Ask why a society punishes wrongdoing and the answers arrive quickly, confidently, and in open contradiction with one another. Some say punishment exists to deter future crime; others, that it exists to give offenders what they deserve, independent of any effect on the future; still others, that its purpose is simply to remove a dangerous person from the streets, or to give someone the chance to become someone who would not offend again. Each answer sounds reasonable in isolation. The trouble begins when two equally culpable offenders, guilty of an identical crime under identical circumstances, turn out to deserve different sentences depending on which of these answers is doing the work -- a fact that ought to unsettle anyone who assumes punishment rests on a single, coherent justification rather than several incompatible ones dressed in the same legal language.

Consider deterrence taken at face value. If the purpose of a sentence is to discourage future offenders, the correct sentence for any given crime is whatever sentence most effectively reduces future occurrences of that crime -- a calculation that has little to do with how much any individual offender actually deserves. A first-time offender who poses almost no risk of reoffending might, on pure deterrence logic, warrant an unusually harsh sentence if making a visible example of him would discourage others considerably, while a repeat offender unlikely to be deterred by anything might warrant a lighter one. Almost no one, pressed with this implication directly, actually endorses it; the idea of punishing a person more harshly than his own conduct warrants, purely to influence the behaviour of strangers, strikes most people as using him as a tool for someone else's benefit rather than answering for what he himself did.

Retribution answers this discomfort directly: an offender should be punished because and only to the extent that he deserves it, entirely apart from whatever effect the punishment has on anyone else. This has real appeal -- it treats the offender as a responsible agent answerable for his own choice, not as raw material to be shaped for the benefit of future statistics. But retribution creates a puzzle of its own the moment it is pressed for specifics: desert may tell us that a worse crime deserves a worse punishment than a lesser one, but it supplies no obvious unit for translating a wrong into a term of years, a fine, or a specific penalty. Two retributivists can agree entirely that a crime deserves punishment and disagree by a considerable factor about how much, with no shared method for settling which of them is right.

Rehabilitation and incapacitation supply yet a third and fourth vocabulary, oriented not backward toward desert or forward toward general deterrence but toward this particular offender's own future conduct -- can he be changed, and if not, must he simply be kept away from opportunity. In practice, most real sentencing systems borrow from all four accounts at once, with guidelines built from a mix of retributive proportionality, deterrent severity for particularly harmful crimes, incapacitation for offenders judged likely to reoffend, and rehabilitative programming layered on top of whichever sentence the other three produce. This is often criticised as theoretical incoherence, a system unable to decide what it is actually for.

But the incoherence may be doing real work rather than merely disguising confusion. A system run on deterrence alone would periodically demand punishing the innocent-seeming for the sake of the guilty-seeming, wherever the numbers favoured it; retribution alone provides no method for calibrating severity and no language for a system that might, at real cost, choose to make an offender better rather than merely settle a score; incapacitation alone would justify indefinitely detaining anyone statistically likely to reoffend, whether or not they had yet done anything to deserve it. Each theory, taken to its own logical extreme, produces a result almost no one actually wants. What holds in practice is not a single correct theory finally identified, but an arrangement in which each rationale is permitted to operate only within limits the others impose on it -- deterrence bounded by proportionality, proportionality softened by a concern for what an offender might become, incapacitation restrained by the requirement that punishment answer for a wrong already committed rather than a crime merely predicted. The incoherence is not a flaw awaiting a philosopher's correction; it is closer to the point.