To address agents' moral hazard over effort, incentive contracts impose risk on the agents. As performance measures become noisier, the conventional agency analysis predicts that principals will reduce the incentive weights assigned to such measures. However, prior empirical results (Prendergast 2002) frequently find the opposite, i.e., incentive weights are larger (agents bear more risk) in more uncertain environments. This paper provides new evidence on the association between the extent of uncertainty and the level of risk imposed on agents. In the context of contracts between managed care organizations and physicians, we examine the effect of task characteristics and the legal liability environment on the extent of risk that physicians bear. We derive the optimal weighting of multiple performance measures in a model of a physician's choice of revenue-generating and cost-control efforts. The model predicts that physicians who face less task uncertainty bear more cost risk in their contracts, as predicted by the conventional moral hazard model. Likewise, the model predicts that as the association between task uncertainty and legal liability uncertainty becomes stronger, physicians bear less cost risk in their contracts. Our empirical results generally support these predictions. We offer an explanation for why these results tend to be consistent with the conventional moral hazard analysis, contrary to empirical results in a number of previous studies.