A sample of 953 individuals was randomly selected from the population of drunken drivers convicted in 1991 in Sweden (n=20,235). The analysis is confined to men, since the number of women were too few for a reliable analysis. The variables were coded on the basis of records from the preliminary investigations and official records on prior criminality. A path analysis yielded two main conclusions. First, alcohol misuse has a strong effect on BAC, but it does not affect the frequency of drunken driving. Second, criminality has a strong effect on the frequency of drunken driving, but not on BAC. It thus seems fruitful to keep these two aspects of drunken driving conceptually and empirically apart. The findings of the analysis thus support the hypothesis that how often an individual drives while intoxicated depends on his criminal involvement, and the degree of intoxication depends on his involvement with alcohol. That driving while intoxicated would be an ingredient in a wider context of social deviance or illicit behavior seems compelling even though the issue of causal attribution cannot be definitely settled here. Considering the potential bias of cross-sectional data, particularly in terms of selection effects, this proposition should be tested in other ways, so that the exogenity of the criminality factor can be better ascertained. 7 tables, 1 figure, and 17 references