Many social policies rely on assessments of violence potential in their implementation. Decisions concerning such civil matters as involuntary institutionalization and level of institutional custody, and criminal matters such as type or length of sentence and the granting of bail often rely on estimates of the potential that an individual will be violent in the future. Unfortunately, the prediction of violence is a difficult task and predictions are often found to be inaccurate. Research on the prediction of juvenile violence has identified a number of "postdictors" of violent behavior. It is questionable whether those criteria can be applied accurately to the prediction of violent behavior. Practicing psychologists and psychiatrists (N=64) were asked about the procedures they followed when making assessments of violence potential. The majority of respondents (78%) reported an evaluation of past assaultive behavior to be principle variable used in making assessments of future assaultive potential. This finding suggests that clinicians are using the most current knowledge about predictions. Perhaps conclusions about an individual's potential for violence should be re-framed from an absolute "yes" or "no" statement to an estimate of the probability that the particular individual will become violent in a given set of circumstances within a specified period of time. Although the imperfections in the ability to predict violence may result in injustice, some degree of injustice may not invalidate the policies which caused it. (NB)