Results from 15 years of research based on samples of substantial size are reviewed. These studies favor a hypothesis that the phenomena referred to as alcoholism are by no means unitary. The term alcoholism is best regarded as merely a rubric under which several distinct conditions can be grouped. Results from studies of the covariations among self-reported symptoms and other drinkingrelated variables indicate some 15 psychometrically independent first-order factors, the interrelationships among which indicate five broad second-order dimensions. These factors, as such, and combinations of these factors, as in second-order dimensions, provide reliable operational definitions of what people refer to when they use the word alcoholism. These different operational definitions correspond to different concepts (i.e., constructs). They have different relations to variables of etiology (as in a person's background and current conditions), variables of the total personality, variables of outcome following different kinds of treatment, and variables of life-span development. They provide a basis for diagnosing different problems associated with the use of alcohol, and for designing effective programs of therapy. In the future, improved theory, treatment, and social understanding of the use and abuse of alcohol should be premised on multivariate measurement.