Prisons were supposedly created for the purpose of the tertiary prevention of violence (i.e., reducing the frequency and severity of future violence on the part of people who have already become violent). However, there is much evidence that this method of attempting to prevent violence is often, though not always, either ineffectual or counterproductive, in which case it is either a waste of money or actually exacerbates the problem it was ostensibly intended to solve. This article reviews evidence concerning those questions including an analysis of the effect of punishment (one of the main purposes of prisons) on violent behavior. Punishment--the infliction of pain--will be distinguished from restraint (incapacitation, separation from the community). Successful examples of violence prevention in unconventional prison programs, emphasizing therapy and education rather than punishment, and restorative rather than retributive justice, will be summarized, together with evidence that these programs reduce re-incarceration rates so substantially that they actually save the taxpayers more money than they cost, in addition to enhancing the safety of the general public. The position is taken that traditional prisons provoke more violence than they prevent and are so fundamentally flawed that they cannot be reformed; we argue that they should instead be abolished and replaced by "anti-prisons," that is, locked, secure residential colleges, therapeutic communities, and centers for human development. Prisons will come to be seen as a well-meaning experiment that failed, rather like the use of leeches in medicine.