The U.S. Supreme Court in Jurek v. Texas (1976) affirmed that capital juries are able to identify those capital offenders who will commit serious violence in the future. The capability of capital juries to accurately make these judgments as a means of deciding which capital offenders should receive the death penalty has been widely endorsed in both statute and case law, as well as embraced by jurors. A growing body of research on rates and correlates of prison violence, however, points to this confidence being misplaced. Prior investigations of the accuracy of these capital jury predictions, though limited in number, have found alarming error rates. The current study retrospectively reviewed the post-trial (M = 5.7 years) prison disciplinary misconduct of federal capital offenders (N = 72) for whom juries considered "future dangerousness" as an aggravating factor at sentencing. These jurors' predictive performance was no better than random guesses, with high error (false positive) rates, regardless of the severity of the anticipated violence. In light of prior studies, it is concluded that juror predictions of future violence lack sufficient reliability to play a role in death penalty determinations.