Ten surveys of drinking in the U.S.A. were combined, producing a sample size of 12,603 cases. The present study uses only non-Hispanic Black and White races, bringing the total number down to 9891. Eight drinking categories were used, the heaviest being those who drank at least eight drinks at least three times a week. Classification by sex, age, race and socio-economic status produced 72 social subgroups. The distribution of drinking patterns varied a great deal in these subgroups; frequent light drinkers, for example, were 0% in one subgroup and 42% in another; abstainers were 0% in one group and, in another, 76%. The type of classification of ‘heavier’ drinkers so often used in surveys of small sample size, where the lower limit of amount of drinking is unrealistically low, may give a false impression of the distribution of drinking patterns. We found that the proportions of respondents falling into our categories 1, 2, 3 and 4 (corresponding to many definitions of ‘heavier’ drinkers) are not good predictors of frequent drunkenness, or of serious problems.