Self-reports of alcohol consumption account for approximately 50 % of the reported sales of alcohol. In the absence of a gold standard, it is not known how accurately different methods of measurement reflect actual consumption and whether under-reporting varies among different populations. The objective of the present study was to compare the consumption reported by the widely used quantity-frequency questionnaire (QFQ) with that reported in a cross-check dietary history interview (DHI), which has higher face validity. In 171 male and 197 female subjects of the Amsterdam Growth and Health Longitudinal Study (mean age 36 years), alcohol consumption was assessed by both the QFQ and the DHI. Most subjects reported a moderate consumption of alcohol by both measures. Spearman correlation coefficients were high (0·77 and 0·87 in men and women respectively). Overall, greater alcohol consumption was reported using the DHI. The difference between the DHI and QFQ reports was usually greater for wine than for beer. Backward stepwise regression analysis showed that the difference in reporting was positively related to a more irregular drinking pattern, and in wine drinkers to the square of the QFQ report. Sex, drinking alone or with others and the CAGE (acronym for four questions on drinking behaviour) score were not related to the difference in reporting. The precision of DHI estimation from QFQ reports and other factors was low. Serious questions arise as to the validity and precision of alcohol consumption measurements based on the QFQ alone. QFQ information may be improved by incorporating questions on the type of beverage and drinking patterns.