Outcomes of treatment for alcoholism parallel those reported for psychotherapy generally: improvement 62%-68%, deterioration 6%-10%, and "spontaneous" remission 1 %-33%. Several client characteristics and the amount of treatment are related to improvement, other treatment features are not, and therapist effects and client-treatment-therapist interactions have not been studied. Measurement problems and individual variation in drinking behavior tend to inflate improvement. Controlled drinking not only is an appropriate treatment goal for some alcoholics but also has implications for prevention of alcoholism. Future evaluation research should focus on client-therapist interactions related to entering and remaining in treatment, social climate, cost-benefit analyses, and staffing patterns.Evaluation of treatment for alcoholism has a long, if methodologically undistinguished, history. As early as 1942, Voegtlin and Lemere reviewed studies published between 1909 and 1941 that evaluated any form of treatment for alcoholics and concluded, among other things, that no one form of treatment appeared to be more efficacious than any other, but they questioned the quality of much of the literature. Reviewing evaluations of psychotherapy with alcoholics appearing between 1952and 1963, Hill and Blane (1967 refused to summarize findings about treatment outcome because the results might have been misleading, given the ineptness of the evaluations. Instead, they confined themselves to a discussion of the minimal requirements that outcome studies must meet, noting that only 2 of 49 studies reviewed met accepted standards of evaluation research with regard to design, measurement, analysis, and reporting of results; nearly all the rest took the form of gross retrospective surveys that employed superficial, unreliable, and unvalidated measures and inadequate follow-up procedures.More recent studies and reviews reveal desirable advances in methodological sophistication, notably present in research design, measurement of outcome, and statistical analyses. Emrick (1975) reported that in 72 of 384 studies patients were randomly assigned to treatments or were matched on critical variables; over 90% of these studies have appeared since 1965. More precision in the measurement of drinking behavior, a key outcome variable, has been introduced by using self-report techniques adapted from surveys of drinking behavior in the general population (Cahalan, Cisin, &