This article compares five indices of alcohol consumption in a general population survey conducted in 1985 in the Netherlands. Self-reports of consumption were obtained with a prospective diary, a retrospective 7-day recall method, and three summary measures, such as a quantity-frequency index. The coverage of sales data appeared highest for the diary (67%), which suggests a higher validity. Special attention was given to comparisons of quantity and frequency of drinking between the diary, on the one hand, and the weekly recall and summary measures, on the other. It was found that underreporting, relative to the diary reports, was generally higher in the frequency than in the quantity domain. This result, together with the finding from longitudinal studies that intraindividual variation is also higher for drinking frequency, leads to the conclusion that forgetting is a potent source of undercoverage in surveys and to the hypothesis that large differences in overall drinking pattern between populations (e.g., in regularity of drinking) may account for the large differences in coverage rates of sales data. Furthermore, the subjectively assessed probability of drinking by means of a "usual" frequency question appeared a poor predictor of (diary) drinking frequency for respondents reporting a low or moderate frequency. For subjects claiming a high "usual" drinking frequency, a reasonable correspondence between diary and summary measures was found. This mitigates the fear often expressed that heavy drinkers particularly underreport their consumption.