Philosophical Theories of Punishment

Every system of criminal law authorises the state to do things to people that would otherwise be crimes: locking someone in a cell, seizing their property through a fine, or in some jurisdictions ending their life.

That authorisation needs a moral argument behind it, because the bare fact that a legislature passed a statute does not by itself make deliberately inflicting harm on a person just. Jurisprudence calls this the problem of justifying punishment, and it has occupied legal philosophers from Plato's dialogues on justice through to the present day.

A useful way to see why the question is hard is to notice that punishment has two features that normally make an act wrong: it is intentional and it is harmful.

If a private citizen deliberately locked a neighbour in a room for a year because the neighbour had stolen from them, that would itself be a serious crime, false imprisonment layered on top of any claim to be righting a wrong. What changes when the state does the same thing, following a conviction, is precisely what a theory of punishment has to explain.

Philosophers of punishment usually organise their answers along one axis: is the justification backward-looking, pointing at what the offender has already done, or forward-looking, pointing at some future good the punishment will produce. Retributivism is the paradigm backward-looking theory. Utilitarianism is the paradigm forward-looking theory. Almost every subsequent position in this topic, from the just deserts revival to Hart's mixed theory, is best understood as an attempt to combine the two axes rather than a wholly new third answer.

It is worth separating two questions that these theories often blur together: why does a society need a practice of punishment at all, and why is it permissible to punish this particular offender for this particular offence. The two questions can have different answers.

A society might be entitled to run a punishment system for broadly utilitarian reasons, reducing crime and reassuring citizens that wrongdoing has consequences, while still requiring that any individual only be punished because they are guilty and to a degree proportionate to their wrong. That distinction, which Hart later made explicit, is the seed of the mixed theories covered in section six.

Retributivism holds that punishment is justified because the offender deserves it, and for no other reason. The clearest statement of this position comes from Immanuel Kant, in The Metaphysics of Morals (1797).

Kant argued that a person must never be punished merely as a means to promote some other good, whether that good is the offender's own reform or the deterrence of other people, because treating a person as a mere instrument for social ends denies their status as a rational, autonomous agent. Punishment is owed to the offender because of what they did, full stop.

Kant illustrated the point with a famous, deliberately extreme example: even if a society on a remote island decided to disband and disperse forever, it would still be obliged to execute its last convicted murderer before dispersing, because justice demands it regardless of any future benefit. No one will be deterred by an execution on a vanishing island, so the example isolates the retributivist claim from any utilitarian residue. What remains is desert alone.

Modern retributivists soften Kant's absolutism but keep his structure. The central claims are that punishment must be proportionate to the seriousness of the offence and the offender's culpability, that only the guilty may be punished, and that punishing an innocent person can never be justified by any amount of social benefit it might produce.

This last claim is retributivism's strongest intuitive appeal: it explains why we think framing an innocent person is wrong even in a hypothetical where doing so would prevent a riot, because framing treats the innocent person's suffering as a mere means to calm the crowd.

Retributivism's difficulty is the mirror image of its strength. Critics ask what desert actually consists of, since it cannot be measured the way a utilitarian can in principle measure prevented harm. If desert is meant to track moral blameworthiness, retributivism needs an account of culpability that can rank offences and offenders consistently, and that account has to do real work rather than simply restating that serious crimes deserve serious punishment.

The theory is also vulnerable to the charge that it authorises cruelty for its own sake if it is not paired with an independent limit on severity, a worry taken up directly in section four.

Utilitarianism reverses retributivism's direction of justification entirely. Jeremy Bentham, in An Introduction to the Principles of Morals and Legislation (1789), argued that all law and policy should be judged by whether it maximises aggregate happiness, understood as pleasure minus pain across everyone affected.

Punishment inflicts pain on the offender, so on Bentham's own terms it is an evil, justified only when it prevents a greater evil, chiefly the pain that future crimes would cause. Bentham called this the principle of utility, and he tried to formalise the comparison of pleasures and pains through what is often called the hedonic or felicific calculus, weighing factors such as intensity, duration and certainty.

Utilitarian justifications for punishment typically appeal to several distinct mechanisms rather than one. Deterrence covers both specific deterrence, discouraging this offender from reoffending, and general deterrence, discouraging others who observe the punishment. Incapacitation removes an offender's physical opportunity to reoffend, most obviously through imprisonment.

Rehabilitation aims to change the offender so that they no longer wish to offend. All three point forward in time, and a utilitarian sentence can in principle draw on any mix of them depending on which is most cost-effective for a given offender and offence.

The forward-looking structure gives utilitarianism real practical strengths. It naturally justifies rehabilitative and preventive measures that retributivism has no obvious place for, and it fits well with evidence-based sentencing, since deterrent and incapacitative effects are at least in principle measurable and can be tested against reoffending data.

The classic objection is the scapegoat problem, developed most sharply by H.L.A. Hart and other mid-twentieth-century critics. If punishment is justified purely by its future benefits, then punishing a known innocent person could, in some imaginable case, produce more good than harm, for instance by calming a community that mistakenly believes that person guilty and would otherwise riot.

Pure utilitarianism appears to have no principled objection to punishing the innocent in such a case, which strikes most people as an intolerable result. Utilitarians have replied with rule-based versions of the theory, arguing that a general rule against punishing the innocent produces more long-run utility than case-by-case exceptions, but critics reply that this response smuggles in something like a desert-based side constraint, which is exactly what a consistent utilitarian was supposed to avoid needing.

By the mid twentieth century, rehabilitative sentencing built on utilitarian premises had fallen out of favour in several common law jurisdictions, partly because indeterminate sentences aimed at reform were seen as producing unpredictable and sometimes harsher outcomes than a fixed sentence tied to the offence would have.

This disillusionment produced the just deserts revival, most influentially associated with Andrew von Hirsch, whose 1976 report Doing Justice: The Choice of Punishments, written for the Committee for the Study of Incarceration in the United States, argued that sentencing should be anchored to the seriousness of the offence rather than to predictions about the offender's future conduct.

Von Hirsch's version of desert theory put proportionality at its centre as a limiting principle rather than only a positive justification. Proportionality requires ordinal consistency, meaning offences of similar seriousness should receive similarly severe sentences and more serious offences should receive more severe ones, and it requires cardinal restraint, meaning the overall severity of a sentencing scale should not drift upward simply because it is politically convenient.

This gave desert theory a dual role: it explains why we punish at all, and it caps how severely we may punish, closing off the retributivist worry that desert-talk alone offers no ceiling.

Just deserts theory also reframed the debate about sentencing disparity. If punishment is meant to be proportionate to the offence, then two offenders who committed comparable crimes with comparable culpability should receive comparable sentences regardless of their perceived dangerousness, social background or predicted risk of reoffending.

This principle influenced determinate sentencing reforms and sentencing guideline schemes in several jurisdictions from the late 1970s onward, including structured sentencing grids adopted in parts of the United States and sentencing guidelines development in England and Wales.

The revival did not eliminate forward-looking considerations from sentencing law, and von Hirsch himself acknowledged a role for deterrence within proportionality limits. What the just deserts movement achieved was narrower and more durable: it made proportionality, rather than either pure desert or pure prevention, the working currency that legislatures and sentencing councils actually use when setting punishment ranges, which is why proportionality reappears throughout modern sentencing statutes even in systems that never adopt the label retributivist.

A third family of theories treats punishment as fundamentally about meaning rather than only about desert or deterrence. Joel Feinberg, in his 1965 essay 'The Expressive Function of Punishment', argued that what distinguishes punishment from a mere penalty, such as a parking fine or a tax, is that punishment carries an additional element of condemnation.

A fine for illegal parking says only that the behaviour is discouraged; a criminal conviction and sentence says that the community formally condemns the act and its author. On Feinberg's account, this expressive or symbolic function is not incidental to punishment but part of what makes an act count as punishment at all.

Feinberg's expressive theory explains features of punishment that pure desert and pure deterrence struggle with on their own, such as why a formal criminal conviction matters even when the attached penalty is nominal, and why victims and communities often want public acknowledgement of wrongdoing as much as they want the offender harmed or deterred.

It also gives content to the idea of proportionality beyond arithmetic: a sentence should express condemnation proportionate to the wrong, neither trivialising serious offences nor treating minor ones as though they were grave.

Antony Duff developed this insight into a fuller communicative theory of punishment, set out across works including Trials and Punishments (1986) and Punishment, Communication, and Community (2001). Duff argues that punishment should be understood as a form of secular penance: a communicative process addressed to the offender, calling on them to recognise and repent the wrong, to understand its impact, and ultimately to be reconciled with the community whose norms they violated.

On this view punishment is not simply something imposed on a passive offender, whether to make them suffer or to make an example of them, but a dialogue the offender is invited to participate in, which is why Duff gives weight to apology, reparation and restorative processes as parts of a proper response to crime rather than as soft alternatives to real punishment.

The communicative approach has been influential in debates about restorative justice and about what a criminal trial itself, as opposed to only the sentence, is meant to accomplish.

Its critics ask what should happen when an offender refuses to accept the intended message, since if punishment's justification depends on successful moral communication, an unrepentant offender's refusal seems to leave the state's hard treatment without the justification the theory promised, a gap Duff answers by arguing that the attempt to communicate can be independently justified even when the message is not received.

H.L.A. Hart, in the essay 'Prolegomenon to the Principles of Punishment' (1959), later collected in Punishment and Responsibility (1968), offered the most influential attempt to dissolve the apparent standoff between retributivism and utilitarianism by arguing that the two camps were actually answering different questions. Hart distinguished the general justifying aim of a system of punishment, the reason society is entitled to maintain such a system at all, from questions of distribution, meaning who may be punished and how severely.

On Hart's account, the general justifying aim of punishment is broadly utilitarian: societies maintain punishment systems because doing so reduces crime and its costs, chiefly through deterrence. But the distribution of punishment to individuals is governed by retributive constraints: only the guilty may be punished, and the severity of punishment must be proportionate to the offence, regardless of whether punishing more severely, or punishing an innocent person, would produce marginally better deterrent outcomes in a given case.

This is precisely the move that answers the scapegoat problem raised against pure utilitarianism in section three: Hart's theory permits a utilitarian answer to why we punish while flatly ruling out punishing the innocent as a distributive matter.

Hart's mixed theory is not merely diplomatic. It reflects how criminal justice systems actually function. Legislatures typically justify having a criminal law and a sentencing system in broadly consequentialist terms, citing public safety, deterrence and the prevention of harm in policy documents and parliamentary debate.

Courts, when sentencing an individual, typically reason in desert-based and proportionality terms, asking what this offence and this offender's culpability warrant, and applying aggravating and mitigating factors that track blameworthiness rather than only predicted risk. The same sentencing guideline can therefore be defended at the policy level on utilitarian grounds and applied at the case level on retributive ones without contradiction, because the two levels are answering Hart's two different questions.

This is also why most jurisdictions officially list several purposes of sentencing side by side, such as punishment, deterrence, rehabilitation, protection of the public and reparation, rather than committing exclusively to one philosophical theory. Section 57 of the Sentencing Act 2020 in England and Wales, for example, lists exactly this kind of plural purpose for sentencing offenders aged eighteen and over.

In India, the Bharatiya Nyaya Sanhita, 2023, which replaced the Indian Penal Code, 1860, with effect from 1 July 2024, does not name a single philosophy of punishment either, and Indian courts have long treated deterrence, reformation and retribution as complementary aims to be balanced according to the facts of the case rather than as competing theories where one must be chosen exclusively. Hart's mixed theory gives that everyday, pluralist sentencing practice its clearest philosophical explanation.