Using scales that distinguish between relative genetic and environmental loading, cluster analysis was used to identify three subtypes of alcohol dependence in Caucasian men from the Epidemiologic Catchment Area study (n = 911). Although all subjects met DSM-III criteria for alcohol dependence, only the severe subtype showed evidence of substantial genetic influence. When compared on a range of clinical characteristics, the mild subtype (53% of the sample) was typically least adversely affected and the severe subtype (17%) most affected, with the dyssocial subtype (30%) falling between. Severe subtype subjects had significantly greater comorbid drug dependence and were at least four times more likely than mild subjects to have sought treatment for alcohol problems. Ratio of genetic scale score to total symptom count (genetic ratio) was highest for the severe subtype (mean = 0.37), and negatively correlated with age of first alcohol problem (rs = -0.16) and years between first intoxication and first problem (rs = 0.19). No significant correlations were found between these clinical features and genetic ratio for the mild or dyssocial subtypes. Use of these scales and subtypes may improve our ability to detect specific gene effects in genetic linkage studies and to identify environmental influences in behavioral and epidemiological studies.