To estimate variability in ethanol metabolism, eight normal men received oral doses of 95% ethanol (1 ml/kg) after an overnight fast on each of 4 successive weeks. For each subject, slopes of linear decay curves of blood ethanol were highly reproducible--coefficients of variation ranged from 3% for the least variable subject to 12% for the most variable subject (mean, 8%). Compared to this low intraindividual variation, interindividual variation was slightly higher--the mean coefficient of interindividual variation was 14%, with a range from 10% to 17%. A one-way ANOVA with repeat measures showed that on any one of four separate occasions of ethanol administration, the eight subjects differed from one another (p less than 0.01), but that each subject remained similar from one test to another. The consumption of food before ethanol administration increased variability. Fever associated with upper respiratory infection also increased variability, but exercise did not.