An earlier study retrospectively evaluated the effectiveness of six separate substance abuse treatment programs and generated a set of hypotheses for matching patients to the most appropriate programs. In the present study, these predictors and the matching strategy were tested in a prospective design, using the same treatment programs and a new sample of 130 alcohol-and 256 drug-dependent patients. The new group of patients who were treated in their predicted program (matched patients) were compared with those patients from the same sample who were not treated in their predicted program (mismatched patients). Treatment staff were not apprised of the matching criteria or which patients were matched, thus permitting an experimental test of the predictions. Results indicated superior performance during treatment and an average of 19 per cent better 6-month outcomes for the matched patients than for their mismatched counterparts. The matching effect was seen in both the alcohol-and drug-dependent samples and in all treatment programs. The authors discuss the application of these findings to other types of patients and treatments in substance abuse and other fields of psychiatry.