In studies of delayed effects of environmental exposures there is a need for accurate reconstruction of exposure, to the environmental factor under study, for a large number of individuals. There is also a need for similar assessments of exposure to possible confounding factors for the same population. In general, such reconstruction is not possible but it is often possible to make an estimate of exposures as they would have likely occurred. This requires selection of populations and circumstances which allow such estimates to be made. For each environmental exposure and for each disease outcome to be evaluated it is likely that the optimum exposure assessment methodology will be different. In each case these opportunities will have to be identified, tested and validated to the extent possible. Exposures will almost always have to be assessed retrospectively, if for no other reason than that prospective designs would not yield results for long periods of time after the initiation of the study. Examples are presented for the case of residential radon progeny exposure and associated excess lung cancer, where the delay is of the order of decades, and for the case of the possible teratogenic effects associated with living in the proximity of sites in which hazardous wastes are stored. In the latter case the delay is much shorter.