This study examines the hypothesis that treatment is a cumulative process; that is, treatment success is best viewed in terms of the patient's entire treatment history, rather than the index treatment episode. Three-hundred and eight patients with a primary heroin addiction were studied for 2 years posttreatment. Using posttreatment arrests as the dependent variable, the effects of prior treatment were assessed. Those with six or more prior treatment episodes and who had been in treatment for 12 or more months during the most recent episode averaged only 0.2 arrests in the 2 years posttreatment, while those with no prior treatment, but 12 or more months in the recent treatment averaged 0.88 arrests. Logistic analysis found that each prior treatment reduced the probability of a posttreatment arrest by 25%. Based on a linear regression, patients with six or more treatments prior treatments averaged half the number of posttreatment arrests as someone with no treatments before the index episode.