Predictions of dangerousness or nondangerousness use the same data and can be based upon two evaluation methods: the clinical and the actuarial. Each method focuses on different pieces of data and each uses a separate method to turn those data into predictions. The central problem with clinical predictions is their inaccuracy; many more offenders are predicted to be violent than prove to be violent. On the other hand, predictions of nondangerousness are easier to make than predictions of dangerousness, simply because dangerousness is a rare event. Actuarial predictions are more systematic than clinical predictions and are usually more reliable. Still, evidence shows that actuarial predictions are not sufficiently accurate to inform predictions of dangerousness for specific persons. Analysis of future dangerousness has particular relevance to New York. Under its proposed death penalty legislation, evidence of nondangerousness would be admissible as a "mitigating factor" in the penalty phase of capital trials. Empirical evidence indicates that the vast majority of those convicted of capital murder are able to make a satisfactory and nonviolent adjustment to prison life. To overpredict dangerousness and sentence every capital offender to death cannot be justified on the basis of what is known about the probabilities of future violence by convicted murderers. 2 tables and 108 footnotes