The problem of air pollution modeling is treated beginning from a philosophical standpoint, in which a model is viewed as a universal statement and a complementary set of singular statements from which specific cause-effect relationships are deduced; proceeding to the formulation of a specific model from fundamental physical principles. In the course of the analyses, a number of basic issues are examined. These include the types of information that an air pollution model is capable of providing (it is shown that specific events are not predictable, only the set of possible events can be described); the problem of model validation (even with a perfect model and error-free input data and observations, discrepancies will exist between predicted and observed quantities); the character and representation of long-range dispersion (the conventional concepts of transport and diffusion become illdefined when applied within the context of long-range dispersion models); and other topics relevant to the use of models in decision making processes.