In the matter of J (Children)

In a free society, it is a serious thing indeed for the state compulsorily to remove a child from his family of birth. Interference with the right to respect for family life, protected by article 8 of the European Convention on Human Rights, can only be justified by a pressing social need. Yet it is also a serious thing for the state to fail to safeguard its children from the neglect and ill-treatment which they may suffer in their own homes. This may even amount to a violation of their right not to be subjected to inhuman or degrading treatment, protected by article 3 of the Convention. How then is the law to protect the family from unwarranted intrusion while at the same time protecting children from harm?

In England and Wales, the Children Act 1989 tries to balance these two objectives by setting a threshold which must be crossed before a court can consider what order, if any, should be made to enable the authorities to protect a child. The threshold is designed to restrict compulsory intervention to cases which genuinely warrant it, while enabling the court to make the order which will best promote the child’s welfare once the threshold has been crossed. That threshold is defined by section 31(2) of the Act as follows:

“A court may only make a care order or supervision order if it is satisfied –

(a)

that the child concerned is suffering, or is likely to suffer, significant harm; and

(b)

that the harm, or likelihood of harm, is attributable to –

(i)

the care given to the child, or likely to be given to him if the order were not made, not being what it would be reasonable to expect a parent to give to him; or

(ii)

the child’s being beyond parental control.”

Once that threshold is crossed, section 1(1) of the Act requires the court to treat the welfare of the child as its paramount consideration, having regard to the checklist of factors listed in section 1(3). These include “any harm which [the child] has suffered or is at risk of suffering” (section 1(3)(e)). There are therefore three questions to be answered in any care case: first, is there harm or a likelihood of harm; second, to what is that harm or likelihood of harm attributable; and third, what will be best for the child?

It is some indication of the importance of the issues that the apparently simple words of section 31(2) have been considered by the House of Lords and the Supreme Court in no less than six cases: In re M (A Minor) (Care Orders: Threshold Conditions) [1994] 2 AC 424; In re H (Minors) (Sexual Abuse: Standard of Proof) [1996] AC 563; Lancashire County Council v B [2000] 2 AC 147; In re O (Minors) (Care: Preliminary Hearing) [2003] UKHL 18, [2004] 1 AC 523; In re B (Children) (Care Proceedings: Standard of Proof) (CAFCASS intervening) [2008] UKHL 35, [2009] AC 11; and In re S-B (Children) (Care Proceedings: Standard of Proof) [2009] UKSC 17, [2010] 1 AC 678. In re M was concerned with the first limb of the first question: what is meant by “is suffering” significant harm? We are concerned with the second limb of that question: what is meant by “likely to suffer” significant harm?

A child may be protected, not only if he is actually suffering harm as a result of a lack of reasonable parental care, but also if it is likely that he will do so in the future. But how is a court to be satisfied that it is likely that this particular child – the child concerned - will suffer significant harm in the future? It has twice been held in the House of Lords that the mere possibility, however real, that another child may have been harmed in the past by a person who is now looking after the child with whom the court is now concerned is not sufficient. The court has to be satisfied on the balance of probabilities that this person actually did harm that other child: see In re H [1996] AC 563 and In re B [2009] AC 11. But in both those cases, it was not established that the other child had been harmed at all. The issue in this case is whether it makes a difference that another child has indeed been harmed in the past and there is a possibility that this parent was responsible for that harm.

Before turning to the facts and the arguments in this case, it should be emphasised that in the real world the issue hardly ever comes packaged in this simple way. There are usually many readily provable facts upon which an authority can rely to satisfy the court that a child is likely to suffer significant harm unless something is done to protect him. Cases in which the only thing upon which the authority can rely is the possibility that this parent has harmed another child in the past are very rare. As the Court of Appeal pointed out, this case has itself been artificially constructed by the decision to treat the issue as a preliminary question of law: [2012] EWCA Civ 380, [2012] 3 WLR 952, para 81. Who can say what facts the court might have found relevant had the history been fully investigated in the usual way?

There are three children concerned in this case: HJ, a girl born on 20 June 2005, so now aged seven years and seven months; TJ, a boy born on 17 August 2006, so now aged six years and five months; and IJ, a girl born on 19 July 2009, so now aged three years and six months. The two eldest, HJ and TJ are the children of DJ, the second respondent, and his former partner, SC. They have been looked after by their father for the whole of their lives, having remained in his care when their mother left in 2008.

IJ is the daughter of JJ, the first respondent, with whom DJ formed a relationship in 2008. It was originally thought that DJ was the father of IJ, but DNA testing established that her father is SW, with whom JJ had earlier had a relationship. IJ has been part of the family unit with DJ, HJ and TJ for the whole of her life. Her mother, JJ, also formed part of that family unit for 20 months after IJ was born. Then in March 2011, the local authority formed a child protection plan which required her to move out of the family home, which she did. The local authority issued care proceedings in respect of all three children in April 2011. DJ and JJ married in June 2011 and have since had a child together, RJ, who was born on 1 December 2011.

The local authority quite properly took steps to protect the three older children after being made aware by another local authority of the findings of Judge Masterman in earlier care proceedings relating to JJ’s second child, S, who was born on 13 August 2005. Those proceedings were brought because of the death of her first-born child, T-L, when T-L was only three weeks old. T-L was born on 9 March 2004 and discharged from hospital two days later. When T-L died on 29 March 2004, she was found to have multiple fractures to her ribs, caused on at least two occasions, bruising to her left jaw, right side of her face, left shoulder and left inner elbow, all caused non-accidentally, and serious and untreated nappy rash. She had died as a result of asphyxia caused either by a deliberate act or by SW taking her to bed with him and JJ leaving her in SW’s care. Both were held to have colluded to hide the truth. In the circumstances, the judge found that “singling out a likely perpetrator does not help this couple because it must be debateable as to which is worse, to inflict this injury or to protect the person responsible”.

Following Judge Masterman’s judgment of 24 May 2006, JJ and SW withdrew from contact with S and eventually consented to his being adopted outside the family. They remained in a relationship until August 2007, when SW committed an assault upon JJ to which he eventually pleaded guilty at Cardiff Crown Court in June 2008. This was by no means the first occasion known to the local authority upon which SW had been violent towards JJ.

The current care proceedings were transferred to the High Court in September 2011, for determination as a preliminary issue whether the local authority could rely upon Judge Masterman’s findings to cross the threshold in section 31(2) of the Children Act 1989. By the time that this issue came to be tried by Judge Hallam in November 2011, the local authority had conceded that “the only matter that could meet the threshold criteria, at the relevant time, are the findings . . . as to the physical injuries sustained by T-L. They do not seek to bring failure to protect into the equation”. Hence the issue Judge Hallam had to determine was “whether JJ’s inclusion in a pool of perpetrators in earlier proceedings involving a different child and a different relationship can form the basis of the threshold in relation to a subsequent child in later proceedings”.

After examining the earlier authorities, Judge Hallam concluded that “the likelihood of significant harm . . . can only be proved by reference to past facts which are proved on the balance of probabilities”. The only facts available to the local authority had not been proved to that standard. Hence the threshold was not met and the proceedings were dismissed. Following this judgment, JJ returned to the family home, where she has remained ever since. No proceedings have been taken in respect of RJ, who was born shortly after the judgment.

Judge Hallam did remark that she was “aware that the present law does cause consternation for local authorities, professionals involved in the protection of children and academic commentators. However, it is quite apparent that the higher courts have considered those concerns and taken them into account in reaching their decisions”. The local authority appealed to the Court of Appeal, which reached the same conclusion in April 2012. However, the Court of Appeal took the unusual step of itself granting the local authority permission to appeal to this court. Giving the leading judgment, McFarlane LJ commented that “Artificially to limit the judicial exercise in a manner which invites the court to ignore part of the evidence in the case, might well set up the legal point for determination in a clinically clear and legally accessible manner, but it cannot, in my view, represent a proper exercise of the judicial task” (para 81).

While I would not criticise the local authority for agreeing to isolate the issue in this way, that comment does underline the unreality of the question with which this court, in common with Judge Hallam and the Court of Appeal, has been presented. The unreality is further illustrated by the way in which Mr Stephen Cobb QC, on behalf of the appellant local authority, has framed the issue as a choice between two extremes:

“Where a previous court has found that there is a real possibility that one or other or both of two or more carers have perpetrated significant harm on a child in his or her care, is that ‘finding’ a ‘finding of fact’ that may be relied upon in subsequent proceedings relating to only one of the potential perpetrators in support of a conclusion that there is a real possibility or likelihood of a subsequent child in a new family unit of which he or she is part suffering significant harm or is it a ‘finding’ that must be totally ignored in the subsequent proceedings?”

As will be seen, this is not the only way of framing the issue.

The leading authorities are well-known and have been reviewed many times, most recently by this court in In re S-B [2010] 1 AC 678. However, as Mr Cobb correctly observes, they were not reviewed with this precise issue in mind and it is therefore necessary to consider them briefly once more.

The starting point is the decision in In re H [1996] AC 563. The proceedings concerned three young girls whose elder sister alleged that she had been sexually abused by their mother’s partner, her step-father and the father of the two youngest girls, from the age of seven or eight. The judge did not find the allegations proved but held that there was a real possibility that they were true. As this was the only basis upon which it was suggested that it was likely that the other children would suffer harm in the future, the case was dismissed. The decision of the House of Lords is authority for three important propositions: first, that the standard of proof of such allegations is the simple balance of probabilities; second, that “likely” in section 31(2) does not mean “more likely than not”; rather, it means likely “in the sense of a real possibility, a possibility that cannot sensibly be ignored having regard to the nature and gravity of the feared harm in the particular case” (per Lord Nicholls of Birkenhead at p 585F); third, however, “A decision by a court on the likelihood of a future happening must be founded on a basis of present facts and the inferences fairly to be drawn therefrom” (p 590A); unresolved judicial doubts and suspicions “can no more form the basis of a conclusion that the second [likelihood of harm] threshold condition in section 31(2)(a) has been established than they can form the basis of a conclusion that the first [present harm] has been established” (p 589E).

The third proposition was held only by a majority of three to two but has been reaffirmed in two later decisions at this level. In In re B [2009] AC 11, the essential facts were very similar to those in In re H. Care proceedings were brought in respect of three children, because the oldest of the three, a 16-year-old girl, alleged that their mother’s husband, her step-father and the father of the two younger children, had sexually abused her and physically abused both her and her elder brother. The judge could not decide that it was more likely than not that the girl was telling the truth, but nor could he decide that it was more likely than not that the husband was telling the truth. Hence those allegations could not form the basis of a conclusion that the children were likely to suffer harm in the future. On appeal, the children’s guardian, with the support of the local authority, invited the House of Lords to over-turn the decision in In re H in favour of a test that there was a “real possibility” that certain events had happened. The House unanimously and unhesitatingly declined that invitation (per Baroness Hale of Richmond at paras 53-54). The “thoroughly convincing” reasons given by Lord Nicholls were summarised thus (para 54):

“The threshold is there to protect both the children and their parents from unjustified intervention in their lives. It would provide no protection at all if it could be established on the basis of unsubstantiated suspicions: that is, where a judge cannot say that there is no real possibility that abuse took place, so concludes that there is a real possibility that it did. In other words, the alleged perpetrator would have to prove that it did not. Mr Cobb [for the children] accepts that it must be proved on the balance of probabilities that the child ‘is suffering’ significant harm. But nevertheless he argues that those same allegations, which could not be proved for that purpose, could be the basis of a finding of likelihood of future harm. If that were so, there would have been no need for the first limb of section 31(2)(a) at all. Parliament must be presumed to have inserted it for a purpose. Furthermore the Act draws a clear distinction between the threshold to be crossed before the court may make a final care or supervision order and the threshold for making preliminary and interim orders. If Parliament had intended that a mere suspicion that a child had suffered harm could form the basis for making a final order, it would have used the same terminology of ‘reasonable grounds to suspect’ or ‘reasonable grounds to believe”’.

The House also reaffirmed and clarified that the standard of proof was the simple balance of probabilities. No more severe standard was to be applied because of the seriousness of the allegations. The inherent probabilities were only part of deciding what was more likely than not to have happened. As Lord Hoffmann pointed out (at para 15), assaulting children is a serious matter, but if it is clear that the child has indeed been assaulted, it makes no sense to say that neither of the possible perpetrators is likely to have done it. The fact is that one of them did and the task is to decide whether it is more probable that one rather than the other was responsible.

In neither In re H nor In re B had the question of identifying a perpetrator arisen. The “real possibility” under discussion was whether or not the alleged abuse had taken place at all. But the question did arise in In re S-B [2010] 1 AC 678, a judgment of this court to which seven Justices subscribed. The case concerned two children. The older child, J, was found to have sustained non-accidental bruising to his face and arms at the age of only four weeks. The judge found it likely that only one parent had been responsible and there was no question of the other having failed to protect the child. She concluded that neither parent could be ruled out as the perpetrator of the injuries, although she later indicated a 60% likelihood that it was the father. The parents had separated while the proceedings were going on and their younger child, W, was born before the hearing. He had never come to any harm, having been removed from his mother at birth. Nevertheless, the judge concluded that, because there was a real possibility that the mother had injured J, there was also a real possibility that she would injure W in future.

The principal point in the case was the standard of proof to be applied in identifying the perpetrator of injuries which are found to have been non-accidental. The court reaffirmed that this was the simple balance of probabilities, and pointed to the very real advantages of making such a finding where it was possible to do so (paras 36 to 38). The judge had applied too high a standard and it was not possible to treat her later 60% indication as a positive finding that the father was the perpetrator. Hence the case was sent back to be re-heard.

In relation to the younger child, W, there was another reason to remit the case (para 49):

‘The judge found the threshold crossed in relation to [W] on the basis that there was a real possibility that the mother had injured [J]. That, as already explained, is not a permissible approach to a finding of likelihood of future harm. It was established in In re H [1996] AC 563 and confirmed in In re O [2004] 1 AC 523 that a prediction of future harm has to be based upon findings of actual fact made on the balance of probabilities. It is only once those facts have been found that the degree of likelihood of future events becomes the ‘real possibility’ test adopted in In re H. It might have been open to the judge to find the threshold crossed in relation to [W] on a different basis but she did not do so.’

Strictly speaking, that paragraph may be obiter, as the case was to be remitted for re-hearing in any event. However, it did constitute an independent reason for remitting the case in relation to W, a reason which would have applied to him even if the judge had applied the correct standard of proof in relation to the perpetrator of J’s injuries but been unable to decide between the parents. It is that paragraph which has apparently caused “consternation” in some quarters and which is under challenge in these proceedings.

One reason for that “consternation” may be that paragraph 49 is inconsistent with some observations of Wall J in In re B (Minors) (Care Proceedings: Practice) [1999] 1 WLR 238. The case concerned two-year-old twins, CB and JB, one of whom, CB, had suffered shaking injuries on two occasions. The judge found that the mother had shaken the child on the second of those occasions, so (the father being off the scene) there was a proper factual basis for concluding that the threshold criteria were met in relation to both children. Nevertheless, the judge went on to observe that they would have been met even if he had been unable to decide who was responsible for the shaking. His basis for doing so appeared to be that, as “likely” means a “real possibility”, “there must therefore, in my judgment, be a possibility which cannot sensibly be ignored that if JB were left in the care of his parents – or either of them – he too will suffer significant harm” (p 248E). With the greatest of respect to the undoubted wisdom and experience of the judge, that reasoning fails to distinguish between the degree of likelihood required by the word “likely” and the factual findings required to satisfy the court of that likelihood, a distinction which was clearly drawn by the House of Lords in the later case of In re B [2009] AC 11.

It is also suggested that para 49 of In re S-B [2010] 1 AC 678 is inconsistent with the two other cases which, together with In re H [1996] AC 563, make up a “trilogy” of House of Lords cases in which the leading opinion was given by Lord Nicholls. The second in the trilogy is Lancashire County Council v B [2000] 2 AC 147. The House of Lords was not concerned with the basis for predicting the likelihood of future harm for the purpose of section 31(2)(a) but with the proper interpretation of the second part of the threshold test, the “attributability” criterion in section 31(2)(b). A seven-month-old baby had suffered injury from having been violently shaken on at least two occasions. Care proceedings were brought in relation to that child and also in relation to the child of a child-minder who looked after the injured child during working hours. The judge found that the injuries had been caused either by the mother, or by the father, or by the child-minder, but he could not identify the perpetrator. He concluded that the threshold criteria were not met in respect of either child, but the Court of Appeal and the House of Lords held that they were met in respect of the injured child. However, the Court of Appeal held that they were not met in respect of the child-minder’s child and permission to appeal that decision out of time to the House of Lords was refused.

The House held that it was not necessary, for the purpose of section 31(2)(b), to make a finding that an individual parent whose parental responsibility would be curtailed by the order was responsible for the harm suffered. It was enough that the harm was attributable to the care given to the child not being what it ought to have been. What was meant by the “care given to the child”? Did it refer only to the care given by the parents, or by other primary carers, or by anyone who was looking after the child? The House held that it referred principally to the child’s primary carers, but where, as here, the care of the child was shared, it could also embrace those who shared that care (at p 166B-D). This might mean that the attributability condition was satisfied when there was no more than a possibility that the parents were responsible (at p 166F).

Hence, it is now argued, the same should apply to the likelihood of harm condition in section 31(2)(a). If a parent can fall foul of the attributability criterion when there is no more than a possibility that she has harmed the child, why should she not fall foul of the likelihood criterion in the same circumstances?

However, the Lancashire case was not about the likelihood criterion. There is nothing in the House of Lords’ decision in the case to cast doubt upon the decision of the Court of Appeal (albeit reached with no enthusiasm) that the likelihood criterion in section 31(2)(a) was not met in relation to the child-minder’s child. That child had not suffered any harm at all. The risk of future harm could be established only on the basis of proven facts, not mere suspicion. It was clear that the parents of the injured child would play no part in the child’s care in the future. It had not been established on the balance of probabilities that the child-minder was the perpetrator of the injuries (per Robert Walker LJ at p155F-G).

The decision of the Court of Appeal in the Lancashire case is, of course, in line with para 49 of the judgment of this court in In re S-B [2010] 1 AC 678. In In re F (Interim Care Order) [2011] EWCA Civ 258, [2011] 2 FLR 856, the Court of Appeal held, on similar facts, that it was bound to dismiss the local authority’s appeal. The child concerned was a baby, C, born in June 2010, who had come to no harm: indeed the hearing took place only three days after he was born. Proceedings were brought because, in 2004, his father had been party to care proceedings involving his child, J, by a different woman. At the age of six months, J had suffered two fractures to the right leg on separate occasions. The judge found that either the mother or the father was responsible and in relation to the second fracture the other parent had failed to protect J. But (as in this case) the local authority in the current proceedings relied only on the real possibility that the father was the perpetrator of J’s injuries and not upon any failure to protect.

In refusing the local authority permission to appeal to the Supreme Court, Wilson LJ referred both to the Lancashire case and to para 49 of In re S-B. He observed that “the strict status of that passage as obiter carries very little significance in circumstances in which it is all of a piece with a number of earlier, yet also recent, decisions of the House of Lords” (para 14). After referring to the “consternation” caused by para 49, he continued, “No doubt there are hard and worrying cases. But the requirement of proven factual foundation is a bulwark against the state’s removal of a child from his family, which I consider very precious. I also applaud the Supreme Court’s regular acknowledgement of the fact that, although it can depart from its previous decisions, the exercise of departure is highly unsettling for the law and should be undertaken only with great caution” (para 15).

The application for permission to appeal was not renewed before this court, perhaps because the local authority had since launched new proceedings alleging that the threshold was crossed on an entirely different basis (current drug-taking by the parents). Should the threshold be crossed on that basis, the question might arise as to the relevance of the inclusion of the father in a (small) pool of possible perpetrators of the injuries to J at the so-called “welfare” stage of the inquiry, where the court is considering what order if any, will best promote the welfare of the child. Wilson LJ raised the question, which he clearly regarded as causing difficulties for a trial judge, but it was unnecessary to answer it in the circumstances.

The third case in the “trilogy”, In re O [2004] 1 AC 523, was concerned with the welfare stage of the inquiry. There were two separate appeals. In one case, a child had suffered serious injuries; the father admitted responsibility for one of them; but the judge held that neither parent could be excluded as a possible perpetrator of the others. The Court of Appeal held that the case should proceed on the basis that there was no risk, either to the injured child or to a younger child, from the mother. In the other case, a child had died as a result of serious injuries which the judge found had been inflicted by the mother’s partner. A differently constituted Court of Appeal found that there was insufficient evidence that the mother’s partner was the sole perpetrator, so the judge at the welfare hearing would not be able to disregard the risk to the surviving child presented by the mother.

Lord Nicholls re-iterated the principles already established in In re H and Lancashire County Council v B: that in considering the likelihood of future harm for the purpose of section 31(2)(a), the court had to act only on the basis of proven facts (para 17); that “likely” meant a real possibility; and that the care given to the child for the purpose of section 31(2)(b) included the care given by any of the child’s carers so that the threshold could be crossed even though the identity of the perpetrator was not known (para 19). It was in that connection, and not in connection with the “likelihood” criterion, that Lord Nicholls (para 20) referred with apparent approval to what Wall J had said in In re B [1999] 1 WLR 238, 248.

He went on to hold that, once it had been proved, to the requisite standard, that a child had suffered harm or was likely to do so (para 26), the court could take into account at the welfare stage the inclusion of the parent in a pool of possible perpetrators of the child’s injuries. It would be “grotesque” to proceed on the basis that the child was at risk from neither parent (para 27). The importance to be attached to this, as to any other feature, would depend upon the circumstances of the case (para 31).

On the other hand, Lord Nicholls went on to discuss, albeit obiter, whether unproven allegations of harm could be taken into account at the welfare stage and held that they could not (para 38). This accorded with the approach of the Court of Appeal in In re M and R (Child Abuse: Evidence) [1996] 2 FLR 195 (para 39): it would be extraordinary if Parliament intended that evidence insufficient to establish harm for the purpose of section 31(2)(a) should be sufficient to establish harm for the purpose of section 1(3)(e) (para 40).

Once again, therefore, it is argued that, if inclusion in a pool of possible perpetrators can be taken into account (with whatever difficulty in assessing its importance) both for the purposes of the attributability criterion and at the welfare stage of the inquiry, then why can it not be taken into account for the purposes of the likelihood criterion in section 31(2)(a)? Once again, however, there is nothing in In re O to cast doubt upon the necessity for founding a prediction of future harm upon a proven factual basis. Lord Nicholls went out of his way to reiterate the necessity for this. It is also striking that, although both appeals were concerned with a child who had not (yet) been harmed, there was no discussion of the basis upon which the threshold had been crossed in their cases. Thus there was no discussion of whether or not the possible perpetrators remained members of the same household at the relevant time: clearly, if one child has suffered harm at the hands of one of two parents who remain members of the same household, it is possible (though not inevitable) to infer that another child in that household is also likely to suffer harm. Furthermore, the discussion of the relevance of inclusion in the pool at the welfare stage was all in the context of proven harm to the child concerned.

Finally, reference should be made to the case of North Yorkshire County Council v SA [2003] EWCA Civ 839, [2003] 2 FLR 849. This too was not a case about the “likelihood” criterion. The child concerned had suffered serious non-accidental injuries. The case was about the identification of the pool of possible perpetrators for the purpose of the later stages of the inquiry. The court held that a “no possibility” test, ruling out only those people who could not be perpetrators, cast the net far too wide. The court preferred the test of including only those where there was a “likelihood or real possibility” that they were responsible (para 26). In doing so, Dame Elizabeth Butler-Sloss P referred to the opinion of Lord Nicholls in In re O [2004] 1 AC 523, and expressed the view that the “real possibility” test which he had applied at the welfare stage should also be applied at the section 31(2) part of the case (para 21). Once again, however, this was not a case about predicting the likelihood of future harm. It was about identifying the pool of possible perpetrators at the attributability and welfare stages. Its approach at that stage was expressly approved by this court in In re S-B [2010] 1 AC 678, at para 43.

It is worth remembering that all of these cases, in which the judge (or the Court of Appeal) found it impossible to identify a perpetrator, were decided before the decision of the House of Lords in In re B [2009] AC 11. That includes the decision of Judge Masterman in the earlier proceedings which led to this case. As was apparent in In re B, until that case was decided, the opinion of Lord Nicholls in In re H [1996] AC 563 had frequently been misinterpreted so as to require a higher standard of proof where the allegations made were serious. The nostrum had taken hold that “the more serious the allegation, the more cogent the evidence needed to prove it”: In re B [2009] AC 11, para 64. Reference had been made to In re H in two House of Lords cases which were concerned with two quite different statutes: see B v Chief Constable of Avon and Somerset Constabulary [2001] 1 WLR 340 (concerning sex offender orders) and R (McCann) v Crown Court at Manchester [2003] 1 AC 787 (concerning anti-social behaviour orders). These had led to suggestions that in care proceedings there was a “heightened standard of proof” and even that the difference between the criminal and civil standards was “largely illusory”. Those suggestions were firmly rejected by the Court of Appeal in In re U (A Child) (Department for Education and Skills intervening) [2005] Fam 134, and equally firmly by the House of Lords in In re B [2009] AC 11.

Hence, in In re S-B [2010] 1 AC 678, this court re-iterated that the simple balance of probabilities was the test for the identification of a perpetrator. There were cases where the judge might find this difficult, and he was under no obligation to do so if he could not. But there were particular benefits in doing so if he could. It would promote clarity in future planning for the child (para 36). It would also enable the professionals to work with the family on the basis of the judge’s findings (para 37). And it would help the child in due course to understand and come to terms with what had happened and why he might have been removed from his family (para 38).

It may well be that a misunderstanding of the standard of proof has been responsible for some of the difficulties. Trial judges have felt themselves unable to decide which parent was responsible for harming a child. This did not matter too much at the attributability and welfare stages, although it was bound to make the task of deciding what would be best for the child much more difficult, especially if the parents had separated after the harm was done. It clearly did matter at the stage of deciding whether another child was likely to suffer significant harm, but trial judges will no doubt have taken comfort from the dictaof Wall J in In re B [1999] 1 WLR 238 until the basis for predicting likelihood came before the House once more in In re B [2009] AC 11 and In re S-B [2010] 1 AC 678.

It is also argued that the approach established in the care cases, culminating in In re S-B, is out of step with the modern approach in criminal and civil law where it is clear that harm has been suffered but not clear who has caused it. Section 5 of the Domestic Violence, Crime and Victims Act 2004 (as amended in 2012) creates the offence of causing or allowing a child (or vulnerable adult) to die or suffer serious physical harm. If a child dies or suffers serious physical harm as a result of the unlawful act of a member of the child’s household who has frequent contact with the child, and there was at the time a significant risk of serious physical harm being caused to the child by such a person, then such a person is guilty of an offence if either (i) he or she was the person whose act caused the death or serious physical harm or (ii) he or she was or ought to have been aware of the significant risk, failed to take such steps as he or she could reasonably have been expected to take to protect the child from the risk, and the act occurred in circumstances of the kind that he or she foresaw or ought to have foreseen. The prosecution does not have to prove whether it is (i) or (ii), thus getting round the well-known problem that if a child was seriously injured or died at the hands of one of his parents, but it could not be proved which, both had to be acquitted: see, for example, R v Lane (1986) 82 Cr App R 5.

It will be immediately apparent that child care law already protects the child who is injured in such circumstances. He has suffered significant harm attributable to the care given to him not being what it would be reasonable to expect a parent to give to him. As the Lancashire case [2002] 2 AC 147 made clear, it is not necessary to identify the perpetrator of the harm in order to cross the threshold. Furthermore, child care law will protect another child who is a member of the same household where the other child was injured.

The criminal law is only concerned with punishing those who bear some responsibility, either directly or through a failure to protect from predictable harm, for a death or serious injury which has actually taken place. It does place a duty upon those looking after a child to protect him from the foreseeable risk of serious harm caused by others in the household. But child care law already does the same. A parent who fails to protect her child not only stands to lose that child but also stands to have that failure taken into account in predicting risks to other children. The criminal law is not addressing the question with which we are concerned, which is how the likelihood of future harm can reliably be predicted.

Developments in the civil law are even less in point. In Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32, the House of Lords created an exception to the general principle that a person who has suffered harm as a result of a breach of duty can only recover damages against a person who can be shown to have caused or contributed to that harm. That exception applies only where the claimant has contracted mesothelioma, has been exposed to asbestos from more than one source, and it cannot be shown which source was responsible for the disease. Any defendant who has wrongfully exposed the claimant to asbestos is liable. But in Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 AC 229, the Supreme Court made it clear that this is a special rule, created only because of the special difficulty of proving causation in mesothelioma cases. It can scarcely be described as the herald of a brave new world in which people may be held liable for harm whether or not they have caused it.

In any event, like the criminal law, the civil law is looking to the past, not the future. It seeks to compensate for harm already done, not to predict whether harm is likely in the future.

As this review has shown, there is no inconsistency between para 49 of In re S-B [2010] 1 AC 678and any of the earlier House of Lords authorities. On the contrary, although it addresses a different factual situation from those addressed in In re H [1996] AC 563 and In reB [2009] AC 11, it is entirely consistent with them in principle. Nor is it inconsistent with developments in the criminal and civil law. In my view, it represents a correct statement of the law, and should be followed in preference to the obiter dictaof Wall J in Re B [1999] 1 WLR 238, 248.

Time and again, the cases have stressed that the threshold conditions are there to protect both the child and his family from unwarranted interference by the state. There must be a clearly established objective basis for such interference. Without it, there would be no “pressing social need” for the state to interfere in the family life enjoyed by the child and his parents which is protected by article 8 of the ECHR. Reasonable suspicion is a sufficient basis for the authorities to investigate and even to take interim protective measures, but it cannot be a sufficient basis for the long term intervention, frequently involving permanent placement outside the family, which is entailed in a care order.

That view is supported by the legislative history. The single threshold contained in the Children Act 1989 replaced the many separate criteria for taking or keeping children away from their families which were contained in the previous law. Most care proceedings were brought in juvenile courts under section 1 of the Children and Young Persons Act 1969. This contained a list of specific grounds for making care orders, two of which catered for the likelihood of future harm: where it was probable that this child would suffer harm in future, either because another child in the same household had already been harmed, or because a person who had been convicted of harming another child had joined the household. During the 1970s, however, local authorities increasingly resorted to taking children into care by making them wards of court: the High Court (along with the divorce courts) had power, “in exceptional circumstances making it impracticable or undesirable” for the child to be cared for by his parents or any other individual, to commit a ward to the care of a local authority: section 7(2) of the Family Law Reform Act 1969.

The Review of Child Care Law: Report to Ministers of an Interdepartmental Working Party (1985) (DHSS), which led to the 1989 Act, rejected a broad welfare test, because it would not offer the degree of statutory protection against unwarranted interference which the preceding House of Commons Select Committee Report on Children in Care had considered essential (para 15.10). The Review also rejected the “exceptional circumstances” criterion, because it would add little to the broad welfare criterion and “leave too much to subjective interpretation” (para 15.11). Instead, it proposed a criterion of actual or likely harm (paras 15.12 to 15.25). Much of the discussion focussed on the definition of harm – the Review contemplated a substantial deficit in the standard of health, development or well-being which could reasonably be expected of the particular child, whereas the Act speaks of “significant” harm. The Review did not discuss the meaning of “likely” or how such a likelihood is to be predicted. It is clear, however, that the threshold was designed to be more precise than the previous wardship criterion and to focus upon what was seen as the key objective justification for state interference – that the child was suffering or likely to suffer harm.

The threshold comes in two limbs and each has two distinct components. In the first limb, the court must be satisfied (a) that the child is suffering significant harm, and (b) that that harm is attributable to the care given to him not being what it would be reasonable to expect a parent to give him. In the second limb, the court must be satisfied (a) that the child is likely to suffer significant harm, and (b) that that likelihood is attributable to the care likely to be given to him if the order is not made not being what it would be reasonable to expect a parent to give to him. It would, as Lord Nicholls pointed out in In re H [1996] AC 563, 591, be odd if the first limb had to be proved to the satisfaction of the court but the basis for prediction in the second limb did not.

Since In re H, a clear distinction has been drawn between the degree of likelihood required by the second limb and the basis upon which the court can be satisfied of that likelihood. It was held in In re H that “likely” did not mean “more likely than not”, but a “real possibility”, the degree of possibility required depending upon the seriousness of the harm which was feared. The House might have held that “likely” meant “probable”, but it adopted a more flexible test. It is worth noting that Lord Nicholls did not think that adopting this more flexible test would significantly prejudice the parents’ interests: “so far as the parents are concerned, there is no particular magic in a threshold test based on a probability of significant harm as distinct from a real possibility” (p 585D).

Having adopted a flexible test of likelihood, it became all the more important to hold that an objective factual basis was required from which to draw the inference that future harm was likely. This was controversial in In re H [1996] but has been firmly established ever since and for very good reasons. Care courts are often told that the best predictor of the future is the past. But prediction is only possible where the past facts are proved. A real possibility that something has happened in the past is not enough to predict that it will happen in the future. It may be the fact that a judge has found that there is a real possibility that something has happened. But that is not sufficient for this purpose. A finding of a real possibility that a child has suffered harm does not establish that he has. A finding of a real possibility that the harm which a child has suffered is “non-accidental” does not establish that it was. A finding of a real possibility that this parent harmed a child does not establish that she did. Only a finding that he has, it was, or she did, as the case may be, can be sufficient to found a prediction that because it has happened in the past the same is likely to happen in the future. Care courts need to hear this message loud and clear.

In In re S-B [2010] 1 AC 678, the “real possibility” that the mother had harmed J was the only basis upon which the judge concluded that it was likely that W would suffer harm in the future. There was nothing else. J had suffered bruises and all bruising to a tiny baby must be taken seriously. But they had probably been caused on one occasion by one parent. It could not be suggested that the other parent had failed to protect him. What was impermissible, as stated in para 49, was to make this, and this alone, the basis for predicting that the mother was likely to harm W in the future. It may well be that when the case was re-heard, facts emerged from which it was possible to make such a prediction.

Cases such as In re S-B are vanishingly rare. As McFarlane LJ pointed out in the Court of Appeal [2012] 3 WLR 952, para 109, the Lancashire case [2000] 2 AC 147, in respect of the child-minder, was “truly a one-point case. There were no other adverse findings made against the childminder” (he says “other” but he must mean no adverse findings) (para 108). Likewise, In re S-B “was [a case] of a one-off (‘whodunit’) injury, there was no question of failure to protect and no finding of collusion” (para 111). Even in In re F [2011] 2 FLR 856, there were no adverse findings against the father (para 112). Most care cases are not “one-off whodunit” cases. They come with a multitude of facts.

It is, of course, a fact that a previous child has been injured or even killed while in the same household as this parent. No-one has ever suggested that that fact should be ignored. Such a fact normally comes associated with innumerable other facts which may be relevant to the prediction of future harm to another child. How many injuries were there? When and how were they caused? On how many occasions were they inflicted? How obvious will they have been? Was the child in pain or unable to use his limbs? Would any ordinary parent have noticed this? Was there a delay in seeking medical attention? Was there concealment from or active deception of the authorities? What do those facts tell us about the child care capacities of the parent with whom we are concerned?

Then, of course, those facts must be set alongside other facts. What were the household circumstances at the time? Did drink and/or drugs feature? Was there violence between the adults? How have things changed since? Has this parent left the old relationship? Has she entered a new one? Is it different? What does this combination of facts tell us about the likelihood of harm to any of the individual children with whom the court is now concerned? Does what happened several years ago to a tiny baby in very different circumstances enable us to predict the likelihood of significant harm to much older children in a completely new household?

Hence I agree entirely with McFarlane LJ when he said that In re S-B is not authority for the proposition that “if you cannot identify the past perpetrator, you cannot establish future likelihood” (para 111). There may, or may not, be a multitude of established facts from which such a likelihood can be established. There is no substitute for a careful, individualised assessment of where those facts take one. But In re S-B is authority for the proposition that a real possibility that this parent has harmed a child in the past is not, by itself, sufficient to establish the likelihood that she will cause harm to another child in the future.

It follows that I would dismiss the local authority’s appeal on the single issue which has been identified for the decision of this court and the courts below. But what should be the consequence of that?

As McFarlane LJ pointed out, there were several facts found by Judge Masterman which might have been relevant to an assessment of whether it was likely that this mother would harm children in the future. There was “(a) gross and substantial collusion expressly designed to prevent the court identifying the perpetrator; (b) failure to protect T-L; (c) deliberately keeping T-L away from health professionals in order to avoid the detection of injury” (para 109). The local authority have chosen not to rely upon these. They acquiesced in the decision to treat this as a one point case. The result was that this mother returned to the household where she had previously been looking after the three subject children for some time without (as far as we know) giving any cause for concern. She has now been looking after her new baby for more than a year, also without (as far as we know) giving any cause for concern.

In those circumstances it would be most unfair to the whole family, not only to this mother, but also to her husband and all the children, for these proceedings to continue further. If the local authority wish to make a case that any of these children is likely to suffer significant harm in the future, they will have to bring new proceedings. The current application must remain dismissed.

I agree that the appeal should be dismissed. I will set out the reasons for my agreement in all save the final three of the following paragraphs of this judgment. Then I will identify an issue between myself and most other members of the court which, although important, does not affect the disposal of the appeal.

Where a child has suffered significant harm but the court is unable to identify its perpetrator or perpetrators, it will consign all those whose perpetration of it remains a real possibility to a pool of possible perpetrators. One of the most vexed issues in the public law relating to children surrounds the legal consequences of the consignment of a person, let us say X, to a pool of possible perpetrators. To be specific: at what stage or stages, and in what way, does a court appraise the risk posed to a child by X when his or her perpetration of significant harm to that child or to another child has been adjudged to be a real possibility yet no more than a real possibility?

It is agreed that, when the threshold to the making of a care or supervision order set by section 31(2) of the Children Act 1989 (“the Act”) has been crossed and the court proceeds to consider, under section 1(1) of the Act, whether it would serve the welfare of the child to make such an order, the consignment of a proposed carer to a pool of possible perpetrators of harm to that child or to another child must, with whatever degree of difficulty, be weighed in the balance: for it is relevant to the harm which the child is at risk of suffering (section 1(3)(e)) and to the capacity of X to meet his or her needs (section 1(3)(f)). So the court will, in particular:

It seems that occasionally, probably in order to aid the difficult task which in the future he or another judge may face of weighing, at the welfare stage, the significance of his consignment of X, along with another person (let us say Y), to a pool of possible perpetrators, a judge has proceeded to suggest that the real possibility that Y perpetrated the injuries is stronger than the real possibility that X did so. In a case in the Court of Appeal, namely In re T (Care Proceedings: Appeal) [2009] EWCA Civ 1208, [2010] 1 FLR 1325, I suggested, at para 62, that a judge who had made such a suggestion might have been trying to dance on the head of a pin. Three weeks later, in In re S-B (Children) (Care Proceedings: Standard of Proof) [2009] UKSC 17, [2010] 1 AC 678, para 44, the Supreme Court advised judges to be cautious about making such a suggestion. In my view judges would at least have to ask themselves (for, if they did not do so, an appellate court, as in In re T, might be invited to ask itself) whether their reasons for considering that the real possibility of perpetration was stronger in relation to Y than in relation to X were not, instead, reasons for concluding that the former had probably perpetrated the injuries and that therefore there should be no pool at all.

But is the consignment of a person to a pool of possible perpetrators also relevant to the issue whether the threshold set by section 31(2) of the Act is crossed?

The threshold has two parts: (a) that the child is suffering, or is likely to suffer, significant harm; and (b) that the harm or likelihood of harm is attributable either to the care given, or likely to be given, to him not being what it would be reasonable to expect a parent to give to him or to his being beyond parental control.

The phrase “not being what it would be reasonable to expect a parent to give” in section 31(2)(b) covers every case in which a court concludes that the care provided by one or other of the parents or by anyone else who shared the child’s care fell short of a reasonable standard, even if, provided that he must have been within those categories, the person whose care thus fell short cannot be identified: Lancashire County Council v B [2000] 2 AC 147. It is therefore only in few cases that subsection 2 (b), known as the attributability condition, presents any significant extra hurdle to the crossing of the threshold.

We may thus put the attributability condition to one side and ask, more narrowly, whether the consignment of a person to a pool of possible perpetrators is also relevant to whether the child is suffering, or is likely to suffer, significant harm. Indeed we may further narrow the question by also putting to one side the child who is suffering significant harm. For in such a case, and subject to the attributability condition, the threshold is crossed irrespective of whether the perpetrator of the harm to the child can be identified.

So this appeal surrounds the unharmed child, whom a local authority alleges to be likely to suffer significant harm at the hands of a person because she (or he) has been consigned to a pool of possible perpetrators of harm to another child.

The Court of Appeal directed substantial criticism at the appellant local authority, Stockton-on-Tees Borough Council, for having presented this application for care orders to the judge on the basis that the only fact upon which it relied for the crossing of the threshold was that the first respondent, JJ, had been consigned to a pool of possible perpetrators of injuries to her deceased child, T-L. In making this confined presentation Stockton took a deliberate decision not to rely on the findings of Judge Masterman that, had she not been the perpetrator of the injuries, JJ had failed to protect T-L, had kept her away from health professionals in order to avoid detection of her injuries and had colluded with T-L’s father in attempting to prevent the court from identifying their perpetrator.

The most powerful of the criticisms made in the Court of Appeal emanated from Lord Judge CJ. He suggested that if, on Stockton’s confined presentation, the ultimate verdict of the courts was to be that the threshold had not been crossed but if, on a wider presentation, the threshold would have been crossed and the court would have proceeded to make care orders, Stockton’s decision might turn out irresponsibly to have left the children at risk.

The Lord Chief Justice made a powerful point. It is no more than partly undermined by a provisional conclusion, derived from the limited material before the court, that it seems improbable that the welfare of the children would be served by their being taken into care or even by obliging JJ’s husband, namely DJ, to separate from her as a condition of the court’s declining to remove some or all of them from his care. Nevertheless I decline to join in the condemnation of Stockton. The relevance or otherwise to the threshold of a carer’s presence in a pool of perpetrators is a question which has rightly concerned local authorities and other child care professionals for years: see In re F (Interim Care Order) [2011] EWCA Civ 258, [2011] 2 FLR 856, para 15. Local authorities need to understand the parameters of their ability to obtain care and supervision orders. The only decision of the Supreme Court directly to have addressed the question, namely In re S-B (Children) (Care Proceedings: Standard of Proof), cited above, was primarily reached on another basis, which related to the standard of proof. The present appeal at last confers upon this court the opportunity, indeed imposes upon it the obligation, to decide the question, in effect once and for all. For Stockton’s severe pruning of its case has ensured that the decision cannot branch out in another direction.

Nevertheless in my view leading counsel for JJ and DJ are right to submit that it is a “vanishing rarity” for a past finding of a person’s (say a mother’s) possible perpetration of injuries to a child not to be accompanied by findings that, in any event, she had culpably ill-treated or neglected the child. The most usual such findings, which, as I have indicated, were all made by Judge Masterman in relation to JJ, are that, even assuming in her favour that she was not the perpetrator of the injuries, she had culpably failed to protect the child from them, had failed to seek medical attention for them and, by lying, had sought to disable the court from identifying her then partner as their perpetrator. Findings of such a character can be profoundly serious because they appear to betoken a mother’s willingness to sacrifice the elementary interests of a child to be safe and free from injury on the altar of some adult relationship. On reflection, I am not at all sure that such findings become irrelevant just because the mother is now living, as JJ is now living, with another partner who does not represent a risk to a subject child and from whom there is therefore no specific need for the child to be protected: for no doubt the child will continue to need protection from a variety of situations and from persons other than the new partner. More widely, such findings raise grave concerns about that mother’s entire capacity for responsible care; and, if marshalled by a local authority as the factual foundation for the crossing of the threshold, they would need most carefully to be weighed against such evidence as indicated an improvement in her capacity for responsible care as at the relevant date.

Stockton does not dispute that a prediction of likelihood of significant harm must be founded on proven facts. There is, of course, no express statement of such a requirement in the Act itself. It arises as a result of judicial interpretation at the highest level, namely in In re H (Minors) (Sexual Abuse: Standard of Proof)) [1996] AC 563; and the interpretation has been specifically reaffirmed in In re B (Children) (Care Proceedings: Standard of Proof) (CAFCASS intervening) [2008] UKHL 35, [2009] AC 11. I respectfully disagree with the view, which Lord Reed at para 93 describes as attractive, that the interpretation is over-complicated. At all events Stockton argues, instead, that a person’s consignment to a pool of possible perpetrators of harm to a child is itself a finding of fact which can found a prediction that a second child is likely to suffer significant harm; or, even if not, that, when added to a finding (being on any view a finding of fact) that the first child did suffer harm, it provides the requisite foundation upon which it would be open to the court to predict the likelihood.

Is a bald statement that there is a real possibility that X caused the injuries a statement of fact? I personally do not find the answer easy. But there is no need for an answer. For the consignment of two (or more) persons to a pool of possible perpetrators goes further than that: it is a statement not only that there is a real possibility that X caused the injuries and alternatively that Y caused them but also that no one else caused them, i.e that one or other of them did cause them. I am clear that such a conclusion does amount to a finding of fact. But is it a fact relevant to the threshold? Where X and Y remain together as a unit and are putting themselves forward as carers for an unharmed child, it is certainly relevant and might well suffice in enabling the threshold to be crossed in relation to that child; for in those circumstances the fact is that somebody in the child’s proposed home did perpetrate injuries to another child. But the difficulty arises in the case, reflective of the facts in this appeal, in which X and Y no longer remain together as a unit and in which only one of them, X, is put forward as a carer. The consignment of X to a pool is not a finding that X did cause the injuries. So the question arises whether in that situation such a finding of fact is relevant.

Here is the crux of this appeal. Stockton’s case is deeply illogical.

Stockton argues that the children are at risk of suffering significant harm because of the presence in their home of JJ. So the requisite foundational facts must relate to JJ. To point to no more than the fact that T-L suffered grave non-accidental injuries is, on any view, insufficient. A second fact is needed such as relates to JJ. Stockton seeks to make the link with her by relying on her consignment to the pool of possible perpetrators of the injuries. But her consignment to the pool is not a relevant fact because it falls short of ascribing their perpetration to her. The result is that there is no relevant second fact. In both In re H and In re B, cited above, it was clear that, if in each case the girl had suffered harm, it had been at the hands of her step-father; in the event, however, the court’s conclusion was no more than that there was a real possibility that she had suffered the harm. Stockton does not challenge the principle established by those cases that such a conclusion was insufficient to found a likelihood that the step-father would cause significant harm to other children. The present appeal presents the precisely obverse situation; there was no doubt that T-L had suffered harm and the only issue, which in the event was unable to be resolved save in terms of a real possibility, related to the identity of the perpetrator. There is in my view no basis for departing from that principle in this precisely obverse, yet analogous, situation. The harm and the person’s responsibility for it are the two planks on which any conclusion about likelihood must rest and they must be equally sturdy.

If and insofar (so Stockton proceeds to submit) as the basis of the appeal is illogical, then so be it; for otherwise legalism would triumph over child protection. But - in my view - logic is the blood which runs through the veins of the law: allow it to escape and ultimately the edifice collapses. Nor is the rigid approach to the factual foundation properly categorised as legalism. In In re H, cited above, Lord Nicholls commented, at p 592, on the unanimous conclusion of the House that, when it turned from looking at the past to look at the future, a likelihood of significant harm meant only a real possibility of it. “Therein”, said Lord Nicholls, “lies the protection for children.” He continued:

“But, as I read the Act, Parliament also decided that proof of the relevant facts is needed if this threshold is to be surmounted. Before the section 1 welfare test and the welfare ‘checklist’ can be applied, the threshold has to be crossed. Therein lies the protection for parents. They are not to be at risk of having their child taken from them and removed into the care of the local authority on the basis only of suspicions...”

My view remains that the need for the local authority to prove the facts which give rise to a real possibility of significant harm in the future is a bulwark against too ready an interference with family life on the part of the state. And, subject to the caveat that the court received no argument on the impact of article 8 of the European Convention on Human Rights, I incline to the view that nothing less than a factual foundation would justify such grave interference with the rights of the child and the parents thereunder to respect for their family life: see Olsson v Sweden (1989) 11 EHRR 259, in which, at paras 67 and 68 (which it has cited with approval on many subsequent occasions), the European Court of Human Rights stressed that a child’s removal into care was justified only if it was necessary in a democratic society in the sense that it corresponded to a pressing social need and was based on reasons which were relevant and sufficient.

In a final argument which found some favour with McFarlane LJ in the Court of Appeal, but which, with respect, I find puzzling, Stockton urges that the interpretation of the alternative limb of section 31(2)(a) of the Act should be no less flexible than the modern approach of the criminal law reflected in section 5 of the Domestic Violence, Crime and Victims Act 2004, as amended. Its general effect is that, where a child has died or been seriously harmed as a result of an unlawful act, a member of his household is guilty of an offence even if the jury can conclude no more than that either he perpetrated the act or he was or ought to have been aware that the child was at significant risk of harm and failed to take reasonable steps to protect him from it. In my view, however, there is no disparity of principle in this respect between the criminal law and the proper approach to the likelihood of significant harm in the subsection. In paras 103 to 109 of his judgment McFarlane LJ reverted to the findings of Judge Masterman that, if she did not perpetrate the injuries to T-L, JJ was at any rate culpable in the respects summarised at para 67 above; and the Lord Justice suggested that the effect of the dismissal of the appeal would be that such findings of culpability “must be ignored” in evaluating the likelihood of significant harm to another child. But they are to be ignored only because, in setting up the present proceedings as a test case, Stockton has invited the court to ignore them.

I therefore conclude that the mother’s consignment to the pool of possible perpetrators of the injuries to T-L is irrelevant to whether the three subject children are likely to suffer significant harm.

The issue between myself and most other members of this court arises out of the following observations:

First, I hasten to make clear that I respectfully agree with a different point made by Lord Hope at para 87, namely that, in addressing the likelihood that an uninjured child will suffer significant harm, the court can treat X’s consignment to a pool of possible perpetrators of injury to a second child as information which invites further inquiry. It would therefore go too far to say that, at the threshold inquiry, the consignment should be ignored. It might enable the court to find relevant facts established. There might, for example, be a disputed allegation that X had perpetrated the injuries which a third child had suffered; and the real possibility that X had perpetrated the injuries to the second child might, when added to the other evidence, lead the court to find the allegation established. X’s perpetration of the injuries to the third child would then be a relevant fact. By no alchemy, however, could X’s possible perpetration of the injuries to the second child then become a relevant fact.

My disagreement with most of my colleagues relates to their suggestion, made most explicitly by Lord Hope, that although X’s consignment to a pool cannot alone constitute a factual foundation for a prediction of likely significant harm, it can, if weighed together with other facts which are on any view relevant, figure as part of the requisite factual foundation. I feel driven to the conclusion that their suggestion is illogical; and that if, for the purpose of the requisite foundation, X’s consignment to a pool has a value of zero on its own, it can, for this purpose, have no greater value in company. I can only hope that their suggestion, highly authoritative though it will be, will not destabilise the requisite foundation; will not in practice lead to a prediction about a real possibility in the future being founded, in part, on no more than a real possibility about what happened in the past; and will not, by the back door, lead to the ascendancy over the majority view in In re H (Minors) [1996] AC 563 of the minority view which, so Lord Reed (with whom Lord Clarke and Lord Carnwath agree) indicates, would in principle have held some attraction for him.

It is not the function of the court to give rulings on points of law arising from a hypothetical state of facts. But I do not see this appeal, or the approach that has been taken by the local authority in pruning the case as it has done in order to present us with a single issue on which authoritative guidance is said to be needed, as open to that criticism. From time to time cases of the kind that require that sort of treatment do come up, and it is in the public interest that the court should respond to them in as helpful a way as possible. That is especially so where the problem to which its attention has been drawn is said to be due to a difficulty in applying the way judges have interpreted the provisions of a statute to a given set of facts, which need to be isolated from their usual surroundings in order to obtain a definitive ruling on the point.

The court will, of course, need to be satisfied that the issue which has been raised is one which does genuinely require attention, and is not itself an artificial one. It is, no doubt, highly unusual for cases under section 31(2) of the Children Act 1989 to be decided on such a narrow basis. The question whether the threshold has been crossed must almost always be decided on a consideration of all the facts. Had it not been for the need for guidance on the point that has been raised, that would certainly have been the proper way to approach this case. But I am entirely satisfied that the issue that has been presented to us in this appeal is one of general public importance and that we should deal with it.

I also think that such facts as we have at our disposal suggest that it would be wrong to criticise the local authority for having acted irresponsibly. JJ’s relationship with SW, with whom she was living at the time of T-L’s death, ended more than five years ago. Since then she has formed a new relationship with DJ, to whom she is now married, and she has been looking after the three children who are the subject of these proceedings for about four years and her new baby for more than a year. So far as we know, there has been no cause for concern. It would be open to the local authority to take the appropriate action at any time if there were grounds for thinking that the children were suffering or were likely to suffer significant harm. I am prepared to assume in the local authority’s favour that it would not have approached this case in the way it has done if there were reasons for thinking that it was putting the children at risk of suffering significant harm by doing so.

Turning to the issue that has been presented to us, the golden rule must surely be that a prediction of future harm has to be based on facts that have been proved on a balance of probabilities. I would respectfully endorse Lord Wilson’s observation in para 75, repeating what he said in In re F (Interim Care Order) [2011] EWCA Civ 258, [2011] 2 FLR 856 that this is a bulwark against too ready an interference with family life on the part of the state. One has only to reflect for a moment on the tragic consequences of some of the notorious cases of two decades ago to appreciate the wisdom of this remark. They were referred to by the First Division of the Court of Session in L, Petitioners (No 1) 1993 SLT 1310, 1313: see also Rochdale Borough Council v A [1991] 2 FLR 192; Sloan v B 1991 SC 412 (the Orkney case). L was a case where eight children from three separate families in Ayrshire were made the subject of place of safety orders under Part III of the Social Work (Scotland) Act 1968 on inadequate evidence. It was several years before those orders could be set aside, and it took many months of careful and skilled work after that to bring the families who had been split up in traumatic circumstances together again.

These were extreme examples. Many other cases of unjustified removal may be expected to occur if the golden rule is not adhered to. Unjustified removal can itself result in significant harm to the children – the very thing that section 31 was designed to prevent. It is not just the immediate distress of separation. Long-term relationships between parents and their children can be damaged too, as the court found in the Ayrshire case in the series of unreported hearings that it took to restore the situation as closely as possible to what it had been before the removal took place.

To identify what is and is not a “finding of fact” for the purposes of the threshold, which looks to the future, is only the first step in the analysis. Views may differ as to what is truly a “fact” for this purpose, but I do not think that one need dwell on this question. Anything that is proved on a balance of probabilities may be taken into account to see what can be made of it. The crucial steps are to identify what is relevant and what is not, and then to determine whether what is found to be relevant is sufficient for a finding that the threshold has been crossed. If it is not, the question will be whether there are other facts and circumstances that can be brought into account to satisfy this requirement.

Let us assume that the question is whether the child would be at risk while in the care of X. A finding that X is in the pool of two or more possible perpetrators of harm that a child sustained in the past is a finding of fact. It means that, because X is in the pool, it is possible that X was the perpetrator. If the perpetrators are still together, it will be relevant, and may on its own be sufficient, to show that the threshold has been crossed. If the parties have separated and X is the carer, I decline to say that a finding that X was in the pool will no longer be relevant. That is so for two reasons: first, because it is information which invites further inquiry as to whether the subsequent child is likely to suffer harm while in the care of X; and, second, because, in combination with other facts and circumstances that the inquiry reveals about X’s attitude or behaviour, it may help to show that this threshold has been crossed. It may have a bearing on the weight of the evidence when looked at as a whole, including an assessment of the balance of probabilities.

Each case will turn on its own facts, and the context in which the finding is being examined will vary from case to case. It is not possible, dealing with an issue about the risk of future harm in the abstract as we are here, to be more precise than that. So I do not, with respect, agree with Lord Wilson that the finding can never be relevant. But it will not on its own be sufficient. The crucial point is this: it cannot, and must not, be treated on its own as a finding of fact that it was X who caused or contributed to the injuries.

The question which has been put to us, as set out in the Statement of Facts and Issues, is whether (i) a finding that a child has suffered harm while in the care of more than one person and (ii) a finding that one or both of the carers have perpetrated that harm are findings of fact which may be relied on in subsequent proceedings relating to only one of the potential perpetrators, in support of a conclusion that a subsequent child is likely to suffer significant harm in a new family unit of which that potential perpetrator is part.

The answer which I would give, applying the test set out in para 49 of In re S-B (Children) (Care Proceedings: Standard of Proof) [2010] AC 678, is that these findings may be relied on only to the extent that they may be relevant to the issue the court has to decide. But to find that this information is relevant does not go far enough. This is because such findings would not be sufficient, on their own, to establish that a child in the new family unit was likely to suffer significant harm. If they are the only findings that are available, they must be disregarded in the assessment for lack of sufficiency. A prediction of future harm based on what has happened in the past will only be justified if one can link what has happened in the past directly and unequivocally with the person in the new family unit in whose care the subsequent child is living or will now live.

For these reasons, and for those given by Lady Hale, Lord Wilson (except in his final three paragraphs) and Lord Reed I too would dismiss the appeal.

I share the misgivings of Lord Wilson about the suggestion that something which has not been proved to the requisite standard and therefore has no weight of its own at the threshold stage, can nevertheless add to the weight of something else that has been proved. I therefore agree with the observations at paragraphs 78-80 of his judgment. With that reservation, I agree with the judgment of Lady Hale and would, like the rest of the court, dismiss the appeal.

I see the attraction of the view that the approach to the interpretation of section 31(2)(a) of the Children Act 1989 adopted by the majority in In re H (Minors) (Sexual Abuse: Standard of Proof) [1996] AC 563 was over-complicated and has had unfortunate consequences. As Lord Lloyd of Berwick said, at p 581:

“Parliament has asked a simple question: Is the court satisfied that there is a serious risk of significant harm in the future? The question should be capable of being answered without too much over-analysis.”

The number of subsequent appeals concerned with the interpretation of section 31(2)(a) suggests that the approach adopted in that case may have raised problems of interpretation or application calling for repeated clarification. This is unfortunate, especially in an area of the law which has to be applied on a daily basis by courts at all levels, and in which clarity is therefore of particular importance. Nevertheless, it is too late to reconsider the decision in In re H. The view of the majority has been followed in two later decisions at the highest level, as Lady Hale has explained, and it must in my view be treated as settled law.

The case of In re H was itself concerned with the question whether, where a child alleged that she had been sexually abused and that allegation was the sole cause for concern, the allegation had to be proved before the court could intervene. As Lord Nicholls of Birkenhead explained at p 591, the conclusion that it did was consistent with the terms of section 31(2)(a), which permit intervention where “the child concerned is suffering, or is likely to suffer, significant harm”. Why, it was asked, would Parliament have provided the former of those alternatives, if the likelihood of harm (construed as meaning a real possibility) would be sufficient in a case where previous harm was alleged but not proved?

As Lady Hale has explained at paras 48-50 of her judgment, if In re H established that a real possibility of past harm was an impermissible basis for inferring a real possibility of future harm, the decisions of the Court of Appeal in Lancashire County Council v B [2000] 2 AC 147, and of this court in In re S-B (Children) (Care Proceedings: Standard of Proof) [2009] UKSC 17; [2010] 1 AC 678,established that a real possibility that X was the perpetrator of past harm was an insufficient basis for inferring a real possibility that X might perpetrate future harm. This conclusion does not cause any difficulty in circumstances where all the possible perpetrators of the past harm are also responsible for the future care of the child in question: where for example a child is living with X and Y, proof that either X or Y previously harmed another child may provide a basis for inferring that either X or Y might harm the child in question, and that there is therefore a real possibility of that child’s being harmed. The court cannot on the other hand draw such an inference, following Lancashire County Council and In re S-B, where it is proved that either X or Y harmed another child but it is only X who is now responsible for the care of the child in question. In other words, where the person who harmed a child cannot be identified, the threshold cannot be met in relation to another child solely on the basis that a possible perpetrator of the harm is involved in the care of that child unless all possible perpetrators are so involved.

In substance, as it appears to me, the court is saying that, as a matter of law, a real possibility that X harmed another child in the past is not by itself a basis upon which the court can properly be satisfied that there is a likelihood that X will harm the child in question in the future. A basis for such a rule can be found, ultimately, in the language of section 31(2)(a): as Lady Hale has explained, the decision in In re H reflected the terms of that provision, and that decision, together with the subsequent decisions of the House of Lords and the Supreme Court, including the decision in In re S-B, form a coherent body of law. Support for such a rule can also be found in the contrast between section 31(2)(a) and section 43(1)(a), which applies where the court is satisfied that the applicant “has reasonable cause to suspect that the child is suffering, or is likely to suffer, significant harm”. The approach adopted in In re S-B also derives support from the legislative history of section 31(2), as Lady Hale has explained.

I am, with respect, less convinced, at least in the absence of any argument on the point, by Lord Wilson’s suggestion that it is an approach which the court is compelled to adopt in order to comply with article 8 of the European Convention for the Protection of Human Rights and Fundamental Freedoms. Article 8 is undoubtedly relevant to the circumstances in which a care order should be made, but there seems to me to be room for argument as to whether it is engaged by a provision which merely confers upon the court the jurisdiction enabling it to make such an order. As Lord Clyde explained in the Lancashire County Council case [2000] 2 AC 147, 170, section 31(2) merely opens the way to the possibility that an order may be made. A care order will only be granted by the court after careful consideration of the case in accordance with section 1, including consideration of the matters specified in section 1(3) and the requirement in section 1(5) that an order is only to be made if the making of it would be better for the child than the making of no order at all. The need for caution and restraint in the making of an order is further underlined by the requirement that any order be proportionate in order to comply with article 8 of the Convention.

Finally, I would observe that if, as has been said, the current law is causing consternation, that appears to me to be an over-reaction. It is important to emphasise, as Lady Hale has done at paras 52-54, that the court’s inability to establish whether X was the perpetrator of harm to a child in the past does not necessarily mean that the threshold set by section 31(2)(a) cannot be met in relation to a child now being cared for by X. It means however that some other cause for concern, besides the possibility that X was the perpetrator of the harm, must be established. The onus thereby imposed is, in a case of that kind, one which should ordinarily be capable of being discharged where substantial causes for concern currently exist. In practice, in the great majority of cases where a child has been harmed by one of its primary carers but it has not been possible to identify which of them was responsible, and only one of them is now responsible for the care of another child, it will be possible to establish facts on the basis of which a prognosis as to the future risk of harm can be made. The case at hand would itself appear to have been such a case, if the evidence before the court had not been deliberately restricted.

For these reasons, and those given by Lord Hope and Lady Hale, I agree that the appeal should be dismissed.