I am grateful to the respondents for their thoughtful and challenging papers, which give me much to think about. In this short reply I cannot deal adequately with, or even address, all the issues they raise: all I can do here is discuss some of the general, more methodological, issues about the approach that I took in my original paper – issues clustering around the question of whether we should talk of the presumption of innocence only as a specific legal doctrine that operates within the criminal process; or more broadly of presumptions of innocence that also function in other criminal law contexts, and outside the law in our civic dealings with each other and with the state. My respondents show that my appeal to this broader conception of the PoI in my original paper was at least insufficiently clear; it remains to be seen whether it was fundamentally misguided.

• 1 In Favour of a Narrow Presumption of Innocence

  One motive for suggesting that we should talk not of ‘the presumption of innocence,’ but of multiple presumptions of innocence operating in different contexts, with different effects, and defeasible in different ways, was to try to sidestep disputes between advocates of broad and narrow readings of ‘the PoI.’ I hoped to do justice to the concerns of those who favour a narrow reading, which gives ‘the PoI’ a tolerably determinate meaning in a clearly specifiable context, by identifying the specific PoI that operates in the criminal process; but also to the concerns of those who favour a broader reading, which gives ‘the PoI’ a much wider but also therefore less determinate role in a range of less easily specifiable contexts, by talking of the different presumptions of innocence that operate in such different contexts. Thomas Weigend resists this suggestion for reasons with which I sympathize:¹xThomas Weigend, ‘There is Only One Presumption of Innocence,’ in this issue. whether we talk of a single, wide PoI that operates well beyond the criminal process, or of multiple PoI operating in different contexts, we are liable to replace a determinate principle that can do substantive work with a rhetorical device that expresses, rather than rationally grounds, our normative preferences.²xSee Lonneke Stevens, ‘The Meaning of the Presumption of Innocence for Pre-trial Detention. An Empirical Approach,’ in this issue, pp. 246-7. We could say, indeed, that in the context of the criminal trial the PoI is what Magnus Ulväng would count as a ‘dogmatic rule,’ rather than merely a ‘rule of thumb’:³xMagnus Ulväng, ‘Presumption of Innocence Versus a Principle of Fairness. A Response to Duff,’ in this issue, pp. 208-10: Ulväng argues that ‘a PoI ought to be described as a rule of thumb,’ rather than as a ‘rule of obligation’ or a ‘principle’; but if we treat ‘has not been proved guilty according to law’ as a ‘condition of application’ of the rule, this PoI seems to fit his conception of a ‘rule of obligation.’ if a person has been charged with a criminal offence but has not been proved guilty according to law,⁴xFor this wording, see ECHR, Art. 6(2). officials of the criminal justice system shall presume him to be innocent. But once we move beyond that specific context, it becomes progressively less clear just what the, or a, PoI demands of whom.

  Alwin van Dijk offers a conception of the PoI that is broad even by the standards of those who favour a wide reading: we should count as ‘a PoI interference,’ he argues, ‘any act that might convey to a reasonable actor that he is not presumed innocent of a punishable offence’; the more important issues then concern the conditions under which this PoI can be justifiably infringed.⁵xAlwin A. van Dijk, ‘Retributivist Arguments against Presuming Innocence’, this issue, p. 250. It is infringed when, for instance, a defendant is convicted after his guilt has been proved beyond reasonable doubt; but, we think, that infringement is justified. Now Van Dijk rightly emphasizes that his PoI, like those which I discuss, is a matter of action, rather than of belief or feeling: it concerns not what we, or criminal justice officials, believe or feel about a person, but how we, or they, behave towards him – whether he is treated as an innocent, or as someone who is guilty, or suspected.⁶xSee too Geert Knigge, ‘On Presuming Innocence: Is Duff’s Civic Trust Principle in Line with Current Law, Particularly the European Convention on Human Rights?’, in this issue, pp. 236-7. He also rightly notes that ‘innocence’ in this context is not a matter of being ‘pure, wholesome, fulfilling, natural,’⁷xAs Ulväng describes the ‘orthodox’ idea of innocence in this issue, p. 216. As Ulväng also notes, there are complications even in the idea of being ‘innocent’ of a criminal offence, since a person might satisfy some of the conditions of criminal liability but not others: I cannot discuss these complications here. but simply of not being guilty of what the law defines as a criminal offence. However, his PoI is too broad to be helpful. In particular, by so defining it that it is interfered with whenever a person is treated as guilty, he elides an important distinction between two kinds of case in which we may justifiably treat a person as other than innocent. We can see this distinction by looking more carefully at the nature of presumptions.

  As Weigend notes, ‘presumptions’ in law are often conclusions that courts are permitted or required to reach given proof of certain facts:⁸xWeigend, this issue, pp. 193-4. given proof that p, the court may, or shall, presume that q. Such presumptions count as legal presumptions, rather than simply principles of ordinary reasoning, only if proof that p would not normally constitute proof that q: they allow or require courts to reach conclusions that the evidence by itself does not warrant. They are typically rebuttable, by the provision of evidence that not-q;⁹xIndeed, if they are not rebuttable, they are not so much presumptions as substantive rules of law that define the relevant offence: see Duff, ‘Strict Liability, Legal Presumptions, and the Presumption of Innocence’, in Appraising Strict Liability, ed. A.P. Simester (Oxford: Oxford University Press, 2005), 125, at 130-2. in adversarial courts, they serve to shift the probative burden from prosecution to defence. The PoI is not that kind of presumption: it specifies not a conclusion that the court may reach given certain evidence, but the position from which it must start. It is nonetheless worth calling it a presumption: for it specifies a legally relevant proposition, that the defendant is innocent of the offence charged, that the court must take for granted; and that proposition is usually not one that we would be empirically justified in taking for granted. If this presumption was irrebuttable, it would then always provide a reason against treating a person as guilty:¹⁰xCompare Ulväng, this issue, pp. 209-10, on rules of thumb. such treatment, which includes conviction and punishment, would always interfere with or infringe the PoI; it would always infringe the defendant’s rights, since the PoI expresses a right (indeed, a human right). But must we really say that to convict and punish someone who has been duly proved guilty is to infringe his right to be presumed innocent, albeit justifiably so?¹¹xCompare D.N. Husak, Overcriminalization: The Limits of the Criminal Law (Oxford: Oxford University Press, 2008), 92-103, on the ‘right not to be punished,’ which he takes to be infringed even when a person who has been duly proved to be guilty is punished; but it seems better to say that the right is simply a right not to be punished unless proved guilty of a crime. For an explanation and critical discussion of the idea of ‘infringing’ a right, see J. Oberdiek, ‘Lost in Moral Space,’ Law and Philosophy 23 (2004): 325. If we infringe someone’s rights, we owe them something – an apologetic explanation, perhaps some compensation. But we owe no such thing to a person who has been proved guilty: while an actually guilty person has the right to be presumed innocent in the absence of proof of his guilt, he has no such right once his guilt is proved; such proof entitles us to treat him as guilty. By contrast, if we subject a person who has not been duly proved guilty to pre-trial detention, to prevent offences that we fear he will commit if left free, we infringe the PoI, and his rights: if the danger is great enough that he will commit crimes that are serious enough, we might be justified in doing so; but we have failed to give him his due, and therefore now owe him an apologetic explanation and compensation.¹²xSee my ‘Who Must Presume Whom to be Innocent of What?,’ in this issue, p. 12: but contrast Stevens, this issue; Knigge, this issue, p. 227; Weigend, this issue, p. 197. More generally, it seems plausible to say that the criminal law’s officials have a duty to treat us, and we have a right that they treat us, as presumptively innocent – that their treatment of us should start from that position; but we have no right, not even one that can be justifiably infringed, that they continue to treat us as innocent after our guilt has been proved.

  We should thus understand the PoI as expressing the right to be presumed innocent not come what may, but ‘until proved guilty according to law.’ Even if we should therefore reject Van Dijk’s unqualified PoI, however, this is not yet to say that we should confine the PoI to the criminal trial or the criminal process. But I note here two further reasons, reflecting points made by my respondents, in favour of a narrowly understood PoI.

  First, such a narrow PoI is clearly a matter of law: it is a presumption with which courts, and other criminal justice officials, are legally required to operate; it can ground legal appeals from those whose innocence has not been properly presumed. Once we begin to talk of the, or a, PoI outside that context, however, its legal status becomes much less clear – or clearly non-existent. I talked, for instance, of a PoI that protects us against over-hasty prosecution – the requirement that prosecutors take us to be innocent of a crime until there is evidence strong enough to constitute a case to answer:¹³xIn this issue at n. 20. but it is not clear whether I took this to be a matter of legal duty, a legal right that those too hastily prosecuted could assert; or rather a civic duty that lacks the force of law. Even more obviously, when I talk of the ‘civic trust’ that, I claim, we owe to each other, and of the civic PoI that I take to inform such civic trust, I cannot mean that we have a legal duty to presume each other to be innocent. Geert Knigge accuses me of a ‘confusion of fact and law.’¹⁴xKnigge, in this issue, pp. 233-4. As a matter of sociological fact, civic trust is essential for ‘a well-functioning society’; as a matter of sociological fact, we play a range of different roles in relation to the criminal law: but such trust and such roles are not matters of law, and this undercuts my attempt to give them a normative status. I must plead guilty to a failure to keep separate the question of what legal rights and responsibilities we have (or should have) as citizens, defendants, or offenders, from the question of what non- or extra-legal civic rights and duties we have as a matter of civic or political morality: I certainly did not mean to claim that civic trust, and the rights and duties that I ascribe to such roles as those of defendant and offender, all are or all should be matters of law; I meant to claim only that they figure in the civic morality of a well-functioning democratic polity. This is not to say, however, that they are simply matters of sociological fact. Knigge implies that we can ascribe ‘normative status’ to such roles or expectations only if they are matters of law:¹⁵x‘In Duff’s analysis, the “normative role” operates as a vehicle for the introduction of all kinds of responsibilities, which have no clear basis in the law’: in this issue, p. 234. but my concern (as I ought to have made clearer, to myself as well as to readers), is with political morality rather than with sociological fact; the normative status I ascribe to these roles and to the demand of civic trust is political-moral rather than legal. There is then the further question of whether any of the rights or responsibilities that attach to those civic roles should be recognized and enforced as a matter of law; but my primary concern is with the political morality that underpins the law.

  Second, however, once I move away from the PoI as a legal principle, to treat it instead as a principle of civic morality, it becomes harder to give it any determinate role, content or foundation. The meaning and implications of the PoI are notoriously uncertain even if we see it only as a legal principle operating within the criminal process; but we can make progress in discussing what it means, and what it should mean, as a matter of law. Whether our aim is the interpretation or the reform of our existing law, we will need to connect the law to underlying principles of political theory concerning the proper role of the state and the law; but our focus will be on a reasonably determinate, or determinable, legal doctrine. That is indeed where my argument began, with the meaning of the PoI as a legal doctrine within the criminal process. As both Knigge and Stevens point out,¹⁶xSee Knigge, this issue, pp. 230-33; Stevens, this issue, p. 247. however, I soon abandon that grounding in existing law: the presumptions that I posit either lie outside the law, or are flagrantly inconsistent with the PoI as it is interpreted by the courts. I am thus not analysing the PoI as an existing legal principle; I cannot claim that the rights and duties I attach to the various presumptions that I identify should all be legal rights and duties; I am not engaged in sociological analysis of the values or expectations that inform our civic lives: I am instead, it seems, gesturally sketching a range of norms that lack determinate content or grounding. But when I talk of such a variety of presumptions, with such different implications, defeasible in different ways, they lose the ‘institutional support’ and ‘substantive significance’ that genuine principles need: they are not recognized in the law, nor do they help to explain more specific rules and doctrines.¹⁷xSee Ulväng, this issue, pp. 213-4.

  There is force to these criticisms: I at least failed to make my strategy clear, and failed to provide a clear enough explanation of just what status I claim for these presumptions, or just what roles they should play. Maybe I should therefore abandon these large ambitions; put the PoI back in its place as a legal principle operating within the criminal process; and engage in a separate discussion of the various issues that I tackle outside that context, without trying to (ab)use the rhetorical power of the PoI in ways that stretch it beyond its useful meaning.

  Maybe, but …. I suspect that the question of whether we should talk of the presumption, or presumptions, of innocence beyond the criminal process is a presentational one: nothing of real substance hangs on it; the question is whether it is illuminating to use this phrase in this broader way – or so I will suggest.

• 2 On the Other Hand

  We begin, but cannot end, with the PoI as it figures in the criminal process once a person is formally charged with a criminal offence.¹⁸xSee Knigge, this issue, p. 226; and Weigend, this issue, p. 198, on the meaning of ‘charged with a criminal offence.’ Already questions arise about what is to be thus presumed by whom, to what effect. For instance, must a prosecutor presume the innocence of the person she charges? Knigge suggests that she need not: it is her job is ‘to maintain at the trial that the defendant is guilty.’¹⁹xKnigge, this issue, p. 227; this reminds us of the importance of examining roles (see further below). Weigend suggests that she must: she must ‘[a]ssume that the suspect did not commit the crime, (…) and then ask yourself whether you can do to him what you intend to do’;²⁰xWeigend, this issue, p. 196. she can assume that he is innocent of the crime, while insisting that he must face trial given the strength of the evidence of his guilt. More precisely, since the PoI concerns actions rather than beliefs, the question is whether charging someone with a crime, and seeking in court to prove his guilt, is consistent with the proposition that he is innocent of that crime. The same question arises about pre-trial detention and other kinds of restriction on defendants: can a court impose such restrictions without implicitly denying that proposition? So far, however, the question is limited: are such measures consistent with the presumption that the defendant is innocent of the crimes with which he has been charged? Thus to justify such restrictions as being necessary to prevent the future offences (failing to appear for trial, interfering with the criminal process, committing other offences) that the defendant might otherwise commit does not violate this PoI, so long as those predictions of future offending are not based on a pre-judgement that he is guilty of the offence(s) charged.²¹xSee Stevens, in this issue, on how Dutch judges actually make decisions about pre-trial detention; see also Weigend, this issue, p. 197, on the importance of whether defendants are required to appear for trial.
  

  If we want to understand even this limited PoI, however, we must go further and deeper than this – whether the understanding we seek is analytical (how is the PoI as it figures in our current legal practice best rationalized?) or normative (what role should what kind of PoI play in our law?). We must identify the values in which the PoI is grounded, to ask which version of the PoI best expresses them. We must offer an account of the proper aims of the criminal process: to understand what it is to presume innocence, and why that should matter, we must understand the process in which that presumption is to figure. But to develop an account of the criminal process, we must understand its role as part of a system of criminal law, and the role of the criminal law as part of the apparatus of the state; and this must then be set within a larger account, in political rather than purely legal theory, of the proper aims of the state, and of the relationship between state and citizen (and between citizens, since the state must claim to be acting in the name of the citizens).

  This is not yet to say that we must talk of presumptions of innocence beyond the specific context of the post-charging criminal process. But once we look beyond that context, we must notice that issues of criminal innocence or guilt (or suspected guilt) arise in other contexts: in various ways our civic conduct towards each other, and the conduct of state officials towards us, display judgements of our criminal innocence or guilt – our innocence or guilt of specific past offences, or of offences in general, or of predicted future offences. It would be strange if there was no normative connection between the PoI as it operates as a formal legal principle in the criminal process, and the ways in which we treat each other, and officials treat us, as innocent or guilty (or suspected) outside that context; between the PoI as a legal principle and the values that should guide these other types of conduct. It is worth asking whether and how far our civic conduct towards each other, and our treatment by state officials, in these various other contexts should be structured by a conception of each other as innocent of past or future crimes; and whether and how far it can properly express a conception of each other as guilty, or as not determinately innocent (as suspect).

  In tackling such questions, I suggested, two strategies are helpful. One is to think about civic trust, as a practical demeanour towards fellow citizens. This is not a legal duty; it is not essentially a matter of beliefs; it is not simply a matter of sociological fact. It is a normative ideal of how we should behave towards each other as fellow members of a polity: our starting point should be to treat each other as law-abiding – as innocent of both past crimes and future crimes.²²xAlthough Weigend rejects my account of civic trust, its content might not in the end be very different from that of the ‘civil liberties and fundamental freedoms’ to which he appeals: this issue, p. 202. Such a demeanour is part of what it is to be a citizen, and to recognize others as my fellow citizens; it partly defines the distinctive role of citizen in a democratic polity. Second, we must then ask about the conditions given which such trust may be undermined, qualified, or defeated. In that connection it is useful to think about the different roles that we can play in relation to the criminal law, in particular such roles as suspect, defendant, convicted offender and ‘ex-offender.’ These roles are defined by the rights and responsibilities, the benefits and burdens, that attach to them: some of these might be matters of law, others matters of social expectation; in both cases we need to ask not just how the roles are in fact defined in this or that society, but how we should define and understand them. One central dimension of these roles is the way in which the role-bearer’s innocence is put into doubt, challenged or denied. So we must ask about the conditions under which innocence is properly challenged or denied, and about the normative implications of such challenges or denials: what difference do they make to the person’s rights and responsibilities, to how others (especially state officials) may treat him, and to what he can be expected or required to do.

  This is, I still believe, a fruitful approach to understanding the proper role of the criminal law in a democratic polity: to ask how the law should address and treat the citizens of such a polity (which is also to ask how they should treat each other); to examine the roles that they may play in relation to the criminal law, and the rights and responsibilities that define those roles; and to look at the ways in which their innocence of past or future crime should be taken for granted, challenged or denied, and the implications of such takings for granted, challenges or denials. I thought it useful to talk here of different ‘presumptions of innocence’ related to these various roles: not in order thereby to conclude debates about how these roles should be understood, or to provide distinct normative foundations for accounts of those roles, but to characterize what is at stake, and to highlight the connections between the various issues that I discussed, and the connections between those issues and the role played by the traditional PoI in the criminal process. If the use of that phrase, ‘presumption of innocence,’ turns out to confuse or obscure rather than to illuminate, I should drop it: what matters is the substantive approach that I have recapitulated here – to examine the ways in which criminal innocence and guilt figure in the various roles that we may play as citizens of a democratic polity.

Noten

• 1 Thomas Weigend, ‘There is Only One Presumption of Innocence,’ in this issue.

• 2 See Lonneke Stevens, ‘The Meaning of the Presumption of Innocence for Pre-trial Detention. An Empirical Approach,’ in this issue, pp. 246-7.

• 3 Magnus Ulväng, ‘Presumption of Innocence Versus a Principle of Fairness. A Response to Duff,’ in this issue, pp. 208-10: Ulväng argues that ‘a PoI ought to be described as a rule of thumb,’ rather than as a ‘rule of obligation’ or a ‘principle’; but if we treat ‘has not been proved guilty according to law’ as a ‘condition of application’ of the rule, this PoI seems to fit his conception of a ‘rule of obligation.’

• 4 For this wording, see ECHR, Art. 6(2).

• 5 Alwin A. van Dijk, ‘Retributivist Arguments against Presuming Innocence’, this issue, p. 250.

• 6 See too Geert Knigge, ‘On Presuming Innocence: Is Duff’s Civic Trust Principle in Line with Current Law, Particularly the European Convention on Human Rights?’, in this issue, pp. 236-7.

• 7 As Ulväng describes the ‘orthodox’ idea of innocence in this issue, p. 216. As Ulväng also notes, there are complications even in the idea of being ‘innocent’ of a criminal offence, since a person might satisfy some of the conditions of criminal liability but not others: I cannot discuss these complications here.

• 8 Weigend, this issue, pp. 193-4.

• 9 Indeed, if they are not rebuttable, they are not so much presumptions as substantive rules of law that define the relevant offence: see Duff, ‘Strict Liability, Legal Presumptions, and the Presumption of Innocence’, in Appraising Strict Liability, ed. A.P. Simester (Oxford: Oxford University Press, 2005), 125, at 130-2.

• 10 Compare Ulväng, this issue, pp. 209-10, on rules of thumb.

• 11 Compare D.N. Husak, Overcriminalization: The Limits of the Criminal Law (Oxford: Oxford University Press, 2008), 92-103, on the ‘right not to be punished,’ which he takes to be infringed even when a person who has been duly proved to be guilty is punished; but it seems better to say that the right is simply a right not to be punished unless proved guilty of a crime. For an explanation and critical discussion of the idea of ‘infringing’ a right, see J. Oberdiek, ‘Lost in Moral Space,’ Law and Philosophy 23 (2004): 325.

• 12 See my ‘Who Must Presume Whom to be Innocent of What?,’ in this issue, p. 12: but contrast Stevens, this issue; Knigge, this issue, p. 227; Weigend, this issue, p. 197.

• 13 In this issue at n. 20.

• 14 Knigge, in this issue, pp. 233-4.

• 15 ‘In Duff’s analysis, the “normative role” operates as a vehicle for the introduction of all kinds of responsibilities, which have no clear basis in the law’: in this issue, p. 234.

• 16 See Knigge, this issue, pp. 230-33; Stevens, this issue, p. 247.

• 17 See Ulväng, this issue, pp. 213-4.

• 18 See Knigge, this issue, p. 226; and Weigend, this issue, p. 198, on the meaning of ‘charged with a criminal offence.’

• 19 Knigge, this issue, p. 227; this reminds us of the importance of examining roles (see further below).

• 20 Weigend, this issue, p. 196.

• 21 See Stevens, in this issue, on how Dutch judges actually make decisions about pre-trial detention; see also Weigend, this issue, p. 197, on the importance of whether defendants are required to appear for trial.

• 22 Although Weigend rejects my account of civic trust, its content might not in the end be very different from that of the ‘civil liberties and fundamental freedoms’ to which he appeals: this issue, p. 202.