The validity of a risk assessment can be no better than that of the exposure assessment upon which it is based. The general paucity of relevant exposure data, combined with the limited appreciation by most risk assessors of the critical dimensions and metrics of exposure, often leads to an overreliance on exposure models of questionable validity. The problems of identifying and interpreting relevant metrics of exposure for epidemiologic studies and risk assessments are illustrated through the presentation of three case studies. The first examines the effects of ozone on respiratory mechanical function and demonstrates that the appropriate averaging time is > or = 6 hr, rather than 1 hr, as is implied by the current ambient air quality standard. The second case study examines the effects of sulfur oxides and particulate matter in ambient air on morbidity and mortality. It indicates that the effects are most closely associated with the acidity of the aerosol, providing a basis for an index of exposure more relevant than those currently used, i.e., sulfur dioxide and nonspecific gravimetric mass concentration of particulate matter. The third case study examines the effects of lead on blood pressure. It shows that blood lead in concentrations below 35 microgram/dL correlates with blood pressure in both humans and animals independently of other known casual factors for blood pressure elevation. It also examines the variable relations between levels of lead in blood and in environmental media to illustrate the potential problems which can arise from the use of biological markers, such as lead in blood, as indices of exposure.