This study, based on data drawn from the responses of 18,323 males and 25,440 females to the 1988 National Health Interview Survey, a nationally representative, multistage probability sample of the United States, attempts to define more precisely the level of drinking at which the relationship between heart disease and alcohol consumption is a protective one. Its attempt at precision derives from (1) using drinking categories that represent various points within the range of moderate drinking (1-6 drinks) defined in the literature as protective; (2) adjusting for underreporting that commonly occurs in population surveys by using consumption at time of heaviest drinking; and (3) controlling for age, body mass, smoking, former drinker, and former smoker status, duration of drinking, and sociodemographic factors. It also examines whether the relationship derived from these levels conforms to the U-shaped curve that demonstrates the protective effect of moderate drinking when abstainers are not used as the reference group. Relative to infrequent drinkers (less than 1 drink per day), men report more heart disease at the level of more than five drinks per day. However, black men also report more heart disease, relative to infrequent drinkers, at the greater than two drinks per day level; and women report more heart disease at the level of more than two drinks per day at the time of their heaviest drinking. Former drinkers of both genders, considered as an independent variable in the regression analysis, were more likely to report having heart disease. Abstainers, light drinkers, and infrequent drinkers were not significantly different in their reports of heart disease. Our results are consistent with studies that suggest protection from heart disease occurs only at lower levels of drinking.