Clomipramine, haloperidol, and placebo were compared with baseline in the treatment of autism, and overall outcome, specific symptoms, and side effects were examined. It was hypothesized that clomipramine would be better tolerated than haloperidol and prove superior on a measure of stereotypy. Individuals with a DSM-IV diagnosis of autistic disorder (mean age, 16.3 years; range, 10-36 years) were randomly assigned, by using a Latin square design, to the following 7-week trials: placebo, clomipramine (mean daily dose, 128.4 mg; range, 100-150 mg), or haloperidol (mean daily dose, 1.3 mg; range, 1-1.5 mg). Data on 36 subjects were analyzed and taken together; the results favored haloperidol. In those patients who were able to complete a full therapeutic trial, clomipramine proved comparable to haloperidol in terms of improvement compared with baseline. However, significantly fewer individuals receiving clomipramine versus haloperidol were able to complete the trial (37.5% vs. 69.7%, respectively) for reasons related to both side effects and efficacy or behavior problems. In the intent-to-treat sample, which is perhaps more clinically relevant, only haloperidol proved superior to baseline on a global measure of autistic symptom severity, as well as specific measures for irritability and hyperactivity. Clomipramine did not seem more effective on a measure of stereotypy, nor was it better tolerated.