Although quantity-frequency measures of alcohol consumption have been successfully used in epidemiologic studies of alcohol use, other measures may enhance detection of alcohol-disease associations. We examined two measures of alcohol consumption (quantity-frequency and an item eliciting history of heavy drinking) in a population-based longitudinal study of community-dwelling elderly men (baseline n = 1155). The primary utility of the quantity-frequency measure was in separating those who never drink, current drinkers, and former drinkers. The health and functional status of former drinkers was typically poorest. When persons who had consumed alcohol in the month prior to interview were stratified by quartile of total amount of alcohol, there were few systematic differences among the strata. The screening question on heavy drinking, however, was successful in detecting persons who were at markedly increased risk of morbidity and mortality. Thus although quantity-frequency measures are analytically appealing, other measures in some circumstances may serve a supplementary role in detecting persons at risk for adverse outcomes and for understanding the effects of alcohol consumption.