We review epidemiologic studies of particulate air pollution and mortality in U.S. cities with respect to important methodologic issues. Many of these studies suffer from serious deficiencies in their control of the confounding effects of other pollutants. As a consequence, the small risks reported to be associated with the particulate component of air pollution could easily be attributed to residual confounding by co-pollutants. Most studies, moreover, have not considered modification of air pollution effects by seasonal factors, making the interpretation of the estimated risks difficult. We use a new analysis of mortality in Philadelphia that considers four pollutants simultaneously as well as seasonal effects to illustrate the methodologic issues raised in this paper. Air pollution, which is a complex mixture, appears to be associated with mortality even at the generally low levels of pollution in U.S. cities, but currently neither the statistical tools nor the biological understanding of mechanisms exists to tease out the contribution made by each component of this mixture. We conclude that it is not possible with the present evidence to show a convincing correlation between particulate air pollution and mortality.