It is suggested that the admission of calculations of dangerousness into sentencing decisions will lead to its overuse and to overconfinement. While a number of strategies have been proposed for limiting predictive sentencing, it is unlikely that they would be effective if employed individually. While the use of a combination of exclusionary controls, just desert limits, and procedural requirements might present a significant barrier to the overuse of dangerousness in sentencing decisions, these alone are insufficient to avoid abuses that could occur when dangerousness is perceived in terms of moral culpability and overconfidence exists in predictive judgments. What is required to achieve limitations in the proper use of predictions of dangerousness is an active and troubled conscience. A system that maintains a healthy skepticism about predictive capacities has less need to impose restrictions on the use of dangerousness and has less need to fear overuse than one in which overconfidence prevails. 2 tables and 91 footnotes.