The task of improving the ability of clinicians to predict which of their patients will be violent has come to be seen as one of establishing the relative merits of actuarial and clinical prediction. The meaning of these terms is unclear. 'Clinical' is usually defined by exclusion, that is, as something other than actuarial. The term 'actuarial' is often used to refer to the techniques of risk prediction in financial services. In the psychiatric and psychological literature relating to the assessment of dangerousness, three further meanings have emerged. That whereby actuarial refers to any mathematical means of combining information is the most widely accepted. Whichever definition is employed, the conclusion of most reviews has been that the future is actuarial. It is argued here that, while mathematical approaches have been successful in showing that risk factors for violence in the general population apply also to the mentally disordered, important questions remain unanswered. Mathematical methods address only one form of probability, that which arises from chance. A development of another form of probability, that which arises from causes, offers the prospect of improved risk assessment in psychiatry. It also offers a definition of clinical prediction that is not based on exclusion.