Numerous studies have examined the natural time course of human lung function growth and decline throughout life. In most of these studies the investigators used statistical models that required a priori assumptions concerning the underlying form or structure of the lung function data, thus introducing possible biases. In this study we used recently developed nonparametric regression (spline) techniques to describe the evolution of lung function measures with age. This procedure yields an optimally fitted smooth curve through the data and estimates of the process velocity and does not require assumptions concerning the underlying shape of the data curves. The lung function growth-velocity curves are used to estimate the age of growth cessation. This technique was applied to the FVC, FEV1, and the FEV1/FVC ratios of 1,295 females and 1,230 males who were tested in at least one of the first nine surveys of the Tucson epidemiologic study of airway obstructive diseases. Data were analyzed stratified according to gender, smoking status, and respiratory symptoms or diseases. The results indicate large differences between the fitted FEV1 and FEV1/FVC smoothed curves of the various subgroups compared with asymptomatic nonsmokers. These differences were most pronounced in the adult symptomatic smokers, who had higher rates of lung function loss that also began at earlier ages, for both sexes. No significant differences were observed between asymptomatic and symptomatic nonsmokers, most likely because of the reduced number of symptomatic nonsmokers, particularly among the males.