There is convincing experimental evidence that Expected Utility fails, but when does it fail, how severely, and for what fraction of subjects? We explore these questions using a novel measure we call the uncertainty equivalent. We find Expected Utility performs well away from certainty, but...