• 1 Introduction

  Antony Duff has chosen to portray the presumption of innocence (PoI) as a broad and general rule or principle. In his paper, he gives this essentially procedural norm an extensive, explanatory and normative function far outside its usual scope, i.e., the trial. According to Duff, some sort of PoI is engaged at the pre-trial stage as well as in the later face when punishments are executed. The same is the case even after a convicted criminal has served his sentence, i.e., when he is restored to civic life.

  Since the scope of application of a PoI, as Duff portrays it, is very broad, he correctly assumes that we are perhaps not working with a single norm, but rather with many (different) versions of a PoI. Therefore, the definition of the presumption might vary according to the different contexts where it is at work. Its implications at a pre-trial stage are likely to be somewhat different from the effects at the trial stage or after a sentence has been served. It is also likely that the meaning of the term ‘innocent’ will vary depending on the context. Nevertheless: throughout the text, Duff continuously refers to the normative standard as (a) presumption of innocence.

  I am essentially sympathetic to the idea of a wider notion of justice and/or fairness within the whole sphere of the criminal law in a wide sense. Nevertheless, I doubt that it is actually a (version of) PoI that provides the moral anchoring points Duff is looking for. My hypothesis is that the PoI is one – among several – emanation of some other, deeper principle that connects different moral assumptions in the complex normative framework known as ‘criminal law.’ This obviously implies the question of what that principle of procedural norms in fact consists.

  My paper will mainly focus on the following two issues. First I will try to say something on the ‘ontological’ differences between presumptions, principles and rationales (section 2). Even though they all reflect some aspect of some sort of a rule (or a norm), it is my belief that they have different weight (or standing) within a normative system and consequently that they have different tasks to fulfil. Much of what Duff considers in his paper relates to fundamental (moral and perhaps legal) principles or the underlying rationales of a criminal law system. It remains to be seen whether a presumption of a more practical sort can produce or summarize the reasons that together make up the set of values which, according to Duff, ought to pervade the use of criminal law.

  Next, I will say something about the notion of ‘innocence’ and how it ought to be understood (section 3). To argue that there is an actual rule requiring that judges must presume (i.e., believe) that defendants are innocent – or at least act as if so is the case – is obviously problematic from a normative point of view. And if there is such a rule it is probably not upheld in a considerable number of cases.

  My suggestion will be that the concept of ‘innocence’ should be perceived in a less far-reaching and more normative sense, i.e., as something different from ‘entirely without responsibility or guilt’ of a crime. One way to understand Duff is to say that a notion of innocence implies an obligation to treat everyone who is being scrutinized within the institutions of criminal law fairly. This would entail a practice where any branch of criminal law (in a broader sense) is applied without unjustified prejudice, where each case is considered on its own merits and not too much is read into facts that are not verified, etc. Put in different terms: to be fair implies having good reasons for any measures taken and reflects a principle of caution; if the reasons are not convincing, then we ought to abstain from engaging criminal law. Such an account of a PoI could offer new dimensions in understanding what it actually means to treat someone morally and legally as ‘not responsible’ (for something).

• 2 The Ontology of a Presumption

  Duff does not offer a full definition of the presumption in question. In its most familiar context – the trial – the PoI requires courts to

  ‘presume, of the defendant, that he is innocent of the offence for which he is being tried.’

  However, this is uncontroversial and consistent with the general, and almost universal, descriptions given in works of legal doctrine, as well as ECHR Article 6.¹xSee, e.g., A.P. Simester & G.R. Sullivan, Criminal Law: Theory and Doctrine, 3rd ed. (Oxford and Portland: Hart publishing, 2007), 51ff.; Asbjørn Strandbakken, ‘In dubio pro reo,’ in: Uskyldspresumsjonen (Bergen; Fagbokforlaget, 2003), ch. 1; Karol Nowak, Oskyldighetspresumtionen (Stockholm; Norstedt Juridik, 2003), ch. 2. See also UN Universal Declaration of Human Rights Art. 11. In this context, it should be noted that even though a PoI seems to be deeply rooted as a robust and vital norm in the exercise of legal power in criminal law, one rarely finds any comprehensive definitions of it.²xThe Latin maxim in dubio pro reo, which often is said to capture PoI, is not codified in Swedish law. The maxim is nevertheless valid as a legal principle and can thus be invoked in legal reasoning, etc. Swedish doctrine on procedural law contains many references to a PoI (mainly ECHR Art. 6), but surprisingly little has been said about what it actually means to presume innocence and how strong such a duty really is. Normally one only finds references to meta-rules or mediating principles, concerning:

  • a principle of objectivity (mainly for judges and juries, but also to a certain extent for prosecutors and other officials);

  • a right against self-incrimination (and corresponding right to remain silent);

  • a burden of proof (placed on the prosecution);

  • a high epistemic standard of proof (beyond reasonable doubt);

  • an obligation to acquit defendants if the standards are not met (in dubio pro reo);

  • a general protection against court rulings written in an insulting or chicane manner;

  • a general obligation to treat defendants with respect (e.g., by addressing them as the ‘defendant’ and not as a guilty criminal).

  None of these standards fully captures the presumption of innocence, and in any case they all concern the practice at the trial stage. As indicated above, Duff’s ambition is to make a contribution to the legal discourse by adding a much broader version of the presumption in question; such a PoI stretches far beyond the realm of the courts.³xSimilar ideas have to some extent been put forward by others, cf. Patrick Tomlin, ‘Extending the Golden Thread? Criminalisation and the Presumption of Innocence,’ The Journal of Political Philosophy (2012): 44-66 and Magnus Ulväng, ‘Criminal and procedural fairness – Some challenges to the presumption of innocence’, Criminal Law and Philosophy (2013). These are, however, not as comprehensive as Duff’s normative account.

  The wider version of PoI is given some characteristics, but the details are not further specified. So much is clear, however, that a distinguishing feature is that all parties involved have rights as well as responsibilities. Duff’s depiction of criminal law and of its different actors thus resembles a game with different roles. If you play the game according to the rules, then you will enjoy its benefits. If you do not play by the rules you run the risk of losing the trust that society prima facie has in everyone. A suspect, defendant, or even a convicted person, takes on a distinctive normative character and is supposed to collaborate in the process of assessing responsibility. Correspondingly, the state has certain responsibilities towards the suspect/defendant/convicted criminal/ex-offender: as a main rule, he is to be trusted and restored to civic life. Thus, any consequences comparable to punishments must be motivated by strong reasons for why the measure or intrusion is meant to be justified. It is an open question if – and if so, for how long a time – the trust in the offender is forfeited (e.g., the cases of probation and revoked licenses). Trust or presumed innocence can be regained depending on the person’s behaviour.

  There is nothing controversial in portraying the criminal process in terms of a game. Even in legal systems where plea bargain,⁴xCf. Andrew Ashworth & Mike Redmayne, The Criminal Process, 3rd ed. (Oxford and New York: Oxford University Press, 2005), ch. 10. Absprachen im Strafprozess,⁵xHeike Jung & Kathrin Nitschmann, ‘Das Bekenntnis zum Schuldbekenntnis – zur Einführung des plaider coupable im französischen Strafprozess,’ ZStW, vol. 116, no. 3 (2004): 785-96. etc. do not exist, it is commonly known that defendants who, to the best of their ability, attempt to prevent, remedy or limit the harmful consequences of the crime or give themselves up, are offered extenuation. Co-operation can thus be rewarded.

  In the same manner, evaluating evidence is an activity that is governed by the efforts of the parties. Facts are to be compared with counter-facts; evidence is thus given a relative weight. A ‘false’ burden of proof can sometimes arise during a trial, e.g., if the prosecutor presents conclusive evidence; the defendant is then implicitly required to rebut the facts in question. Such a practice does not necessarily violate a presumption of innocence (in its narrow sense). There is no (ontologically) fixed limit on when the requirement ‘beyond reasonable doubt’ is met. During a trial, the burden of proof can therefore in practice be shifted – or rather the balance can be tipped – if either of the sides in a trial submits relevant facts that affect the evaluation of the evidence as a whole. That this kind of balancing of arguments during the submission of evidence is much more complex than simply stating that the prosecutor always has the burden of proof.

  Consequently: if the PoI provides both procedural safeguards for the defendant (PoI in its narrow version) and certain obligations on the parties to conform to the practice of the procedure (PoI in its broader version), then what does it means to presume innocence? More specifically: what is a presumption?

  2.1 Presumptions as a Sort of a Rule

  The fact that (the broader version of) a PoI seems to confer rights and impose duties/responsibilities implies that we are dealing with some kind of a rule. A judge (prosecutor, civil servant at a prison, etc.) is obliged to presume an accused, etc. as innocent; likewise suspects, accused or convicted persons have responsibilities to participate in the process.

  In this section I will say something about how I understand the concept of ‘a rule’ and why I believe that a PoI ought to be described as a rule of thumb and not as a rule of obligation or even a principle. A tentative conclusion is that rules cannot be viewed either as dogmatic rules of obligation or as guiding rules (or principles); such a description is too basic. In the following I only discuss normative rules that aim at regulating humans and their conduct.⁶xNormative rules are here contrasted to causal rules or ‘rules’ concerning ascertainable social regularities.

  There are many different kinds of rules and numerous ways of systematizing them or dividing them up, e.g., rules that qualify relations between entities (rules of qualification), rules that confer powers (rules of competence) and rules that create obligations (rules of conduct). The last type is without doubt the most important for my inquiry. The PoI (in any form) aims at conducting the behaviour of others (mainly officials).

  A common feature of all rules that create obligations for others is that they contain a described pattern of action: [If X, then Y]. ‘X’ describes the conditions of application and ‘Y’ indicates the norm act.⁷xSee, e.g., Georg Henrik von Wright, Norm and Action. A logical enquiry (London: Routhledge and Kegan Paul, 1963), 70ff., Magnus Ulväng, ‘Straffrätt och principer,’ in Katedralen – Tre texter om straffrätt, ed. Petter Asp et al. (Uppsala: Iustus förlag, 2009), and Nils Jareborg, ‘A Lecture on Principles,’ in Festschrift für Heike Jung, eds. Müller-Dietz et al. (Baden-Baden; Nomos, 2007), 32. Furthermore: rules are normally not valid for everyone in all contexts. The conditions of application tell us who is subjected to the rule (the addressee) and under which circumstances. In order to structure a normative rule, the action pattern must further be supplemented with a deontic operator that reflects the relative strength of the reasons for following the proscribed norm act. In other words, a rule needs to be operated by a modal term: ‘shall,’ ‘must,’ ‘ought,’ ‘shall-not,’ ‘must-not,’ ‘ought-not,’ etc. The modal term ‘may’ can also serve as an operator, but if the may-operator is used in combination with a negation, we can easily transform a may-not rule into a shall-rule. [You may not do X] is equivalent to [You shall do not-X]. In the same manner any crime description instructing a judge to convict a defendant under certain conditions (shall-rule), could be formulated as a prohibition (may not-rule).⁸xA statute (addressed to the judge) saying ‘A person who takes the life of another shall be sentenced for murder to imprisonment for ten years or for life’ can easily be transformed into a rule (aimed at people in general) stating that ‘you shall not take the life of another.’ Cf. the Swedish Penal Code, ch. 3 s. 1.

  From this basic conception of a rule, we can start to distinguish between different sorts of rules. For example: [If X, you shall Y] suggests strong reasons for complying with the action pattern, whilst [If X, you ought to Y] reflects relatively less strong reasons for doing Y. At the same time, there is a crucial difference between rules of obligation that engage the modal term ‘shall’ and the modal term ‘ought’. Any addressee of a shall-rule cannot refrain from following the rule without changing the formulation of the rule (the conditions of application).⁹xA judge who decides to not sentence a defendant who has intentionally taken the life of another to imprisonment will certainly create an exception to the rule ‘you shall not take the life of another.’ Any justification or excuse in criminal law doctrine displays such exceptions that have helped amend the rule and its scope of application. The rule is so to say dogmatic. To comply with an ought-rule is equivalent to say that there are prima facie good reasons to follow the action pattern [If X, then Y], but not following this kind of rule does not necessarily mean that you violate the rule. If there are ‘better reasons’ to do not-Y in a specific case or different context, then the addressee can choose to prioritize these reasons without changing the content of the norm.

  From what has been described above, we discern the difference between on the one hand dogmatic rules and, on the other, guiding rules/guidelines (e.g., rules of thumb, technical instructions but also principles). The latter provide instructions to an addressee concerning what he generally has good reasons to do. However, a guiding rule is not conclusive. Whether the addressee ought to follow the norm act depends on the context and whether the reasons for not-Y override the reasons for doing Y. In this sense guiding rules, e.g., principles, are not conditional but instead categorical.¹⁰xNorms of this sort are categorical in the sense that they – in a specific context and for the relevant addresses – always, and unconditionally, produce good reasons for doing something. This is not to say that there cannot be other norms (principles, etc.), equally categorical, which can override the first norm. Provided that someone belongs to the relevant group of addressees, a guiding rule provides generally, synthesized, standing, good reasons for doing X (or not-X). They are thus (usually) not promulgated by a legislator, but rather formulated by others as summaries of the normative reasons for acting according to certain values or goals within specific institutions of law. A principle and other guiding rules are normally (re)constructed from a body of dogmatic rules. As will be elaborated below, they are the result of legal reasoning within a certain context, i.e., the framework of, e.g., the criminal law as provided by the legislator. Thus, guiding rules metaphorically evolve from within the system and revolve around the promulgated rules. What remains unanswered is what kind of rule a presumption of innocence might be.

  To presume something is according to any dictionary roughly the same as to take something for granted or as a starting point.¹¹xSee, e.g., Oxford Dictionary: Presumption: ‘The action of taking for granted or presuming something; assumption, expectation, supposition; an instance of this; a belief based on available evidence.’ A presumption in law represents an inference of something universally applicable to particular circumstances. Engaging a presumption thus seems to supply lawyers with a rule that offers a solution to a problem in cases where we do not have (strong) reasons to think otherwise. Should this count as a dogmatic ‘shall-rule’ or a guiding ‘ought-rule’?

  If what I have said so far is an adequate description of the ontology of rules, then it seems fair to say that a PoI is not a dogmatic rule of obligation. The description pattern is categorical since it states that everyone is to be presumed innocent (or worthy of trust, etc.), but the conditions of application are not. The meaning of the rule is without doubt to [do X, unless you have better reasons for doing not-X]. Hence, the rule is not dogmatic, and choosing not to follow the pattern description does not automatically mean that the rule is thereby altered, which, as stated above, would have been a clear sign of a dogmatic rule. Consequently, it does not necessarily have to be a PoI that confers obligations on officials. The right against self-incrimination (and to remain silent), a duty not to address the defendant as guilty (in speech or in legal rulings) before guilt has been established, a mandatory standard that the evidence presented must be persuasive enough to meet with the epistemic standard ‘beyond reasonable doubt,’ etc. are separate (often dogmatic) rules that stem from the same normative source as a PoI. Nevertheless, the latter cannot be the rule that actually confers these duties/rights on the parties. The formulation lacks the necessary elements of a dogmatic rule that the rules just mentioned possess.

  If a presumption, of this sort, does not belong to the category of dogmatic rules, can it instead be classified as an ought-rule? If so, what kind of an ought-rule? Could the PoI in itself be a principle or does it merely reflect a principle?

  We are still a long way from a feasible description of what a presumption ‘is.’ Below I will try to answer some of these questions adopting the idea that rules exist on different levels in a normative system.

  2.2 A Metaphor

  One problem in trying to capture and qualify different kinds of rules is the ambiguity as to which form a specific rule actually takes. We can often find norms expressed in legal sources, but unless they are officially promulgated by an authoritative, competent body (e.g., a parliament), the described action patterns and conditions of application may vary significantly according to the descriptions.

  As said above, the PoI is just one example of norms that lack an authoritative definition (unless one settles with the formulation found in ECHR Art. 6(2)). Its closer content is derived from various formulations found in legal writings, precedents, travaux préparatoires, etc. Rules, principles, guidelines, general clauses – or fragments of such – are constantly (re)constructed and (re)formulated linguistically in the stream of products that emanate from everyday legal practice. In one sense it is therefore perfectly true that footprints of a rule are found on a ‘surface level’ of the law as linguistically formulated norms. They are part of the turbulent, constant flux we call ‘the law.’ The objectified content of such a rule can with some certainty be established by looking at the elements of everyday practice. All that would be recognized as valid rules or other norms belongs to the reproduced law.¹²xA PoI does not exist in Swedish law in the form of a (single) rule. The PoI expressed in ECHR Art. 6(2) could nevertheless easily be re-interpreted as a rule of obligation applicable in Swedish courts: ‘If a judge tries a case, then he shall presume the person charged as innocent until otherwise is proven.’ There is nothing controversial in trying to analyse meta-rules, principles, etc. in this positivistic sense. Cf. Hart’s rule of recognition, see H.L.A. Hart, The Concept of Law, 2nd ed. (Oxford and New York: Oxford University Press, 1997), 100ff., 263ff. See also Neil MacCormick, H.L.A. Hart (Stanford: Stanford University Press, 1981), 40ff., 48ff. On the other hand: such a rule does not tell us very much and as I concluded above, it cannot be regarded as a dogmatic rule of obligation.

  If we try to explain the PoI as a principle or other form of guiding rule, we need to address the following questions: What does it then mean to ipso facto presume something (e.g., innocence)? When are the reasons sufficient to ‘prove otherwise’? Who is addressed by the rule? When would a judge (or other) violate the rule? Who enacts them (since they do not seem to be promulgated by a legislator)?, etc.

  The point that I will try to make below is that it is too primitive to view rules as either dogmatic rules of obligation or as guiding rules (or principles). ‘Rules’ – no matter how broad, general or detailed they are – are elaborated by underlying purposes, which are expressed by principles. These principles link surface level material (e.g., rules) to underlying goals or values, thus rationalizing the different rules in a coherent system.¹³xSee also Neil MacCormick & Ota Weinberg, An Institutional Theory of Law. New Approaches to Legal Positivism (Dordrecht: Reidel, 1986), 73. As a consequence principles exist simultaneously on different levels of a normative system. It can either refer to surface level material, e.g., someone’s formulation/definition of a rule or principle or to the normative reasons that justify the surface level material. Hence, the content of ‘a rule’ and its implications are somewhat different depending on whether we are analysing the ‘black letter law’ expressions as they are displayed in statues, court decisions or mere statements of legal science, or looking at the normative force of the rule embedded in underpinning reasons (on a deep structure level).

  A rule that obliges an official to presume someone innocent until he has better reasons to do otherwise is not especially conclusive when it comes to conducting people’s behaviour. Its normative force is fairly weak. The impact of such a rule can nevertheless be rather strong if one focuses on the reasons behind the expressed surface-level rule. We thus need to recognize deeper layers of the law, i.e., levels where reasons, values and desirable goals exist and are organized to explain, justify and underpin the results that we see on the surface.¹⁴xThe idea of analysing normative hierarchies in terms of a multi-structured system of different levels is not new. See Kaarlo Tuori, Critical Legal Positivism (Hants: Ashgate, 2002). Cf. also Hart, The Concept of Law. Chomsky has introduced similar metaphors in his theory of language. Such a level represents a legal culture that provides rationality and explanatory force to the rudimentary rules.¹⁵xThis underlying level might need to be supplemented with a deep structure where all values, goals, etc. are stored, but I will not deal with this question here. Cf. Tuori, Critical Legal Positivism, 183ff.

  Within this layer, the normative reasons for a certain practice are gathered as a coherently organized bridge between deep structure values and surface level expressions. Typical deep structure material are the Rechtsstaat principles and Human rights; they serve as guidelines and thus make up the normative connection between a rule of obligation or other expressions on the surface and the desired deep structure value. Such deep structure norms can be considerably more important and have greater normative force within the institutions of law compared to an explicit rule (on the surface) that prescribes a certain conduct.

  Tuori describes the process vividly:

  ‘With respect to the continuous alterations at the law’s surface level, the legal culture represents the memory of the law and keeps alive the connection to the past of the law; when the legal culture encounters new legislation, the past encounters the future, and this cannot but leave traces in the law, primarily oriented towards the future. As the memory of the law, as sedimented history, the legal culture calms down the storms of the surface.’¹⁶xTuori, Critical Legal Positivism, 163 (with obvious reference to Foucault).

  MacCormick comes very close to this:

  ‘Principles express the underlying purposes of detailed rules and specific institutions, in the sense that they are seen as rationalising them in terms of consistent, coherent and desirable goals. Thus legal principles are the meeting point of rules and values.¹⁷xMacCormick & Weinberger, An Institutional Theory of Law, 73.

  In sum: normative principles – but presumably also other forms of ought-rules – represent the normative binding material between deep structure values and surface level norms. They consist of sedimented, organized reasons and have the function of reminding the addressees of the fundamental goals and values with a practice.¹⁸xSee also Jareborg, A Lecture on Principles, 365. In this sense they guide human behaviour.

  2.3 How to Identify a Principle

  The next questions are: how do we actually know that a principle ‘exists,’ and how can we establish that it is valid?

  Tuori and Tolonen suggest that two criteria be used to identify principles as valid. Similar to the identification of legally valid rules (through, e.g., a rule of recognition) principles must first have substantive significance or value, and, secondly, institutional support.¹⁹xSee Tuori, Critical Legal Positivism, 178f. and Hannu Tolonen, ‘Principles and Goals,’ in: Scandinavian Studies in Law (Stockholm: Stockholm Institute for Scandinavian Law, 1991): 275ff. It is difficult to make any sharp distinctions between the requirements for institutional support (or ‘official’ behaviour and acceptance from the internal point of view). The difference between the recognition of rules and principles hence lies foremost in the requirement of substantive significance. Whilst a rule only has to be accepted, a principle must add some normative significant justification to a practice. The latter requirement is fulfilled if the principle is recognized in sources of law, e.g., as a (basic) right in a statute or a bill of rights, or through references in travaux préparatoires, preambles, or precedents, etc. Secondly, the principle must explain and/or justify some substantive rules by supplying them with reasons, and thereby creating a link between practice and value.

  If a broader version of PoI is to be understood as a legally valid normative principle, then we must first find references to it in sources of law.²⁰xBy this I do not suggest that there has to be actual (or literal) references to the principle in question in order to be recognized as valid. The underlying values and goals, which give the principle normative value, must instead be recognized as valid within the normative system. Consequently, the fact that a principle is developed or constructed in legal doctrine does not alone meet with the requirement of institutional support. See also Tuori, Critical Legal Positivism, 179. Nevertheless, as soon as the principle is used or referred to in other sources of law – or constructed from a set of coherent valid deep structure reasons – then a principle can be said to be ‘discovered’ and the necessary conditions are fulfilled. This criterion presents no problem. There are numerous references to a PoI in legal sources, at least in its narrower sense. The first criterion is nevertheless more of an obstacle. We need to be rather precise with what kind of practice the proposed principle is supposed to justify. Here we encounter several problems.

  The first is that it is difficult to establish a normative, explanatory link between the values that a PoI aims at honouring and a specific practice. The underlying deep structure values in Duff’s version of PoI seem to concern human dignity, conditions for civic life and a certain attitude of state officials to maintain objectivity and impartiality but also trust. The surface layer practice that is supposed to ensure society a cautious, humane and parsimonious criminal law cannot directly be derived from a practice of treating people (subjected to criminal law) as ‘innocent.’

  The identified values seem to justify or explain a practice of treating the accused fairly and with due prudence for the fact that a judge, jury (or some other assessor) might be wrong, etc. Nevertheless, this does not necessarily imply that an accused or convicted person ought to be treated as innocent (not guilty, not responsible, not in lack of trust, etc.). Hence, the institutions that are rationalized through a connection to these values are more diverse and complex than what can be captured by a PoI.

  Secondly, if Duff’s PoI(s) is supposed to be applicable in several different contexts (i.e., pre-trial inquest, trial, execution of punishment, restoration back to civic life, etc.), then it is likely that we have to connect different deep structure values with different institutions on the surface level in order to create a principle. Whatever reasons that are being synthesized in a principle, we can almost certainly assume that they will vary depending on what institution that they are supposed to rationalize. Henceforth, we will have different principles applicable in different contexts.

  The disparity in formulation of the principles will put a strain on the attempts to describe them as valid, at least if we are looking for a single principle. The distinct formulation might be recognized as having institutional support and substantive significance to a different degree depending on which institution of law is concerned.

  It is likely that any principle derived from a practice of a PoI would be equally diverse. The substantive worth of a PoI when elaborating the institution of a criminal trial would most likely be considerable. The human rights on a deep structure level help to explain and justify many of the procedural rules engaged in practice. On the other hand: if we look at the practice of enforcing a sentence, the principle would probably have little institutional support and a disputable significance when it comes to explaining this particular practice. Keeping a prisoner incarcerated whilst rehabilitating him into becoming a better person has very little to do with a presumption of innocence. Duff is most certainly correct in his assumption that we are working with different principles. Consequently, we encounter problems if we try to identify the broader version of PoI as a normative deep structure principle with explanatory force.

  2.4 Conclusions

  Engaging the metaphor of a multi-layered structure into the analysis has shown that norms can be said to exist simultaneously on different levels of a normative system. A rule or a guideline can thus have different normative standing and different meaning. E.g., a principle can have strong normative force as a mediating principle on a deep structure level even though its reflections on the surface level sometimes tend to be norms that appear rather fragmented, weak or nebulous. The important conclusion of all this is that if Duff’s PoI exists, then it is probable that a rule of this sort is rather weak on the surface level. A norm of this kind would most likely lack clear conditions of application as well as a described norm act to follow. Its stronger force ought to be found on a deeper level; the PoI is thus the outgrowth of a stronger principle.

  This would explain why a specific definition or formulation of a procedural rule like PoI seems to have little meaning outside the proper scope of application – i.e., the trial – but at the same time might have a fundamental significance when it comes to creating reasons for how we ought to treat persons who are subjected to the actual application of criminal law in a broad sense (including all stages from pre-trial investigation, trial and execution of punishment). This illustrates one of the problems with Duff’s PoI. As a surface level norm the broader version of PoI lacks institutional support. Perceived as a deep structure norm, it has obvious difficulties in linking the deep structure values with a specific practice. The surface level in Duff’s version is much too turbulent.

  Whether Duff’s PoI ought to be called a principle remains an open question. If (some version of) a PoI can be said to express the underlying purposes of detailed rules and specific institutions, then it could be viewed as a principle that fulfils the same tasks as other deep structure norms. I believe that this is what Duff is trying to show us with his comprehensive version of the presumption. On the other hand: if we focus exclusively on the narrow version of a PoI, then its core message more resembles some sort of a practical rule that provides guidance as to how to achieve a certain objective. A rule that tells us what to do as an empirical generalization can properly be labelled a rule of thumb or a technical instruction.²¹xSee, e.g., William Twining & David Miers, How to Do Things with Rules: A Primer of Interpretation, 4th ed. (London and Edinburg: Butterworths, 1999), 124 and Joseph Raz, Engaging Reasons: On the Theory of Value and Action (Oxford and New York: Oxford University Press, 1999), 227. To assume innocence in cases of uncertainty proscribes a prima facie direction of what attitude we should take when we engage criminal law. Thus, a PoI provides reasons that are reflections of values, but the norm act does not really connect this practice with coherent and desirable deep structure values. It is rather an expression of what follows from a principle than a principle itself.

• 3 Innocence

  3.1 Innocence Lost …

  Whether the suggested critique of Duff’s PoI is considered to be relevant or not ultimately depends on what we put into the proscribed norm act – ‘presume innocence’ – and how we define ‘innocence.’

  The notion of being innocent could imply many different normative assertions. It could refer to a complete exoneration and acquittal, e.g., in a situation where an allegedly responsible, accused or charged, person (A) is cleared from all suspicions, guilt or responsibility. However, a statement of innocence does not necessarily have to be so absolute. It could convey a more cautious message that A is (or could be) partly responsible for something (X), which represents harm (H), but for which there are other reasons not to hold A responsible. Compare the following examples:

  • A did not do X which caused H (rather, it was B who was the perpetrator).

  • A did do X, which caused H, but he does not deserve censure since he lacked guilt, capacity, etc.

  • A did do X, but H cannot reasonably be said to have been caused by X since H is a too remote consequence.

  • There is strong and reliable evidence that A did do X (which caused H), but due to the fact that there is a small shred of doubt of A’s guilt, he will not be held formally liable for H.

  • A did do X and is to blame for causing H. We shall therefore punish him accordingly.

  • A did X and thereby caused H1 and for this he is punished. Nevertheless, we ought not to hold him responsible for any harms (H2) or consequences that have not yet materialized from H1.

  Duff’s broader version of PoI does not imply any specified definition of the adjective ‘innocence.’ This is partly because the presumption is supposed to be applicable in many different contexts.

  With an orthodox version of the concept of innocent, it could be understood as an ideal for a humane attitude. Innocence then represents something pure, wholesome, fulfilling, natural, and individualistic. Surely, it is not in such romantic terms that we think of ‘innocence,’ and it is not in this manner that a PoI – in its broad or narrow sense – is perceived in legal sources.

  The rather few references to a PoI that can be found in Swedish legal doctrine refer to some technical procedural rights that are linked to each other, e.g., the right to an impartial tribunal, the right to objective rulings in cases of acquittal, a privilege against self-incrimination and other rights securing a fair trial. None of these (other) rules expressly state that someone is to be treated as innocent. It has more to do with controlling the state and its officials in order to keep high epistemic standards regarding good reasons for any measures taken. High standards of this sort minimize the risk of errors created by prejudice, corruption or incompetence. In no way does this presuppose that the suspect, accused or convicted person is presumed to be innocent.

  In my view a presumption of innocence can never be understood as a normative claim to treat anyone subjected to the criminal law system as being innocent in its pure or absolute sense. It would actually be rather absurd if defendants were brought to trial if judges – and other officials – were bound by a rule of obligation forcing them to believe that the defendants are innocent. It would be even more bizarre if we accepted an institution such as a prison if every official involved in the execution of punishment were supposed to act as if the convicted prisoners actually were innocent (ex post). The notion of innocence must therefore have a rather different meaning. Thus, the whole formulation of the PoI – as it is found in ECHR Article 6 – is misconceived. No one is seriously assuming anyone to be innocent. On the contrary: there must be rather good reasons to believe that a suspect or an accused is guilty. Otherwise the police or a prosecutor would be committing an offence merely by initiating legal proceedings.²²xCf. Swedish Criminal Code, ch. 15, s. 6 (false or unjustified prosecution) and ch. 15, s. 7 (false or unjustified accusation). The point of assuming innocence is to minimize the risk that defendants in dubio draw the short straw. I.e., it is a thumb-rule on how we are to deal with the uncertainty that always exists in decision-making: the defendant gets the benefit of the doubt.

  Questions of guilt or responsibility are thus always a matter of degree. The meaning of these concepts also varies depending on the context. A PoI is never created beforehand and then placed in a certain context. First the institutions are invented (trial, pre-trail investigation, imprisonment, parole, probation, etc.); only then do we create rules for how to engage and assess reasons within such a system.

  Thus, it is normally useful to work with different meta-rules that create standards for a practice that is thought to be justified and desirable (e.g., who is supposed to present justification or other reasons for what, to what extent something must be proven, how do we handle uncertainty and what kind of language we ought to use within a certain institution, etc.). When we review the different institutions it becomes apparent that these were not created to establish anyone’s innocence. In fact, the opposite is the case. The whole idea of many of the institutions rests on a presumption of guilt. Thus, being a defendant is a very special role where people are more or less forced to co-operate. The meta-rules that stem from PoI have to play the role of restraining the use of power with which the different institutions of criminal law have been entrusted.

  The PoI can be viewed as a procedural safeguard that aims at being a reminder that all rules involving the imposition of a burden – detention, incarceration, supervision, disqualification, etc. – ought not to be applied on persons (a suspect, an accused, a convict or an ex-offender) unless there are good reasons for it. It obliges courts and others to take a certain attitude to the defendant and the practice. Thus, we are not necessarily presuming innocence, but rather acknowledging the possibility that we might be wrong (in imposing criminal law sanctions). Any judgments on what can be justified to do within the process of implementing criminal law (in a broad sense) is retrospective or backward-oriented; the cases are evaluated ex post and any measures taken from the trial stage and thereafter are justified because of what the defendant has done.

  The main purpose of pre-trial measures and the trial itself is to determine whether the reasons to invoke criminal law sanctions (in some form) are sufficient. This is true even if the purpose or goal of having the institution itself might be crime prevention or something alike and thus forward-looking. We might argue that we incarcerate arrested or convicted people on essentially preventive grounds. However, the reason that justifies a court (or other competent body) in taking measures in a specific case is always the existence of a certain event in the past for which the culprit is held responsible.

  What has been said above is enough to refute the idea that a PoI, in a criminal law context, has something to do with the concept of innocence, at least in stricto sensu.²³xOne alternative version of the PoI could be that the presumption refers to the reasons for not holding someone responsible, i.e., the many exemptions from criminal liability expressed within the concept of a crime. The PoI would then partly function as a meta-rule clarifying that criminal liability not only requires that basic act requirements are fulfilled, but also that every meta-rule or principle that could exonerate a defendant must be taken into consideration (fault-requirement and absence of excusatory defence, etc.). In my view, this interpretation of a POI will not work either. It would turn the PoI from being a procedural principle into a principle of substantive criminal law. There is nothing to support the supposition that a PoI is to function as a safe-guard that aims at ensuring minimum standards for criminal liability. Strict liability as well as some forms of presumptions (of guilt) are accepted by the court within certain limits. See, e.g., the rulings of the ECtHR in the case Salabiaku v. France (1988). Regardless of which: it is arguable whether it is correct to label someone as innocent in cases where he or she has performed an act that corresponds to a specific offence but later is acquitted because of, e.g., lack of fault-requirement or an excusatory defence. A defendant who is actually ‘innocent’ in the original meaning of the word will naturally benefit from the practice of a PoI, but it is not a precondition for treatment according to this norm that someone actually is purely innocent; nor are we really assuming that everyone who is acquitted from suspicions, allegations or charges really is innocent. In fact: it seems as if the adjective ‘innocent’ has little meaning in a criminal law context. The moment that an official (police officer, prosecutor, judge, etc.) can present sufficient reasons in order to take measures within his field of responsibility, then innocence (at least in its pure form) is lost. A court, tribunal, state official, etc. will decide whether there are sufficient reasons to take further measures, but as long as there are good reasons to try a case it cannot be argued that any rights of an innocent person have been violated. The same ought to be true in cases where the police or a prosecutor present sufficient reasons for simply investigating an alleged crime at a pre-trial stage.

  A pre-trial investigation or a trial is not a search for an objective and material truth. The formulation of the epistemic standard of what will suffice as ‘proved guilty’ reflects what level of risk that we are willing to take when it comes to being wrong.²⁴xWhat risks we are willing to take when it comes to wrongful convictions is also reflected in the standards of what is necessary for an already convicted (innocent) person to get a new trial. Convictions, acquittals, as well as any other measures taken by authorities within the institutions of criminal law, are seldom a matter of assessing the likelihood of innocence.

  3.2 … and Found?

  I began my response to Duff by saying that I am basically sympathetic to his idea, and that I myself have argued for a similar viewpoint. I shall therefore try to offer a better interpretation of Duff’s account than the one I have submitted so far. This will not create any problems since I think it is clear that he never intended to build his normative theory (expressed in his paper) on any narrow version of the presumption in question.

  An alternative reading of the broader PoI would be to assume that this kind of standard/meta-rule aims at regulating the assessment of how good the reasons for holding someone responsible are. The standard could also be viewed as a normative principle within a legal culture, connecting surface level material (e.g., narrow versions of PoI in different contexts) with deep structure values and (moral) goals.

  If so, Duff is not at all discussing PoI as a legally binding rule that obliges courts or other officials to declare people’s innocence. His broader version of PoI is rather to be comprehended as a plea for a morally decent use of coercion and repressive measures within the institutions of criminal law. A PoI might work as a reflection of the reasons there are for treating people according to certain standards. If such an interpretation of Duff’s PoI is feasible, a statement that ‘X is presumed to be innocent of H’ would have the normative meaning ‘X should not reasonably be held responsible for H,’ i.e., ‘there are not sufficient reasons for holding X responsible for H.’ If this interpretation is accepted, then innocence might be restored.

  I will from now on only view the PoI as a normative statement/standard in order to see if this kind of norm can have any function outside its usual context, i.e., as a procedural norm that regulates or summarize certain aspects of the criminal trial.

• 4 Reconciling Duff’s Account with Doctrinal Principles of Criminal Law

  4.1 Summary

  So far, I have refuted the idea that PoI (in any version) really has anything to do with the concept of innocence, at least not in its pure sense. I have furthermore questioned the ontological status of a presumption of this kind. In my view this sort of norm – with regard to its formulation and prescribed norm act – does not meet the requirements of a rule of obligation. Thus, it cannot be said to confer any rights or obligations on the parties. Furthermore: it is arguable whether a PoI fulfils the criteria of a normative principle. On the surface level it does not have the form of a principle. Within the legal discourse, it is difficult to connect certain fundamental values with the broad practice in which Duff suggests that his PoI applies. As explained above in section 2, it is true that much of what the different PoIs represent or reflect can be derived from fundamental values or goals on a deep structure level. The PoI – or a version of this – is thus just a ripple on the surface, echoing some kind of normative principle. However, as a deep structure norm it does not suffice in order to justify or explain the different expressions on the surface. It might be that PoI is but a rule of thumb or a guideline telling judges and other officials how to balance reasons and what to do if there are no reasons (at all).

  4.2 Fairness – Innocence in Another Disguize

  Below I shall try to advocate a similar normative (and/or moral) way of reasoning as Duff, but at the same time downplay the importance of (any version of) a PoI.

  On a general level I think that Duff and I agree on everything essential, including the idea of an instrumental form of civic trust. My reason for suggesting that PoI should be abandoned as the main rationale of the theory discussed in Duff’s paper is that a procedural norm of this kind can easily be misconstrued if it is transferred to a context where it (or they) does not really belong. Fortunately, there are alternative principles that one could adhere to.

  My impression of what Duff is trying to introduce into both criminal and procedural law doctrine is a certain moral attitude. This attitude could best be described as some sort of ambition to treat people fairly or in accordance with good reasons combined with a moral claim of frugality in the use of sanctions (i.e., a principle of humanity or parsimony). Such moral claims are concerned with treating people in general (including criminals) with proper respect for their capacity as moral agents to understand and respond to censure. This moral point of view would be well recognized within doctrinal works and other sources of criminal law.²⁵xSee, e.g., Andrew Ashworth, Principles of Criminal Law, 5th ed. (Oxford and New York: Oxford University Press, 2006), 97ff., Magnus Ulväng, Påföljdskonkurrens – problem och principer (Uppsala: Iustus förlag, 2005), 102ff.; Nils Jareborg, ‘Criminalization as Last Resort (Ultima ratio),’ Ohio State Journal of Criminal Law (2005): 521-34. In any legal system that is governed by a rule of law and honours the principles of legality, guilt, conformity and proportionality, one can detect an ulterior goal to treat citizens (including criminals) according to a principle of fairness. Duff’s ambition therefore is both commendable and likely to fit an account of the normative framework of the criminal law in a Rechtsstaat.

  The basic message of a criminal law, based on Rechtsstaat principles, could be summarized in the following way: When we engage criminal law (or the law of criminal procedure), and take measures of a repressive and censuring kind, we ought to be aware of the considerable risk that we (as assessors of a particular situation) are wrong in our judgments/assessments. Thus, we need to calculate with a wide margin of error. As a direct consequence we ought to be careful not to rely too much on assumptions about what might have occurred in the past (and why it happened), just as we should be cautious when we try to predict the future. Blaming someone for something without the sufficient (good) reasons or punishing a person to a greater extent that he or she deserves is as problematic as punishing a completely innocent person. The censure becomes unintelligible for the person who is being blamed without sufficient reasons and the condemnation is consequently often unfounded. Hence, we sense the kinship between a principle of fair imputation and a PoI. They both seek to avoid condemnation that would be unjustified. If one wishes to explain the deep structure values of a practice of criminal law, a principle of fairness has many advantages compared to a broader version of PoI.

  Firstly, we are undoubtedly talking about a normative principle that would be easy to identify through its institutional support and which expresses the underlying purposes of detailed rules and specific institutions. It is not an empty rule of thumb. Instead it already pervades the whole of criminal law. One needs only to refer to lateral principles of proportionality, guilt, legality, treating equal cases alike, etc. to see the common ground. I will return to this below.

  Secondly, the moral claim of a principle of fairness has the advantage of working both in a retrospective direction (e.g., assessing responsibility and desert) and a future oriented perspective (e.g., criminalizing deeds, taking other preventive measures, e.g., collateral consequences of punishments).

  Thirdly, rationalizing a principle of fairness does not have to explain or utilize the notion of innocence. Pointing out what is unjustified or unfair is sufficient. For this we only need to identify what ought not to be punishable, or what that should be sufficient as a justification or excuse, etc.

  4.3 Elaboration of a Principle of Fairness

  If we compare what has been said above about a principle of fairness with Duff’s plaidoyer for a broader PoI, there are many similarities and an obvious overlap.

  Civic trust as a PoI at a pre-trial stage²⁶xSee Duff, ‘Who Must Presume Whom to be Innocent of What?,’ this issue, 174 ff., 180 f. is a way of expressing a normative, moral base for requiring good reasons to take action against a person whom we a priori ought to trust. The same is true for accepting any collateral consequences of punishment.²⁷xSee ibid., 185ff. Duff’s PoI aims at restraining the use of intrusive measures on vague suspicions as well as different forms of exclusion from civic life based on prejudice. Any practice of this sort must respond to (good) reasons.

  I think Duff is correct on this point. Anyone remotely interested in Rechtsstaat principles or fundamental rights ought to be concerned with mediating principles that help to restrain the use of criminal law. My only objection is that it does not presuppose the use of a PoI, which is most commonly thought of as a retrospective norm (mainly used in a different context). It seems more adequate to analyse these principles as an expression of a moral ambition to restrain the use of criminal law through principles of retributive justice and proportionality. Deviating from such principles always indicates a risk of assigning someone’s burdens unfairly (i.e., according to something other than what is deserved) since we ought to assume that people are responsible agents, capable of moral deliberation. Consequently, we must not presume anything about their future behaviour unless we have good reasons for it. All else would be unfairly imputed on the person in question.

  Such a proposition is valid regardless of whether we focus on pre-trial detention or collateral consequences of punishment. A principle advocating that any measures taken against a person must be justified and deserved is also similar to a principle of prospective proportionality, which stresses that any action taken within a specific area must not go beyond what is necessary to achieve the objectives. A principle of fairness can thus be prospective as well as retrospective.
  If Duff’s only claim is to argue for a reason-responsive use of intrusive measures, then there ought to be nothing controversial in his claim (except the fact that he wants to put the PoI into the equation). Most of what he proposes is already captured within valid normative principles that underpin many criminal law systems built on the Rechtsstaat ideals. There is, however, a possibly important semantic difference in how different measures are justified. With a broader PoI, any suspect or defendant is given a normative status that imposes duties on him.²⁸xSee ibid., 175f., 182ff. The same is true for the ex-offenders. In Duff’s rhetoric they are expected to collaborate in order to regain the trust that we otherwise would have in them. If they do not comply, one gets the impression that they might not deserve the restored trust and equal treatment. As I have discussed above in section 2, such effects are likely to occur. As an empirical statement it is thus true that suspects or defendants sometimes must collaborate in order to be able to refute allegations or charges that are put forward against them.
  From a normative point of view, it seems a lot more problematic to put such duties on the suspect or defendant in order for him to be treated fairly. People who are subjected to the criminal law and its institutions are confronted with allegations based on reasons. These are reasons for believing certain things (e.g., that A is guilty, that A is dangerous, that A will interfere with an investigation, that A will flee the country) and reasons for doing things based on our beliefs (e.g., arrest A, impose restrictions on A’s free movement, subject A to supervision, revoke A’s license). If the reasons are strong enough, the state might take any measures that are thought necessary, regardless of whether the person subjected to the events submits to it or not. Obliging defendants and others to collaborate in order to be ‘fully trustworthy’ citizens has a different overtone, which is associated with welfare state policies. This could pose a problem to Duff’s PoI even though much of what it entails is fully compatible with ‘my’ proposed Rechtsstaat principle of fairness. My point here is that there is a big difference between social engineering and institutional consequences. The first approach is rather dangerous and entails the risk that fair treatment is only conveyed to the ones loyal to the state, but not to others. Being a defendant (suspect, convict or ex-offender) is a very special situation. Co-operation is much or less forced upon the person. By this I am not suggesting that Duff is advocating a communitarian welfare state version of PoI. My only reservation is that a duty to collaborate with the state or its officials ought never to be part of the normative prerequisites of a theory of justice or fairness.

• 5 Final Remarks and Reservations

  Everything that has been said so far about Duff’s account of a morally decent practice of pre-trial measures, convictions and execution of punishment is compatible with the fundamental principles of criminal law within a Rechtsstaat. Consequently I am convinced that Duff is on to something very important when it comes to finding common traits or characteristics of a criminal law based on the rule of law.

  The main linguistic and conceptual difference between my arguments for a principled criminal law based on Rechtsstaat values and Duff’s PoI is that I do not have to engage the concept of innocence. According to the expressed critique above I believe this to be an advantage. In my view it is not necessary to engage the notion of innocence in order to create safeguards against an unfair use of criminal law measures. In the same way it is not fully appropriate to approach the problems from a strictly procedural point of view. Personally I think that we have everything to gain, and nothing to lose, in recognizing the kinship between procedural rules/principles of fairness (e.g., captured in a PoI) and criminal law principles that strive towards ensuring that any punishment or censure is fairly attributed.

  I am not pretending that shifting the perspective from a PoI to doctrinal principles of criminal law would solve every problem when we try to find normative standards that could help to explain the relationship between deep structure values and ripples on the surface level in the form of presumptions, rules, opinions, concepts, etc. I am certainly not denying that a principle of fairness can be equally problematic as a presumption of innocence, being as it is too broad, general and unspecific and thus more or less vacuous. It could also be reproached for being too diverse. Certainly, we must – just like Duff’s broader version of a PoI – assume that there are (many) different principles (of fairness). Normative principles in law, which have a strong moral character, often suffer from such deficiencies. However, such shortfall could (at least partially) be mended by specifying how fairness is to be understood in particular situations. In this sense I think that the concepts of fairness and/or justification are less problematic than the notion of innocence and its implications.

Noten

• 1 See, e.g., A.P. Simester & G.R. Sullivan, Criminal Law: Theory and Doctrine, 3rd ed. (Oxford and Portland: Hart publishing, 2007), 51ff.; Asbjørn Strandbakken, ‘In dubio pro reo,’ in: Uskyldspresumsjonen (Bergen; Fagbokforlaget, 2003), ch. 1; Karol Nowak, Oskyldighetspresumtionen (Stockholm; Norstedt Juridik, 2003), ch. 2. See also UN Universal Declaration of Human Rights Art. 11.

• 2 The Latin maxim in dubio pro reo, which often is said to capture PoI, is not codified in Swedish law. The maxim is nevertheless valid as a legal principle and can thus be invoked in legal reasoning, etc.

• 3 Similar ideas have to some extent been put forward by others, cf. Patrick Tomlin, ‘Extending the Golden Thread? Criminalisation and the Presumption of Innocence,’ The Journal of Political Philosophy (2012): 44-66 and Magnus Ulväng, ‘Criminal and procedural fairness – Some challenges to the presumption of innocence’, Criminal Law and Philosophy (2013). These are, however, not as comprehensive as Duff’s normative account.

• 4 Cf. Andrew Ashworth & Mike Redmayne, The Criminal Process, 3rd ed. (Oxford and New York: Oxford University Press, 2005), ch. 10.

• 5 Heike Jung & Kathrin Nitschmann, ‘Das Bekenntnis zum Schuldbekenntnis – zur Einführung des plaider coupable im französischen Strafprozess,’ ZStW, vol. 116, no. 3 (2004): 785-96.

• 6 Normative rules are here contrasted to causal rules or ‘rules’ concerning ascertainable social regularities.

• 7 See, e.g., Georg Henrik von Wright, Norm and Action. A logical enquiry (London: Routhledge and Kegan Paul, 1963), 70ff., Magnus Ulväng, ‘Straffrätt och principer,’ in Katedralen – Tre texter om straffrätt, ed. Petter Asp et al. (Uppsala: Iustus förlag, 2009), and Nils Jareborg, ‘A Lecture on Principles,’ in Festschrift für Heike Jung, eds. Müller-Dietz et al. (Baden-Baden; Nomos, 2007), 32.

• 8 A statute (addressed to the judge) saying ‘A person who takes the life of another shall be sentenced for murder to imprisonment for ten years or for life’ can easily be transformed into a rule (aimed at people in general) stating that ‘you shall not take the life of another.’ Cf. the Swedish Penal Code, ch. 3 s. 1.

• 9 A judge who decides to not sentence a defendant who has intentionally taken the life of another to imprisonment will certainly create an exception to the rule ‘you shall not take the life of another.’ Any justification or excuse in criminal law doctrine displays such exceptions that have helped amend the rule and its scope of application.

• 10 Norms of this sort are categorical in the sense that they – in a specific context and for the relevant addresses – always, and unconditionally, produce good reasons for doing something. This is not to say that there cannot be other norms (principles, etc.), equally categorical, which can override the first norm.

• 11 See, e.g., Oxford Dictionary: Presumption: ‘The action of taking for granted or presuming something; assumption, expectation, supposition; an instance of this; a belief based on available evidence.’

• 12 A PoI does not exist in Swedish law in the form of a (single) rule. The PoI expressed in ECHR Art. 6(2) could nevertheless easily be re-interpreted as a rule of obligation applicable in Swedish courts: ‘If a judge tries a case, then he shall presume the person charged as innocent until otherwise is proven.’ There is nothing controversial in trying to analyse meta-rules, principles, etc. in this positivistic sense. Cf. Hart’s rule of recognition, see H.L.A. Hart, The Concept of Law, 2nd ed. (Oxford and New York: Oxford University Press, 1997), 100ff., 263ff. See also Neil MacCormick, H.L.A. Hart (Stanford: Stanford University Press, 1981), 40ff., 48ff. On the other hand: such a rule does not tell us very much and as I concluded above, it cannot be regarded as a dogmatic rule of obligation.

• 13 See also Neil MacCormick & Ota Weinberg, An Institutional Theory of Law. New Approaches to Legal Positivism (Dordrecht: Reidel, 1986), 73.

• 14 The idea of analysing normative hierarchies in terms of a multi-structured system of different levels is not new. See Kaarlo Tuori, Critical Legal Positivism (Hants: Ashgate, 2002). Cf. also Hart, The Concept of Law. Chomsky has introduced similar metaphors in his theory of language.

• 15 This underlying level might need to be supplemented with a deep structure where all values, goals, etc. are stored, but I will not deal with this question here. Cf. Tuori, Critical Legal Positivism, 183ff.

• 16 Tuori, Critical Legal Positivism, 163 (with obvious reference to Foucault).

• 17 MacCormick & Weinberger, An Institutional Theory of Law, 73.

• 18 See also Jareborg, A Lecture on Principles, 365.

• 19 See Tuori, Critical Legal Positivism, 178f. and Hannu Tolonen, ‘Principles and Goals,’ in: Scandinavian Studies in Law (Stockholm: Stockholm Institute for Scandinavian Law, 1991): 275ff. It is difficult to make any sharp distinctions between the requirements for institutional support (or ‘official’ behaviour and acceptance from the internal point of view). The difference between the recognition of rules and principles hence lies foremost in the requirement of substantive significance. Whilst a rule only has to be accepted, a principle must add some normative significant justification to a practice.

• 20 By this I do not suggest that there has to be actual (or literal) references to the principle in question in order to be recognized as valid. The underlying values and goals, which give the principle normative value, must instead be recognized as valid within the normative system. Consequently, the fact that a principle is developed or constructed in legal doctrine does not alone meet with the requirement of institutional support. See also Tuori, Critical Legal Positivism, 179. Nevertheless, as soon as the principle is used or referred to in other sources of law – or constructed from a set of coherent valid deep structure reasons – then a principle can be said to be ‘discovered’ and the necessary conditions are fulfilled.

• 21 See, e.g., William Twining & David Miers, How to Do Things with Rules: A Primer of Interpretation, 4th ed. (London and Edinburg: Butterworths, 1999), 124 and Joseph Raz, Engaging Reasons: On the Theory of Value and Action (Oxford and New York: Oxford University Press, 1999), 227.

• 22 Cf. Swedish Criminal Code, ch. 15, s. 6 (false or unjustified prosecution) and ch. 15, s. 7 (false or unjustified accusation).

• 23 One alternative version of the PoI could be that the presumption refers to the reasons for not holding someone responsible, i.e., the many exemptions from criminal liability expressed within the concept of a crime. The PoI would then partly function as a meta-rule clarifying that criminal liability not only requires that basic act requirements are fulfilled, but also that every meta-rule or principle that could exonerate a defendant must be taken into consideration (fault-requirement and absence of excusatory defence, etc.). In my view, this interpretation of a POI will not work either. It would turn the PoI from being a procedural principle into a principle of substantive criminal law. There is nothing to support the supposition that a PoI is to function as a safe-guard that aims at ensuring minimum standards for criminal liability. Strict liability as well as some forms of presumptions (of guilt) are accepted by the court within certain limits. See, e.g., the rulings of the ECtHR in the case Salabiaku v. France (1988). Regardless of which: it is arguable whether it is correct to label someone as innocent in cases where he or she has performed an act that corresponds to a specific offence but later is acquitted because of, e.g., lack of fault-requirement or an excusatory defence.

• 24 What risks we are willing to take when it comes to wrongful convictions is also reflected in the standards of what is necessary for an already convicted (innocent) person to get a new trial.

• 25 See, e.g., Andrew Ashworth, Principles of Criminal Law, 5th ed. (Oxford and New York: Oxford University Press, 2006), 97ff., Magnus Ulväng, Påföljdskonkurrens – problem och principer (Uppsala: Iustus förlag, 2005), 102ff.; Nils Jareborg, ‘Criminalization as Last Resort (Ultima ratio),’ Ohio State Journal of Criminal Law (2005): 521-34.

• 26 See Duff, ‘Who Must Presume Whom to be Innocent of What?,’ this issue, 174 ff., 180 f.

• 27 See ibid., 185ff.

• 28 See ibid., 175f., 182ff.