The largest U.S. environmental health risk is cardiopulmonary mortality from ambient PM2.5. The concentration-response (C-R) for ambient PM2.5 in the U.S. is generally assumed to be linear: from any initial baseline, a given concentration reduction would yield the same improvement in health risk. Recent evidence points to the perplexing possibility that the PM2.5 C-R for cardiopulmonary mortality and some other major endpoints might be supralinear: a given concentration reduction would yield greater improvements in health risk as the initial baseline becomes cleaner. We explore the implications of supralinearity for air policy, emphasizing U.S. conditions. If C-R is supralinear, an economically efficient PM2.5 target may be substantially more stringent than under current standards. Also, if a goal of air policy is to achieve the greatest health improvement per unit of PM2.5 reduction, the optimal policy might call for greater emission reductions in already-clean locales-making "blue skies bluer"-which may be at odds with environmental equity goals. Regardless of whether the C-R is linear or supralinear, the health benefits of attaining U.S. PM2.5 levels well below the current standard would be large. For the supralinear C-R considered here, attaining the current U.S. EPA standard, 12 μg m-3, would avert only -17% (if C-R is linear: - 25%) of the total annual cardiopulmonary mortality attributable to PM2.5.