A method for setting air quality standards for long-term cumulative exposures of a population based on epidemiological studies has been developed. It uses exposure estimates interpolated from monitoring stations to zip code centroids, each month applied to zip code by month residence histories of the population. Two alternative cumulative exposure indices are used--hours in excess of a threshold, and the sum of concentrations above a threshold. The indices are then used with multiple logistic regression models for the health outcome data to form dose response curves for relative risk, adjusting for covariates. These curves are useful for determination of at what exposure amounts and threshold levels, effects which have both statistical and public health significance begin to occur. The method is applied to a ten year follow-up of a sub cohort of 7,343 members of the National Cancer Institute-funded Adventist Health Study. Up to 20 years of residence history was available. Analysis for prevalence of symptoms was conducted for four air pollutants--total oxidants, sulfur dioxide, nitrogen dioxide, and total suspended particulates. For each pollutant, cumulated exposures were calculated above each of five different thresholds. Statistically significant effects were noted for total suspended particulates, total oxidants, sulfur dioxide, past and passive smoking.