Rethinking the relationship between reverse burdens and the presumption of innocence

Criminal lawyers regard burdens of proof placed on the accused with deep suspicion. Recently, this suspicion has spurred an interest in how to reconcile these so-called ‘reverse burdens’ with the rule that it is for the prosecution to prove guilt beyond a reasonable doubt in a criminal trial. Though views on this differ among commentators, all reach their conclusions by reference to the presumption of innocence (PoI). Unfortunately, such analysis frequently falls prey to a serious error. Namely, the existing literature fails to adequately distinguish the thin conception of the PoI (a trial rule) from a thick PoI (a general norm of the criminal law) or ignores the distinction entirely. In either case, failure to appreciate this distinction and attend to its consequences raises significant doubt that existing analyses of reverse burdens are sound. This article addresses this failure and offers a fresh approach to reconciling reverse burdens and the PoI.


Introduction
To date, the typical approach to analysing reverse burdens of proof has been to do so solely in terms of the presumption of innocence (PoI). 1 The rationale typically offered for this is that every reverse burden is prima facie a breach of the PoI because it introduces the possibility that the accused can be convicted despite reasonable doubt as to their guilt. This incompatibility between reverse burdens and the PoI has long troubled academic commentators, which has led to a considerable volume of literature on reverse burdens' relationship with the PoI (Ashworth and Blake, 1996;Dennis, 2005;Hamer, 2007;Picinali, 2014;Roberts, 1995Roberts, , 2002aRoberts, , 2002bRoberts, , 2014Stumer, 2010;Tadros and Tierney, 2004). This article will argue that this approach requires considerable refinement if it is to be of any value in analysing the PoIcompatibility of reverse burdens.
Below, I will defend the claim that existing PoI approaches are deficient because they fail to distinguish between two different meanings of the phrase 'presumption of innocence'. Specifically, it will be argued that the distinction between 'thick' and 'thin' conceptions of the PoI has received inadequate attention in existing case law and literature on reverse burdens in England & Wales and the United States. 2 As a result, proponents of existing approaches to reverse burdens invariably resort to one of two undesirable and untenable propositions: either they reduce the entire PoI to its thin form, or they merge the thick and thin PoI into one entity.
The former is unhelpful for analysing reverse burdens because it presents at best a partial explanation of how reverse burdens and the PoI can be reconciled. Meanwhile, any insights gained from the latter approach will always be suspect, as they fail to take account of the analytical distinction between the thick and thin PoI. All of this should worry anyone who wants to understand how reverse burdens can be used in a way which is compatible with the PoI. This suggests that what is needed is a fresh approach to the whole issue.
This article will remedy the deficiencies described above and offer a new approach to determining how to allocate a persuasive burden to the accused in a way which is consonant with the PoI. To do so, I will first explain and build upon the distinction between thick and thin PoI and argue that a more complete account of reverse burdens must engage with the relatively abstract values of the thick PoI without merely reducing to the thin PoI. This can be done, I will argue, by distilling more concrete 'guiding principles' from the values of the thick PoI. I will then take on this task, drawing on a range of Anglo-American scholarship and case law. The result will be a set of three guiding principles which can help lawmakers who are tasked with deciding whether or not a reverse burden can be imposed in a way which is PoI-compatible.
To accomplish the goals set out above, this paper will proceed in six further sections. The first section will set out in greater detail the reasoning behind rethinking how the PoI is used to analyse reverse burdens. In this section, I will argue that the distinction between thin and thick forms of the PoI has not been sufficiently appreciated in existing analyses of reverse burdens and then flesh out the nature of this distinction and its impact. Next, I will give an overview of a new approach to reverse burdens, one which takes account of a thicker notion of the PoI but derives from it specific principles. In the following three sections, I will then describe and explain three 'guiding principles' for reverse burden compatibility, based on case law and legal scholarship from Anglo-American law. Finally, I will conclude the article and offer a vision of how the guiding principles might be operationalised.

The need to rethink the relationship between the PoI and reverse burdens
The nature of the relationship between the PoI and reverse burdens is a contentious subject among criminal lawyers. Some see the PoI as prohibiting reverse burdens only with respect to the formal elements of an offence, with anything labelled as a defence being beyond the scope of the presumption's 2. The jurisdictions of the US and England & Wales have been chosen because they have the best developed and most accessible literature on reverse burdens and the PoI, and because they are indicative of the broader issues faced by common law systems trying to reconcile reverse burdens and the PoI.
protection (Picinali, 2017;Roberts, 1995Roberts, , 2002aSchwikkard, 1998). Conversely, others interpret the PoI as prohibiting reverse burdens outright, on the grounds that they allow for an accused to be convicted in spite of reasonable doubt as to guilt. 3 Between these two positions, a series of intermediate views also exists, each with different interpretations of how reverse burdens can be compatible with the PoI. 4 Theorising on reverse burdens is therefore fairly well-trodden ground in criminal law scholarship. Unfortunately, however, the prevailing approaches to reverse burden compatibility share an important defect: none has paid sufficient attention to the subtle but crucial distinction between two different meanings to the term 'presumption of innocence', each of which has different implications for reverse burdens.
In scholarship on reverse burdens and the PoI, there exist already two prominent ways of categorising different conceptions of the PoI. For reasons which will shortly be made clear, I eschew both of the existing means of categorising the PoI here. The first approach is to categorise different ways of understanding the PoI as either 'procedural' or 'substantive' (Lippke, 2016: 20-21;Picinali, 2014: 244-248;Stumer, 2010: ch. 3). On this account, the former understanding of the PoI is that it is purely 'procedural in character' and 'concerns only the proof of facts at trial' (Lippke, 2016: 20;Picinali, 2014: 245). Alternatively, the latter type of PoI is more robust and turns on a substantive theory of criminalisation or punishment: anything that leads to the conviction of someone for conduct that ought not to be criminalised or punished is a violation of the PoI on this account (Picinali, 2014: 245). As such, a substantive PoI entails broad normative claims which are centred on some thicker notion of what conduct is and is not justly criminalisable. 5 Crucially, as Picinali has observed, one apparently cannot know whether the imposition of a reverse burden breaches the PoI, without first knowing whether the PoI is substantive or procedural (Picinali, 2014(Picinali, : 244-245, 2020. This claim is important and will be revisited shortly. A second approach, sometimes used alongside or instead of the procedural/substantive labels, is to speak of the PoI being either 'narrow' or 'wide'. 6 Depending on how these terms are used, they sometimes correspond to roughly the same distinction as 'procedural' and 'substantive'. On a narrow account of the PoI, the presumption is 'more procedural', and concerned 'primarily with the duties of state actors during the trial and pre-trial process' (Farmer, 2018: 69). By contrast, a wide PoI would be a substantive principle with 'implications for the definition of criminal wrongs' (Farmer, 2018: 68;citing Husak, 2008: 92-103). As is clear from this brief discussion, there appears to be considerable overlap between the two approaches. Yet the existence of these two sets of labels, has in my view brought little additional clarity to debates about the rationale and scope for the PoI, or to the contemporary discussions on the use of reverse burdens.
In fact, one consequence of the increasingly jargon-filled debate about the PoI and its limits has been a profound sense of confusion when discussing the application of the PoI to reverse burdens. 7 To see this confusion in action, one need only revisit Picinali's claim that whether or not a reverse burden is PoIcompatible cannot be known without first committing to either a substantive or proceduralist interpretation of the PoI (Picinali, 2014: 244-245). Picinali reaches this conclusion by reference to a hypothetical example, which is worth reproducing in full here: 3. This is the position in Canada, where reverse burdens are virtually banned. See R v Whyte [1988] 2 SCR 3 (Supreme Court of Canada). The view has also been endorsed in England and Wales by the Law Commission's predecessor: Criminal Law Revision Committee, Evidence (General), (Cmnd 4991, 1972), para. 140, and by Williams (1988). 4. Two examples are Stumer, who argues that the compatibility of a reverse burden depends on whether there is an actual risk of wrongful conviction, and Tadros and Tierney, who would prohibit any reverse burden which effectively criminalises behaviour which was not originally intended to be criminalised (Stumer, 2010;Tadros and Tierney, 2004). 5. A prime example of a so-called 'substantivist' interpretation of the PoI can be found in Tadros and Tierney (2004). 6. For examples of authors who prefer the terminology of wide vs narrow, see Ashworth (2006); Duff (2013);Farmer (2018, sec. 4). 7. For more on the confusion caused by use (and misuse) of jargon surrounding the presumption of innocence generally, see Roberts (2020).
David is charged with possessing an imitation firearm that is readily convertible into a firearm 'so designed or adapted that two or more missiles can be successively discharged without repeated pressure on the trigger'. The offence is constructed by combining section 5(1)(a) of the Firearms Act 1968 with sections 1(1) and 1(2) of the Firearms Act 1982 . . .
However, s. 1(5) of the 1982 Act states that 'it shall be a defence for the accused to show that he did not know and had no reason to suspect that the imitation firearm was so constructed or adapted as to be readily convertible into a firearm' . . .
Does the provision amount to a violation of the presumption of innocence? (Picinali, 2014: 244-245 (footnotes omitted).) After setting up this hypothetical, Picinali moves on to claim that there is no settled answer to this question, 'the answer being a function of how the presumption of innocence is understood' (Picinali, 2014: 245). Yet this seems to miss the point. If the answer is a function of the PoI, and the PoI is either procedural or substantive, then we must commit to one or the other understanding in the abstract before any progress on reverse burdens can be made.
This leaves us in the unenviable position of choosing to understand the PoI solely as a trial rule about usual allocation of the burden of proof (procedural or narrow PoI) or as a catholic theory of punishment, criminalisation and citizenship (substantive or wide PoI). But it is not obvious why it must be one or the other, nor is it clear how one could plausibly endorse either conception of the PoI without some recognition of the other. It is perhaps telling in this respect that even the most ardent advocate of a minimalist, purely procedural PoI, Paul Roberts, also states that 'properly conceived . . . the presumption of innocence is a complex doctrine of political morality, from which Woolmington [and] other important procedural rules and principles are derived' (Roberts, 2014: 318). Perhaps with this in mind, Roberts has recently advocated for greater methodological pluralism in discussions of the PoI and cautioned against conducting a 'tournament among rival conceptions' of the presumption, in search of one 'outright winner' (Roberts, 2020: 2). This reveals the truth of the matter, which is that the choice presented by Picinali rests on a false dichotomy.
In fact, the relevant distinction is not between a wide, narrow, procedural or substantive PoI, but simply between what I call the 'thick' or 'thin' conceptions of the PoI. It is these labels that I will use here. This choice reflects my commitment to the very kind of methodological pluralism Roberts has sought to encourage: as my task here involves normative theorising about the proper connection between the PoI and reverse burdens I aim to cast my net as widely as possible, understanding the PoI as both a trial rule and a general norm of the criminal law. Because the existing labels mentioned above do not neatly track this distinction, and because they seem to be set up as mutually exclusive terms, I choose instead to adopt the labels of 'thick' and 'thin' PoI.
This choice reflects my contention that these two ways of understanding the PoI are not wholly different concepts. Rather, the distinction is relevant in the analytical sense, and the analytical task at hand is to understand how to allocate the burden of proof in a way which is PoI-compatible. As such, I contend that both the thick and thin conception of the PoI are relevant to reverse burdens. I am conscious that this simply risks adding more jargon to an already jargon-laden area of law. But I believe that the terminology of thick and thin is clearer than existing labels, and, importantly, comes without the baggage of the 'substantive vs procedural' or 'wide vs narrow' debates. For these reasons, I eschew the existing terminology in favour of the 'thick' and 'thin' notions of the PoI. With this in mind, I aim now to flesh out the meaning of the thick and thin PoI.
Under the thin PoI, the presumption is nothing more than a trial rule allocating the burden of proof to the prosecution to the standard of beyond reasonable doubt, subject to certain exceptions (such as those laid out in Woolmington). Once courts have found that a statute imposes a persuasive burden on the accused, the thin PoI operates in roughly the following way: First, the court considers whether the burden pertains to an element of the offence or to an affirmative defence. If the reverse burden requires the accused to disprove an element of the offence, it is automatically deemed an unjustifiable breach of the thin PoI on any account. If it pertains to a defence, the approach differs between English and US law. In the US, a reverse persuasive burden on an affirmative defence will never amount to a breach of the PoI under Winship (In re Winship, 397 U. S. 358 (1970) at 364 et seq.). However, in England and Wales, the question becomes one of whether or not imposing a persuasive burden is proportionate in the circumstances, with each case ultimately turning on its own facts (Sheldrake v DPP [2005] 1 A.C. 264).
The conception of the PoI which has historically found the most favour with common lawyers has been the thin PoI. For example, in English law, the canonical reverse burden case is the oft-quoted judgment of Viscount Sankey LC in Woolmington, in which the Lord Chancellor memorably describes the PoI as a 'golden thread' running '[t]hroughout the web of English Criminal Law' (Woolmington v DPP [1935] AC 462 at 481). As such, Viscount Sankey held that the PoI required the prosecution to prove the guilt of the accused beyond reasonable doubt, subject to the exceptions of the insanity defence and any statutory exception (Woolmington at 481-482). On this basis, the House of Lords in Woolmington held that the reverse burden in that case was not compatible with the accused's PoI and quashed the conviction (Woolmington at 481). Similarly, in the United States the Fourteenth Amendment to the US Constitution is interpreted as requiring 'proof beyond reasonable doubt of every fact necessary to constitute the crime' in question (In re Winship at 364). This leaves it open to individual American states (who are primarily responsible for creating and enforcing the criminal law) to define crimes as they see fit, with only those facts classed by the legislature as elements of an offence being constitutionally protected by the PoI. These approaches are paradigmatic examples of what I call the 'thin PoI'.
The difficulty with trying to analyse reverse burdens by reference only to the thin PoI is that, in the case of the US, many reverse burdens (i.e. those attached to defences) are simply beyond its very limited scope. Meanwhile, under English law, the thin PoI seems able to accommodate analysis of reverse burdens on affirmative defences by employing the notion of proportionality. Yet, upon closer examination, it is clear that this simply raises the question: what determines whether or not a reverse burden is proportionate? No satisfactory general answer to this question can be found in the case law (Dennis, 2005: 936). More importantly, the thin PoI is of no help at all in answering such a question, as it gives only a rule of thumb about the default allocation of the burden of proof. Thus, relying solely on the thin PoI leaves the more difficult issue of which reverse burdens are or are not allowed both unresolved and unresolvable.
It is clear that the thin PoI alone cannot be relied upon to explain how to employ reverse burdens in a criminal trial in a way which is PoI-compatible. The US version of the thin PoI gives no guidance as to how to treat reverse burden defences, and the English version proves to raise more questions than it answers. The thin PoI therefore has only a limited role to play. This is not to diminish the clearly important role of the thin PoI, which, by prohibiting reverse burdens on elements of an offence, remains a non-trivial restriction on the structure and operation of the criminal law. Nevertheless, it is insufficient to answer the broader question of how reverse burdens can be used in a way which coheres with the PoI, especially in the context of defences. Faced with this problem, some commentators have instead chosen to couch their theories about reverse burdens in terms of a thick PoI.
The thick PoI can be thought of as a presumption of innocence operating at the level of general principle. Unlike the thin PoI, which is restricted to the trial itself, the thick PoI is much wider, more general, and more abstract. As Duff puts it, the thick PoI is 'an expression of deeper values that should structure the state's dealings with its citizens ' (2013: 171). I contend that these deeper values must include at least the following three elements: (i) a preference for false acquittals over false convictions; 8 8. Sometimes expressed as the 'Blackstone ratio' that it is better for 10 guilty people to go free than for one innocent person to go to prison; see Blackstone (1765: Book IV, ch. 27).
(ii) the aim of accurate fact-finding; and (iii) the principle of liberty or minimum state intervention. 9 Together, these values at least partially ground the thick PoI, which combines them into one general principle requiring that the accused be generally presumed innocent. Couched in these general terms, the thick PoI can then be characterised more precisely as a norm which requires the state to treat all of its citizens as innocent of any criminal wrongdoing. In my terminology, then, the thick PoI is what Roberts is referring to when he says that the presumption is properly understood as 'doctrine of political morality' (Roberts, 2014: 318).
The thick PoI has an important role to play in the overall project of the criminal law. Specifically, it entails normative claims about how the state ought to treat its citizens, or even about how citizens ought to treat one another (Duff, 2012(Duff, , 2013. Such claims are certainly important in articulating the values of the criminal law, and even in helping to structure the criminal process. 10 However, they are also very abstract compared to the relatively specific question of how to allocate the burden of proof in a criminal trial. This level of abstraction means that the thick PoI, though relevant to the reverse burden question, is not suited to answering the question directly. What is needed, then, are more concrete principles which can give a sense of priority to the competing values in the specific context of reverse burdens. This situation leaves anyone seeking to understand how reverse burdens can be used in a way which is coherent with the PoI in a quandary, forced to choose between a thick or thin conception of the PoI before addressing the burden of proof issue at all. Faced with this problem, scholars have tended to take one of two flawed approaches. The first option, preferred by those who tend to favour 'proceduralist' interpretations of the PoI, is to simply embrace the thin PoI as the only relevant form of the presumption linked to the issue of reverse burden compatibility. Paul Roberts and Federico Picinali's scholarship both typify this approach (Roberts, 2002a;Picinali, 2020Picinali, , 2014. Roberts is characteristically forthright about accepting the consequences of adopting a thin or procedural PoI for the purposes of evaluating how to allocate the burden of proof. On his view, the PoI is purely procedural, and requires only that the accused not be made to bear the persuasive burden on any element of an offence, as formally defined by the legislature or common law (Roberts, 2002a(Roberts, , 2002b: sec. IV). If a fact is labelled a defence, Roberts' conception of the PoI would simply not apply, no matter the consequences. He makes this concession in order to preserve the 'conceptual purity of the law of evidence' (Roberts, 2002a: 35). Though I am sympathetic to this motivation, I do not find it persuasive for the task at hand.
In seeking to better understand how reverse burdens can be used PoI-compatibly, as is the goal of this article, it would be a mistake to reduce the PoI to only a (thin) trial rule, as Roberts has done. To see why, it is important to appreciate the fact that Roberts did not have this goal in mind when laying out his approach to reverse burdens. Rather, he was engaged in a kind of doctrinal-conceptualist analysis of English case law, critiquing specific reverse burden cases as they were decided. 11 My task here is different: I aim to answer broader questions about how the burden of proof can be allocated in a way which is PoI-compatible. I do not claim to be providing a definitive theory of what the PoI entails, or ought to entail, under English law, and thus there is no need to restrict my analysis as Roberts did.
Another advocate of the thin PoI, Federico Picinali, has recently sought to 'deflate' the PoI, reducing it only to a rule which requires the state to bear the burden of proving the defendant's guilt (Picinali, 2020). Notably, Picinali does not argue that this burden requires only that the state prove the formal elements of the offence; rather he introduces the notion of 'constitutive facts' which the prosecution must always prove (Picinali, 2020: 14). Any fact which is not constitutive could therefore be the subject of a reverse burden on Picinali's account, just like anything labelled a formal defence is beyond the 9. The latter two principles are adopted from Roberts and Zuckerman's five foundational principles of the law of evidence (2010: 18). 10. For example, Tomlin suggests that we use the (thick) PoI as a benchmark for deciding what conduct should or should not be criminalised (Tomlin, 2013). 11. The term 'doctrinal-conceptualist' is in fact borrowed from Roberts (2011: 384-387). scope of Roberts' PoI. Perhaps wishing to avoid the criticisms associated with Roberts' approach, Picinali explains that these constitutive facts are 'not coextensive' with the formal elements of an offence, but rather refer to 'the facts necessary for someone to be responsible for a given crime ' (2020: 14). This allows him to maintain a commitment to a barebones conception of the thin PoI, which he holds out as the only relevant norm for assessing the PoI-compatibility of reverse burdens.
Picinali's approach, however, still raises the same issue as Roberts': it is of little use to an enquiry into how the burden of proof can be allocated in a broadly PoI-compatible way. It is no answer to say that this depends only on whether or not some fact in question is a 'constitutive' one, because this simply begs the question of which facts should be constitutive. Again, Picinali himself would be likely to simply embrace this criticism, viewing the substantive question of criminalisation as being wholly outwith the scope of the PoI. In light of all this, I reject the view, implicit in the proceduralist approach of Roberts and explicit in that of Picinali, that the thin PoI is the only PoI relevant to reverse burdens.
I am not alone in rejecting the thin PoI as disposing of the compatibility issue entirely. Indeed, several scholars have also rejected this approach, arguing instead for a substantive human right to be presumed innocent. A prominent example of commentators who take such a view can be found in the work of Tadros and Tierney (2004). Unfortunately, this work demonstrates the second error referred to above: collapsing the distinction between thick and thin PoI. Tadros and Tierney present an original conception of the PoI based on the 'gravamen' or purpose of an offence, and apply it to reverse burdens at some length (2004: 416-422). They argue that reverse burdens will infringe the PoI unless the following four conditions are met: i. Based on the available evidence, it is known beyond a reasonable doubt that the accused falls within the gravamen of the offence in question; ii. Even where this is known, it may still be the case that the accused is in fact innocent; iii. Proof on the balance of probabilities that the accused falls within some exception will generate a reasonable doubt that they fall within the gravamen of the offence; iv. Anything less than proof on the balance of probabilities will not create such a reasonable doubt (Tadros and Tierney, 2004: 420).
On this view, the prosecution must prove beyond reasonable doubt not just that the accused satisfies all of the elements of the offence, but also that their conduct falls within the gravamen of the offence. This would effectively reshape the process of criminalisation, leaving it to courts to infer what the gravamen of an offence is and to presume that the accused does not fall within it, rather than simply requiring the prosecution to prove the actual elements of the offence.
What is important here is the method, rather than the substance, of their argument. Specifically, the question must be asked: which PoI justifies these radical conclusions? Surprisingly, Tadros and Tierney state that they are concerned not with what would here be called the thick PoI, but with 'general interpretation of Article 6(2) [of the ECHR] in the context of criminal procedure ' (2004: 405). In other words, they are in fact offering up a very expansive conception of a PoI which is still only a trial rule. This is despite the fact that the European Court of Human Rights (ECtHR) has made it clear in their jurisprudence on the PoI that Article 6(2) is only engaged when someone is charged with a criminal offence (Oztürk v Germany (1984) 6 E.H.R.R. 409). It is difficult to see how this could justify the stringent requirements put forward by Tadros and Tierney, given that these would constrain the process of criminalisation by reference to a PoI which is not engaged until someone has been charged with a crime.
This example is indicative of what happens when the distinction between thick and thin is collapsed. In fact, it seems likely that Tadros and Tierney were in fact appealing to a thicker, more substantive notion of the PoI than that which is encapsulated in Article 6(2) of the ECHR. There is nothing wrong in principle with such an appeal, but it is important that it is made transparently. In this regard, I do agree with Roberts' insistence on conceptual clarity. The thick PoI is relevant to the issue of allocating the burden of proof, but if it is to plausibly ground any claims about how the burden of proof should be allocated this needs to be done explicitly. Otherwise, there is a risk of confusion about the proper ambit of both the thick and thin PoI, which risks undermining the force of both of them.
In light of all the foregoing discussion, I submit that the best option is to construct a framework which will address the PoI compatibility of reverse burdens specifically, whilst also linking back to the thick PoI. This is the strategy which will be pursued below: I will build such a framework based on three guiding principles which have been espoused (if not always adhered to) by courts and by theorists analysing the issue of how to allocate the burden of proof in a PoI-compatible way. The merits of this approach are twofold. First, it avoids the problems described above. Unlike relying on the thin PoI, a new framework allows for a more wide-ranging analysis of the allocation of the burden of proof in the criminal trial. Unlike appealing to the thick PoI in the abstract, this new framework does have a sufficiently close link to reverse burdens, but still makes reference to the thick PoI's deeper values. Second, the approach advocated for here ensures that the conceptual clarity of the PoI (both thick and thin) can be preserved, which is itself a desirable outcome.

Three guiding principles for reverse burdens
The framework set out below is the result of a survey of case law and scholarly literature on burdens of proof, the PoI, and fair trials more generally. The form of this new framework is best thought of as a 'threshold consideration', followed by a series of three general principles. The first hurdle which must be cleared is the thin PoI, which I describe as a threshold consideration because it represents a clear minimum standard which must not be breached. This requires that any reverse burden be attached to a defence, rather than an element of the offence. Next, the first guiding principle provides guidance on how to resolve ambiguity in statutes which allocate some sort of burden to the accused, without indicating whether it is a persuasive or evidential burden. The second guiding principle then addresses the process of rationalising the imposition of a persuasive burden, strictly limiting the role of rationales borrowed from civil law. Finally, the third guiding principle applies specifically to offences which impose a specific type of responsibility on the accused and identifies when such offences can compatibly employ reverse burdens.
There are three guiding principles which have been drawn from Anglo-American case law and scholarship on reverse burdens. Guiding Principle 1 (GP1) is adapted from the English case of Hunt, drawing on the judgment in the Scottish case of Nimmo (Nimmo v Alexander Cowan & Sons Ltd [1968] A.C. 107; R v Hunt [1987] A.C. 352). GP1 can be stated succinctly as follows: Where the law is ambiguous as to the locus of the burden of proof, there is a rebuttable presumption against imposing a reverse persuasive burden. The second guiding principle (GP2) then describes how to rationalise allocating a persuasive burden to the accused. GP2 provides that: In the absence of some independent justification, rationales based on the allocation of the burden of proof in civil law do not apply to reverse burdens in the criminal context. Finally, Guiding Principle 3 (GP3) gives guidance as to the approach to offences which impose a specific type of responsibility on the accused, and can be stated as follows: Imposing a reverse persuasive burden will not infringe the PoI if it attaches to a 'secondary' responsibility which stems from the regulatory duty. To understand how these three principles operate together, we must first examine each individually.

Guiding Principle 1
Where the law is ambiguous as to the locus of the burden of proof, there is a rebuttable presumption against imposing a reverse persuasive burden. At trial, the defendant was charged under s. 5(1) of the MDA 1971 after being found in possession of a powder which was confirmed to contain morphine. However, Hunt made a submission of no case to answer, on the basis that the prosecution had not proved that the morphine in his possession was of the relevant concentration required by the regulations. This submission was rejected by the trial judge (Hunt at 367). Subsequently, the rejection of Hunt's submission was upheld in the Court of Appeal, on the basis that the accused bore the persuasive burden to show that he fell within the exception to 5(1) and that he would not have been able to discharge this burden anyway (R v Hunt [1986] Q.B. 125, 135-136 (Court of Appeal)). It therefore fell to the House of Lords to decide where the burden of proving the concentration of the morphine should ultimately lie: on the prosecution, or on the accused?
For the court in Hunt, the issue surrounding the burden of proof was treated as one of statutory construction. The approach taken by the House of Lords in this case was therefore to look to the mischief at which the Act was aimed and the ease with which one party or the other could discharge the persuasive burden in order to determine where the burden of proof should lie (Hunt at 374 (Lord Griffiths); Hunt at 386 (Lord Ackner)). Lord Griffiths-with whom Lord Keith, Lord Mackay and Lord Ackner agreedthus drew on the judgment in Nimmo, but this time extended the principle, holding that: Parliament can never lightly be taken to have intended to impose an onerous duty on a defendant to prove his innocence . . . and a court should be very slow to draw any such inference from the language of a statute (Hunt at 374).
It is clear from the above passage that the court in Hunt placed a particular emphasis on ensuring that the defendant had some safeguards to stop courts from freely imposing reverse burdens by implication. 12 However, it is not the reasons relied upon in the court's construction of the statute which is of the most interest for this article. Rather, the key principle to extract from Hunt is that a presumption against imposing reverse persuasive burdens could act as a check on the courts' ability to create new and possibly unintended, reverse burdens. It is of note that GP1 and the extract from Hunt above are not actually equivalent. Lord Griffiths in Hunt was emphasising that the difficulty of discharging the burden should be a factor taken into account when construing the statute. Thus, the emphasis was on the onerousness of the burden. By contrast, GP1 is not concerned with the ease of discharging the persuasive burden, but with the incidence of the burden in the first place. 12. It was open to the court to rule that reverse burdens could never be imposed by implication, which would have provided a much more effective safeguard for the accused and pre-empted many of the criticisms which the case later received. However, their Lordships declined to do so; see Roberts (2011: 369-370 (Lord Griffiths)). For critique of Hunt see Mirfield (1988) and Williams (1988), but cf. Birch (1988: n. 7) who defended the judgment.
It is unfortunately the case that legislatures often do not draft legislation with burden of proof issues in mind (for criticism of this tendency, see Criminal Law Revision Committee, 1972). This often leaves the task of determining where the burden of proof should lie with the judiciary, who must construe the statutes in question. As a result, reverse burdens have proliferated in recent decades, with hundreds of statutory offences imposing persuasive burdens on the accused (Ashworth, 2006;Ashworth and Blake, 1996). It is true that, by rejecting the suggestion that reverse persuasive burdens be limited only to those imposed expressly by Parliament, the judgment in Hunt paved the way for this proliferation of reverse burdens. Yet the case also provides a blueprint for how to rein in the number of implied reverse burdens, by creating a presumption against imposing them. This is what GP1 would do.
Moreover, such a rebuttable presumption walks a middle path between the extremes of Hunt (largely unlimited implied reverse burdens) and a flat ban on implied reverse burdens. The main benefit of this approach is that it gives judges flexibility when construing statutes, but also requires that they give reasons if they are to rebut the presumption against encumbering the accused with a reverse burden. Moreover, these reasons would be reviewable on appeal, providing a further check. This is desirable because ultimately reverse persuasive burdens are deviations from the norm that the prosecution should bear the burden of proof in a criminal trial. As such, they require some justification if they are to be imposed. In cases of express reverse burdens there is at least prima facie justification, in that it was intentional on the part of the legislature. In implied reverse burden cases there is no such justification, even in a prima facie sense, and so it is preferable to require courts to provide one. The presumption in GP1 accomplishes this and could operate in much the same way as the presumption of mens rea in English law.
The power of GP1 to act as a check on implied reverse burdens will be determined, at least in part, by what is required to rebut the presumption against imposing a persuasive burden. Here, the analogy to the presumption of mens rea may be helpful. Under this rule, it is generally presumed that criminal statutes are to be construed as if they require mens rea, unless Parliament has indicated otherwise (Sweet v Parsley [1970] A.C. 132). This can only be rebutted by an Act of Parliament expressly stating that the offence does not require mens rea, or if this is a necessary implication of other parts of the Act (B (A Child) v DPP [2000] 2 A.C. 428). The presumption in GP1 would function similarly, requiring either express words or necessary implication for a court to read an ambiguous statute as imposing a persuasive burden on the accused. GP1 can therefore act as guidance to both judges and legislators.
For judges, the principle provides a directly applicable and clear standard against which they can assess ambiguous statutes: if there is no express provision regarding the type of burden, and no necessary implication that it should be persuasive, then it will be construed as evidential only. For legislators, GP1 is still indirectly applicable, in the sense that although it is not directly aimed at them, its existence will put legislators on notice that they must be explicit if they wish to impose a persuasive burden on the accused. Again, this is similar to the presumption of mens rea, in that it is aimed at courts but still affects the content of the substantive criminal law.
The result of this is that GP1 would make the law more consistent and transparent, both by aiding statutory construction and by requiring Parliament to be more explicit in its legislation. 13 This is beneficial in terms of increasing legal certainty-an end in itself-but is also firmly rooted in a value which is closely connected to the thick PoI: the preference for false acquittals.
GP1 will lead to fewer wrongful convictions because it will make it more difficult for the accused to be saddled with a persuasive burden. Given that reverse persuasive burdens increase the probability of wrongful conviction, it follows that fewer reverse persuasive burdens would decrease the number of wrongful convictions. 14 GP1 therefore reduces the risk of this (highly undesirable) situation 13. This could also lead to increased political scrutiny of reverse persuasive burdens in the course of drafting and debating legislation, which would be very welcome considering the current state of the law. 14. Reverse persuasive burdens have this effect because they make conviction easier (by relieving the Crown of some of its burden), which, ceteris paribus, leads to more convictions. More convictions in general leads to a higher number of erroneous convictions, but the inverse is also true. For more, see Hamer (2014).
materialising, by requiring a more stringent test for when an ambiguous statute can be interpreted in a way which saddles the accused with a probative burden. Moreover, GP1 provides a definitive statement of which of the three values of the thick PoI takes priority in a case of statutory ambiguity: namely, a presumption against imposing a persuasive burden prioritises the preference for false acquittals over the commitment to accurate fact-finding. But rather than doing this in the abstract, and potentially undermining a general commitment to accurate factfinding, GP1 is limited only to cases of ambiguity, due to the increased risk of false conviction being amplified by the fact that this conviction may have come about through nothing more than misunderstanding of Parliamentary intent. GP1 reduces the risk of this (highly undesirable) situations materialising, by requiring a more stringent test for when an ambiguous statute can be interpreted in a way which saddles the accused with a probative burden.

Guiding Principle 2
In the absence of some independent justification, rationales based on the allocation of the burden of proof in civil law do not apply to reverse burdens in the criminal context. GP2 relates to a different aspect of reverse burdens' role in the criminal process. Specifically, GP2 imposes a constraint on the imposition of a reverse persuasive burden in the criminal law, restricting the role of analogies to civil law. More precisely, GP2 generally precludes reasoning by analogy to principles of civil law in order to justify imposing a reverse burden, unless there is some sort of contextspecific justification for making such an analogy. The reason for this, as will be discussed below, is that there is no principled reason for making such analogies. Moreover, there is in fact good reason not to assume that civil law and criminal law should necessarily have identical rules regarding the burden of proof. Namely, criminal law and civil law are distinct areas of the law, each with different goals, functions and consequences. The significance of these differences means that basing the allocation of the burden of proof in the criminal law on the parallel rules of the civil law is irrational, and therefore undesirable, unless there is some specific justification for doing so in certain contexts.
In order to flesh out the reasoning underpinning GP2, I propose to examine a prominent example of an analogy to civil law, and the pitfalls that have accompanied it in English criminal law. The example I have in mind is found in s. 101 of the Magistrates' Courts Act 1980 ('s. 101'). Section 101 provides that: Where the defendant . . . relies for his defence on any exception, exemption, proviso, excuse or qualification, whether or not it accompanies the description of the offence or matter of complaint in the enactment creating the offence or on which the complaint is founded, the burden or proving the exception [etc] shall be on him; and this notwithstanding that the information or complaint contains an allegation negativing the exception, exemption, proviso, excuse or qualification.
Notably, similar provisions exist in the legal systems of many common law jurisdictions including Scotland, Northern Ireland, the Republic of Ireland, New Zealand and some Australian states (Smith, 1987: 227). As such, it seems clear that the problem is by no means confined to English law, and I will take it to be indicative of a more general phenomenon.
Although s. 101 applies only to summary offences, its influence has since been extended by the House of Lords in Hunt to indictable offences in England and Wales. 15 A brief examination of the pedigree of s. 101, however, reveals that it is an oddity in the criminal law, and one which is rooted in the logic of civil, rather than criminal law. The problem with this analogy to civil law, however, is that it is inappropriate and was made without any justification. Section 101 therefore represents a paradigmatic case of the type 15. In other jurisdictions, such as Scotland, it expressly extends to trials on indictment; see the Criminal Procedure (Scotland) Act 1995, Schedule 3, para 16.
of reasoning which GP2 seeks to preclude. To see why, and therefore to understand what GP2 would require in practice, it is necessary to go back to the beginning of s. 101's legislative history in England & Wales.
What is now known as s. 101 was first enacted in English law as s. 14 of the Summary Jurisdiction Act 1848 (Jervis's Act) (11 & 12 Vict. c. 43). Unlike the modern s. 101, s. 14 of the 1848 Act was lodged in permissive, rather than mandatory terms, relieving the prosecution or complainant in a magistrates' court from the burden of proving a negative proposition contained in an exemption, exception, proviso, etc. 16 This also gave the defendant the opportunity to disprove such a negative proposition, but it did not require such proof from the defendant-in other words, it did not create a reverse burden. However, when s. 14 of the 1848 Act was re-enacted as s. 39(2) of the Summary Jurisdiction Act 1879, the reverse burden was introduced and has remained ever since, despite the rule being repealed and re-enacted several times.
The history and the status of s. 101 with respect to indictable offences was considered at some length by Lord Griffiths in Hunt. According to Lord Griffiths, the original provision in the 1848 Act was Parliament's attempt to state 'the common law rule as to proof that the judges then applied to trials on indictment' (Hunt at 371). In turn, this rule regarding burdens of proof appears to have evolved from 18th-century case law, which established that the burden of pleading was on the prosecution to show that the accused did not fall within an exception to a statutory offence. The classic statement comes from the case of Jarvis, where it was held that if an exception was built into 'the enacting part of a law' (i.e. the same provision as the one which created an offence) then it was up to the plaintiff or prosecutor to plead that the accused did not fall within the exception (Rex v Jarvis (1756) 1 Burrow 148, at 155; 97 E.R. 239 at 243). By contrast, if the exception to a general prohibition was contained in a separate proviso then the accused would bear the burden of pleading this is as defence (Jarvis at 241-243).
In the later 18th century, this rule continued to be applied with respect to the burden of pleading (Jelfs v Ballard (1799)  Yet, by the early 19th century it seems that the courts sporadically chose to extend these rules regarding the burden of pleading, to the allocation of the burden of proof. This is evident in cases like Taylor v Humphries, where a publican was relieved of the burden of proving that he was exempt from an offence of unlawfully selling alcohol before 12.30 on a Sunday because an exception (opening early to serve travellers) was found in the wording of the same provision which created the offence (Taylor v Humphries (1864) 17 C.B. N.S. 539 at 549; 144 E.R. 216 at 220). A subsequent case interpreting the same statute also refused to interpret the rule in Jarvis regarding the burden of production as imposing a burden of proof on the accused, citing the unfairness this would cause to anyone caught by the reverse burden (Davis v Scrace (1868-1869) L.R. 4 C.P. 172 at 175-177).
At this stage there are two important observations to be made about s. 101 and its history. The first is that the modern enactment of s. 101, combined with its extension in Hunt to trials on indictment, is now quite a long way from its origins as a common law rule regarding burdens of pleading. Rather than simply requiring certain parties to specify certain facts in their written pleas to the court, as did the rule in Jarvis, the modern s. 101 imposes a full persuasive burden on the accused and, in doing so, allocates the risk of non-persuasion to them. The result is that a rule which originally sought only to specify who must plead what (with or without proof) has now been extended to justify a broad requirement on defendants to prove their innocence. All of this, it must noted, is based simply on the language of statute and whether a proviso is separate from, or included in, the wording of an offence. Moreover, the fact that all of this and more was rehearsed by Lord Griffiths in his examination of the history of s. 101 (Hunt at 370-374) is a serious disappointment. Unsurprisingly, his reasoning was sharply criticised by contemporary commentators (Mirfield, 1988;Smith, 1987;Williams, 1988). Perhaps the only saving grace of s. 101, then, is that it has been 'studiously ignored' in practice (Smith, 1987: 231).
One might be inclined, then, to ask why and how Lord Griffiths could have made the leap that a rule regarding mere burdens of pleading should apply even more forcefully to the matter of allocating the burden of proof. The answer reveals the second observation to be made here about s. 101, which is that the provision's pedigree, combined with an analogy to the rules of pleading in civil law, were central to Lord Griffiths' reasoning in Hunt. Specifically, Lord Griffiths relied on Lord Wilberforce's judgment in Nimmo-a civil law case-to reach the conclusion that s. 101 did no more than to 'merely state the orthodox principle (common to both the criminal and civil law) that exceptions, etc. are to be set up by those who rely on them' (Nimmo at 109). Yet, as Glanville Williams asked, 'Why should the civil and criminal law be the same on the burden of proof? '(1988: 291) There is no satisfactory answer to this question.
Indeed, the rationale of Lord Griffiths in Hunt and Lord Wilberforce in Nimmo appear to boil down to little more than an unjustified analogy between two very different areas of the law. However, as Williams observed, there are serious reasons to doubt that the rules which apply to burdens of proof in the civil law necessarily extend to the criminal context (1988: 291). The criminal law has different objectives and very different consequences to that of the civil law. Modern criminal law seeks 'to do justice on the facts as they appear at trial' and therefore eschews the complex rules of pleading which are common to the civil law (Williams, 1988: 291). Also, criminal law carries with it much more severe consequences than any civil action: in the UK a criminal trial can end in life imprisonment, and in some other countries it can condemn the defendant to death. This is a sharp contrast from, say, the award of damages for breach of contract in a civil court.
The upshot of all of this is that there is simply no reason to assume that the criminal law ought to follow the civil law when it comes to allocating the burden of proof. Though the discussion of s. 101 provides just one example, it is illustrative of the dangers inherent in such careless analogising. As such, my contention is that if one seeks to rationalise the imposition of a persuasive burden on the accused by analogy to civil law, then this can only be done with some independent justification. For example, if there is some persuasive reason in a particular case to analogise between civil law rules of procedure and the context of the criminal law then GP2 would not preclude this. Instead, it would simply require that this justification be made, rather than drawing baseless parallels between the civil and criminal law. Put simply, GP2 makes it clear that there is no automatic analogy between civil and criminal law when it comes to justifying where the burden of proof will lie. As the example of s. 101 shows, the dangers of importing civil law principles into the criminal law unthinkingly are all too real.
Like the first guiding principle, GP2 is also grounded in the thick PoI. Recall that the thick PoI espouses a general principle of the criminal law, providing that the state should treat all of its citizens as legally innocent. It should also be noted that the thick PoI is context specific; that is to say, the thick PoI has specific implications for the criminal process, reflecting what is at stake in a criminal case. GP2 is effectively a restatement of this idea, but with respect to the process of justifying the imposition of a persuasive burden. In other words, GP2 foregrounds the importance of the distinctiveness of the criminal process, which is glossed over and effectively ignored by relying on analogies to cases like Nimmo. Moreover, GP2 serves as a reminder to legislators tasked with drafting new criminal legislation that they must keep the differences between civil and criminal law firmly in mind. In this sense, it can help to guide the legislator into making a more principled, and reason-driven justification for imposing a persuasive burden, rather than simply imposing reverse burdens indiscriminately or based on convenience. This would further the rationality of the criminal law and also bring it more in line with the values of the thick PoI.

Guiding Principle 3
Imposing a reverse persuasive burden will not infringe the PoI if it attaches to a 'secondary' responsibility which stems from the regulatory duty.
GP3 applies to the imposition of a persuasive burden on the accused with respect to a certain subset of offences. Specifically, GP3 provides that there will be no infringement of the PoI if the accused is required to prove that they have fulfilled some sort of 'secondary prospective responsibility' related to a regulatory duty. 17 This guiding principle draws on two key sources: Duff's theory of strict criminal responsibility and Picinali's justificatory theory of regulatory offences (Duff, 2007, ch. 10;Picinali, 2017). Although nominally focused on different topics, these two theories share an important similarity: both offer an explanation for reverse persuasive burdens in the limited context of offences which impose a secondary responsibility on the accused. For that reason, they will be relied upon as the foundation of GP3. This guiding principle explains why reverse persuasive burdens on what are typically considered regulatory offences are not infringements of the PoI, provided certain conditions are met. However, GP3 is concerned with the imposition of secondary responsibility, rather than with any formal label of an offence as 'regulatory' or otherwise.
The meaning of the phrase 'regulatory offence' is not without its controversy (Norrie, 2014: 104-109). Indeed, there are many possible ways to define what constitutes a regulatory offence, though no single definition seems entirely satisfactory. 18 At its narrowest, the phrase may refer only to offences which arise under a statutory scheme designed to regulate a socially useful activity, as opposed to 'truly' criminal laws which regulate behaviour with no social utility (Ogus, 2014: 29). At the other end of the spectrum, the broadest conception of regulatory offences would include all those which have 'the intention or effect of controlling, ordering, or influencing . . . behaviour' (Lacey, 2004: 147). Between these two extremes there exists a range of possible meanings which could be ascribed to regulatory offences. However, I will not fully adopt any such definition here.
Instead, I will simply label any offence which attracts what Duff calls 'strict prospective criminal responsibility' as a regulatory offence (Duff, 2007: 242). Thus, for the present purposes, the term regulatory offence should be read as meaning 'any offence which imposes secondary criminal responsibility'. On my account, regulatory offences will typically be those that regulate activities which 'involve special risks' above and beyond those of everyday life, 'and against which it is appropriate to demand that [an] agent take special precautions' (Duff, 2007: 243). Because of the special risks attached to such activities, the law justifiably imposes two types of prospective criminal responsibility on those who engage in these activities (Duff, 2007: 242-243). In this context, 'criminal responsibility' is an umbrella term that refers to conduct which the criminal law gives the agent good reason not to do (by criminalising it). 19 Prospective criminal responsibility, then, is a responsibility imposed by the criminal law, on an agent to do (or refrain from doing) something in the future (Duff, 2007: 30-36). 20 Regulatory offences, in the sense in which we have defined them here, impose two types of prospective criminal responsibility on agents who are subject to the criminal law. 21 The first is primary responsibility, i.e. the responsibility to take special precautions when participating in a regulated activity. What is of particular interest here, however, is that regulatory offences also impose secondary 17. Briefly, secondary responsibility refers to a responsibility to show that some other (primary) responsibility has been fulfilled.
This term is borrowed from Duff and will be unpacked below (Duff, 2007: 242). 18. Picinali gives a brief but useful summary of some of the competing ways of defining the term 'regulatory offence' in Picinali (2017: 683-684). 19. As Duff puts it, 'I am responsible for that for which I must answer, and I must answer for that which there was reason for me not to do: I must answer . . . criminally for doing what, according to the criminal law, there was legal reason for me not to do' (Duff, 2007: 22). 20. See Duff (2007: 30-36). 21. The subjects of the criminal law include (at least) all the citizens and guests of the polity in the jurisdiction of the criminal law.
prospective responsibilities on the subjects of the criminal law. This means that, in addition to fulfilling their primary responsibilities, agents must also 'make sure that [they] will be able to show that [they have] discharged' this primary responsibility (Duff, 2007: 242). Often this will take the form of due diligence and record keeping to demonstrate compliance with the relevant regulatory duty. Crucially, failing to fulfil this secondary responsibility may itself be proof that the accused has not discharged their primary responsibility (Duff, 2007: 243). As such, Duff contends that it is justifiable to impose an evidential burden on the accused to show that they have fulfilled their secondary responsibility (2007: 244-245). However, Duff would not extend his argument to justify imposing a persuasive burden on the accused (2007: 245).
Instead, Duff would impose a requirement that the evidence adduced 'suffices to cast reasonable doubt on the charge' in question (2007: 245). Furthermore, he would also add a 'rider' to any such evidential burden, stating that 'normally only proof on at least the balance of probabilities will suffice to create a reasonable doubt ' (2007: 245). The reason for taking this approach seems to be that Duff is reluctant to transfer the risk of non-persuasion from the prosecution to the defence simply to ensure compliance with a secondary responsibility. However, it is difficult to see how this would differ significantly, if at all, in practice from simply imposing a persuasive burden on the accused. 22 Moreover, there is good reason to think that Duff's argument in fact leads to the conclusion that the existence of a regulatory duty actually justifies imposing either an evidential or persuasive burden on the accused. In what remains of this section I aim to show that this is the case, and therefore to arrive at GP3.
To understand why regulatory offences which carry reverse persuasive burdens do not infringe the PoI, it is necessary to make a brief digression to Picinali's theory of regulatory offences. This theory offers an explanation for why regulatory offences are often accompanied by a lack of the procedural safeguards traditionally associated with criminal trials, and are more likely to entail reverse burdens or even strict liability. 23 Picinali asserts that there are two principal grounds on which we might justify denying a defendant a safeguard in the context of regulatory offences. The first of these grounds, which he calls 'deontic contradiction', is the one of interest here. 24 Deontic contradiction essentially refers to the case where a safeguard and a regulatory duty are in direct contradiction with each other. In this scenario, Picinali argues, there is good reason for the duty to take precedence over the safeguard: namely, if an agent has undertaken a regulated activity, they must be taken to have voluntarily accepted the responsibilities which accompany this activity (2017: 686-687). This argument makes intuitive sense and can be used to demonstrate why reverse burdens, whether persuasive or evidential, are justified in the context of regulatory offences.
Applying Picinali's line of reasoning to reverse burdens and the PoI is fairly straightforward. The relevant safeguard, the thin PoI, and the relevant duty, that of an agent to prove they upheld their primary responsibility under a regulation, are flatly contradictory. For example, the owner of a factory has a legal obligation to ensure that they take all reasonable steps to 'ensure . . . the health safety and welfare' of their employees at work (their primary responsibility). 25 However, the owner also has a further duty to keep some record of the steps they have taken to ensure the safety (their secondary responsibility). This is because failure to do so will usually be evidence that they have not discharged their primary responsibility in the first place (Duff, 2007: 242-243). Assuming that they have voluntarily engaged in the regulated activity-here, owning and running a factory-they have implicitly agreed to uphold certain responsibilities while doing so. 26 As such, the factory owner cannot simultaneously enjoy a right which 22. In fairness, Duff appears to concede this point, saying that 'in most cases' where an accused bears an evidential burden 'she would also be able to discharge a persuasive burden' (Duff, 2007: 245). 23. On Picinali's account the PoI is one such procedural safeguard (Picinali, 2017: 683). 24. The second ground, 'unreasonable complaint', will not be discussed as it is not necessary for the present argument, but for an overview see Picinali (2017: 686). 25. Health and Safety at Work Act 1974, s. 2. 26. It is worth emphasising that this agreement is implicit from their taking part in the regulated activity; it is irrelevant whether or not the accused in fact agreed to accept the responsibility. effectively presumes they have discharged their responsibilities, because this would completely undermine their obligation under the law. The contradiction between their duty and the safeguard therefore must be resolved in favour of the duty. This is justified by their voluntary participation in the regulated activity. It is clear that the imposition of some burden on the accused in the context of regulatory offences can be justified for the reasons above. I return now to Duff's conclusions about the limitations of this, namely, that only an evidential burden on the accused can be justified. Duff's rationale for extending his argument only to evidential burdens on the accused appears to stem from a preoccupation with a hypothetical situation in which records, though studiously kept, are destroyed or otherwise unavailable (2007: 245). As a result, he concludes that secondary responsibility should entail a type of unconventional evidential burden. Such a burden would require the accused to adduce evidence 'that suffices to create a reasonable doubt', with the caveat that this can normally only be done by proving on the balance of probabilities that the secondary responsibility was discharged. There are two very odd moves in this argument, both of which betray the fact that imposing a reverse persuasive burden will be justified in the context of a regulatory offence.
First, Duff's new evidential burden is unconventional in the sense that it has a standard of proof required in order for it to be met: the accused must adduce evidence which creates a reasonable doubt. It is unusual to speak of an evidential burden as entailing a standard of proof at all, as nothing is proved even if an evidential burden is discharged. 27 Moreover, if the defendant bears a burden which is nominally evidential, but which entails a standard of proof (however low), the result is that the risk of non-persuasion now lies with them. Despite the nomenclature adopted, it therefore appears that Duff is actually arguing in favour of a reverse persuasive burden with a lower standard of proof. This is borne out by the second unusual aspect of his argument, where Duff claims that this new evidential burden should include a 'rider' stipulating that normally a reasonable doubt can be created only by proof on the balance of probabilities (2007: 245). Again, this simply gives the impression of imposing a persuasive burden on the accused but with an exception for unforeseeable situations such as the destruction of records in a fire.
It is now clear that the better view is that reverse burdens of either kind can be justified with respect to regulatory offences. The reasoning for this, as has been shown above, is roughly as follows. Certain offences stem from regulatory duties which impose both a primary and secondary responsibility on people taking part in regulated activities. When someone voluntarily takes part in such a regulated activity, the duties imposed on them by these responsibilities will often contradict safeguards such as the PoI. As Picinali's theory demonstrated, this deontic contradiction should be resolved in favour of the duty in the context of regulatory offences. Extending this argument to the specific circumstance of reverse burdens, it can be seen that the logical conclusion is that in principle it can be justified to impose either an evidential or a persuasive burden on the accused in cases involving regulatory offences. This is what GP3 provides.
Unlike GP1, the situations to which GP3 applies justify giving priority to accurate fact-finding over the preference for false acquittals. In circumstances where the law imposes secondary responsibility on an individual, it would be very strange, and indeed wholly counterintuitive, to impose no requirement on the accused to prove that they have fulfilled this responsibility. This is clear when one considers the alternative, which is that the prosecution would have to prove that the accused had not fulfilled this responsibility-a difficult task, given the problems inherent to proving a negative proposition. Yet rather than extrapolating from this that there is some general duty on the accused to prove any negative proposition (a conclusion which would generally privilege the preference for false acquittals over the aim of accurate fact-finding), GP3 instead limits this to cases where the failure to discharge on responsibility is itself prima facie evidence of guilt. The upshot of this is that there is a clear justification for prioritising accurate fact-finding, as the risk of a false acquittal is minimised. Importantly, this also means that GP3 coheres with the thick PoI, because the justification is rooted in the concept of secondary responsibility and therefore still holds even if we presume the accused is innocent.

Conclusion
This article set out to develop a new method of assessing the role of reverse burdens in the criminal process, one based on establishing a more concrete link between the thick PoI and reverse burdens. This approach sought to move beyond the thin PoI, which was shown to be too narrow to be helpful by itself, whilst also avoiding the problem of becoming too abstract. As such, I developed three guiding principles above, which connected the specific issue of reverse burdens to the more abstract thick PoI. The target audience for these principles was specified as lawmakers, referring primarily to legislators who are tasked with deciding where to allocate the burden of proof when creating a criminal offence. The principles may also be used by judges in their capacity as lawmakers, such as when construing ambiguous legislation (as in GP1).
The principles espoused above can now be assembled into a kind of framework, which it is hoped will demonstrate one way in which they can be of practical value. Drawing on the discussion above, one way to conceptualise this framework might be as a threshold consideration, followed by a series of principles which lawmakers could consider: Threshold Consideration: The accused cannot be made to bear a persuasive burden on any element of an offence.
(Thin PoI) Guiding Principle 1: Where a statute is ambiguous as to the locus of the burden of proof, there is a rebuttable presumption against imposing a reverse persuasive burden.
Guiding Principle 2: In the absence of some independent justification, rationales based on the allocation of the burden of proof in civil law do not apply to reverse burdens in the criminal context.
Guiding Principle 3: Imposing a reverse persuasive burden will not infringe the PoI if it attaches to a 'secondary' responsibility which stems from the regulatory duty.
The first step is the threshold consideration provided by the thin PoI. Though limited in scope, this provides some important protection to the accused. This consideration must be taken into account by the legislature when drafting an offence, or by a judge when considering, for example, whether a 'defence' which the accused must prove is actually nothing more than the logical equivalent of an element of the offence. 28 In this sense, the thin PoI acts as a floor, rather than a ceiling, limiting which facts may be subject to reverse burdens of proof.
Next, GP1 is a reminder to legislators that any ambiguity on their part will trigger a rebuttable presumption favouring the accused. This principle will therefore provide an incentive to legislate clearly and transparently with regard to the burden of proof. In the case of judges as lawmakers, GP1 will also guide their decision making when construing ambiguous statutory provisions. GP2 then provides a very general rule regarding what reasons lawmakers can legitimately rely on when imposing a reverse persuasive burden. The upshot of GP2 is that the distinctiveness of the criminal process (as compared to the civil law) means that any analogy to civil law rules on the burden of proof should be treated with caution. Specifically, such analogies should only be relied upon to the extent that the civil and criminal law are similar. Finally, GP3 provides a broad exception to the normal application of the PoI to offences which impose secondary responsibility. This will be of use to legislators when considering whether or not to explicitly impose secondary responsibility, and therefore to justify imposing a reverse persuasive burden on that basis. Judges may also rely on GP3 if they are asked to decide where the burden of proof lies with respect to the breach of a regulatory duty.
This new method of evaluating reverse burdens represents progress in unravelling the complex relationship between the individual's right to be presumed innocent and the state's broad power to criminalise behaviour. More to the point, the framework described above does not require a radically revisionist rethink of reverse burdens. Instead, it invites us to think in terms of the PoI's deeper values, and to reflect on how we might use them to chart a way forward for assessing how to allocate the burden of proof in a way which is compatible with the PoI. It is hoped that this will be seen as more appealing than endlessly trying to remould the thin PoI in order to make sense of putting probative burdens on the accused. The end result is more nuanced and complex than many existing approaches to reverse burdens in the literature and in the courts. It is also decidedly less radical than proposals to ban all persuasive burdens on the accused. But perhaps this more nuanced character is no defect. On the contrary, it is hoped that the approach espoused here will bring about a more well-rounded understanding of the role of reverse burdens in the criminal law, and provide a sensible resolution to this seemingly intractable contradiction.