The prevailing features in the modern theory of punishment were developed by analytic philosophers half a century ago. The theory in the Anglo-American philosophical world was and still is governed by a small handful of basic conceptual distinctions, self-consciously deployed by virtually all theorists no matter what substantive views they also hold about punishment. The terminus a quo of these ideas are the influential writings of H.L.A. Hart (1959) in England and John Rawls (1955) in the United States. Though both Hart and Rawls pass muster as centrist liberals, they believed these analytic distinctions to be ideologically neutral.
Defining the concept of punishment must be kept distinct from justifying punishment. A definition of punishment is, or ought to be, value-neutral, at least to the extent of not incorporating any norms or principles that surreptitiously tend to justify whatever falls under the definition itself. To put this another way, punishment is not supposed to be justified, or even partly justified, by packing its definition in a manner that virtually guarantees that whatever counts as punishment is automatically justified. (Conversely, its definition ought not to preclude its justification.)
Justifying the practice or institution of punishment must be kept distinct from justifying any given act of punishment. For one thing, it is possible to have a practice of punishment — an authorized and legitimate threat system — ready and waiting without having any occasion to inflict its threatened punishment on anyone (because, for example, there are no crimes or no convicted and sentenced criminals). For another, allowance must be made for the possibility that the practice of punishment might be justified even though a given act of punishment — an application of the practice — is not.
Justification of any act of punishment is to be done by reference to the norms (rules, standards, principles) defining the institutional practice — such as the classic norms of Roman law,nulla poena sine leges and nulla poena sine crimen (no punishments outside the law, no punishments except for a crime). Justification of the practice itself, however, necessarily has reference to very different considerations — social purposes, values, or goals of the community in which the practice is rooted. The values and considerations appropriate to justifying acts are often assimilated to those that define judicial responsibility, whereas the values that bear on justifying the punitive institution are akin to those that govern statutory enactments by a legislature.
The practice of punishment must be justified by reference either to forward-looking or to backward-looking considerations. If the former prevail, then the theory is consequentialistand probably some version of utilitarianism, according to which the point of the practice of punishment is to increase overall net social welfare by reducing (ideally, preventing) crime. If the latter prevail, the theory is deontological; on this approach, punishment is seen either as a good in itself or as a practice required by justice, thus making a direct claim on our allegiance. A deontological justification of punishment is likely to be a retributive justification. Or, as a third alternative, the justification of the practice may be found in some hybrid combination of these two independent alternatives. Recent attempts to avoid this duality in favor of a completely different approach have yet to meet with much success (Goldman 1982, Hoekema 1986, Hampton 1984, Ten 1987, von Hirsch 1993).
Acknowledgment of these distinctions seems to be essential to anything that might be regarded as a tolerably adequate theory of punishment.
Two substantive conclusions have been reached by most philosophers based in part on these considerations. First, although it is possible to criticize the legitimacy or appropriateness of various individual punitive acts — many are no doubt excessive, brutal, and undeserved — the practice of punishment itself is clearly justified, and in particular justified by the norms of a liberal constitutional democracy. Second, this justification requires some accommodation to consequentialist as well as to deontological considerations. A strait-laced purely retributive theory of punishment is as unsatisfactory as a purely consequentialist theory with its counter-intuitive conclusions (especially as regards punishing the innocent). The practice of punishment, to put the point another way, rests on a plurality of values, not on some one value to the exclusion of all others.