Consistent maximum-likelihood heritability estimates of consumption of alcoholic beverages were observed at three separate times during a 14- to 18-year period in adult twin males initially aged 42-56 years in 1969-1973. Log transformation of the average number of drinks/week of the returnees to all three examinations was examined relative to potential covariates representing both antecedents of drinking alcohol and consequences of alcohol consumption. Significant relationships were noted for 38 of the covariates at one or more of the separate examinations, including positive correlations with smoking, coffee consumption, high-density lipoprotein cholesterol, mean corpuscular volume, systolic blood pressure, uric acid and behavioral measures, and negative correlations with blood urea nitrogen, red blood cell count, tea consumption, and tricep skinfolds. Analysis of the average alcohol consumption adjusted for nine independent covariates selected from multiple stepwise regression resulted in a modest decline in maximum-likelihood heritability estimates compared with unadjusted data, but little difference from heritability estimates obtained when abstainers from alcohol (no alcoholic beverages consumed at all three examinations) were excluded. The most striking effect of omitting abstainers from alcohol was the decline in the intraclass correlations in dizygotic twins. Bivariate analyses of alcohol and individual covariates revealed the phenotypic correlation between alcohol consumption and a measure of hostility was primarily environmental, that for high-density lipoprotein, smoking and coffee drinking with alcohol was primarily genetic, and the phenotypic correlation between alcohol consumption and mean corpuscular volume had both significant genetic and environmental correlations.(ABSTRACT TRUNCATED AT 250 WORDS)