Objective: Many trials have demonstrated the effectiveness of cognitive behavioral interventions for alcohol dependence, yet few studies have examined why particular treatments are effective. This study was designed to evaluate whether drink refusal training was an effective component of a combined behavioral intervention (CBI) and whether change in self-efficacy was a mechanism of change following drink refusal training for individuals with alcohol dependence. Method: The present study is a secondary analysis of data from the COMBINE study (COMBINE Study Research Group, 2003), a randomized clinical trial that combined pharmacotherapy with behavioral intervention in the treatment of alcohol dependence. The goal of the present study was to examine whether a drink refusal skills training module, administered as part of a 16-week CBI (n = 776; 31% female, 23% non-White, average age = 44) predicted changes in drinking frequency and self-efficacy during and following the CBI, and whether changes in self-efficacy following drink refusal training predicted changes in drinking frequency up to 1 year following treatment. Results: Participants (n = 302) who received drink refusal skills training had significantly fewer drinking days during treatment (d = 0.50) and up to 1 year following treatment (d = 0.23). In addition, the effect of the drink refusal skills training module on drinking outcomes following treatment was significantly mediated by changes in self-efficacy, even after controlling for changes in drinking outcomes during treatment (proportion mediated = 0.47). Conclusions: Drink refusal training is an effective component of CBI, and some of the effectiveness may be attributed to changes in client self-efficacy. (Contains 1 table, 2 footnotes, and 3 figures.)