Perhaps the most compelling evidence for an effect of low levels of airborne particulate matter (PM) on health comes from timeseries regression analyses that link daily fluctuations in PM and mortality. The evidence is descriptive; the studies were retrospective, so variables could not be manipulated to test specific hypotheses. The approach, although powerful statistically, has rested on several premises: a) that a limited number of fixed outdoor monitors of air quality, a circumstance universal to these analyses, provided reliable estimates of exposure of the population-at-risk; and that adequate adjustments were made for both b) potential confounders, such as collinear air pollutants and changes in weather, and for c) cyclic factors that might modify mortality rate, such as day of the week and seasonal trends. Daily and seasonal changes in weather may affect mortality rate both directly and indirectly, the latter by influencing the physicochemical properties and ground-level concentrations of PM and of air pollution in general. The soundness of these underlying premises remains a matter of debate.Is the association causal? Descriptive epidemiologic studies alone generally are viewed as insufficient to establish causation, especially if the estimated effect is small. They may, however, be used to infer causation. To justify the inference, a number of criteria should be met (1-3). Perhaps the most elusive of these criteria has been a biologic explanation or mechanism for the association (4). Indeed, weak biologic plausibility has been cited by Vedal (5) as "the single largest stumbling block to accepting the association as causal" (p. 558). Our objective (accepting the weight of the evidence as supporting a casual association) is to propose an explanatory hypothesis that, in accordance with Popperian principal, is refutable. We first briefly review salient features of the time-series evidence and their implications.First, the association has been found among communities and nations that differed in their principal sources and composition of air pollution, including PM, as well as in climate. This would imply that no specific attribute of air pollution is an essential proximate cause of death and that an undefined number of such attributes may qualify as sufficient causes. Other forms of environmental stress, including meteorologic variables, may also qualify as sufficient causes (6-8).Second, the association appears independent of population size and density. It has involved chiefly elderly persons with one or more chronic diseases, usually of the heart, blood vessels, or lungs. This would imply that failing health, attributable to aging or illness, largely defines the population at risk. As a corollary, the rest of the population is probably not at risk, especially at current levels of air pollution. And because the rate of physiologic decline can vary markedly among individuals, chronologic age is not a reliable index of vulnerability. However, one must be cautious regarding attributed causes of death. Death ...