We recorded maximum expiratory flow-volume curves in 3046 healthy persons, blacks and whites, age 7 and over--a representative population of lifetime nonsmokers except for some black adult males, who were healthy smokers or ex-smokers. We computed regression equations for lung function measurements (FVC, FEV1.0, FEV10/FVC, PEF, MEF 50% and MEF 25%) as a function of age, height and weight terms for eight subgroups (by sex and race, and for children or adults). Objective statistical criteria were used to select the optimal equations. Simple linear regressions on age and height are inaccurate, in particular for young adults and for the elderly. Weight affects most function measurements: lung function first increases with weight ('muscularity effect') and decreases with further increases in weight ('obesity effect'). The regression equations allow more accurate prediction of normal lung function. In addition, the lower 95% confidence limits are closer to the predicted values and are valid regardless of height, weight and age within each subgroup.