United Kingdom Supreme Court

You are here: BAILII >> Databases >> United Kingdom Supreme Court >> S-B Children, Re [2009] UKSC 17 (14 December 2009)
URL: http://www.bailii.org/uk/cases/UKSC/2009/17.html
Cite as: [2010] 2 WLR 238, [2010] PTSR 456, [2010] Fam Law 231, [2010] 1 All ER 705, [2010] 1 AC 678, [2010] AC 678, [2010] 1 FCR 321, [2009] UKSC 17, [2010] 1 FLR 1161

[New search] [Printable PDF version] [Buy ICLR report: [2010] PTSR 456] [Buy ICLR report: [2010] 1 AC 678] [Buy ICLR report: [2010] 2 WLR 238] [Help]

Court Order

Michaelmas Term
[2009] UKSC 17
On appeal from: [2009] EWCA Civ 1048

JUDGMENT

S-B Children

before

Lord Hope, Deputy President
Lord Rodger
Lady Hale
Lord Brown
Lord Collins
Lord Kerr
Lord Clarke

JUDGMENT GIVEN ON

14 December 2009

Heard on 25 and 26 November 2009

Appellant Anthony Hayden QC Magdalen Case (Instructed by Dawson Cornwell)		Respondent Susan Grocott QC Sasha Watkinson (Instructed by Trafford Borough Council Legal and Democratic Services)

		2nd Respondent Frances Judd QC Alexander Kloss (Instructed by Rowlands Solicitors)

LADY HALE

This is the judgment of the court.

This case is about the proper approach to deciding who has been responsible for harming a child in proceedings taken to protect that child, and others in the family, from harm. It raises profound issues: on the one hand, children need to be protected from harm; but on the other hand, both they and their families need to be protected from the injustice and potential damage to their whole futures done by removing children from a parent who is not, in fact, responsible for causing them any harm at all. The facts of this case present us with that dilemma in an unusually stark form.

The facts

Because we have decided to allow this appeal and send the case back to be decided afresh, we should say only enough about the facts to explain how the dilemma arises. We shall use pseudonyms for the two children concerned, one who has been harmed and one who has not. Jason was born on 19 May 2007. On 15 June 2007, when he was just four weeks old, he was found to have bruising on his arms and face, which the doctors immediately thought was caused non-accidentally and not, as the mother suggested, by the baby pinching himself or sleeping on his dummy. Jason has not lived with his family since then, although he has had frequent and good quality contact with his mother.

Jason was living with his mother and father at the time and described by the doctors as "thriving". Both parents said that it was the father who had got up to attend to the baby when he woke up on the morning when the bruises were noticed. The mother took the baby to the clinic that morning and pointed them out to the health visitor. It was not possible to give precise timing for the bruises but it was not suggested that they were old or of different ages. They could have been inflicted by both parents, but the judge found it more likely that only one of them had inflicted them. The bruises had not been there for so long, nor would they have caused the baby such pain and distress, that the other parent must have known that he was being harmed. This was not, therefore, a case where one parent had failed to protect the child from harm caused by the other. It was, colloquially, a pure "whodunit".

The other child is William, born on 12 July 2008, while the proceedings to protect Jason were in train. By then the parents had separated, although they were still in touch with one another. The father had stopped visiting Jason, had withdrawn from co-operation with the social workers and with his solicitors, and played no further part in the proceedings. He has parental responsibility for Jason but not for William. William was removed from his mother shortly after birth and placed with the same foster carer as his brother. He has never been harmed. The case for removing him from his mother rests on the likelihood of his being harmed in the future if he is returned to her.

The law

In this country we take the removal of children from their families extremely seriously. The Children Act 1989 was passed almost a decade before the Human Rights Act 1998, but its provisions were informed by the United Kingdom's obligations under article 8 and article 6 of the European Convention on Human Rights. These affect both the test and the process for intervening in the family lives of children and their parents.

As to the test, it is not enough that the social workers, the experts or the court think that a child would be better off living with another family. That would be social engineering of a kind which is not permitted in a democratic society. The jurisprudence of the European Court of Human Rights requires that there be a "pressing social need" for intervention and that the intervention be proportionate to that need. Before the court can consider what would be best for the child, therefore, section 31(2) of the 1989 Act requires that it be satisfied of the so-called "threshold conditions":

"(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."

The leading case on the interpretation of these conditions is the decision of the House of Lords in Re H (Minors) (Sexual Abuse: Standard of Proof) [1996] AC 563. Three propositions were established which have not been questioned since. First, it is not enough that the court suspects that a child may have suffered significant harm or that there was a real possibility that he did. If the case is based on actual harm, the court must be satisfied on the balance of probabilities that the child was actually harmed. Second, if the case is based on the likelihood of future harm, the court must be satisfied on the balance of probabilities that the facts upon which that prediction was based did actually happen. It is not enough that they may have done so or that there was a real possibility that they did. Third, however, if the case is based on the likelihood of future harm, the court does not have to be satisfied that such harm is more likely than not to happen. It is enough that there is "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).

Thus the law has drawn a clear distinction between probability as it applies to past facts and probability as it applies to future predictions. Past facts must be proved to have happened on the balance of probabilities, that is, that it is more likely than not that they did happen. Predictions about future facts need only be based upon a degree of likelihood that they will happen which is sufficient to justify preventive action. This will depend upon the nature and gravity of the harm: a lesser degree of likelihood that the child will be killed will justify immediate preventive action than the degree of likelihood that the child will not be sent to school.

The House of Lords was invited to revisit the standard of proof of past facts in Re B (Children) (Care Proceedings: Standard of Proof) [2008] UKHL 35, [2009] AC 11, where the judge had been unable to decide whether the alleged abuse had taken place. The suggestion that it would be sufficient if there were a "real possibility" that the child had been abused was unanimously rejected. The House also reaffirmed that the standard of proof of past facts was the simple balance of probabilities, no more and no less.

The problem had arisen, as Lord Hoffmann explained, because of dicta which suggested that the standard of proof might vary with the gravity of the misconduct alleged or even the seriousness of the consequences for the person concerned (para 5). He pointed out that the cases in which such statements were made fell into three categories. In the first were cases which the law classed as civil but in which the criminal standard was appropriate. Into this category came sex offender orders and anti-social behaviour orders: see B v Chief Constable of Avon and Somerset Constabulary [2001] 1 WLR 340 and R (McCann) v Crown Court at Manchester [2002] UKHL 39, [2003] 1 AC 787. In the second were cases which were not about the standard of proof at all, but about the quality of evidence. If an event is inherently improbable, it may take better evidence to persuade the judge that it has happened than would be required if the event were a commonplace. This was what Lord Nicholls was discussing in Re H (Minors), above, at p 586. Yet, despite the care that Lord Nicholls had taken to explain that having regard to the inherent probabilities did not mean that the standard of proof was higher, others had referred to a "heightened standard of proof" where the allegations were serious. In the third category, therefore, were cases in which the judges were simply confused about whether they were talking about the standard of proof or the role of inherent probabilities in deciding whether it had been discharged. Apart from cases in the first category, therefore, "the time has come to say, once and for all, that there is only one civil standard of proof and that is proof that that the fact in issue more probably occurred than not" (para 13).

This did, of course, leave a role for inherent probabilities in considering whether it was more likely than not that an event had taken place. But, as Lord Hoffmann went on to point out at para 15, there was no necessary connection between seriousness and inherent probability:

"It would be absurd to suggest that the tribunal must in all cases assume that serious conduct is unlikely to have occurred. In many cases, the other evidence will show that it was all too likely. If, for example, it is clear that a child was assaulted by one or other of two people, it would make no sense to start one's reasoning by saying that assaulting children is a serious matter and therefore neither of them is likely to have done so. The fact is that one of them did and the question for the tribunal is simply whether it is more probable that one rather than the other was the perpetrator."

Lady Hale made the same point, at para 73:

"It may be unlikely that any person looking after a baby would take him by the wrist and swing him against the wall, causing multiple fractures and other injuries. But once the evidence is clear that that is indeed what has happened to the child, it ceases to be improbable. Someone looking after the child at the relevant time must have done it. The inherent improbability of the event has no relevance to deciding who that was. The simple balance of probabilities test should be applied."

None of the parties in this case has invited the Supreme Court to depart from those observations, nor have they supported the comment made in the Court of Appeal that Re B "was a 'sweeping departure' from the earlier authorities in the House of Lords in relation to child abuse, most obviously the case of Re H" ([2009] EWCA Civ 1048, para 14). All are agreed that Re B reaffirmed the principles adopted in Re H while rejecting the nostrum, "the more serious the allegation, the more cogent the evidence needed to prove it", which had become a commonplace but was a misinterpretation of what Lord Nicholls had in fact said.

Re B was not a new departure in any context. Lord Hoffmann was merely repeating with emphasis what he had said in Secretary of State for the Home Department v Rehman [2001] UKHL 47, [2003] 1 AC 153, at para 55. A differently constituted House of Lords applied the same approach in Re D (Secretary of State for Northern Ireland intervening) [2008] UKHL 33, [2008] 1 WLR 1499.

In Re B, the House also declined an invitation to overrule the decision of the Court of Appeal in Re M and R (Minors) (Sexual Abuse: Expert Evidence) [1996] 4 All ER 239. This was concerned with the stage after the court is satisfied that the threshold has been crossed. The court has then to decide what order, if any, to make. The welfare of the child is the paramount consideration: 1989 Act, s 1(1). In deciding whether or not to make a care or supervision order, the court must have regard in particular to the so-called "checklist" of factors: 1989 Act, s 1(3), (4). These include "(e) any harm which he has suffered or is at risk of suffering".

In Re M and R, the Court of Appeal determined that section 1(3)(e) should be interpreted in the same way as section 31(2)(a). The court must reach a decision based on facts, not on suspicion or doubts. Butler-Sloss LJ said this:

"[Counsel's] point was that if there is a real possibility of harm in the past, then it must follow (if nothing is done) that there is a risk of harm in the future. To our minds, however, this proposition contains a non sequitur. The fact that there might have been harm in the past does not establish the risk of harm in the future. The very highest it can be put is that what might possibly have happened in the past means that there may possibly be a risk of the same thing happening in the future. Section 1(3)(e), however, does not deal with what might possibly have happened or what future risk there may possibly be. It speaks in terms of what has happened or what is at risk of happening. Thus, what the court must do (when the matter is in issue) is to decide whether the evidence establishes harm or the risk of harm."

In agreeing with this approach in Re B, at para 56, Lady Hale commented that in such a case, "as indicated by Butler-Sloss LJ …, the 'risk' is not an actual risk to the child but a risk that the judge has got it wrong. We are all fallible human beings, very capable of getting things wrong. But until it has been shown that we have, it has not been shown that the child is in fact at any risk at all". Re M and R was also approved by Lord Nicholls in Re O and another (Minors) (Care: Preliminary Hearing) [2003] UKHL 18, [2004] 1 AC 523, a case to which we shall return.

The House in Re B also recognised that courts and local authorities have different roles to play in protecting children from harm. It is worth re-emphasising this, given the understandable concerns in the wake of the "Baby P" case that social workers and other professionals were not being sufficiently active in their protective role, and the resulting increase in the numbers of care proceedings. Social workers are the detectives. They amass a great deal of information about a child and his family. They assess risk factors. They devise plans. They put the evidence which they have assembled before a court and ask for an order.

Article 6 of the European Convention on Human Rights requires that "In the determination of his civil rights and obligations, . . . everyone is entitled to a fair and public hearing within a reasonable time by an independent and impartial tribunal established by law". The court subjects the evidence of the local authority to critical scrutiny, finds what the facts are, makes predictions based upon the facts, and balances a range of considerations in deciding what will be best for the child. We should no more expect every case which a local authority brings to court to result in an order than we should expect every prosecution brought by the CPS to result in a conviction. The standard of proof may be different, but the roles of the social workers and the prosecutors are similar. They bring to court those cases where there is a good case to answer. It is for the court to decide whether the case is made out. If every child protection case were to result in an order, it would mean either that local authorities were not bringing enough cases to court or that the courts were not subjecting those cases to a sufficiently rigorous scrutiny.

The "whodunit" problem

So far the position is plain. But the threshold criteria do not in terms require that the person whose parental responsibility for the child is to be interfered with or even taken away by the order be responsible for the harm which the child has suffered or is likely to suffer in the future. It requires simply that "the harm, or likelihood of harm, is attributable to … 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". Clearly, the object is to limit intervention to certain kinds of harm – harm which should not happen if a child is being looked after properly. But is it also intended to limit intervention to cases where the person whose rights are to be interfered with bears some responsibility for the harm?

It cannot have been intended that a parent whose child has been harmed as a result of a lack of proper care in a hospital or at school should be at risk of losing her child. The problem could be approached through the welfare test, because removal from home would not be in the best interests of such a child. However, because of the risk of social engineering, the threshold criteria were meant to screen out those cases where the family should not be put at any risk of intervention. Hence attention has focussed on the attributability criterion. In the case confusingly reported in the Law Reports as Lancashire County Council v B [2000] 2 AC 147, but in the All England Law Reports as Lancashire County Council v A [2000] 2 All ER 97, the House of Lords considered what is meant by "the care given to the child". Does it mean only the care given by the parents or primary carers or does it mean the care given by anyone who plays a part in the child's care? Lord Nicholls, with whom Lord Slynn, Lord Nolan and Lord Hoffmann agreed, found that it referred primarily to the former. But if, as in that case, the care of the child was shared between two households and the judge could not decide which was responsible for the harm suffered by the child, the phrase "is apt to embrace not merely the care given by the parents or other primary carers; it is apt to embrace the care given by any of the carers" (p 166). Thus the criteria were satisfied in respect of a child, A, who had been injured, even though this might have been attributable to the care she had received from her childminder rather than from her parents.

Lord Clyde put the test in this helpful way, at p 169C, with the same result:

"That the harm must be attributable to the care given to the child requires that the harm must be attributable to the acts or omissions of someone who has the care of the child and the acts or omissions must occur in the course of the exercise of that care. To have the care of a child comprises more than being in a position where a duty of care towards the child may exist. It involves the undertaking of the task of looking after the child."

However, it is worth noting that the Court of Appeal had confirmed that the criteria were not satisfied in respect of the childminder's child, B, because he had not been harmed at all. The only basis for suggesting that there was any likelihood of harm to him was the possibility that his mother had harmed the other child and that had not been proved: Re H applied. The local authority did not appeal against this.

Re O and another (Minors)(Care: Preliminary Hearing) [2003] UKHL 18, [2004] 1 AC 523 was concerned with the more common problem, where the child has been harmed at the hands of one of his parents but the court cannot decide which. The attributability condition was satisfied. Furthermore, when considering the welfare test, the court had to proceed on the basis that the child was at risk. Lord Nicholls, with whom all other members of the Committee agreed, said this, at para 27:

"Quite simply, it would be grotesque if such a case had to proceed at the welfare stage on the footing that, because neither parent, considered individually, has been proved to be the perpetrator, therefore the child is not at risk from either of them. This would be grotesque because it would mean the court would proceed on the footing that neither parent represents a risk even though one or other of them was the perpetrator of the harm in question."

Lord Nicholls went on, at para 32, to give the following guidance, on the assumption that the hearing would be split into a "fact-finding" and a "disposal" stage and that each might be heard by a different judge:

". . . the judge at the disposal hearing will take into account any views expressed by the judge at the preliminary hearing on the likelihood that one carer was or was not the perpetrator, or a perpetrator, of the inflicted injuries. Depending on the circumstances, these views may be of considerable value in deciding the outcome of the application: for instance, whether the child should be rehabilitated with his mother."

In Re B, Lady Hale commented as follows at para 61:

"The decisions in In re H, Lancashire County Council v B [2000] 2 AC 147, and In re O [2004] 1 AC 523 fit together as a coherent whole. The court must first be satisfied that the harm or likelihood of harm exists. Once that is established, . . . ,the court has to decide what outcome will be best for the child. It is very much easier to decide upon a solution if the relative responsibility of the child's carers for the harm which she or another child has suffered can also be established. But the court cannot shut its eyes to the undoubted harm which has been suffered simply because it does not know who was responsible. The real answers to the dilemma posed by those cases lie elsewhere – first, in a proper approach to the standard of proof, and second, in ensuring that the same judge hears the whole case. Split hearings are one thing; split judging is quite another."

We are told that practice has now changed and that, barring accidents, the same judge does conduct both parts of a split hearing. Nevertheless, the main object of splitting the hearing is to enable facts to be found. If the threshold is not crossed, the case can be dismissed at that stage. If it is crossed, the professionals can base both their assessments and their further work with the family upon the facts found. It is not at all uncommon for parents to become much more open with the professionals when faced with the judge's clear findings based upon what the evidence shows. Hence there should always be a judgment to explain his findings at that stage.

These proceedings

It was necessary to give the above account of the development of the law in order to understand what happened in these proceedings. The case was originally identified as suitable for a split hearing; then it was decided to hold a composite hearing; but for regrettable practical reasons, the hearing was split once more. By that stage, the father was playing no part, but for some unknown reason the local authority decided not to issue a witness summons to require his attendance. That is regrettable because the judge might well have found it easier to make clear findings had he given evidence. The mother played a full part in the proceedings and in the assessments, but only accepted that the bruises were non-accidentally caused after the possibility of a blood disorder had effectively been ruled out.

The judge heard evidence over three days in January 2008 and three further days in March. She handed down a detailed judgment in note form on 3 April. This was before the House of Lords' decision in Re B. At the outset, under the heading 'Test', she directed herself as follows:

"The test I have applied in relation to these findings is that set out in the House of Lords case of [Re H] of 1996. The standard of proof I apply is on the balance of probability. The allegations in this case are very serious indeed and in many respects are also very unusual. When I apply the appropriate standard of proof, it has to be based on evidence of reliability and cogency equivalent to the gravity of the allegations."

She then listed five questions, three of which are relevant to the issue before this Court: first, whether the child had suffered non-accidental injury; second whether the perpetrator could be identified; and third "even if the perpetrator cannot be identified, can either of the parents be excluded as a perpetrator?" However, having concluded that the injuries were non-accidental, she did not in terms ask herself whether she could identify the perpetrator. She simply listed the various factors which she took into account in relation to each parent. She indicated at the outset of her list relating to the father that "there is a high index of suspicion in relation to the father" and concluded that he could not be ruled out. There was no such index in relation to the mother but for a variety of reasons the judge also concluded that the mother could not be ruled out.

The final hearing was listed for 5 June but could not proceed. As suggested in Re O, the judge was invited to give an indication of the relative likelihood of father or mother being responsible for the injuries, in order to assist with the assessment process. In oral exchanges she indicated that it was more likely that the father was the perpetrator than the mother. In a written "Adjunct to Judgment" she explained that "Invidious though it is to be too specific, but to help further assessments, I am prepared to say that I feel it 60% likely that the father injured the child and 40% likely that it was the mother."

The final hearing eventually took place before the same judge in December 2008 with judgment in January 2009. Part of the reason for the delay was that the mother had been unwell following the birth of her second child, William, in July. At the final hearing, the judge was invited to revisit her findings in the light of Re B, in which judgment was given on 11 June 2008. She declined to say that her finding meant that the father was the perpetrator of the injuries. She observed that:

"When one is deciding these issues, a judge frequently reluctantly comes to the conclusion that he cannot decide who is to blame between two parents or among more than two people who have had care of the child over the relevant period. However, although unable to form a definitive decision to the requisite standard, a judge can still have an impression, falling short of a finding, that the propensity of the parties and the surrounding circumstances make it more likely that it was one party than another."

Hence the mother was not "absolved as a really possible or likely perpetrator". This meant that the threshold was crossed, not only in relation to the child who had suffered harm, but also in relation to the child who had not. The fact that there was a real possibility that she had caused the injuries to Jason meant that there was a real possibility that she would injure William.

After considering the welfare factors she concluded that the mother's vulnerable personality was such that she would need therapy in order to make the necessary changes so that she could provide a safe and stable upbringing for the children. Their lives could not be put on hold in the meantime. Hence the judge approved the care plan to place them both for adoption and made care and placement orders in respect of both children. She did, however, give the mother permission to appeal but this was not included in the original order drawn up by the court.

Lord Justice Wall also gave permission to appeal, observing that the case "provides a useful opportunity for the Court of Appeal to resolve a point which has arisen following the decision of the House of Lords in Re B, namely (1) if only parents are 'in the frame' for having injured a child but (2) the judge cannot as between parents identify the perpetrator of the injuries, can that judge (3) apportion likely responsibility between them?" Before the Court of Appeal, however, this was not the main issue. It was argued that, following Re B, the test for identifying the perpetrator was the balance of probabilities and that the effect of the "Adjunct to judgment" was that this judge had in fact identified the father. The appeal was dismissed: [2009] EWCA Civ 1048.

Identifying the perpetrator: the standard of proof

The first question listed in the statement of facts and issues is whether it is now settled law that the test to be applied to the identification of perpetrators is the balance of probabilities. The parties are agreed that it is and they are right. It is correct, as the Court of Appeal observed, that Re B was not directly concerned with the identification of perpetrators but with whether the child had been harmed. However, the observations of Lord Hoffmann and Lady Hale, quoted at paragraph 12 above, make it clear that the same approach is to be applied to the identification of perpetrators as to any other factual issue in the case. This issue shows quite clearly that there is no necessary connection between the seriousness of an allegation and the improbability that it has taken place. The test is the balance of probabilities, nothing more and nothing less.

Of course, it may be difficult for the judge to decide, even on the balance of probabilities, who has caused the harm to the child. There is no obligation to do so. As we have already seen, unlike a finding of harm, it is not a necessary ingredient of the threshold criteria. As Lord Justice Wall put it in Re D (Care Proceedings: Preliminary Hearings) [2009] EWCA Civ 472, [2009] 2 FLR 668, at para 12, judges should not strain to identify the perpetrator as a result of the decision in Re B:

"If an individual perpetrator can be properly identified on the balance of probabilities, then . . . it is the judge's duty to identify him or her. But the judge should not start from the premise that it will only be in an exceptional case that it will not be possible to make such an identification."

There are particular benefits in making such a finding in this context, especially where there is a split hearing. Miss Frances Judd QC, on behalf of the children's guardian in this case, has stressed that the guardian would rather have a finding on the balance of probabilities than no finding at all. There are many reasons for this. The main reason is that it will promote clarity in identifying the future risks to the child and the strategies necessary to protect him from them. For example, a different care plan may be indicated if there is a risk that the parent in question will ill-treat or abuse the child from the plan that may be indicated if there is a risk that she will be vulnerable to relationships with men who may ill-treat or abuse the child.

Another important reason is that it will enable the professionals to work with the parent and other members of the family on the basis of the judge's findings. As the Court of Appeal said in Re K (Non-Accidental Injuries: Perpetrator: New Evidence) [2004] EWCA Civ 1181, [2005] 1 FLR 285, at para 55:

"It is paradigmatic of such cases that the perpetrator denies responsibility and that those close to or emotionally engaged with the perpetrator likewise deny any knowledge of how the injuries occurred. Any process, which encourages or facilitates frankness, is, accordingly, in our view, to be welcomed in principle."

Often, it is not only the parents, but the grandparents and other members of the family, who may be the best resource to protect the child in the future but who are understandably reluctant to accept that someone close to them could be responsible for injuring a child. Once that fact is brought home to them by a clear finding based upon the evidence, they may be able to work with the professionals to keep the child within the family.

Re K also suggested, at para 56, that there would be long term benefits for the child, whatever the outcome of the proceedings:

". . . we are also of the view that it is in the public interest that children have the right, as they grow into adulthood, to know the truth about who injured them when they were children, and why. Children who are removed from their parents as a result of non-accidental injuries have in due course to come to terms with the fact that one or both of their parents injured them. This is a heavy burden for any child to bear. In principle, children need to know the truth if the truth can be ascertained."

If the judge cannot identify a perpetrator?

The second and third questions in the statement of facts and issues ask whether judges should refrain from seeking to identify perpetrators at all if they are unable to do so on the civil standard and whether they should now be discouraged from expressing a view on the comparative likelihood as between possible perpetrators. These appear to be linked but they are distinct.

As to the second, if the judge cannot identify a perpetrator or perpetrators, it is still important to identify the pool of possible perpetrators. Sometimes this will be necessary in order to fulfil the "attributability" criterion. If the harm has been caused by someone outside the home or family, for example at school or in hospital or by a stranger, then it is not attributable to the parental care unless it would have been reasonable to expect a parent to have prevented it. Sometimes it will desirable for the same reasons as those given above. It will help to identify the real risks to the child and the steps needed to protect him. It will help the professionals in working with the family. And it will be of value to the child in the long run.

In North Yorkshire County Council v SA [2003] EWCA Civ 839, [2003] 2 FLR 849, the child had suffered non-accidental injury on two occasions. Four people had looked after the child during the relevant time for the more recent injury and a large number of people might have been responsible for the older injury. The Court of Appeal held that the judge had been wrong to apply a "no possibility" test when identifying the pool of possible perpetrators. This was far too wide. Dame Elizabeth Butler-Sloss P, at para 26, preferred a test of a "likelihood or real possibility".

Miss Susan Grocott QC, for the local authority, has suggested that this is where confusion has crept in, because in Re H this test was adopted in relation to the prediction of the likelihood of future harm for the purpose of the threshold criteria. It was not intended as a test for identification of possible perpetrators.

That may be so, but there are real advantages in adopting this approach. The cases are littered with references to a "finding of exculpation" or to "ruling out" a particular person as responsible for the harm suffered. This is, as the President indicated, to set the bar far too high. It suggests that parents and other carers are expected to prove their innocence beyond reasonable doubt. If the evidence is not such as to establish responsibility on the balance of probabilities it should nevertheless be such as to establish whether there is a real possibility that a particular person was involved. When looking at how best to protect the child and provide for his future, the judge will have to consider the strength of that possibility as part of the overall circumstances of the case.

As to the third question, times have changed since Re O. Barring unforeseen accidents, the same judge will preside over both parts of the hearing. While it is helpful to have a finding as to who caused the injuries if such a finding can be made, the guardian's view is that it is positively unhelpful to have the sort of indication of percentages that the judge was invited to give in this case. Lord Justice Thorpe suggested, [2009] EWCA Civ 1048, para 17, that judges should be cautious about amplifying a judgment in which they have been unable to identify a perpetrator: "better to leave it thus". We agree.

The unasked question

If the judge can identify a perpetrator on the balance of probabilities, what is to be done about the risk that he may be wrong and that some-one else was in fact responsible? We are indeed all fallible human beings. We can make mistakes, however hard we try to pay careful attention to the quality of the evidence before us and reach findings which are rationally based upon it.

However, once the court has identified a perpetrator, the risk is not a proven risk to the child but a risk that the judge has got it wrong. Logically and sensibly, although the judge cannot discount that risk while continuing to hear the case, he cannot use it to conclude that there is a proven risk to the child. But all the evidence (if accepted by the judge) relating to all the risk factors that the judge has identified remains relevant in deciding what will be best for the child. And he must remain alive to the possibility of mistake and be prepared to think again if evidence emerges which casts new light on the evidence which led to the earlier findings. It is now well settled that a judge in care proceedings is entitled to revisit an earlier identification of the perpetrator if fresh evidence warrants this (and this Court saw an example of this in the recent case of Re I (A Child) [2009] UKSC 10). The guardian also submits that the professionals will find it easier to work with this approach.

It is important not to exaggerate the extent of the problem. It only really arises in split hearings, which were not originally envisaged when the Children Act was passed. In a single hearing the judge will know what findings of fact have to be made to support his conclusions both as to the threshold and as to the future welfare of the child. Moreover, cases rarely come as neatly packaged as this one does. In most cases, the injuries are such that, even if one parent was not responsible for causing them, she was undoubtedly responsible for failing to protect the child from the person who did cause them. In many cases, there are other risks to the child besides the risk of physical injury. The evidence which is relevant to identifying the perpetrator will also be relevant to identifying the other risks to the child and to assessing what will be best for him in the future. But clearly the steps needed to protect against some risks will be different from the steps needed to protect against others. And the overall calculus of what will be best for the child in the future will be affected by the nature and extent of the identified risks. There are many, many factors bearing upon the child's best interests and the identification of risks is only one of them.

The conclusion in this case

We have every sympathy for the judge, who was only repeating the mantra which many other judges at every level had repeated in the past. But it is clear that she did misdirect herself on the standard of proof at the fact-finding hearing. Because she later said that she had simply been unable to decide, we do not think that we can accept the invitation of Mr Anthony Hayden QC, on behalf of the mother, to treat her "Adjunct to judgment" as a finding that the father was the perpetrator. That was not what she thought she was doing. However, that was an ex post facto rationalisation on her part. We cannot know what finding she would have made had she directed herself correctly in the first place. It is only right, for the sake of these children and their mother, that they should have the whole case put before a different judge who can decide the matter on the right basis.

There is a further reason to remit the case. The judge found the threshold crossed in relation to William on the basis that there was a real possibility that the mother had injured Jason. That, as already explained, is not a permissible approach to a finding of likelihood of future harm. It was established in Re H and confirmed in Re O, 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 Re H. It might have been open to the judge to find the threshold crossed in relation to William on a different basis, but she did not do so.

The case may look very different now that the mother's life has moved on and in the mean time, thankfully, the children have been well protected from harm. The appeal is therefore allowed and the case remitted for a complete rehearing before a different judge.