Identification of people who will benefit most from brief interventions is an important research challenge in the study of addictive disorders. The current study investigated predictors of response to correspondence interventions for alcohol abuse. We examined both subject retention and alcohol intake over a 12-month period. The primary focus was on the predictive utility of self-efficacy, stages of change and alcohol dependence. Self-efficacy performed relatively well in the study, predicting both retention and later consumption. When predicting 12-month consumption from pretest assessments or examining subject retention over the last 6 months, self-efficacy offered a significant contribution to multivariate analyses. However, in some other predictions a significant effect of self-efficacy was eliminated after the entry of other variables. Stages of change significantly predicted mid-way through treatment, but did not provide an independent prediction of overall retention or treatment response. Neither the degree of alcohol dependence nor level of alcohol problems figured in any of the predictions. Older subjects stayed longer in the study, and those with lower intake and higher pretest self-efficacy had the lowest consumption at 12 months. Results are compared with previous research on prediction of outcomes in addictive disorders.