Treatment matching research is predicated on heterogeneity among subjects and their differential response to treatments. The sine qua non of a treatment matching study is the integrity of the treatment variable, since detection of client-treatment interactions requires delivery of treatments that are highly specific, consistent and distinct. Matching research thus presents particular challenges in treatment implementation, as greater heterogeneity in subjects may generate a broader array of problems than study treatments are designed to address, leading to several potential threats to treatment integrity. Moreover, as practiced outside of research settings, treatments for alcoholism are marked by ideological heterogeneity and a lack of purity across approaches. In this article we describe the strategies used in Project MATCH to protect treatment integrity while treating a large and heterogeneous sample of alcoholics in a number of geographically distant sites. These include: strategies for treating a variety of alcoholics within a single treatment approach; development of clinical care guidelines and clinical deterioration criteria; specification of treatments in manuals with minimization of overlapping active ingredients; selection criteria for therapists intended to enhance both generalizability of findings as well as treatment integrity; and extensive therapist training and monitoring.