Aims. Two randomized controlled trials of residential drug abuse treatment programs found the programs to be equally effective, based on outcomes among those assigned to the treatments. This study aimed to compare the relative efficacy of the programs, based on outcomes among those who received the specific treatment program as planned. Design. Secondary analyses of data from two concurrent randomized controlled trials, with stratification by actual length of stay. Setting. Two residential drug abuse treatment facilities in the United States. Participants. Six hundred and twenty-eight clients were enrolled over a 2–year period, representing 85% of all clients admitted, 91% of all eligible clients, and 95% of those asked to participate. Interventions. At one facility, clients were randomized to 3-month or 6-month versions of a traditional therapeutic community program. At the second facility, clients were randomized to 3-month or 6-month versions of a modified therapeutic community program that emphasized relapse prevention and health education. Measurements. Time from admission to first drug use (except alcohol); and Addiction Severity Index (ASI) composite scores for severity of drug, alcohol, legal, and employment problems. Findings. Five hundred and thirty-nine clients (86%) completed a follow-up interview at least 16.5 months after admission. In the relapse prevention trial, benefits of the 6-month program were generally limited to those who stayed at least 40 days. In the therapeutic community trial, among those who stayed at least 171 days, the 12-month program had a beneficial effect on employment. Otherwise, there were inconsistent differences between the 6- and 12-month programs. Conclusions. On average, clients who stayed in treatment at least 80 days benefited from continuing in treatment for up to 6 months, but not beyond. Conversely, those admitted to programs of longer planned duration who dropped out of treatment early had worse outcomes than those who dropped out of shorter programs. Thus, although longer planned duration of treatment may be efficacious, it is not effective.