The paper is concerned with methodological problems relating to the scientific study of the effectiveness of Alcoholics Anonymous (AA). Studies of the effectiveness of AA fall into two categories--longitudinal and cross-sectional, and may be criticized on the basis of over-simple criteria of success. The particular problems of the requirement for control groups in studies of effectiveness are pointed out. The problems arising out of the whole process of affiliation and disaffiliation and their implications for scientific study are discussed. The requirement for sound statistical analysis is stressed, and inadequacies in the techniques of follow-up are indicated. The studies which have some bearing on AA as a treatment facility are reviewed. These include multivariate studies involving AA attendance as one factor, studies in which AA was the main variable in a hospital programme, and longitudinal studies of AA from within the organization. Particular difficulties in using cross-sectional surveys for the purpose of estimating efficacy are pointed out. The conclusion reached is that because of the methodological difficulties the totality of these studies dose not add significantly to the knowledge concerning AA which we possess from clinical experience. Moreover, without a change in the unusual characteristics of AA when considered as a treatment facility it seems unlikely to be possible to assess its effectiveness in a scientific manner.