Is the civil higher standard of proof a coherent concept?

Transcription

1 Law, Probability and Risk (2009) 8, Advance Access publication on May 14, 2009 doi: /lpr/mgp011 Is the civil higher standard of proof a coherent concept? ENNIS MCBRIDE Department of Philosophy, School of Humanities, University of Nottingham, Nottingham NG7 2RD, UK The standard of proof used in criminal cases is that of proof beyond reasonable doubt, whereas in civil cases it is that of the balance of probabilities. It is commonly accepted that these standards are coherent ones but a problem arises where criminal conduct has to be proved in civil proceedings: which standard of proof should be used then? The courts have attempted to resolve this problem by rejecting the use of a third, intermediate standard of proof for such cases, while at the same time requiring a higher standard of proof in them. This article argues that consideration of this enhanced standard of proof demonstrates a problem with the concept and questions whether it is a coherent one, with reference to the various interpretations of the higher standard commonly proposed to account for legal decision making. In particular, a critique is presented of two recent decisions of the House of Lords, purporting to settle the issue definitively. Keywords: standards of proof; higher standard of proof; increased scrutiny; degree of belief; degree of probability. In Section 1, I detail some basic concepts in standard probability theory to act as a reference point for the discussion of legal probability which follows 1. Section 2 sets out an historical overview of the formulation and rationale of the higher standard of proof (HSP) as laid down by the House of Lords in Re H (Minors) a standard supposed to lie between the simple balance of probabilities standard used in civil cases and the beyond reasonable doubt (BRD) standard used in criminal cases. In Section 3, I consider Redmayne s argument that confusion has arisen from a failure to discern that there are not one but two such standards of proof. In Section 4, I give examples of various other types of HSP notably that of the anxious scrutiny test, showing how it developed as a seemingly separate concept from the HSP of Re H. In Section 5, I discuss the legal effect of two House of Lords cases decided on 11 June 2008, which purported to lay to rest misunderstandings about the HSP and show how they merge the concepts of anxious scrutiny and HSP. Lastly, in Section 6, I consider whether the HSP, as formulated in Re H and as amended by the recent House of Lords cases, is a coherent standard of proof. I conclude that the current notion of an HSP is legally inconsistent and philosophically incoherent. 1. Introduction axioms and theorems In discussing the HSP in Section 2, I canvass the possibility of representing its tenets in terms of axioms and theorems. The aim of such an axiomatic approach is partly to invoke and invite comparison apxecjm@exmail.nottingham.ac.uk 1 Style note: for the sake of consistency with references to lower courts, I generally refer to the UK s highest judicial tribunal the Appellate Committee of the House of Lords as a court and to the opinions of its members as judgements c The Author [2009]. Published by Oxford University Press. All rights reserved.

2 324 E. McBRIDE with the framework of standard probability theory and partly to reap the benefits de Finetti (1972, p. 67) saw in the process, namely to form a starting point for a complete and rigorous construction and development of the theory and to attain greater precision in discussing it. 1.1 Axioms and theorems The axioms of a probability system are statements that its proponent believes are self-evident and that should be accepted as such by everyone else (see Anderson et al., 2005, p. 253). The four standard probability axioms (based on the so-called Kolmogorov axioms ) are, in ordinary language, the following: (A1) Probabilities are numbers between and including 0 and 1. (A2) The probability of a tautology or, as Anderson et al. (2005, p. 251) put it, a sure event (one certain to happen) is equal to 1. (A3) The probability that one or other of two mutually exclusive events will happen is equal to the sum of their separate probabilities. (A4) The probability that A will happen given that B has happened (the conditional probability of A given B) is equal to the probability that A and B will both happen divided by the probability that B will happen. The theorems of any axiomatic system are just those statements derivable from its axioms by the logical or mathematical rules laid down in the system. Two well-known theorems for present purposes are the following: (T1) the principle of negation: the probability of a statement A is equal to one minus the probability of the negation of that statement and (T2) Bayes theorem, which expresses the probability of A conditional on B in terms of the probability of B conditional on A Axiom, definition or theorem? Despite the references one reads to the standard axioms of probability, there is variation as to the form in which these are presented. Thus, Howson & Urbach (1996) present axioms (A2) (A4) as above, but their first axiom is the probability of an event is greater than or equal to zero and they then derive (A1) above as a theorem. 3 Gillies, on the other hand, presents (A1) and (A2) as one axiom. His second axiom then presents (A3) above as the sum of the probabilities of n mutually exclusive and exhaustive events is equal to 1. He then shows that this is equivalent to (A3) as above. 4 Both sources note in discussing the axioms that Kolmogorov himself presented (A4) as a definition rather than an axiom and discuss the merits of such an approach. Indeed, in his book for student engineers, Johnson (1994, pp. 59, 71) presents (A1) (A3) as above but has (A4) as a definition. For our purposes, it suffices to observe that these differing approaches result in logically equivalent systems of probability just as, e.g. there are many different but compatible systems of propositional calculus. This not only bears out the remark of Wittgenstein (2003, Section 142) that 2 See Howson & Urbach (1996) for the derivations of these theorems. 3 See op. cit. at p See Gillies (2004) at p. 59.

3 IS THE CIVIL HSP A COHERENT CONCEPT? 325 it is not single axioms that strike me as obvious, it is a system in which consequences and premises give one another mutual support but also mirrors the extent to which the choice of axioms in the field of legal probability is a normative one, where it is the judges who write the textbooks. 2. Historical overview of the HSP before Re B and Re Doherty In two judgements given on 11 June 2008 (In Re B and In Re Doherty), differently constituted panels of the House of Lords attempted to lay to rest the question of the meaning of the Re H HSP and in doing so made trenchant criticisms of what they saw as the errors that had crept into forensic reasoning on the subject of the HSP. I will argue in this article that these recent cases do not in fact clarify the law but leave it in a more confused state. First, however, it is necessary to discuss the history of the HSP and of the problems the concept gives rise to and in particular to consider some detailed criticisms Redmayne makes of the concept. What is the HSP? In courts in England and Wales, two standards of proof are used: in criminal cases (tried by a judge and jury or by magistrates), proof must be BRD, but in civil cases (usually tried by a judge alone but in prescribed cases by a judge and jury), proof is on the balance of probabilities (OBOP). A problem arises where, in a civil case, an accusation of criminal conduct must be proved: if a bank clerk is accused of theft from the vault in a criminal case, he can be convicted of the offence of theft only if it is proved BRD that he stole the money in question. However, if the bank sued the clerk for stealing the money in a civil court, the bank would need only to prove its case OBOP in order for the clerk to be found liable for stealing the money and for the bank to win an order for compensation from him. But what standard of proof should the civil court adopt in order to properly recognize the seriousness of the accusation? By way of contrast: if the bank were merely saying their loss of money was due to the clerk s carelessness say, he dropped the key to the safe in the street when he went out to lunch it would sue him for negligence and the standard of proof would uncontroversially be OBOP. The answer of the courts to this question has been to say that they reject the adoption of an intermediate standard of proof between BRD and OBOP and instead to insist that the same standard of proof be applied to all cases in civil courts, but that where allegations of criminal conduct are made, more cogent evidence of such conduct is required. The issue arose directly in Re H & Ors (minors) (1995), a case concerning the so-called threshold criteria in section 31(2) of the Children Act 1989 (which says that a court may make a care order only if it is satisfied that the child concerned is suffering, or is likely to suffer, significant harm 5 ). One of several children alleged sexual abuse against her stepfather. The judge found that while there was a real possibility her account was correct, he did not find it proved OBOP. He therefore found that the threshold criteria were not made out. The Court of Appeal agreed. On appeal to the House of Lords, it was agreed that the conclusion that a child is likely to suffer significant harm could be merely to the standard of a real possibility but the question was, to what standard must a court find the facts needed to form the basis for such a prediction that they occurred OBOPs or merely that there is a real possibility that they took place? Lord Lloyd formulated the issue as follows (paragraph 26): In so far as it is either relevant or necessary in proceedings under the Act to prove an allegation of sexual abuse, is the standard of proof required (i) a standard higher than 5 In practice, and despite the apparent disjunction between is and is likely to, the issue at this stage is what can be proved to have happened in the past: there is no question of proving the probability of a probability as to what will happen in the future.

4 326 E. McBRIDE the ordinary civil standard though falling short of the criminal standard of proof, (ii) the balance of probabilities, but so that the more serious the allegation the more convincing is the evidence needed to tip the balance in respect of it or (iii) the simple balance of probabilities. On this issue, Lord Nicholls (with whom Lords Goff and Mustill agreed) preferred answer (ii). Lord Lloyd and Lord Browne-Wilkinson preferred (iii). The latter two were also in the minority on the main issue, as they considered that in any event, the court did not have to find facts proved OBOP at all in order to find the threshold criteria satisfied. In two passages in his judgement, Lord Nicholls provided an overview of the principles involved in formulating an HSP. He first summarized the position as follows (paragraph 73): The balance of probability standard means that a court is satisfied an event occurred if the court considers that, on the evidence, the occurrence of the event was more likely than not. When assessing the probabilities the court will have in mind as a factor, to whatever extent is appropriate in the particular case, that the more serious the allegation the less likely it is that the event occurred and, hence, the stronger should be the evidence before the court concludes that the allegation is established on the balance of probability. Fraud is usually less likely than negligence. Deliberate physical injury is usually less likely than accidental physical injury. A stepfather is usually less likely to have repeatedly raped and had non-consensual oral sex with his underage stepdaughter than on some occasion to have lost his temper and slapped her. Built into the preponderance of probability standard is a generous degree of flexibility in respect of the seriousness of the allegation. Although the result is much the same, this does not mean that where a serious allegation is in issue the standard of proof required is higher. It means only that the inherent probability or improbability of an event is itself a matter to be taken into account when weighing the probabilities and deciding whether, on balance, the event occurred. The more improbable the event, the stronger must be the evidence that it did occur before, on the balance of probability its occurrence will be established. Ungoed-Thomas J expressed this neatly in In re Dellow s Will Trusts [1964] 1 WLR 451, 455: The more serious the allegation the more cogent is the evidence required to overcome the unlikelihood of what is alleged and thus to prove it. This approach also provides a means by which the balance of probability standard can accommodate one s instinctive feeling that even in civil proceedings a court should be more sure before finding serious allegations proved than when deciding less serious or trivial matters. Lord Nicholls (paragraph 76) went on to explain the reason for not adopting an intermediate standard of proof: The law looks for probability, not certainty. Certainty is seldom attainable. But probability is an unsatisfactorily vague criterion because there are degrees of probability. In establishing principles regarding the standard of proof, therefore, the law seeks to define the degree of probability appropriate for different types of proceedings. Proof beyond reasonable doubt, in whatever form of words expressed, is one standard. Proof

5 IS THE CIVIL HSP A COHERENT CONCEPT? 327 on a preponderance of probability is another, lower standard having the in-built flexibility already mentioned. If the balance of probability standard were departed from, and a third standard were substituted in some civil cases, it would be necessary to identify what the standard is and when it applies. Herein lies a difficulty. If the standard were to be higher than the balance of probability but lower than the criminal standard of proof beyond reasonable doubt, what would it be? The only alternative which suggests itself is that the standard should be commensurate with the gravity of the allegation and the seriousness of the consequences. A formula to this effect has its attraction. But I doubt whether in practice it would add much to the present test in civil cases, and it would risk causing confusion and uncertainty. As at present advised I think it is better to stick to the existing, established law on this subject. I can see no compelling need for a change. How should the key elements of this approach one might call them the axioms and theorems of the HSP best be summarized? As noted in Section 1, it is worth considering this question in some detail as it elucidates the thinking that underpins the judicial concept of probability. The first part of Lord Nicholls argument produces the following scheme: (A1) The OBOP axiom The OBOP standard of proof means that an event (at issue) occurred if and only if a court is satisfied on the evidence before it that the event s occurrence is more likely than not (i.e. that the probability of its occurrence is greater than 50%). (A2) The seriousness axiom The more serious the allegation the less likely it is that the event occurred. (A3) The standards of proof axiom There are degrees of probability but the law seeks to define the degree appropriate for finding facts proven in different types of proceedings by establishing just two standards of proof appropriate for them, namely (i) proof BRD (the higher standard) for criminal proceedings and (ii) proof on a preponderance of probability (the lower standard) for civil proceedings. From (A2) follows: (T1) The cogent evidence theorem The stronger (or more cogent) should be the evidence before the court concludes that a serious allegation is established on the balance of probability (as we shall see, there is an alternative theorem considered which I set out here for convenience) or: (T2) The careful scrutiny theorem The more careful the scrutiny of the evidence required before conclusions are reached on it. From (T1) or (T2) and (A3) follows: (T3) The flexibility theorem Where a serious allegation is in issue in civil proceedings and thus more cogent evidence or more careful scrutiny are called for, the standard of proof required is not a higher one than OBOP but instead represents a degree of flexibility within that standard. 6 Three matters should be noted here: first, whether (T1) in fact follows from (A2) is a moot point but Lord Nicholls links the two by hence so he clearly thinks it does and second, while he qualifies (A2) in the examples he gives ( fraud is usually less likely than negligence ), the qualification is of no account here, as it simply acknowledges the notorious reference class problem: of course, deliberate injury is not less likely than accidental injury in the boxing ring and so on. This just shows 6 To paraphrase a well-known advertising slogan, this makes the standard your flexible friend, the increased credit card.

6 328 E. McBRIDE that it is always possible to restrict the reference class of a probability statement until the statement no longer holds true but this does not affect the truth of the statement as a general statement. Third, note Lord Nicholls appeal to the policy argument that one has an instinctive feeling that one should be more sure before finding serious allegations proved. As will be seen, Baroness Hale did not share this view in Re B, at least as far as children cases are concerned. The three questions to be addressed then are (i) does the set of principles above correctly encapsulate a single HSP; (ii) are the individual principles defensible and (iii) do they, taken as a set, amount to a coherent whole? 3. One higher standard or two? The first question is whether the approach to the HSP adopted by the court in Re H is the only such approach or whether case law shows that different interpretations have been followed. Redmayne (1999) discusses this question in detail and argues that the courts here have in most cases actually adopted two approaches to the HSP. Redmayne (1999, p. 176) calls these the flexible standard and prior probability approaches and finds their source in, respectively, Bater v. Bater and Hornal v. Neuberger Products Ltd. Redmayne distinguishes these approaches on the basis that under the first, the standard of proof varies according to the seriousness of the issues, while under the second, the standard remains fixed but the degree of evidence needed to satisfy it varies because more serious events are said to be less probable. In the terms of the schema given above, Redmayne argues that (T3) the flexibility theorem is actually an axiom of one schema and that it is not present in the other. In the case of Bater, Redmayne (1999, p. 175) quotes in support of his distinction Lord Denning s dictum that many great judges have said that, in proportion as the crime is enormous, so ought the proof to be clear and his references to there being degrees of probability within the OBOP standard. As far as Hornal is concerned, Redmayne (1999, p. 176) relies not on the judgement of Lord Denning (who sat in both cases) but on the dicta of Morris and Hobson LLJ that, respectively, the very elements of gravity become a part of the whole range of circumstances which have to be weighed in the scale when deciding as the balance of probabilities and the balance of probability may be more readily tilted in one case than in another. He finds support for his view in the reference by Ungoed-Thomas J, in Re Dellow s Will Trusts, to more cogent evidence being needed to overcome the likelihood of serious allegations being true, a reference made by the judge following his quotation of part of the judgement of Morris LJ in Hornal. Redmayne (1999, p. 177) states that subsequently both flexible standard and prior probability approaches have been approved by the House of Lords, the first in Blyth v. Blyth and the second in Re H, and he (1999, p. 179) concludes from this that there is confusion over what the HSP actually is and whether there is a distinction between the flexible standard and prior probability approaches. His argument for saying there is such a distinction in the face of judicial dicta to the effect that it is purely academic or semantic is that one approach raises the standard of proof; the other is an observation about the quantum of evidence required to satisfy the standard of proof and the difference that this makes in practice is exemplified by cases such as Re M (A Minor) and Re W (Minors). In the first case, Redmayne says that the Court of Appeal disapproved of a lower court s use of the flexible standard approach, preferring the prior probability approach, and in the second case, the court distinguished between the standard of proof required to show that a father had abused his child, as opposed to that required to show simply that the child had been abused. Redmayne

7 IS THE CIVIL HSP A COHERENT CONCEPT? 329 sees this as a significant difference between the two approaches, as on the prior probability view, once it is proved that a child has been abused, the prior probability that the father was responsible is quite high so easier to prove, whereas on the flexible standard view, an allegation of paternal abuse ratchets up the standard of proof thus making it harder to prove. Is Redmayne right on this first issue, whether there are two distinct approaches to the HSP? The Court of Appeal (before whom counsel cited Redmayne s article) thought not, in R v. Secretary of State for the Home Department ex p A.N. This was a case in which a Mental Health Review Tribunal was faced with competing expert opinions as to whether A.N. was suffering from a psychopathic disorder and could thus continue to be compulsorily detained under the Mental Health Act A.N. contended that the standard of proof in such a serious matter should be akin to that of a criminal matter but the Court of Appeal did not consider that more is required than that the decision is based on cogent evidence which is accepted as correct on the balance of probabilities. It is not necessary, for example, for one body of evidence to have a much higher degree of cogency before it can be accepted on the balance of probabilities. The court considered the differing terminology used by courts considering the HSP issue but took the view that while there were differences in the language used and the rationalisations given over time, the essential point that runs through the authorities is that the civil standard of proof is flexible in its application and enables proper account to be taken of the seriousness of the allegations to be proved and of the consequences of proving them (paragraph 59). The court held that the HSP test is, as laid down in Re H, that the civil standard is one single standard, namely proof on the balance of probabilities (or preponderance of probability). The other standard is the criminal standard of proof beyond reasonable doubt. There is no intermediate standard, nor is the civil standard to be broken down into sub-categories designed to produce one or more intermediate standards (paragraph 60). The court distinguished between the single standard of the HSP and the fact that its application was to be flexible, the rationale behind this finding being that: the flexibility of the standard lies not in any adjustment to the degree of probability required for an allegation to be proved (such that a more serious allegation has to be proved to a higher degree of probability), but in the strength or quality of the evidence that will in practice be required for an allegation to be proved on the balance of probabilities (paragraph 62). The court conceded, however, that while the seriousness of an allegation is a function of the seriousness of its consequences, and vice versa, there will be cases where proof of an allegation may have serious consequences even though it cannot be said that the matter alleged is inherently improbable, but held that this would not affect the standard of proof to be used (paragraph 64). On the central question, the court here seems to be correct. To recall, Redmayne s supposed distinction between the flexible standard and cogent evidence approaches is that the first raises the standard of proof, while the second is an observation about the quantum of evidence required to satisfy the standard of proof. This is on the face of it a valid distinction: if you maintain the standard of proof at a fixed level, then if the inherent probability of evidence is seen as smaller in serious cases, you will require more of it to satisfy your fixed standard of proof in those cases. But what the court in Re A.N. said is that the second is not a different approach from the first, but is rather a manifestation of it: the single civil standard of proof is flexible insofar as and only insofar as it calls for more cogent evidence. The court then is saying that it is not the standard of proof that is actually flexing but rather the cogency of the evidence required to satisfy it.

8 330 E. McBRIDE This solution is supported by consideration of the cases which Redmayne claims as examples of the two approaches. As far as Re M (A Minor) is concerned, while it is true that the Court of Appeal preferred the phrase prior probability to flexible standard to describe the HSP, it specifically stated that we emphasise that the difference is one of expression only (p. 67). Equally, in Re W (Minors), where Redmayne sees a difference between the prior probability and cogent evidence approach to proving child abuse by a father (as opposed to by an unknown person), the court expressly stated that it did not need to consider this distinction, as the father was the only candidate for the role of perpetrator, in which case a finding that sexual abuse has occurred will inevitably lead to a finding as to the identity of the perpetrator. In those circumstances the standard must be the single, higher standard (p. 425). Furthermore, even if one goes to the case cited in Re W in which the distinction was relevant, it is noteworthy that the court held that a higher degree of probability is required to satisfy the court that the father has been guilty of some sexual misconduct with his daughter than would be needed to justify the conclusion that the child has been the victim of some such behaviour of whatever nature and whoever may have been its perpetrator (Re G at p. 21). In other words, the court did not approach the task at hand in two stages (first finding that the child had been abused and then deciding who the abuser was) at all, but rather considered the issue as a single one, namely to ascertain whether the father had abused the child. There was thus no question of applying differing standards of proof at differing stages of the inquiry. Lastly, if one recalls Lord Denning s dictum in Bater ( many great judges have said that, in proportion as the crime is enormous, so ought the proof to be clear ) which Redmayne claims as the fons et origo of the flexible standard approach, is not the notion of clear proof just another way of expressing that of cogent evidence? As far as the cases go then, the court s view in Re A.N. seems convincing and we can look at Re H as being the definitive statement of a single HSP, applied in both Bater- and Hornal-type cases. Having said that, however, the fact remains that Redmayne s approach is correct in that it reminds us that the adoption by courts of a purportedly uniform concept of proof may mask important distinctions between elements of it. There are in fact other approaches to an HSP which remain of relevance, not least because the House of Lords in the June 2008 cases introduced two of them the close scrutiny and quasi-criminal standards into the attack on Re H. In order to understand how these standards arise in the context of an HSP, it is necessary to briefly consider them and how they have been adopted independently of cases such as Re H. 4. Other HSPs 4.1 Quasi-criminal cases The courts have held that certain proceedings, though civil in nature, involve issues so serious that they call for a standard of proof no less than the criminal one of BRD. Examples are B v. Chief Constable of the Avon and Somerset Constabulary (2001) (the making of sex offender orders), Gough v. Chief Constable of the Derbyshire Constabulary (2002) (the making of football banning orders) and R (on the application of McCann) v. Crown Court at Manchester (2003) (the making of anti-social behaviour orders). In the McCann case (heard jointly with, and sometimes cited as, Clingham v. Royal Borough of Kensington and Chelsea [2002]), the prosecution argued that an application in the magistrates court for an anti-social behaviour order was a civil proceeding which attracted the flexible OBOP standard, having due regard to the inherent improbability test. Lord Hope (with

9 IS THE CIVIL HSP A COHERENT CONCEPT? 331 whom all members of the court agreed) accepted that the proceedings were civil but held that when allegations are made of criminal or quasi-criminal conduct which, if proved, would have serious consequences for the person against whom they are made (paragraph 82), then account should be taken of the seriousness of the matters to be proved and the implications of proving them... if this is done the civil standard of proof will for all practical purposes be indistinguishable from the criminal standard (paragraph 83). The point which calls for attention here is that this line of cases is consistent with the axioms and theorems of the HSP suggested in Section 3 above only when considered as a situation where the seriousness of the matters involved is too great to be contained within the flexible level of the OBOP at all. As there are only two levels of proof, one has to shift to the next higher level of BRD. What results is thus not really an HSP at all (in the sense of being a heightened level of proof within the OBOP) but a special case of the straightforward BRD. The assumption that the seriousness factor justifies adopting an HSP is criticized by some members of the court in the June 2008 cases as will be discussed in Section Miscellaneous standards of proof We should mention briefly here some alternative formulations of the HSP which Redmayne discusses such as the clear and convincing evidence test in Addington v. Texas and the reliably shown or convincingly shown tests as used in the context of the European Court of Human Rights. The court in Re A.N. considered these and concluded that they may all lead to much the same result in practice as the flexible application of the balance of probabilities standard (paragraph 65), so they will not be discussed in any more detail here. Note should also be taken of what is a genuinely different approach, namely that adopted at least in the USA of assigning specific probabilities to the HSP. So, e.g. Tillers & Gottfried (2006, p. 140, n. 17) note the view of a Federal Judge that the higher clear and convincing evidence standard of proof equates to 80%. In this vein, Anderson et al. (2005, p. 230) construct a table of probabilities in which they set a cogent standard of proof at 80%, while overwhelming is set at 90% and more likely than not at 60%. The difficulty with this approach is that courts here do not in general deal with figures of probability: the most precise formulation to be found is of the order of that discussed in the Cream Holdings case. In paragraph 116, e.g. it is noted that the word likely is most commonly used to indicate a degree of probability above 0.5 on the scale of 0 to 1 but the court also noted that this is not always the case and that indeed the word may not even imply any particular belief in the person using it. At most, one can discern in the language of measurement discussed above a quasi-ordinal scale of probabilities, ranging at the lowest from mere suspicion, then reasonable suspicion, 7 then real possibility, 8 may well 9 and, moving above 50% likelihood, more likely than not. 10 I use the phrase quasi-ordinal because, as a referee has observed, the scale as it stands cannot be purely ordinal as it contains a numerical term (50%) in the middle. In order to remove any anomaly, it may then be that the scale should be characterized as measuring two things: belief (above the 50% mark) and non-belief (below it) with the centre 7 R v. Gough [1993] AC 646, p. 667, where Lord Goff distinguishes two lower levels of likely than real possibility, namely mere suspicion and reasonable suspicion. 8 In Re H (Minors) [1996] AC Black v. Sumitomo [2002] 1 WLR Bailey v. Rolls-Royce (1971) Limited [1984] ICR 688.

10 332 E. McBRIDE mark being considered as a state of being non-committal rather than accorded the benefit of a numerical measure. 11 Nevertheless, there is no evidence that this scale of measurement can be tied to the interval level of measurement 12 that is represented by the percentage scale indeed, all the levels of proof noted here have been described in the cases cited as coming within the meaning of likely! Of course, one of the main reasons for this aversion to the use of mathematical probabilities is that cases are simply not decided on the basis of naked statistical evidence here, though in the USA, there has been more use made of statistical evidence see, e.g. the cases discussed by Tillers (1997, 2005). 4.3 The increased scrutiny cases Redmayne (1999, p. 177, n. 47) notes a further standard, where it is stated in a few cases that more careful scrutiny is required. He observes that this approach might seem initially appealing, but might in fact imply that judges need not be too careful when scrutinizing evidence in less serious cases. This point seems a valid one as far as the forensic setting is concerned. 13 However, as a referee has pointed out, there is one important sense in which this notion is borne out in a forensic setting, namely in the difference in approach to trial procedure between the comparatively informal magistrates court with its lay justices and the highly formal crown court with its professional judges. Indeed, the average conviction rate after a not guilty plea for at the Crown Court was about 35.6% (HO 2005, p. 36) but at the magistrates court was more than almost double, at 75.67% (op. cit., p. 28). There may be numerous reasons for this, but it is clear that an increased level of scrutiny given by a judge and jury is both perceived to be and is a factor (see Darbyshire, 1997, for a discussion). 11 Achinstein (2003, p. 104) considers a similar scheme, except it has belief and disbelief as positive and negative degrees along the same axis this begs the question what they are degrees of. 12 Statisticians distinguish between levels of measurement, the three important ones for present purposes being nominal, ordinal and interval. Each level of measurement is characterized by its formal properties and by the admissible operations which exist in respect of it. The three levels relevant here are the following: Nominal the weakest level, it merely uses numbers or symbols to classify objects. Its only formal property is that of equivalence, in that members of each group must be equivalent to each other as regards the criteria for placing them in that group. An example would be classification by occupation. The only operation admissible here is that of one to one transformation say each occupation were ordered alphabetically and assigned a number in that order, then reversing the order of numbers would not affect the classification of the objects in question. Ordinal this scale possesses the formal property greater than, in addition to that of equivalence. Thus, if we took the nominal scale of occupations and ranked them by the esteem in which they are held, we would have an ordinal scale. The admissible operation at this level is that of monotonic transformation : the numbers used to classify the groups can be changed, as long as they remain in the same order in relation to each other. Note that even if the underlying scale is thought of as continuous, since we cannot be sure of the distance between any two points on the scale, we cannot perform arithmetical operations on it thus, just because lawyers might be rated 10 on the scale, waitresses 5 and doctors 1, it does not follow that we can say that waitresses are twice as esteemed as lawyers. Interval on this scale, however, the distance between each group is known exactly, as with, e.g. temperature. In addition to the properties of equivalence and greater than, this scale has the property that the differences between the positions on the scale are isomorphic to the structure of arithmetic. This means that the admissible operations upon this scale are those of arithmetic (addition, multiplication and so on) so that one can talk of one measurement being twice as large as another or being the same distance from a fixed point as another measurement. For present purposes, it is of note that the standard probability scale achieves interval-level measurement (in fact, it achieves a higher level, namely that of ratio measurement, because it has a true zero but the difference is not material here). See Siegel (1956, pages 21 ff) for a discussion. 13 Though anyone who has been subject to an Inland Revenue inspection of their accounts, or a personal search at an airport, will know the difference between scrutiny and careful scrutiny, involving as it can the loss of one s shirt, either metaphorically (in the case of a tax inspection) or literally (in the case of an airport search).

11 IS THE CIVIL HSP A COHERENT CONCEPT? The anxious scrutiny cases Beyond this sense of careful scrutiny, the standard of proof arrived at through what is described by the term of art anxious scrutiny is applied in a more substantial number of cases, in England (e.g. Re Officer L), Northern Ireland (Re Surgula), Scotland (Re Secretary of State for the Home Department) and Eire (Sweetnam v. An Bord Pleanala). While this standard is most typically applied in asylum cases where the lives of the asylum seekers in question are at risk, it has been applied generally to cases where fundamental human rights are in issue even in, e.g. planning cases (see Sweetnam v. An Bord Pleanala). The common issue running through these cases is whether a court on judicial review should go beyond its usual role of reviewing procedural propriety on the Wednesbury principle and actually review a decision on its merits. The answer has been that it should not, but (echoing the Re H test) that the anxious scrutiny test requires that The more substantial the interference with human rights, the more the court will require by way of justification before it is satisfied that [a] decision is reasonable (per Bingham MR in ex parte Smith). Indeed, in a further echo of the Re H test, the Court of Appeal in ex parte Mahmood (paragraph 19) referred to a sliding scale of the intensity of judicial review in anxious scrutiny cases. Furthermore, in R v. DPP ex parte da Silva (judicial review of the decision not to prosecute any police officers as a result of the killing of Jean Charles de Menezes), the Court of Appeal found (paragraph 49) that there was no material distinction between the careful and anxious scrutiny tests and decided that on either test, a court should not apply any standard more rigorous than that of the Wednesbury test. So what is the practical effect of the test? An example is given by the Court of Appeal in ex parte AK (Afghanistan), where an assessor had refused to take into account evidence adduced by an asylum seeker contained in an affidavit from a family member on the grounds that it could not add probative or corroborative weight to the asylum claim. The court held (paragraph 28) that this was a misdirection, as there was no trace here of the writer pausing to consider what an independent tribunal might (not would) make of the affidavit and giving anxious scrutiny to that question. Similarly, in WM (DRC), the Court of Appeal held that while an asylum decision is only capable of being impugned on Wednesbury grounds, it will be irrational if it is not taken on the basis of anxious scrutiny, a test which applied in respect of the evaluation of the facts and in respect of the legal conclusions to be drawn from those facts (paragraphs 9 11). The issue here was whether new evidence submitted following the rejection of an asylum claim could constitute a fresh claim in that it created a realistic prospect of success if it were to be referred to an adjudicator for a re-hearing. The court found that the decision maker should ask not whether he himself considered the new evidence credible, but whether an adjudicator could do so on a re-hearing and give anxious scrutiny to that question (paragraphs 24 and 26). So while courts tend to identify two issues (the would versus should question and the level of scrutiny needed to answer it), the anxious scrutiny test in fact appears to lower the standard of proof to be applied in that it requires a decision maker not to reject evidence simply because they do not find it convincing, but to ask rather whether a tribunal might find it convincing. A similar result was arrived at in R v. Lord Saville ex parte A, where the court quashed the decision of the Saville Inquiry into Bloody Sunday to withhold anonymity from soldiers giving evidence. The court observed that Anxious scrutiny is not judicial rhetoric, but an established doctrine with a discernable shape and direction (paragraph 77). In this case, the court held that the Inquiry had used the wrong test by asking whether there was concrete evidence of specific threats to soldiers giving

12 334 E. McBRIDE evidence, where they should have asked whether the soldiers had a reasonable fear of such reprisals (see paragraphs 31, 81 and 106). Lastly, it is noteworthy that the asylum and immigration adjudicators themselves submitted to a Parliamentary Inquiry that... whilst case law requires adjudicators to give anxious scrutiny to the cases before them, it is often difficult to do so within an adversarial system.... the adjudicator cannot give anxious scrutiny... without taking a more active approach to the issues and they argued for the adoption of a more interventionist procedure in asylum and immigration appeals (Select Committee on Constitutional Affairs Second Report, 2004, paragraph 27). What all these cases seem to indicate is firstly that the argument about anxious scrutiny revolves as much around policy issues of reasonableness and proportionality as around issues of levels of proof. Second, despite talk of sliding scales and increased requirements of proof, the real issue seems to turn not on the level of proof needed to establish facts, but on what facts should be established before findings adverse to litigants are made. Thus, it is not that some sort of special test is needed before deciding whether, say, there would be a threat to a soldier s life if he gave evidence to the Saville Inquiry, but rather that what has to be established, using the ordinary test, is whether there could be such a threat. Similarly, it is not that a family member giving evidence in support of an asylum seeker is subject to any type of extraordinary ordeal before their evidence can be accepted (such as the ducking stool or rack of medieval times), but rather that the evidence is to be evaluated on the basis of whether it could be true. As a referee has pointed out, since our law insists that there is no room for finding that a fact might have happened (see Lord Hoffman s speech in Re B discussed below), this cannot be a statement about the level of proof of facts required by the concept of increased scrutiny. 14 It seems clear then that, historically, the anxious scrutiny standard and the Re H HSP represented different concepts with their own development. However, as will be seen in Section 5, the two concepts have now become confused following Re B, in which both were expressed by different judges to underly the HSP. 4.5 Practical objections to the HSP Before considering any conceptual difficulties inherent in the notion of an HSP, I consider here some objections which Redmayne raises (for the sake of consistency, I ignore here his perceived distinction of the Re H approach into two contrary tests the flexible standard and prior probability as I have argued above against the existence of such a distinction). The chief objections are as follows: (a) having a flexible standard of proof is too vague, as there are no rules for assigning particular levels of proof to individual cases 15 ; (b) there is no empirical support for the prior probability approach, as we do not know enough about the prevalence of fraud, arson, child abuse and so on to claim that serious wrongs are less common than less serious wrongs 16 ; (c) this approach also ignores the possibility that allegations of serious but rare events may be more likely to be made when untrue (e.g. claims of child abuse) or on the other hand, since they are unlikely to happen, it is unlikely 14 The relationship between levels of scrutiny and policy considerations is made explicit in the concept of variable scrutiny adopted by the U.S. Supreme Court. Thus, strict scrutiny shifts the burden of proof to the government, requires the government to pursue a compelling state interest and demands that the regulation promoting the compelling interest be narrowly tailored, intermediate scrutiny requires the government to establish that its actual purpose substantially promotes an important government interest and ordinary scrutiny upholds all action that is a rational means to accomplish a legitimate government purpose (Siegel, 2006, pp. 6 8). 15 See Redmayne (1999) at p Op. cit. at p. 184.

13 IS THE CIVIL HSP A COHERENT CONCEPT? 335 that someone will falsely say they did 17 ; (d) on the other hand, if all this approach implies is that the inherent improbability of an event is itself a matter to be taken into account when deciding if it happened, it is a statement of the obvious and it would give cause for concern if judges needed to be reminded that improbable occurrences are less likely to occur than probable ones. 18 As far as (a) is concerned, the lack of rules for assigning standards of proof to cases is no more a problem than it is in any policy-based area of law (thus, there is no mechanical rule for deciding whether a defendant was under a duty of care to a plaintiff or whether exemplary damages are available to a plaintiff, consideration of the latter issue causing the court to comment in Cassell v. Broome that the life of the law often lies not in logic but in experience (paragraph 40)). (b) and (c) are points I come back to in Section 6. (d) is a point that the court made in Re B, to which I now turn. 5. The House of Lords attack on Re H 5.1 Re B The judgements in Re B and Re Doherty were given on the same day by differently constituted panels of the House of Lords. The first case was an express attempt by the guardian ad litem of children to have the decision in Re H overturned. As in Re H, the judge below found only that there was a possibility of abuse of children by their father having occurred, but that it was not more probable than not that he had abused them. The judge therefore proposed, in considering the likelihood of future harm, to deem the allegations against the father unproven. The House of Lords, in dismissing an appeal by the guardian, unanimously held that this was the correct approach. In the course of doing so, the court also revisited Re H and the other cases discussed above in an attempt to give a definitive ruling on the cogent evidence question. Indeed, this aim was expressed in no uncertain terms: Lord Hoffman held that the time has come to say, once and for all, that there is only one civil standard of proof and that is proof that the fact in issue more probably occurred than not (paragraph 13) and Baroness Hale, of the notion that the more serious the allegation, the more cogent the evidence needed to prove it, stated that It is time for us to loosen its grip and give it its quietus (paragraph 64). Furthermore, this seems to have been the reason the House agreed to hear the case at all because Baroness Hale accepted (paragraph 23) that the issue had been authoritatively determined by Re H it was thus nowhere near fulfilling the House s criteria for departing from its own decisions (see Practice Statement (Judicial Precedent) [1966] 1 WLR 1234). Lord Hoffmann s speech (with which Lords Rodger and Walker agreed) was confined solely to discussion of this issue, without stating the facts of the case. He made three interesting points: first, he expressed the nature of forensic proof in explicitly mathematical language: If a legal rule requires a fact to be proved (a fact in issue ), a judge or jury must decide whether or not it happened. There is no room for a finding that it might have happened. The law operates a binary system in which the only values are 0 and 1. The fact either happened or it did not. If the tribunal is left in doubt, the doubt is resolved by a rule that one party or the other carries the burden of proof. If the party who bears the burden of proof fails to discharge it, a value of 0 is returned and the fact is treated as not having 17 Op. cit. at p Op. cit. at p. 185.

14 336 E. McBRIDE happened. If he does discharge it, a value of 1 is returned and the fact is treated as having happened (paragraph 2). Second, he distinguished three groups of cases: (i) those in which the court has classified proceedings as civil but nevertheless thought that, because of their serious consequences, the quasicriminal standard of proof should be applied (e.g. Khawaja), (ii) cases in which it has been observed that cogent evidence is needed to prove inherently improbable events (e.g. Hornal, Re H) and (iii) cases in which judges are simply confused about whether they are talking about the standard of proof or about the role of inherent probabilities in deciding whether the burden of proving a fact to a given standard has been discharged (paragraph 5). Third, Lord Hoffman (paragraph 14) emphasized the italicized words in Lord Nicholls dictum at paragraph 73 of Re H: the court will have in mind as a factor, to whatever extent is appropriate in the particular case, that the more serious the allegation the less likely it is that the event occurred and, hence, the stronger should be the evidence before the court concludes that the allegation is established on the balance of probability. He said this phrase showed that Lord Nicholls was not laying down any rule of law and that common sense, not law, requires that in deciding this question, regard should be had, to whatever extent appropriate, to inherent probabilities. Thus, if a child alleges sexual abuse by a parent, he held that it is common sense to start with the assumption that most parents do not abuse their children but that this assumption could be swiftly dispelled by other compelling evidence. So if, e.g. it was clear that a child was assaulted by one or other of two people, a court could not conclude that neither did so because assault is inherently unlikely, but would have to decide which of the two was the perpetrator. Baroness Hale (with whom all members of the court agreed), accepted (paragraph 69) that some proceedings, though civil in form, required the criminal standard of proof and cited Bater v. Bater, McCann and Khawaja as examples. However, she held that care proceedings did not require this standard of proof, as their purpose is not to punish or to deter anyone but solely to protect children from harm. Indeed, she went further and said she wished to announce loud and clear that the standard of proof in finding the facts necessary to establish the threshold under section 31(2) or the welfare considerations in section 1 of the 1989 Act is the simple balance of probabilities, neither more nor less (paragraph 70). She considered also that neither the seriousness of the allegation nor the seriousness of the consequences should make any difference to the standard of proof to be applied in determining the facts and that the inherent probabilities are simply something to be taken into account, where relevant, in deciding where the truth lies. Indeed, she considered that there is no logical or necessary connection between seriousness and probability, noting that while some crimes such as murder are sufficiently rare to be inherently improbable in most circumstances, other seriously harmful behaviour, such as alcohol or drug abuse, is regrettably all too common and not at all improbable. Regarding the Regent s Park example, she considers that an animal seen outside the zoo on a stretch of greensward regularly used for walking dogs is of course... more likely to be a dog than a lion, whereas if it is seen in the zoo next to the lions enclosure when the door is open, then it may well be more likely to be a lion than a dog.