Wood, David --- "A Retributive Argument Against Punishment" [2011] UMelbLRS 2

Last Updated: 6 April 2011

This article was first published in the Criminal Law and Philosophy, Vol. 5, No. 1, 2011

A RETRIBUTIVE ARGUMENT AGAINST PUNISHMENT

Authors
Greg Roebuck,
University of Melbourne,
Victoria 3010, Australia

David Wood (corresponding author),
Faculty of Law,
University of Melbourne,
Victoria 3010, Australia
e-mail: [e-mail]
telephone: + 61 3 8344 6192
fax: + 61 3 9347 2392

Keywords
punishment, retributivism, proportionality, ordinal, cardinal, commensurability.

Abstract

This paper proposes a retributive argument against punishment, where punishment is understood as going beyond condemnation or censure, and requiring hard treatment. The argument sets out to show that punishment cannot be justified. The argument does not target any particular attempts to justify punishment, retributive or otherwise. Clearly, however, if it succeeds, all such attempts fail. No argument for punishment is immune from the argument against punishment proposed here. The argument does not purport to be an argument only against retributive justifications of punishment, and so leave open the possibility of a sound non-retributive justification of punishment.

Punishment cannot be justified, the paper argues, because it cannot be demonstrated that any punishment, no matter how minimal, is not a disproportionate retributive response to criminal wrongdoing. If we are to hold onto proportionality – that is, proportionality as setting a limit to morally permissible punishment – then punishment is morally impermissible. The argument is a retributive argument against punishment insofar as a just retributive response to wrongdoing must be proportionate to the wrongdoing. The argument, that is, is concerned with proportionality as a retributive requirement.

The argument against punishment is set out on the basis of a familiar version of the ‘anchoring problem’, according to which it is the problem of determining the most severe punishment to anchor or ground the punishment scale. To meet the possible criticism that we have chosen a version of the anchoring problem particularly favourable to our argument, various alternative statements of the anchoring problem are considered. Considering such statements also provides a more rounded view of the anchoring problem.

One such alternative holds that the punishment scale must be anchored not just in the most severe punishment, but in the least severe punishment as well. Other alternatives hold that it is necessary and sufficient to anchor the punishment scale in any two punishments, neither of which needs to be the most or least severe punishment. A further suggestion is that one anchoring point anywhere along the punishment scale is sufficient, because it is possible to ‘project’ from such a point, so as to determine the correlative punishments for all other crimes, and so derive a complete punishment scale. Finally, the suggestion is considered that one can approach the issue of a punishment scale ‘holistically’, denying any distinction between anchoring and derived (or ‘projected’) punishments.

A RETRIBUTIVE ARGUMENT AGAINST PUNISHMENT

1. Introduction

This paper proposes a retributive argument against punishment, where punishment is understood as going beyond condemnation or censure, and requiring hard treatment.[1] A retributive argument against punishment might be thought a strange thing, since retributivism is usually concerned with justifying punishment. The argument sets out to show that punishment cannot be justified. The argument does not target any particular argument or arguments that try to justify punishment, retributive or otherwise. Clearly, however, if it is successful, all such arguments must fail. Punishment cannot both be justified and not justified. No argument for punishment is immune from the argument against punishment proposed here. The argument does not purport to be an argument only against retributive justifications of punishment, and so leave open the possibility of a sound non-retributive justification of punishment.

Punishment cannot be justified, the paper argues, because it cannot be demonstrated that any punishment, no matter how minimal, is not a disproportionate retributive response to criminal wrongdoing. If we are to hold onto proportionality – that is, proportionality as setting a limit to morally permissible punishment – then punishment is morally impermissible. The argument is a retributive argument against punishment insofar as a just retributive response to wrongdoing must be proportionate to (or, at least, in the case of ‘limiting retributivism’,[2] not more than proportionate than) the wrongdoing. The argument, that is, is concerned with proportionality as a retributive requirement.

The paper has two substantive parts. Part Two sets out the paper’s argument against punishment. It does so on the basis of a familiar version of what is called the ‘anchoring problem’. According to this version, it is the problem of determining the most severe punishment[3] to anchor or ground the punishment scale.

Perhaps, however, this version of the anchoring problem is particularly favourable to the argument advanced here. Part Three, then, is concerned with the argument on the basis of alternative statements of the anchoring problem. Considering such alternative statements also provides different perspectives on, and hence a more rounded view of, this problem.

One alternative holds that the punishment scale must be anchored not just in the most severe punishment, but in the least severe punishment as well. Other alternatives hold that it is necessary and sufficient to anchor the punishment scale in any two punishments, neither of which needs to be the most or least severe punishment. A further suggestion is that one anchoring point anywhere along the punishment scale is sufficient, because it is possible to ‘project’ from such a point, so as to determine the correlative punishments for all other crimes, and so derive a complete punishment scale. The suggestion is also considered, that one can approach the issue of a punishment scale ‘holistically’, denying any distinction between anchoring and derived (or ‘projected’) punishments.

2. The Retributive Argument against Punishment

The argument proceeds as follows: all retributivists agree that it is morally impermissible to punish an offender more than he or she deserves. That is, they agree that it is morally impermissible to impose a punishment on an offender that is more severe than the punishment that is proportionate to the seriousness of the offender’s crime. They thus endorse what could be termed the ‘limiting proportionality thesis’:

LPT: It is morally permissible to impose a punishment on an offender only if the punishment is not more severe than the punishment that is proportionate to the seriousness of the offender’s crime.

It is also commonly accepted that it is only morally permissible to deliberately harm a person if it can be demonstrated that the infliction of this harm is morally justified. In other words, there is a presumption that deliberately inflicting harm is morally impermissible, and those seeking to inflict such harm must overcome this presumption. Clearly, punishment involves the deliberate infliction of harm. Combining this presumption with LPT then yields what could be called the ‘demonstrable limiting proportionality thesis’:

DLPT: It is morally permissible to impose a punishment on an offender only if it is possible to demonstrate that the punishment is not more severe than the punishment that is proportionate to the seriousness of the offender’s crime.
More simply and informally, punishment is only morally permissible if it can be demonstrated that it is not disproportionately severe in relation to the seriousness of the offender’s crime.

However, the argument continues, it is never possible to demonstrate that a punishment proposed to be imposed in response to a given crime does not breach the ceiling determined by the seriousness of the offender’s conduct that DLPT sets on permissible punishment. In short, it cannot be demonstrated that a proposed punishment is not disproportionately severe in relation to the seriousness of the crime. It cannot be demonstrated that the ceiling placed on morally permissible conduct by DLPT is not simply the floor of no punishment at all, and hence that it is not breached, no matter how ‘soft’ or insignificant the punishment.

To establish that it is never possible to demonstrate that a proposed punishment does not breach the ceiling that DLPT sets on permissible punishment – that the proposed punishment is not disproportionately severe in relation to the seriousness of the crime – it is necessary to set out the ‘anchoring problem’.[4] This in turn requires introducing a distinction generally drawn between two types of proportionality, ordinal and cardinal. (The distinction is also referred to as that between proportionality simpliciter and commensurability.[5]) This distinction, in any case, is central to any discussion of proportionate punishment.

Ordinal proportionality relates to ‘how severely crimes should be punished relative to each other’ (von Hirsch and Ashworth 2005: 138; emphasis added). It is marked by three requirements: parity (equally serious crimes are to be punished with equal severity[6]); rank-ordering (more serious crimes should receive more severe penalties[7]); and (what is called) ‘spacing’. Parity and rank-ordering are reasonably straightforward, but spacing requires some explanation. As von Hirsch and Ashworth put it:

‘[s]uppose crimes X, Y, and Z are of ascending order of seriousness; but Y is considerably more serious than X but only slightly less so than Z. Then, to reflect the conduct’s gravity, there should be a larger space[8] between penalties for X and Y than for Y and Z’ (von Hirsch and Ashworth 2005: 140; von Hirsch 1993: 18).

One can say more simply that spacing is intended to meet the need for the relative difference in the seriousness of offences to be reflected in the relative difference in the severity of punishments: ‘penalties [should] be...spaced to reflect the relative...spacing of the crim[es]’ (von Hirsch 1993: 39).[9]

Parity and rank-ordering alone are obviously not sufficient for proportionality as ordinarily or pre-theoretically understood (that is, prior to any suggested distinction between ordinal and cardinal proportionality), and hence for what normally would be considered a plausible punishment scale. (Of course, the argument here is to the conclusion that there is no justified punishment scale, and hence that punishment itself cannot be justified.) Parity is merely the corollary of rank-ordering, requiring that equally serious crimes be ranked equally. And rank-ordering is satisfied simply by punishing more serious crimes more seriously. It does not matter, insofar as rank-ordering is concerned, how much more seriously the relevant crimes are punished – ranking is merely one above another. Rank-ordering need in no way reflect the fact that although crime C is more serious than B, and B is more serious than A, C is much more serious than B, than B is than A. A plausible punishment scale must obviously recognise that the comparative differences between the crimes it puts in rank order may differ, perhaps considerably.

What is required, then, is recognition of the spacing of crimes, of the different proportionate space that may exist between crimes, by providing similar or corresponding spacing between the correlative punishments. Spacing is required in a punishment scale, in addition to parity and rank-ordering, so that relative differences in the seriousness of crimes can be reflected in the relative severity of the punishments allocated to them.[10]

To illustrate, suppose a jurisdiction punishes murder (its most serious crime), with twenty years’ imprisonment (whether as a mandatory or maximum sentence), rape (its next most serious crime) with nineteen years’, and littering (its least serious crime) with eighteen years’.[11] The relevant objection to the punishment for littering is not that, in non-comparative terms, irrespective of the punishment for murder and rape, it is extremely severe. (In fact, the punishment scale as a whole may be regarded as extremely severe, although obviously not by proponents of capital punishment or even life imprisonment.) The relevant objection is rather that littering is far less serious than murder or rape – in non-comparative terms, it is not serious at all – and its punishment should likewise be far less severe.

However, observing spacing in a punishment scale is still not sufficient for proportionality in its ordinary, pre-theoretical sense. There is a further problem which the spacing of punishments – or more precisely, adding spacing to parity and rank-ordering – does not solve. Through these three requirements, ordinal proportionality provides at most what could be called ‘relative’ or ‘internal’ proportionality. (Indeed, ordinal proportionality is also referred to by von Hirsch as ‘relative proportionality’, being concerned only with the ‘internal structure’ of a punishment scale (von Hirsch 1992: 76; cf. von Hirsch and Ashworth 2005: 138).) The alternative terminology indicated above[12] is perhaps more appropriate here – these three requirements provide only proportionality as opposed to commensurability. A punishment scale can still, intuitively at least, require disproportionate punishments, and hence be implausible, despite meeting these three requirements.

A punishment scale that meets these requirements can be overall very severe with, for instance, as in the above example, twenty years’ imprisonment for its most serious crime (murder) and eighteen years’ for its least serious crime (littering), punishments for all other offences falling – being spaced – somewhere in between.[13] Punishments, that is, could be ‘bunched’ (as it could be put) at the top of a hypothetical scale that extends from the most severe punishment conceivable to the least severe.

At the other extreme, punishments can generally be very lenient with, for instance, a five dollar fine for murder and a one dollar fine for littering (again, punishments for all other offences being spaced somewhere in between). That is, punishments could be ‘bunched’ at the bottom of the hypothetical scale. Alternatively, punishments could be ‘bunched’ somewhere between these two extremes, with say two years’ imprisonment for murder and one for littering.

Ordinal proportionality, then, with its three requirements of parity, rank-ordering and spacing, is not sufficient for proportionality in its ordinary, pre-theoretical sense. The proportionality provided by ordinal proportionality, including spacing, is purely internal to a punishment scale. The question is how this scale fits or coheres with a real-life, crime-seriousness scale – how it ‘meets the world’, rather than, as von Hirsch puts it, the two scales ‘“float[ing]” independently of each other’ (1985: 92).

The three requirements of ordinal proportionality are not sufficient to ensure an adequate range of punishments to match the range of offences. They do not exclude ‘bunching’, whether at the top or the bottom of the scale, or somewhere in between. As the three examples above demonstrate, ‘bunching’ can still occur even though spacing is satisfied. Neither do the three requirements of ordinal proportionality exclude what could be called the ‘stretching’ of a punishment scale – in the most extreme case, taking the most severe punishment conceivable and the most lenient punishment conceivable as the scale’s anchoring points.[14]

It is necessary, then, to turn to cardinal proportionality (also referred to as ‘nonrelative proportionality (von Hirsch 1992: 77; cf. von Hirsch and Ashworth 2005: 138)), which is concerned with ‘fixing actual (and not just comparative) severity levels’ of punishments (von Hirsch and Ashworth 2005: 141). The problems of ‘bunching’ and ‘stretching’ can only be solved by having a punishment scale with appropriate breadth or range – too little produces ‘bunching’, too much produces ‘stretching’. To provide such a scale, suitable anchoring is required, in addition to the three requirements of ordinal proportionality (of parity, rank-ordering and spacing). Cardinal proportionality is claimed to provide such anchoring. However, the problem then is that the selection of the requisite anchor or anchors (according to whether anchoring it taken to require one or more anchoring points) cannot be justified.

The range of punishments a jurisdiction makes use of (that is, its punishment scale), and so more specifically, the punishment that it regards as proportionate to any particular crime, requires determining the jurisdiction’s most severe punishment (that is, according to the above statement of the anchoring problem; alternative statements are considered in Part Three), the punishment to be allocated to what it regards as its most serious crime. But what should this punishment be? Suppose murder is regarded as the most serious crime. Should the punishment for murder be execution, life imprisonment, twenty years’ imprisonment, ten years’ imprisonment, or a fine? How is one to select from such a range, one punishment as the most severe in an actual sentencing system? Some punishments, for instance, a fine, may appear intuitively far too lenient for murder. Others, such as execution, especially where carried out in particularly painful or degrading ways, may seem intuitively far too harsh. Obviously, intuitions about the appropriate punishment for a given crime can vary greatly (and especially in the case of murder), not just within but across societies (Braithwaite and Pettit, 1990: 179). No doubt, they are influenced by existing, especially long-standing, punishment practices in the jurisdiction in question. However, it is not clear why intuitions or commonly-held beliefs should carry any weight.[15] A punishment scale is required precisely so that we can judge the acceptability of our ordinary intuitions, whether they are too harsh or too lenient or quite reasonable.

There appears, however, to be no, non-arbitrary way of getting beyond such intuitions, to select any one punishment as the most severe in the punishment scale of a particular jurisdiction. Retributivism cannot provide this punishment (Schafer-Landau 1996: 308; 2000: 191) (and does not even claim to do so).

Certainly, there is the hypothetical scale (already referred to) which simply takes the most severe punishment conceivable as its most severe punishment.[16] But an actual scale is required, and this requires an anchoring punishment – without that, one only has the hypothetical scale.

Without an actual scale, it cannot be demonstrated that the imposition of any punishment in response to a crime, no matter how minimal, is not disproportionate to it, and therefore morally impermissible – without such a scale, the retributive proponent of punishment has no answer to the retributive opponent of punishment, who denies that it can be established that a proportionate retributive response need ever be punitive, that solely formal or purely symbolic condemnation or censure does not provide a sufficient such response. Thus, if DLPT holds, it is never morally permissible to impose punishment on an offender.

The most that may be justified is a system of solely formal or purely symbolic censure. Only such a system, according to the argument advanced here, can possibly be justified. If censure is itself a form a hard treatment,[17] or if it requires hard treatment for its effective conveyance or communication (Duff 1999: 51; cf. Wood 2010: 476), then not even censure can be justified.

Note in this context von Hirsch’s reference to ‘a system of formal censure alone (or of censure accompanied by token deprivations)’ or ‘a system of censure with token impositions’ (1993: 38; cf. 1990: 285). Von Hirsch admits (at least, he admitted in Censure and Sanctions) that he now agrees with Nils Jareborg, where previously he had not, that a fine could convey sufficient censure to be the appropriate punishment for murder (1993: 38; cf. 1990: 285-6, 284n). Certainly this would require, on von Hirsch’s view, suitable changes in the conventions whereby censure is conveyed through punitive hard treatment – whether, one could add, these changes occur ‘naturally’ or are brought about in accordance with his ‘decremental strategy’ for deliberately and progressively reducing punishment rates (1993: 40, 45-6; Van Ness 510). The problem, von Hirsch admits, is that ‘more substantial penalty levels may be needed in our tougher, real world’ (1993: 38) for preventive reasons.

However, if this logic were to be taken further, and the requisite changes in conventions occur or are brought about, then not even a fine, but formal censure alone (perhaps ‘accompanied by token deprivations’) would be a sufficient response to murder. Indeed, if formal censure constitutes punitive hard treatment, since it is meant to be experienced as unwelcome,[18] then not even formal censure can be justified. Alternatively, if formal censure only constitutes punitive hard treatment[19] when it is actually experienced as unwelcome – and not, say, laughed at or simply ignored – it is punitive hard treatment and hence not justified when it is so experienced. One way of defending von Hirsh’s idea of a system of solely formal censure (1993: 38) may be to hold that, even if such a system technically counts as a system of punishment on the grounds that such censure is a type of hard treatment, the level of hard treatment it involves is so minor or insignificant that it can be substantively dismissed as punishment on ‘de minimus’ grounds. The term ‘punishment’ would then be reserved for cases where the hard treatment element is ‘over and above the hard treatment already entailed by the most lenient means of communicating censure’ (Tasioulas: 296; and see n. 1).

Certainly, given people’s motivational structures, that they are not ‘angels’ as von Hirsch puts it (von Hirsch 1993: 13; von Hirsch and Ashworth 23), there is the danger of an increase in murder and other crimes. And, von Hirsch concedes, ‘it is preventive concerns that ultimately constrain a deflation of penalty levels, rather than considerations of cardinal desert’ (von Hirsch 1993: 38). It is not, then, just conventions for expressing censure that need to change – people need to change. Any ‘decremental strategy’ would have to be halted if crime rates started to rise (von Hirsch 1993: 40; Van Ness 510).

A further problem is whether a system of formal censure alone could maintain spacing. According to the present suggestion, no offences, irrespective of how serious, would warrant a punitive response, but only formal censure. Presumably, offences would still be censured according to their relative seriousness – more serious offences would receive more severe censure – even though there is no question of such censure being conveyed by different degrees of punitive hard treatment. The question concerning spacing is whether there could nevertheless still be different degrees of formal censure, and furthermore, different degrees sufficient to match the different degrees of seriousness of offences.


3. Alternative Statements of the Anchoring Problem

For reasons of clarity of presentation, Part Two set out the paper’s retributive argument against punishment on the basis of one, familiar statement of the anchoring problem, according to which a punishment scale must be anchored at its most severe punishment. There are, however, alternative statements, and the argument must also be considered on the basis that some such statement is adopted. Otherwise, we are open to the allegation that we have adopted a statement of the anchoring problem particularly favourable to our case. It needs to be shown that our retributive argument against punishment is sound irrespective of how it is thought a punishment scale should be anchored. Considering alternative statements of the anchoring problem also helps provide a more rounded understanding of it.

One alternative statement holds that the punishment scale must be anchored not just in the most severe punishment, but in the least severe punishment as well. For instance, von Hirsch uses ‘anchor’ in relation to both the most and least severe penalty, asking ‘[w]hat should be the scale’s severest and most lenient penalties?’[20] He points out that Kleinig does so too:

‘[t]he scale, [Kleinig] suggests, can be anchored at its upper end by the severest (say, 25 years’ imprisonment [footnote deleted]) and at its lower end by the mildest penalty feasible (say, a very small fine)’.[21]

Other alternatives hold that it is necessary and sufficient to anchor the punishment scale in any two punishments, neither of which needs be the most or least severe punishment. For instance, Kleinig observes that ‘at least two points of contact [i.e. anchoring points] are needed if all arbitrariness is to be removed’.[22]

Consider to start with the claim that a punishment scale must be anchored at both its most severe and most lenient punishments. The argument could be set out as follows, drawing upon (but not following precisely), the example Kleinig himself provides (1973: 118-9). Suppose that there are four equally spaced crimes A, B, C, D, and that they warrant respectively the likewise equally spaced punishments a, b, c, d. That is, spacing is observed if the ratio A:B:C:D equals a:b:c:d. But this is consistent with a, b, c and d being (to mention just two possibilities) four, three, two and one year’s imprisonment respectively, or eight, six, four and two years’ imprisonment respectively.[23] The problem is that these scales are not anchored. However, all choice is not eliminated by providing only one anchoring point. Suppose a punishment scale is anchored at the most severe punishment, in this case, eight years’ imprisonment. This eliminates the first of the two possibilities above (namely, of four, three, two and one year’s imprisonment respectively). But we could still have as the correlative punishments a, b, c and d respectively, say, eight, six, four and two years’ imprisonment respectively, or eight, seven, six and five years’ imprisonment respectively, or eight, seven and a half, seven, and six and a half years’ imprisonment respectively, and so on. However, the argument goes, if the scale is also anchored at the bottom, say at two years’ imprisonment, then the choice is restricted to the first-mentioned range, of eight, six, four and two years’ imprisonment respectively.[24] This example shows (or purports to show), then, in Kleinig’s words, that ‘all arbitrariness’ is removed if two anchoring points are provided (!973: 118[25]).

If, however, a punishment scale must be anchored at its most lenient as well as its most severe punishment, the question arises of where the former anchoring point is to be set. (Part Two, of course, was concerned with this question in relation to the most severe punishment.) According to von Hirsch, Kleinig suggests that the second anchoring point can be ‘the mildest penalty feasible’.[26] (Indeed, one may add, punishments could conceivably ‘fade away’ to a vanishing point, as punishment becomes nothing more than mere censure.[27])

On the other hand, it may be asked why the second anchoring point should be set so low. Surely, for conduct to be justifiably criminalised requires that it satisfies some threshold level of wrongfulness, which makes it properly the subject of criminal law, as opposed only to tort law, or what may broadly be referred to as regulatory law. However, this just raises the same question as with the most severe punishment, of what substantively the least severe punishment should be.

In any case, what is pertinent here is that this alternative statement of the anchoring problem is of no assistance to the critic of the retributive argument against punishment advanced in Part Two. On the one hand, even if the most lenient possible punishment can be accepted as the ‘mildest penalty feasible’, one still has the problem of justifying the most severe anchoring punishment. The question remains of what this punishment should be, of how it is to be determined. On the other hand, if the most lenient punishment cannot be accepted to be the ‘mildest penalty feasible’, but has to be set at some higher level, then there are two problems and not just one. If there is some higher threshold level of wrongfulness of properly criminalised conduct that distinguishes such wrongs from wrongs warranting the attention merely of tort or regulatory law, then in addition to the problem of determining the most severe punishment, there is the further problem of determining the least severe punishment.

Consider now the more general objection that it is necessary and sufficient to anchor the punishment scale at any two points, neither of which need be the most severe or most lenient punishment. The claim appears to be that if one has two anchoring punishments, neither of which are the most severe or lenient punishment, one can ‘project’ from them through what could be called the ‘spacing formula’ they establish, to determine the punishments for all other crimes. Suppose again that there are four crimes, A, B, C and D, and four correlative punishments a, b, c and d. Suppose also that B is anchored at six years’ imprisonment and C at four years’. Suppose also that there is the same space difference between a and b and between c and d as there is between b and c. (This requires supposing in turn that there is the same space difference in crime-seriousness between A and B and between C and D as there is between B and C.) One can then, or so it must be claimed, ‘project’ from the two anchoring punishments, that the punishment for A is eight years’ and for D is two years’.

The claim, then, is that the anchoring punishments need be neither the most nor least severe, if one can ‘project’ from them through the relevant ‘spacing formula’ to determine the punishments for other crimes. But the problem remains of how to select these anchoring punishments. The problem is in principle the same, as where it is held that the anchoring punishment is the most severe (or most lenient) punishment. The argument in Part Two applies (with suitable changes in wording) equally to such anchoring points. Again, our opponent has not made life easier for himself by holding that two anchoring points are required and hence in need of justification, and not just one. If trying to justify one anchoring point is an intractable problem, all the more so is trying to justify two.

There is, furthermore, an additional problem with the space between the supposed anchoring points. This problem concerns the scale by which the ‘spacing formula’ operates, with deciding how spaces in crime-seriousness ‘translate’ into spaces in punishment-severity. Why suppose that the above projections operate according to an arithmetic scale – that spaces in the crime-seriousness scale project arithmetically onto spaces in the punishment-severity scale – as opposed to, say, a geometric or logarithmic scale? Why, for instance, should the spacing formula not require that the rate at which punishments become more severe increases as crimes become progressively more serious. (This matter can in principle be resolved by providing three, or possibly more, anchoring points, to settle the issue of whether an arithmetic, geometric or some other scale is being used.)

Furthermore, it seems to be assumed that, with all these alternatives, it is possible to take a purely objective view of the severity of the punishments that a punishment scale uses. But why suppose, for instance, that twenty years’ imprisonment is twice as severe as ten years’ imprisonment? This may be the case as a matter of abstract arithmetic. However, there is no reason to suppose that life generally, or sentencing in particular, tracks the requirements of abstract arithmetic. On the contrary, the relative severity of prison sentences, it may be claimed, depends on how they are experienced.[28] Perhaps the second ten years of a twenty years’ sentence is less severe than the first ten years, as a prisoner adapts or becomes accustomed to prison life. If this is the case, one should incorporate some principle of diminishing marginal severity, so that, say, twenty years’ imprisonment is considered only one and three quarters as severe as ten years’ imprisonment, not twice as severe. On the other hand, perhaps the second ten years of a twenty years’ sentence is more severe, as impatience and frustration increases, so that, on the contrary, some principle of increasing marginal severity is required, which reflects the reality that twenty years’ imprisonment is, say, two and a quarter times more severe than ten years’ imprisonment.[29]

There is a host of possibly relevant factors. The actual scale may well vary from prisoner to prisoner, some factors pointing to some rough principle of diminishing marginal severity, others to some rough principle of increasing marginal severity. There is obviously the further question of how these (and other) competing factors are to be weighed against each other. To what extent, if any, for instance, do they cancel each other out? There is, furthermore, the underlying question of whether the operation of such factors can be reduced to mathematical formulae, as opposed to their behaving quite erratically.

It may even be suggested that one anchoring point anywhere along the punishment scale – and not necessarily at the top or bottom – is sufficient. This may be suggested on the grounds that it is possible to ‘project’ from one such point through judgments of comparative spacing (such as ‘crime B is three times more serious than anchoring crime A’, ‘crime C is five times more serious than A’, ‘crime D is half as serious as A’), so as to determine the correlative punishments for all other crimes, and so derive a complete punishment scale. Suppose, again, that murder, the most serious crime, is anchored at twenty years’ imprisonment. If it is held that rape is three quarters as serious as murder (not that it is claimed that a punishment scale in practice is formulated in such precise mathematical terms), one can deduce that the punishment for rape is fifteen years’ – that is, ignoring factors such as those just raised, which caste doubt on the use of mathematical formulae. On the other hand, if one knows that rape is anchored at fifteen years’ imprisonment, and one knows on the basis of the spacing formula that, say, murder is one and a third times more serious than rape, then one can deduce that the punishment for murder should be twenty years’. (Again, however, it can be asked why an arithmetic scale should be assumed.)

However, the same problem remains with this statement of the anchoring problem as with that assumed in Part Two, as well as the statements considered earlier in this part. Wherever it is held a punishment scale is to be anchored – whether at the top, bottom, somewhere in between or at two or more points – it is not possible to non-arbitrarily determine any anchoring point.

There is little point, then, in trying to consider any further ways to try to anchor a punishment scale. However, if only for the sake of completeness, one final proposal should be mentioned, even if it seems scarcely plausible. This proposal denies that the anchoring of a punishment scale is required. It holds that any issue of anchoring and anchoring points can be ignored altogether, because one can determine ‘holistically’ – in one fell swoop – a punishment scale as a whole. There is no need, then, for two separate stages, of first determining the anchoring punishment or punishments, and then, through the three ordinal proportionality requirements of parity, rank-ordering and spacing (and in particular, the spacing formula), deriving the punishment for all other crimes from this punishment or these punishments.

However, it is totally unclear how this would work. How could there be any determination of a punishment scale as a whole? This requires not just determining the least and most severe punishments, and correlating them with the least and most severe crimes respectively (or, at least, with what are held to be the least and most severe crimes in the jurisdiction in question). It also requires determining every intermediate punishment, and correlating them with their respective crimes.

This is in substance the suggestion that every sentence be individually determined.[30] As such, it stands in stark contrast with various proposals for only the one anchoring point, and the derivation of a punishment scale from this point through the requirements of ordinal cardinality. One only has a punishment scale as a whole, then, as the sum or end-result of all the individual correlations of crimes and punishments. Rather than ‘holism’, this is rampant particularism! Indeed, perhaps all one has is the individual correlations of crimes and punishments, with no ranking of crimes for relative seriousness or punishments for relative severity, let alone an aligning or correlating of these two rankings to create a punishment scale. In short, perhaps one simply has no punishment scale at all.

There is the further obvious issue of how the individual determinations (the correlations of crimes and punishments) are to be made. Of course, the burden of the argument in Part Two is that there is no satisfactory way of making them. To suggest this can be done on the basis of intuitions just begs the question of whether our existing intuitions are sound – of whether the intuitions we have are the intuitions we should have.

Perhaps the ‘lex talionis’ (Fish 2008: Waldron 1992: Kleinig 1973: 120-3: Shafer-Landau 1996: 299-301; 2000: 193-198) could be conceived as providing correlative punishments for every crime, so that there is no need for some (one or more) punishments to act as anchoring points, from which the punishments for all other crimes can be derived. But it is far too much to expect that this doctrine can provide the degree of detail necessary to determine the fitting punishment for every crime. At most, ‘lex’ can be expected to provide a number of suitable anchoring points (to use current terminology), from which the punishments for all other crimes can be derived. Any advocate of ‘lex’ must, then, adopt the two stage approach just mentioned.


4. Conclusion

This paper has presented a retributive argument against punishment. The argument sets out to show that punishment cannot be justified, because it cannot be demonstrated that any punishment, no matter how minimal, is not a disproportionate retributive response to criminal wrongdoing. Part Two set out this argument on the basis of a familiar version of the anchoring problem, according to which, it is the problem of determining the most severe punishment to anchor or ground the punishment scale. Part Three considered alternative statements of the anchoring problem. Beyond brief references to intuition and the ‘lex talionis’, the paper did not consider possible attempts to respond to this argument.


References
Ashworth, A. (2010). Sentencing and Criminal Justice (5th ed). Cambridge. UK: Cambridge University Press.

Benn, S.I., and Peters, R.S. (1959). Social Principles and the Democratic State. London, UK: George Allen and Unwin.

Braithwaite, J., and Pettit, P. (1990). Not Just Deserts. Oxford. UK: Oxford University Press.

Duff, A. (1999). Punishment, Communication and Community. In Matravers, M., ed., Punishment and Political Theory 48-68. Oxford, UK: Hart.

Duff, A. (2001). Punishment, Communication and Community, Punishment. Oxford, UK: Oxford University Press.

Feinberg, J. (1965). An Expressive Theory of Punishment. Monist 49, 397-423.

Fish, J. M. (2008) An Eye for an Eye: Proportionality as a Moral Principle of Punishment. Oxford Journal of Legal Studies 28, 57-71.

Galligan, D.J. (1981). The Return to Retribution in Penal Theory. In Tapper, C.F.H., ed., Crime, Proof and Punishment: Essays in Memory of Rupert Cross 144-171. London, UK: Butterworths.

Gardner, G. (1958). The Purposes of Punishment. Modern Law Review 21, 117-129.

Kaufman, W. (1973). Without Guilt and Justice: From Decidophobia to Autonomy. New York, USA: Peter H. Wyden, Inc.

Kleinig, J. (1973). Punishment and Desert. The Hague, Netherlands: Nijhoff.

Kolber, A. (2009). The Subjective Experience of Punishment. Columbia Law Review 109, 182-236.

Raz, J. (1982). The Claims of Reflective Equilibrium. Inquiry 25, 307-330.

Sadurski, W. (1985). Giving Desert Its Due. Dordrecht, Netherlands: Reidel.

Shafer-Landau, R. (1996). The Failure of Retributivism. Philosophical Studies 82, 289-316.

Shafer-Landau, R. (2000). Retributivism and Desert. Pacific Philosophical Quarterly 81, 189- 214.

Shaw, W. (1980). Intuition and Moral Philosophy. American Philosophical Quarterly 17, 127-134.

Singer, P. (1974). Sidgwick and Reflective Equilibrium. Monist 58, 490-517.

Tasioulas, J. (2006). Punishment and Repentance. Philosophy 81, 279-322.

Van Ness, D. (1995) Anchoring Just Deserts. Criminal Law Forum 6, 507-517.

Vincent, N. (2010). On the Relevance of Neuroscience to Criminal Responsibility. Criminal Law and Philosophy 4,77-98.

von Hirsch, A. (1985). Past or Future Crimes: Deservedness and Dangerousness in the Sentencing of Criminals. New Brunswick, USA. Rutgers University Press.

von Hirsch, A. (1990). Proportionality in the Philosophy of Punishment: From ‘Why Punish?’ to ‘How Much?’. Criminal Law Forum. 1, 259-290.

von Hirsch, A. (1992). Proportionality in the Philosophy of Punishment. Crime and Justice. 16, 55-98.

von Hirsch, A. (1993). Censure and Sanctions. Oxford, UK: Clarendon Press.

von Hirsch, A. and Ashworth, A. (2005). Proportionate Sentencing. Oxford, UK: Oxford University Press.

Waldron, J. (1992). Lex Talionis. Arizona Law Review. 34, 25-51.

Walker, N. and Padfield, N. (1996). Sentencing: Theory, Law and Practice. London, UK: Butterworths (2nd ed.).

Wood, D. (2010). Punishment: Nonconsequentialism. Philosophy Compass 5, 470-482.

Zedner, L. (1998). Sentencing Young Offenders. In Ashworth, A., and Wasik, M., eds., Fundamentals of Sentencing Theory 165-186. Oxford, UK: Clarendon.

Acknowledgments
We thank Kat Brazenor, Zach Hoskins, Bruce Langtry, Alice Muhlebach, Laura Schroeter, Lucia Zedner and the Journal’s anonymous referees for useful comments and helpful suggestions. Earlier versions of the paper were presented at conferences of the American Law and Society Association, the Australian Society of Legal Philosophy and the North American Society of Social Philosophy. The financial support of the Australian Research Council is gratefully acknowledged.

---

[1] Or perhaps, more specifically, requiring hard treatment over and above the minimal, unavoidable element involved in condemnation or censure. Tasioulas claims that ‘[e]ven purely formal censure constitutes hard treatment, since condemnation is meant to be experienced as unwelcome, a bringing up short of the wrongdoer, a drawing attention to, and denunciation of, his moral wrong-doing’ (2006: 295). He thus refers to ‘the hard treatment...entailed by the most lenient means of communicating censure’ (2006: 296). Likewise, Feinberg observes that ‘[r]eprobation is itself painful, whether or not it is accompanied by further “hard treatment”’ (1965: 400, emphasis added).

On the other hand, though condemnation ‘is meant to be experienced as welcome, a bringing up short of the offender [etc.]’ (emphasis added), it is not hard treatment unless it is so experienced. The wrongdoer may totally ignore the censure, or just laugh at it. As Duff notes (though not specifically in relation to this point), censure ‘can be expressed by a formal conviction, or by a purely symbolic punishment that burdens the offender only insofar as she takes its message of censure seriously’ (2001: 82, emphasis added).

[2] The contrast is with determinative proportionality, according to which it is likewise morally impermissible to impose a punishment on an offender that is less severe than is proportionate to the seriousness of the offender’s crime. See, for instance, von Hirsch 1985: ch. 4.

[3] The word ‘punishment’ is generally used here rather than ‘penalty’, even though the latter may sometimes seem more natural. There is no intention to allude to Feinberg’s distinction between punishment and ‘mere penalties’ (1965: 397). Feinberg thanks Dr. Anita Fritz for alerting him to this distinction (at n. 3).

[4] Braithwaite and Pettit call this the ‘anchoring point’ problem, and conclude: ‘[t]he simple fact is that retributivism cannot supply a unique non-arbitrary way of translating a scaling of crimes into recommendations for levels of sentences’ (1990: 149-150). See also Shafer-Landau 2000: 191, Kaufman 1973: 57, Gardner 1958: 120-1, Walker and Padfield 1996: 111-2, Galligan 1981: 164, Benn and Peters 1959: 188, Kleinig 1973: 116, von Hirsch and Ashworth 2005: 142.

[5] See, for instance, Shafer-Landau 1996: 307-9, Shafer-Landau 2000: 201-4, Walker and Padfield 1996: 111-2, Sadurski 1985: 238-9.

[6] ‘[W]hen offenders have been convicted of criminal conduct of similar seriousness, they deserve penalties of comparable severity’ (von Hirsch and Ashworth 2005: 139; von Hirsch 1993: 18).

[7] ‘Punishing crime Y more than crime X expresses more disapproval for crime Y, which is warranted only if it is more serious. Punishments should thus be ordered on the penalty scale so that their relative severity reflects the seriousness-ranking of the crimes involved’ (von Hirsch and Ashworth 2005: 140; von Hirsch 1993: 18).

[8] And presumably, it should be added, a proportionately larger space. In fact, this would be better expressed as the requirement simply that ‘penalties...be...spaced to reflect the relative...spacing of the crim[es]’ (von Hirsch 1993: 39). This formulation of the sub-requirement is preferable to von Hirsch and Ashworths’ explicit formulation of it. While the latter requires that where X, Y and Z are crimes and ‘Y is considerably more serious than X but only slightly less so than Z’, there be a larger space between the punishments for X and Y than between the punishments for Y and Z, it does not specify how large this difference should be. The simplified ‘spacing’ sub-requirement captures the intuition about ordinal proportionality that the difference in the size of the spaces should reflect the difference in the size of the spaces between X and Y and between Y and Z. This simplified sub-requirement appears to be what von Hirsch in fact intended by the ‘spacing’ sub-requirement: see, for instance, his comment that ‘[s]pacing...depends on how precisely [the] comparative gravity [of crimes] can be calibrated’ (von Hirsch 1993: 18).

Kleinig makes a similar point, stating that ‘[n]ot only do we order deserts, but also (partially) the intervals between them’ (1973: 116).

[9] This sentence is also quoted in n. 8 above.

[10] As pointed out in n. 8 above, this is not provided in von Hirsch and Ashworths’ statement in the text at that n.

[11] Note the harsh sentencing scale of von Hirsch’s Dr. Draco (1993: 36-7; cf. 1990: 283, 284n).

[12] See text at n. 5.

[13] Or consider Shafer-Landau’s example: ‘[i]f our criminal code includes fifty offences, ranked according to severity, and fifty punishments, ranging by one-year increments from a ten-year to a sixty-year incarceration, proportionality [i.e. ordinal proportionality] can readily be satisfied. But few will think that the mildest criminal offence should be met with a ten year deprivation’ (1996: 307-8).

[14] That is, on the assumption – but see Part Three – that two anchoring points, and more specifically, these two anchoring points, are required.

[15] On scepticism about reliance on moral intuitions, see for instance Shaw, Singer and Raz.

It may be objected that without reliance on some intuitions, no normative problem can be solved. While being concerned not to become embroiled in controversial questions of moral methodology, more particularly, of the proper place of moral intuitions in developing a moral theory, it must be made clear what the relevant point is. It is not that moral intuitions about anchors are especially suspect. Rather, it is that getting the anchors right – properly anchoring a punishment scale – is crucial given their fundamental role in determining a sentencing scale. A very heavy justificatory burden is therefore borne by whichever moral intuitions are relied upon to determine these anchors.

[16] von Hirsch 1985, 44n and Kleinig 124.

It may be pointed out that the general problem of anchoring a punishment scale arises with any attempt to establish a scale of rewards and penalties based on past behaviour. For instance, even if a lecturer is confident about how he ranks a group of student essays, how can he know that any paper deserves a particular grade? (Von Hirsch raises this example at 1992: 78, but the following discussion is quite independent of his.) Perhaps every student’s mark should be moved up – or down. (This may be precisely what happens when the marks of different markers are ‘standardised’.)

However, there is an important difference here. Marks are purely conventional, have only relative meaning. Grading is an exercise in ordinal proportionality, and the anchors, if they exist at all, are either a matter of convention, or explicitly specified (e.g 5% of students to receive H1’s, 10% H2A’s, and so on). If every student receives an H1, then an H1 is devalued – unless some case is put forward for the class having been especially good in the year in question, an exception to the class generally. (It is another matter if students always receive an H1, year in and year out.)

That marks are purely conventional is not to deny, of course, that good marks may lead to consequential tangible benefits, such as prizes, scholarships and improved employment opportunities.

With punishment, however, the hard treatment that is suffered is not conventional, but real. It has to be justified in its own right. The student who receives a poor mark is not in the same position as the recipient of punishment in that consequential harms need not follow (the student may turn to some quite different pursuit where his previous marks are irrelevant). This is not to deny, of course, that he may be effectively punished, for instance, by (if he wishes to continue with his studies) being required to take the subject again, or do remedial work.

[17] See n. 1.

[18] See the quotation from Tasioulas in the first paragraph of n. 1.

[19] See the qualification in the second paragraph of n. 1.

[20] von Hirsch 1993: 36. Von Hirsch seems to assume that a penalty scale is to be anchored through these two penalties.

[21] von Hirsch 1993: 38-9, referring to Kleinig 1973: ch. 7. Kleinig himself says, in the example he puts forward (which restricts possible penalties to imprisonment): ‘[l]et us suppose that offensive language is the most trivial wrong, and that one day’s imprisonment is the lightest penalty that we can effectively inflict’ (1973: 118).

[22] Kleinig 1973: 119, cf. 118: ‘...unless there are two points of correlation it will be impossible to remove all arbitrariness...’. Von Hirsch says: ‘[i]f one has decided what the penalty should be for certain crimes, then it is possible to fix the sanction for a given crime, X, by comparing its seriousness with the seriousness of those other crimes’ (1993: 18-19; cf. von Hirsch and Ashworth 2005: 141-3).

[23] If the assumption that the crimes are equally spaced is dropped, and it is held instead that there are different spaces between A and B, B and C, and C and D, then there should likewise be proportionately different spaces between a and b, b and c, and c and d.

[24] The stipulation of equal spacing excludes candidates where the non-anchored punishments differ, for instance, eight, seven, three and two years’ imprisonment respectively. See n. 23.

[25] But see Kleinig 1973: 119, n. 22 above and the accompanying text, where Kleinig says that ‘at least two points of contact are needed if all arbitrariness is to be removed’ (emphasis added).

[26] See text at n. 21 and n. 21 itself.

[27] Perhaps one should add: ‘possessing no more element of hard treatment than is necessary to communicate the censure’. But see n. 1, first paragraph.
[28] As Lucia Zedner pointed out to the authors, this is recognised by retributivists in the promotion of equality of impact over formal equality before the law. This concept stood behind the introduction of unit (or day) fines in many European jurisdictions (Ashworth 2010, ch. 7: ‘Equality before the Law’, esp. 253). On the subjective experience of punishment, see Kolber, and for brief discussion, Vincent 92.

[29] Likewise, a period in prison for a juvenile offender may be more burdensome than the same period for an adult offender (and not least, because the opportunity costs are greater). See Zedner 1998: 173.

[30] ‘Determined’ is preferable to ‘anchored’ here. The suggestion that every sentence be individually anchored, it may be objected, is a contradiction in terms, since the very notion of anchoring presupposes that some punishments are not anchored, but derived from the punishment or punishments that is or are anchored.