Aims. (1) To assess the benefits of matching alcohol dependent clients to three treatments, based upon a priori hypotheses involving 11 client attributes; (2) to discuss the implications of these findings and of matching hypotheses previously reported from Project MATCH. Setting and participants. (1) Clients receiving outpatient therapy (N = 952; 72% male); (2) clients receiving aftercare therapy following inpatient or day hospital treatment (N = 774; 80% male). Intervention. Clients were randomly assigned to one of three 12-week, manual-guided, individual treatments: Cognitive Behavioral Coping Skills Therapy (CBT), Motivational Enhancement Therapy (MET) or Twelve-Step Facilitation Therapy (TSF). Design. Two parallel but independent randomized clinical trials were conducted, one with outpatients, one with aftercare clients. Participants were monitored over 15 months including a 1-year post-treatment period. Individual differences in response to treatment were modeled as a latent growth process and evaluated for 17 contrasts specified a priori. Outcome measures were percentage of days abstinent and drinks per drinking day. Findings. Two a priori contrasts demonstrated significant post-treatment attribute by treatment interactions: (1) outpatients high in anger and treated in MET had better post-treatment drinking than in CBT; (2) aftercare clients high in alcohol dependence had better post-treatment outcomes in TSF; low dependence clients did better in CBT. Other matching effects varied over time, while still other interactions were opposite that predicted. Conclusions. (1) Anger and dependence should be considered when assigning clients to these three treatments; (2) considered together with the results of the primary hypotheses, matching effects contrasting these psychotherapies are not robust. Possible explanations include: (a) among the client variables and treatments tested, matching may not be an important factor in determining client outcomes; (b) design issues limited the robustness of effects; and (c) a more fully specified theory of matching is necessary to account for the complexity of the results.