Steps in the research process are hypothesis formation, sample selection, definitions of alcohol use and crime, and data analysis. The vast majority of alcohol-crime studies have been conducted without clearly defined hypotheses. Because studies are usually couched in terms of bivariate models, they are usually inadequate in addressing such issues as what differentiates criminal events where alcohol is merely present from those where it plays an active role in the crime and what factors distinguish the majority of drinking situations that do not culminate in a crime from those that do. The major source of sample bias is the selection of criminals or alcoholics that have been identified as such by an agency of social control. The evidence suggests that there are a number of different populations of alcoholics, each with a different level and pattern of prior criminality. The considerable variability in definitions of alcohol use and alcoholism in the alcohol-crime literature makes it difficult to compare results among studies. Further, data analysis in much of the literature is confined to a simplistic bivariate model. In examining the role of alcohol in the criminal event, at least four sets of factors should be considered as possible control variables: drinking patterns, personality characteriestics, contextual effects, and norms relating to the frequency and acceptability of drinking among various race, sex, and economic groups. Over one-third of the studies in the literature failed to control for any of these factors. (Author summary modified)