The relationship between abstinence self-efficacy and substance use at 2 years was examined among a sample (N = 470) of persons with substance use disorders and recent incarceration histories. Participants were assigned to residential (therapeutic community/TC or Oxford House) or nonresidential (usual care) conditions. The authors hypothesized abstinence self-efficacy would predict decreased substance use, and residential treatments would moderate this relationship. A conditional effect was observed, with low levels of abstinence self-efficacy predicting significant increases in substance use in the TC and usual care conditions. Supplemental analyses revealed significant decreases in substance use over time among participants in the Oxford House condition, and a significant conditional effect (gender x treatment condition) in relation to substance use. Findings point to the need for researchers to examine factors that mitigate the relationship between abstinence self-efficacy and substance use outcomes, and for treatment providers to consider the Oxford House model for this population.