) the brain mechanism of reversal learning deficits in different types of substance addicts; (2) the influences of substance–related stimuli on addicts’ reversal learning; (3) the effects of the addicts’ impulsivity and compulsivity on their reversal learning; (4) the predictive power of reversal learning on addictive behavior. All in all, understanding the key role of reversal learning in substance addiction is not only rather meaningful for understanding the mechanism of addiction, but also can provide a new direction for treament and prevention of substance addiction.