The optimum level of sulfur pollution control for a coal fired power plant is the point where the sum of societal costs, due to pollution, and control costs is minimized. This basic microeconomic concept has been of limited practical value due to considerable uncertainty in estimating both costs. A probabilistic approach is used to characterize these uncertainties quantitatively for a hypothetical 1000 Mwe plant located near Pittsburgh, Pennsylvania. Only mortality effects within a distance of 80 km of the plant have been included. The results allow explicit consideration of attitude toward risk and appropriate level of investment to prevent deaths. Limitations of the findings are discussed. Implications are described for policy based on alternative sets of values and assumptions