This prospective, observational study investigated factors predicting a lapse to heroin use in 74 heroin-abstinent methadone maintenance patients. After baseline data collection, participants were assessed twice per week for 7 weeks and again at 6 months after baseline. Proportional hazards regression and logistic regression were used to investigate the effects of study predictors on heroin use. A goal of absolute heroin abstinence consistently predicted a lower risk of a lapse, whereas marijuana use was associated with a greater risk. Stress variables were not predictive. The abstinence goal and stress results were consistent with the authors' previous studies of other drug treatment samples. This line of research suggests that factors influencing lapses are similar across drug treatment populations and the role of stress in precipitating relapse remains unresolved.