(Court of Appeal, Pitchford, Black LJJ, Sir Stanley Burnton, 6 February 2014)
The 16-month-old child attended hospital following a seizure and was discharged on the same day. She remained unwell over the following days and investigations disclosed sustained bilateral subdural collections and prominent subarachnoid spaces over both cerebral hemispheres. As a result the father issued proceedings seeking residence of the child and her sibling.
The child was found to have suffered a non-accidental injury caused by the mother's partner based upon several features of the medical evidence and the fact that he had invented traumatic incidents. The mother was found to have failed to protect the children and remained adamant that there was no possibility her partner had caused the injuries. The mother's partner appealed.
The appeal was allowed. The judge had failed to adequately explain how several features of the medical evidence supported a traumatic cause. At best, the medical evidence had not helped a great deal. The judge had not mishandled the non-medical factors which didn't point away from the partner's guilt but the evidence against him was not robust. It seemed that the support the judge found in the medical evidence for a traumatic cause might have propelled her to a finding against the partner that she might not have otherwise made.
An in-depth analysis of this case will appear in a forthcoming issue of Family Law.
|Neutral Citation Number:  EWCA Civ 89|
|Case No: B4/2013/2153|
IN THE COURT OF APPEAL (CIVIL DIVISION)
ON APPEAL FROM TELFORD DISTRICT REGISTRY
RECORDER JUDD QC (SITTING AS DEPUTY HIGH COURT JUDGE)
|Royal Courts of Justice
Strand, London, WC2A 2LL
B e f o r e :
LORD JUSTICE PITCHFORD
LADY JUSTICE BLACK
SIR STANLEY BURNTON
|RE P (FINDINGS OF FACT)|
Miss Samantha King (instructed by Freemans Solicitors) for the Appellant
Miss Victoria Clifford (instructed by Hatchers) for the 1st Respondent
Mr Lee Arnot (instructed by Rodney King & Partners) for the 2nd Respondent
Hearing dates : 13th December 2013
 On 28 June 2013, at the conclusion of a fact finding hearing in proceedings concerning K (a little girl who was born on 21 October 2009), Ms Recorder Judd QC made findings adverse to the appellant. He now appeals against those findings.
 The focus of the fact finding hearing was the events of 30 April 2011. On that day, K had a fit in her cot at home. She was apparently found by the appellant who summoned an ambulance. The ambulance crew noted a red mark (or marks) on the side of K's head and M confirmed the presence of this but there is no note of it having been seen at hospital and no other relevant external marks were found by the doctors. K was discharged the same day but was unwell over the following days, being drowsy and having repeated bouts of vomiting. Eventually investigations were carried out in hospital. An MRI scan which was performed on 17 May 2011 revealed that K had bilateral subdural collections and prominent subarachnoid spaces over both cerebral hemispheres. A small retinal haemorrhage was seen on examination by a consultant ophthalmologist on 20 May 2011; a trainee who had examined K on 18 May 2011 also thought he or she detected some small retinal haemorrhages.
 The judge found that K had suffered a non-accidental injury at some point in the later part of the day on 30 April 2011, caused by the appellant. She was invited to make a finding that there had been another non-accidental injury to K which had caused subdural bleeding which resulted in the subdural collections seen on the 17 May 2011 MRI scan, but she declined to do so.
 An unusual feature of this case was that the local authority had not brought the proceedings, nor were they a party to them. The proceedings were private law proceedings, brought by K's father (F) against K's mother (M), in which each parent sought residence of K and her older brother, R. The children were parties to the proceedings with a guardian appointed for them under Rule 16.4 of the Family Proceedings Rules 2010 and it was the guardian's representatives who drew up the schedule of findings that the court was invited to make. During the course of the proceedings, the appellant was joined as an intervenor. He did not have the advantage of legal representation at any time during the first instance proceedings and therefore appeared in person, although all other parties were represented. Fortunately, by the time of the appeal hearing before us, he was ably represented by counsel. M and F also appeared by counsel. A skeleton argument for the appeal was filed on behalf of the guardian but, quite properly, the guardian was not represented before us.
 For the most part, this judgment will revolve around what happened on and in connection with 30 April but some wider context is required.
 M and F were married in 2005 and eventually separated, acrimoniously, at the beginning of 2011. The children remained living with M.
 In about February 2011, M met the appellant through the internet. By the end of March 2011, they were in a relationship and he had started to stay at her house on occasions. He was himself going through a difficult divorce and there were problems over his contact with his three children by his marriage. His criminal record shows that he was convicted of harassment of his wife on a number of occasions in 2011 and 2012 and that in August 2011 he was imprisoned because of this.
 The appellant stayed overnight with M and the children on the night of 29/30 April 2011 and they went on an outing together during the day on 30 April. It was in the early evening, following their return home, that K had the fit.
 The relationship between M and the appellant continued for some time after the events of April and May 2011 and M gave birth to the appellant's child, Z, in April 2012. The relationship is now over, however. We heard little of Z during the appeal. It appears that he is being cared for by M but is not seeing the appellant. The other two children are living with F and M's contact with them is supervised because of the findings made by Ms Recorder Judd QC. As far as M was concerned, the judge found that she had failed to act in a protective manner towards K and R, resolutely refusing to contemplate the possibility that the appellant had injured K (§70) and joining with him "in trying to minimise and mislead others as to the time he was alone with the children on 30 April" (§71).
M's position in relation to the appeal
 M has not sought to appeal against the judge's findings herself. She accepted that if the findings against the appellant were upheld, she could not challenge the finding that she failed to protect the children. However, she supported the appellant's appeal, arguing that the finding that K had suffered a non-accidental injury on 30 April 2011 was unsafe and that, on the balance of probabilities, K's presentation between 30 April and 20 May 2011 is unexplained.
The evidence in the case: general
 The evidence in the case comprised a significant body of medical evidence plus other more general evidence, including evidence about, and from, the appellant and M.
The medical evidence
 In terms of medical evidence, the court had written and oral expert evidence from Mr Richards (a consultant paediatric neurosurgeon), Dr Anslow (a consultant neuroradiologist) and Mr Elston (a consultant ophthalmic surgeon). Material was available from Dr Hinde (a consultant paediatrician) but he did not attend court. In addition, the papers included material from the various doctors who treated K in April and May 2011.
 The evidence was that subdural collections such as those seen on the first MRI scan can result from disease but can also follow from acute subdural haemorrhages. It was established that when these subdural collections were first seen on 17 May 2011, they were already weeks, if not months, old. The judge proceeded on the basis that, whatever their cause, they predated 30 April and had not resulted from an event on that date. She found that the probable cause was trauma but she did not find that this was non-accidental. She accepted that the trauma could have happened, for example, when K accidentally fell down the stairs in 2010. No criticism was made of the judge in relation to any of this.
 The judge accepted that there was at least one retinal haemorrhage in K's eye as seen on 20 May 2011. Mr Elston dealt with the issue of causation and the judge summarised his evidence at §§32 and 33 of the judgment, upon which I draw for the balance of this paragraph. Mr Elston's view had evolved as the medical picture became clearer as a result of the various doctors giving the case detailed consideration and discussing it. His oral evidence was that the haemorrhage was consistent with a traumatic event on 30 April but that he did not exclude other, unknown causes. He noted that K was not a "normal" child because she had subdural collections at that time and it was possible for those to have caused a rise in intra-cranial pressure which could have led to retinal haemorrhages or for there to have been abnormal dynamics in the brain which could have had an effect on the circulation within the eye. He also noted that any traumatic event on 30 April had not apparently caused any fresh subdural bleeding even though it is known that bleeding into a chronic subdural can occur with minimal trauma.
 The doctors' evidence was that the fit could have had a number of causes. As summarised by the judge at §34, they said that it would be consistent with a traumatic episode which also caused the retinal haemorrhage(s) but it could have been caused by something else. The judge recorded that Dr Hinde "did state that it would be relatively unusual for a child of this age to suffer an afebrile fit for the first time" but she went on to say:
"That a fit in a child such as this is not a specific symptom in itself is clear from what happened at the hospital that night. K was examined, a history was taken and she was discharged."
It seems likely that the judge was here taking account of Mr Richard's view (see the minutes of the experts' meeting at C111) that from a neurosurgical perspective a fit is non-specific and does not in itself indicate a second event.
 The judge reviewed all of the medical evidence together and said that, "as Mr Richards observed, this case does not fit into an easy pattern". She particularly commented that in relation to the events of 30 April, "it is surprising that an episode of trauma sufficient to cause retinal haemorrhaging did not cause subdural bleeding given the pre-existing subdural collections" (§35). She noted Mr Richards' evidence that he thought it unlikely that a fall down the stairs on that day would have left no signs of recent subdural bleeding, but observed that on the other hand ordinary day to day knocks had not led to bleeding into the subdural collections in the weeks before the end of April 2011. She reached the view (§37) that:
"if there was an episode of trauma on 30 April, the medical evidence does not support a finding that the level of force or trauma involved was particularly great. It might have been, but I cannot find on the balance of probabilities that that was so."
She said that in reaching this conclusion, she had taken into account the lack of a noticeable bruise by the time K arrived at hospital that day.
 At §38, the judge said:
"Reading and listening to the medical evidence it is clear to me that the findings are consistent with a traumatic aetiology but also with other causes. I do bear in mind that the cause of the retinal haemorrhage and the fit might not be trauma at all but simply unknown. In this case it is not a theoretical possibility at all but (looking at the medical evidence by itself) a tangible one."
What the judge said in this paragraph ties in with an earlier section of her judgment about the assessment of medical evidence in general (see for example the concluding words of §23) and I have no doubt that the word "it" in the last sentence of the paragraph denotes "an unknown cause" and the sentence should be read as follows: "In this case, [an unknown cause] is not a theoretical possibility but .... a tangible one".
 Returning to the medical evidence later, in the context of her consideration of the evidence as a whole, the judge said, in a passage which I have found of particular importance for this appeal:
"The medical evidence is equivocal and, it seems to me, could point to either a natural or a traumatic cause for the retinal haemorrhages and the fit. The medical evidence does not provide a neat 'fit' or solution for this case but it is clear that the findings are consistent with a traumatic cause, and indeed there are several features that support this as being likely." (§63)
The appellant's and M's submissions to us in relation to the medical evidence
 I think it would be fair to say that in her skeleton argument, rather than focussing on the judge's treatment of the medical evidence, Miss King (counsel for the appellant) principally sought to challenge the judge's deployment of the non-medical evidence as the foundation for a finding that there had in fact been a non-accidental injury. I will return to that aspect of the appeal later.
 However, in her oral argument, Miss King adopted and built upon submissions made by Mr Arnot, counsel for M, in his skeleton argument about the impact that the medical evidence should have had upon the judge's conclusions, namely that the accumulation of medical uncertainties was fatal to the judge's findings which failed to recognise, or take into account, the limitations of the medical evidence.
 The uncertainties enumerated by Miss King and Mr Arnot included:
i) The cause of the subdural collections;
ii) The cause of the prominent subarachnoid spaces and the relationship they had, if any, to the subdural collections and/or the retinal haemorrhage;
iii) How many retinal haemorrhages there were, and possibly whether there were any at all;
iv) Whether the retinal haemorrhaging was superficial, intra-retinal or pre-retinal, which was material to whether it had anything at all to do with what happened on 30 April, because a haemorrhage caused on that day could only have remained visible at eye examinations on 18 and 20 May if it was a pre-retinal haemorrhage;
v) Whether the retinal haemorrhage may have resulted from raised intra-cranial pressure allied to the subdural collections;
vi) The "cause" of the seizure on 30 April and the vomiting between then and 20 May (by which I think is meant the mechanism within the body which produced it), given that there was no cerebral oedema, the retinal haemorrhage would not cause these symptoms, the subdural collections were not likely to be responsible and the experts could not exclude a non-traumatic afebrile seizure;
vii) How it was possible for the appellant to cause retinal haemorrhaging on 30 April extensive enough to leave visible remnants at the eye examinations in May without also causing fresh bleeding into the pre-existing subdural collection and/or any disruption to the brain and/or any external or other injuries;
viii) What it was that the appellant did to provoke injury to K.
 Mr Arnot summarised the position in his skeleton argument in this way:
"the court was left with equivocal medical evidence relating to a single small retinal haemorrhage, recorded 2 - 3 weeks after 30 April, with no cerebral oedema, no external injuries, with a series of unanswered questions, in a child with unexplained and atypical features, including enlarged subarachnoid spaces and chronic subdural collections"
 He said that "[m]edically, there may be nothing to explain", meaning that there may have been no event on 30 April which provoked K's symptoms on that day and thereafter. On his submission, the retinal haemorrhage was the lynch-pin because, if that was not present on 30 April (and the evidence was not sufficient to establish that it was), then all that was left was the seizure which could be explained as an afebrile fit which is not uncommon. He argued that the judge did not go far enough when she categorised the medical evidence as equivocal. He argued that in fact it pointed away from an incident on 30 April although he accepted during argument that the medical evidence did not actually establish on the balance of probabilities that there was no incident on 30 April.
 Miss King argued that it was important to look at the weight that could be put on the component parts of the medical evidence but also to look at it as a whole, which would have revealed that it did not fit together coherently. She submitted that the judge failed to do this and "failed to take cognisance of the inherent improbability that [a non-accidental incident] was the cause of the medical picture on 30 April and mid-May 2011".
Discussion in relation to the medical issues
 The appellant's and M's arguments place considerable reliance on the absence of signs/symptoms, such as fresh bleeding into the subdural collections and firm evidence that the retinal haemorrhage was a pre-retinal haemorrhage. However, an evaluation of their challenge to the judge's treatment of the medical evidence needs to start, perhaps, with a consideration of what signs/symptoms the judge found did exist. Firstly, she proceeded on the basis that K had a seizure on 30 April. There does not seem to have been any real issue about this; the evidence on which it was based came from the appellant himself. Second, the judge accepted there was bleeding in K's eye. As to this, the evidence entitled her to find that a retinal haemorrhage existed on 20 May and she went no further than that, saying that she was "satisfied that there was at least one retinal haemorrhage....as seen on 20 May" (§31). Third, the judge accepted that there were subdural collections which were older than 30 April and could not have been caused by an event on that day.
 As is not uncommon in perplexing cases of alleged non-accidental injury, and as the judge recognised, the medical evidence evolved during the preparation for the hearing. It is important to recognise this when evaluating what the judge said about this aspect of the case.
 Some of the doctors were initially attracted to the theory that the subdural collections, the retinal haemorrhage and the fit were caused by an episode of trauma on or around 30 April (see, for example, Dr Hinde's report of 21 November 2012 and Mr Elston's report of 15 August 2012). This theory had to be abandoned when the neurological evidence established that the subdural collections were too old to have been caused at that time and it was finally despatched at the experts' meeting on 7 December 2012. The written material generated before then has to be read with this in mind.
 Additional questions were put to the doctors after that meeting, in January 2013. One of the questions was whether there was agreement that there was a second event on or around 30 April which caused the fit and the retinal haemorrhage. As I read the replies (C111), the doctors did not go that far. In recording here what they said, I have chosen words that reflect the rather qualified language used in the record of their responses. Mr Richards made the comment (to which I have already referred) that from a neurosurgical perspective, a fit is non-specific and does not in itself indicate a second event; he left it to Mr Elston to comment on the implications of the retinal haemorrhage in this context. Mr Elston said that the retinal haemorrhage could be compatible with being caused by an event on 30 April. Dr Anslow disclaimed the necessary expertise to comment. Dr Hinde said it was slightly unusual for a child to have a single afebrile seizure and that one possible explanation for it was that it could have been associated with a head injury. He said that if Mr Elston excluded all possible causes for the retinal haemorrhage apart from the trauma then it would seem possible that the head injury and the retinal haemorrhage were caused at the same time but he said it was not possible to be definite about this.
 As Dr Hinde did not give oral evidence and did not contribute materially to the answers to further supplemental questions in March 2013, his position rested there. We have not had transcripts of the oral evidence given by Mr Richards, Dr Anslow and Mr Elston but I have set out the judge's account of their positions. I do not detect from it or from what we were told in submissions that they firmed up their views of the causation of the retinal haemorrhage or the seizure beyond them being consistent with a traumatic event on 30 April but also possibly attributable to other causes, identified or unknown.
 The judge observed that the case did not fit into an easy pattern and it is clear at various points in her judgment that she had firmly in mind that there may not have been an incident of trauma at all on 30 April and that the retinal haemorrhage and the fit may have had a natural cause (see, for example §33, the opening words of §37, §38, and §63). She also recognised (§33) that it was notable that there had not been any fresh subdural bleeding although such bleeding can occur with minimal trauma where there is a chronic subdural collection, reverting to this again in §35 where she said that it was surprising that an episode of trauma sufficient to cause retinal haemorrhaging did not cause subdural bleeding given the pre-existing collections.
 With this background, it is no surprise to find the judge saying at §63 that the medical evidence was equivocal and could point to either a natural or a traumatic cause for the retinal haemorrhages and the fit. I do not think exception can be taken either to her comment that the findings were "consistent with a traumatic cause" (§63); there was undoubtedly evidence that the medical picture was compatible with a traumatic incident on 30 April. However, I cannot see upon what basis she added that "indeed there are several features that support this [a traumatic cause] as being likely". She did not go on to identify what she had in mind and I cannot myself readily find any such features in the written medical evidence or in the oral evidence as set out in her judgment, nor did counsel for F invite our attention to any. This aspect of the judgment has caused me considerable concern.
 Miss Clifford for F submitted that it was important to look at the judgment as a whole. She submitted that the judge was alive to the fact that the medical evidence was not clear and was not, at times, a neat fit but it was only part of the totality of the evidence which had to be considered. She argued that the judge showed that she had in mind that there may not have been any trauma at all but properly relied on the fact that trauma was not ruled out either.
 Undoubtedly, the judgment does show that the judge recognised the oddities and uncertainties in the medical picture and was cautious in her approach to it. Miss Clifford is also right to say that the judge was entitled to look at the whole picture as she expressly said in §62 that she did. However, her view that there were several features of the evidence that supported a traumatic cause as likely is either not accurate or not explained. On the material that we have seen, a proper approach would have been to recognise that, at the very best, the medical evidence did not help a great deal. If the judge's words indicate that she derived positive support from it for there having been a traumatic incident on 30 April, and it seems to me that they might well do so, then she approached the non-medical evidence from the wrong angle.
 Before I determine what the implication of this is for the appeal, I will turn to that non-medical evidence.
The appeal in relation to non-medical issues
 The judge's finding against the appellant was heavily dependent upon the fact that he had "invented one traumatic incident that he said occurred on 30 April and seriously embellished another", she thought in order to cover up an inflicted act (§64). The appellant, supported by M, submitted that the non-medical evidence did not bear the weight that the judge put upon it and that it was, for various reasons, insufficient to support the finding made.
 Some factual background is necessary in order to understand the issues that arise in relation to this.
 Between her first presentation at hospital on 30 April 2011 and the conclusion of investigations into her condition in May 2011, K was unwell and M repeatedly consulted doctors about her. After her discharge home on 30 April, she was re-admitted to hospital overnight on 6/7 May 2011 and then admitted again on 15 May 2011. According to social services records, M was advised on 23 May 2011 that the children could not return to her care (E33). On 24 May 2011, the appellant rang social services and told them that he had accidentally knocked K down the stairs two or three weeks before. This was the first mention of any such incident.
 On 26 May 2011, both the appellant and M were interviewed by the police. The appellant gave a long and detailed account of the stair accident. He said that after it K was crying but otherwise there was nothing wrong with her. He said that he had not told M about a stair fall at the time, simply telling her something along the lines of K having fallen through a gate which was in place at the entrance to the front room. He gave various explanations for not having said anything about a fall down the stairs, including that the hospital would ask questions and that his ex-wife would find out and it would damage his case in relation to his children. He also spoke of an incident with the dog when they returned from their outing on 30 April 2011, in which K and the dog head-butted each other and there was a bruise on K's head. He described the bruise at one point as being "quite a big one....like nearly all of the front of her head ....quite a big hefty bruise". He said it was red to start with, "then after a little bit it had come up and it was all like yellow" and that "after a bit you could see it like going you know like the pink-y sort of purple-y colours" although it was not bad when she went to bed. He said that following the dog incident, K looked drowsy and was not herself.
 When M was interviewed later in the day, she did not recollect any incident concerning K in the morning of 30 April. She said she thought that the appellant was lying about a fall down the stairs because he could see how upset she was about losing her children. She acknowledged a head-butting incident with the dog but said there were no marks on K after it.
The judge's treatment of the non-medical evidence
 The judge recorded (§52) that the appellant's evidence was that the account of the fall down the stairs was a lie and that when he told M what he had told the police, she berated him and told him that there was probably a medical explanation for K's injuries. He said they tried to get another police interview for him but the police did not seem to be interested.
 The social services records show that M told the social worker on 27 May 2011 about the appellant's lie about the stair fall. We can see from the police records that the social worker passed this on to the police. The appellant was due to answer bail in relation to harassment charges in July and the police approach, according to the record, was that he could be re-interviewed when he did that.
 The judge considered the lie about the stair fall and the appellant's description of the dog incident at some length. She accepted that the appellant's account of the fall down the stairs had been a lie (§58). His oral evidence was that in saying what he did, he had been trying to help the mother. He said that he had suffered as a result of not seeing his own children and did not wish the same thing to happen to her so stupidly decided to make something up.
 The judge concluded that the appellant was also "untruthful" in his description of the dog incident (§58) in that he had exaggerated it. She said that the appellant had appeared to suggest in oral evidence that the police led him into describing the incident with the dog as more than it was (§54) but she did not accept that that was the case (§53). The transcript of the police interview does not support the appellant's assertion about this and no one has questioned the judge's rejection of it.
 The judge's consideration of why the appellant was untruthful about the two alleged incidents of trauma on 30 April began at §59 where she referred to the case of Lucas, reminding herself that people can lie without being guilty. One possibility that she considered was that the stair fall was invented, as the appellant said, because he was trying to protect M. Another was that he "may well have been frightened, quite reasonably and rightly as it has turned out, that he would be accused of hurting K". A further possibility, and the one which the judge concluded was the probable reason for the appellant's misleading account, was that he was trying to hide another episode of trauma that he had caused that day, probably in the early evening (§60).
 The judge reasoned that the appellant's statement to the police in his interview on 26 May had been preceded by his call to the social worker on 24 May and was not made on the spur of the moment. She considered that he knew how seriously what he was going to say would be taken. She said:
"Although he is emotionally fragile he struck me as an intelligent man (as his final submissions on the medical evidence demonstrate) and it seems to me highly unlikely that he would have made up a story or stories like that when actually there had been no incident of trauma at all. He knew perfectly well that evidence of his failing to watch a toddler so that she received a serious injury would be likely to be used against him in his own family proceedings. I do not believe he would have been prepared to take the blame for something like that if he was innocent of doing anything at all." (§60)
 She thought the most likely explanation for the lie about the stair fall and what she considered to be serious embellishment of the dog incident was "that he was seeking to cover up another traumatic incident which he was too frightened or ashamed to admit to". She had proceeded on the basis that the evidence did not establish that the level of force or trauma involved in K's injury was particularly great, although acknowledging that it might have been (§37) and she said (§64) that "[g]iven my finding about the level of force that is likely to have been involved, and [the appellant's] character generally, I have wondered whether the incident he is concealing is a negligent rather than an inflicted act". However, she concluded that it must have been an inflicted act because "[h]is story about the stairs admitted to negligence and what he is hiding is likely to be more culpable than that" (§64) albeit that she thought it very unlikely that he intended to cause K any harm (§67).
 The judge also referred to both M and the appellant having, as she found, tried to conceal the fact that M was out of the house, probably at the shop, when the fit occurred (§65).
 The judge's reference in the passage from §60 which I have quoted above to the appellant being emotionally fragile is not the only comment in her judgment on his psychological state. She also noted elsewhere that it was plain during his oral evidence that the appellant was "still very emotionally fragile" (§51), that he showed "obvious distress at times" (§51), that he "said on several occasions that he found it difficult to remember things" and that "his evidence around [the dog] incident became very vague" (§54). She took into account that the psychological report noted "his fragile mental state and also that he is a man who acts impulsively" (§59). She recorded that he was receiving counselling and had tried anti-depressants but found them not to be of much assistance and stopped them (§51). The concluding words of her judgment, which were directed to the appellant personally, again acknowledged the emotional state he was in (§75).
The submissions in support of this aspect of the appeal
 Miss King submitted on the appellant's behalf that, with the medical evidence as it was, for the judge to make a finding that there was a non-accidental incident involved a reversal of the burden of proof and also that the judge had failed to take into account the inherent improbability that it was such an incident that had caused the medical picture. It was wrong of the judge, she said, to shore up equivocal medical evidence with the non-medical evidence upon which she relied, namely lying about the stair fall and embellishing the dog incident. The deficits in the appellant's evidence were not capable of proving that he inflicted an injury on K or, if capable, were not sufficient to do so, she submitted.
 Although Miss King accepted that the judge's "Lucas direction" was appropriate, she submitted that the judge nonetheless failed to have due regard to the other reasons why the appellant may have lied, to his psychological profile and his vulnerability, to his position as an unrepresented party and his emotional state during the hearing. The psychological report on him dated 10 May 2012 described him as very restless, distractible and impulsive, he was prone to rigidity of thinking, and in the past he had been prone to irrational behaviour. In these circumstances, the judge should not have relied on the appellant's intelligence to support her view that his lies covered guilt, said Miss King, and she should have given more regard to the impact of his fragile mental state on his ability to provide a consistent account of his actions. Mr Arnot on behalf of M added that the judge should also have taken into account the significant delay between the lies and the fact finding hearing.
 Miss King submitted, furthermore, that the judge had not identified a cogent enough basis for inferring that the lies indicated guilt or set out adequate reasons for doing so. She also submitted that the judge had failed to give adequate consideration to the totality of the appellant's conduct at the time of the fit and in relation to what he said about that day. After discovering the fit, he called the ambulance and, in the aftermath, medical help was regularly sought for K. The lie about the stair fall came, she pointed out, at a time when M and the appellant had just become aware that the children were going to be removed and that deliberate force must be being suggested. It was also relevant, she said, that the appellant very quickly told M what he had said and she told social services and the police. She submitted that the appellant also volunteered information on this to the police on 8 June 2011 and referred us to the police records at F123 to 125 but they do not seem to me to substantiate this.
 I do not accept that the judge's analysis involved any reversal of the burden of proof. In an appropriate case, findings of non-accidental injury can be made even though the medical evidence does not establish that there has been such an injury. The court has to consider the totality of the evidence as the judge did here and to make findings on the balance of probability. I do not wish to assert that a certain type of evidence (for example, lies) can or cannot, in principle, be sufficient to establish that a non-accidental injury has occurred or that any other allegation in family proceedings is made out. So much depends upon the nature and quality of the evidence as a whole.
 I cannot overstate the importance of the opportunity that the trial judge has to consider all of the evidence and to evaluate the witnesses although I do not intend to repeat here what has been said many times before on the subject. Miss Clifford (who represented F) referred in her skeleton argument to Piglowska v Piglowski  1 WLR 1360 and to Re A (Fact-Finding: Disputed Findings)  EWCA Civ 12,  1 FLR 1817 which deal with this question and there are, of course, many other authorities which do so. As Miss Clifford pointed out, those cases also identify how difficult it can be to convey in a judgment the diverse information that has been gathered by the judge in the course of the hearing and that, inevitably, there will be limitations on the expression, in the judgment, of the judge's reasons for his decision. An appellate court should be very cautious about interfering with the trial judge's assessment, particularly when it depends heavily, as it did in this case, upon observation of a party over several days, including in cross-examination directed to the particular point in issue.
 Miss Clifford submitted that it was evident that the judge had indeed thought about the implications of the appellant's lies carefully as she said in §60 of her judgment and correctly followed what Miss Clifford described as the "subtle and delicate process" of assessing the reasons for that deceit, having regard to the totality of the evidence. I accept that submission although, for reasons which will become apparent later, it does not lead to the appeal being dismissed.
 There can be no doubt from the judgment that the judge was acutely aware of the issues in relation to the appellant's mental state. Her repeated express references to it make that clear and she also alluded to examples of his behaviour such as his convictions for harassment of his wife and the fact that he had spoken of standing outside a relative's house in February 2011 with a crossbow thinking of causing harm (see §49 of the judgment). She had obviously made use of the significant opportunity that she had to assess the appellant during the hearing and the concluding words of her judgment show her understanding of his problems and her consideration for him.
 This case is unusual in that the appellant himself said that he had lied and offered an explanation why. In my experience, it is more common for the lies that the judge is evaluating to have been found by the judge himself or herself. That the judge gave careful consideration to the appellant's explanation can be seen from §59 of the judgment. She identified there, referring to the psychological report, features which would support what he said, namely his fragile mental state and impulsivity. She also considered a further explanation not advanced by the appellant, that is that he may have been frightened of being accused of hurting K. In §60 she gave cogent reasons for concluding that the probability was that the appellant was lying to hide another more sinister episode of trauma. These reasons may be described as intellectual or analytical in nature but I have no doubt that, although she does not set it out at this precise point in her judgment, the judge's reasoning will have been shaped more intuitively as a result of her assessment of the appellant as a party and a witness during the hearing.
 The appellant is right in saying that the judge did not expressly say, when considering the import of his lies, that she was weighing in the balance his conduct in the aftermath of the fit or the relatively short life of the stair lie. There can be no doubt that she was well aware that it was the appellant who called the ambulance because it was common ground and there were ambulance records supporting it. We can see from the judgment that she also had it in mind that he and M tried to bring the lie to the attention of the authorities (§52). She knew too that medical help was sought for K in the weeks following 30 April. It cannot be argued, in my view, that any of these facts necessarily pointed away from the appellant's guilt. Like much of the medical evidence, they were equivocal. Some judges would have referred to them expressly as part of their evaluation of the lies and considered in terms whether they might point away from the most sinister explanation but I do not accept that it was fatal, in the context of a relatively concise judgment which confirms that the judge was aware of them, that this judge did not spell them out.
 Why therefore, when I have rejected the criticisms made of the judge's handling of the non-medical factors, have I concluded, as I have done, that the appeal should be allowed?
 This was a difficult and unusual case; it was tricky for the judge to handle and has also given me considerable anxiety. The care with which the judge approached the matter is apparent from her judgment but there is, in my view, a problem with her findings. That problem lies in her treatment of the medical evidence as seen in her §63 with which I dealt above (see particularly §§30 to 34). The non-medical evidence was not particularly robust in nature. It could hardly be said that it pointed inexorably to one end. There was no margin for error when the judge considered its implications in the light of the medical evidence. The support that it appears from §63 she may erroneously have derived from the medical evidence, albeit possibly limited, may have been sufficient to propel her to a finding that she would not otherwise have made against the appellant. Had she acknowledged that the most that could be said of the medical evidence was that it was, in some respects consistent with there having been a traumatic cause, she may have questioned to a greater extent whether there was any event at all on 30 April and may have taken a different view as to the import of the appellant's lies. A contemplation of the absence of medical features which positively supported a traumatic event may also have led to the judge reminding herself of those elements in the medical evidence that did not fit well with a traumatic event. She may, for example, have pursued further the puzzle (to which she alluded at §35) of how retinal haemorrhage was caused without fresh bleeding into the pre-existing subdural collections also arising and was caused by force which she accepted might not have been particularly great.
 It is these matters which, given the finely balanced nature of the evidence, have persuaded me that the appeal should be allowed. We heard some argument about what should occur in these circumstances. I am conscious, however, that it was difficult for counsel to address the issue of whether the case should be remitted for a rehearing without knowing what view this court took of the central issues in the appeal which, of course, affect not only the appellant but also M, given that the findings about her owed something to the finding about him. I have no doubt that decisions as to what the next steps should be need also to be taken bearing in mind the stage that has been reached in relation to arrangements for K and R, for Z, and indeed for the appellant's children. I would therefore invite further submissions in the light of this judgment as to what orders this court should make.
Postscript: I have continued to use the term "non-accidental injury" throughout this judgment notwithstanding the recent decision of this court in Cambridgeshire County Council v PS, JS and S  EWCA Civ 25 because, in contrast to that case, nothing in this case turns upon the meaning of the term.
SIR STANLEY BURNTON:
 I agree.
 I also agree.