The authors evaluated the reproducibility and validity of alcohol consumption measured by a mailed, self-administered questionnaire in two large prospective studies. Subsamples of 173 women and 136 men from these cohorts completed two food-frequency questionnaires and multiple 1-week diet records (four for women and two for men) and provided a fasting blood sample. Intakes of alcohol reported over the previous year by food frequency questionnaire 2 correlated highly with intake assessed by multiple week diet records completed over this period (Spearman r = 0.90 in women and 0.86 in men). Mean daily alcohol intake as assessed by the diet records and the questionnaires was very similar. Serum high density lipoprotein cholesterol levels were significantly correlated with alcohol intakes estimated from the questionnaire (r = 0.40 in women and 0.35 in men) and for the diet records (r = 0.33 in women and 0.38 in men). Four years after completing the diet records, women were asked to report their alcohol intake of 4 years earlier; this estimate was highly correlated with the diet records (r = 0.84) and with the earlier high density lipoprotein cholesterol (r = 0.40). These data suggest that a simple self-administered questionnaire can provide useful estimates of alcohol intake over an extended period of time in subjects participating in prospective epidemiologic studies.