Long-term predictions of dangerousness are used throughout the criminal law in investigation, pretrial detention, bail, sentencing, parole, and early release decisions. Use of such predictions is increasing. Reliance on such predictions raises questions of definition, the ethical limits, and proof of dangerousness. Lawyers have relegated predictions of dangerousness to the psychiatric professions, leaving the moral and evidentiary issues untouched. The appropriate application of these predictions is not a technical question of how well a prediction can be made, nor is it a question of the burden of proof. The use of predictions of dangerousness requires a political judgment balancing the risk and harm to society with the intrusion on the liberty of preventatively detained persons. Not all types of predictions are equally satisfactory. Actuarial predictions are preferable to predictions that rely on intuitive judgments by psychiatric professionals. The use of predictions to alter individual dispositions should be allowed only to the extent that such dispositions would be justified as deserved independent of those predictions. Within the range of punishment or control not undeserved, relative predictions of dangerousness may properly influence dispositional decisions. These principles strike a balance between individual autonomy and State authority. In sentencing, the controlling principle is that base expectancy rate of violence for the criminal predicted as dangerous must be higher than the base expectancy for another criminal with a closely similar record but not predicted to be unusually dangerous, before the greater dangerousness of the former may be relied upon to intensify or extend punishment. Forty-two footnotes and 29 references are provided. (Author abstract modified)