Date: Mar 11, 2026

Appellate Restraint in Non-Accidental Injury Perpetrator Findings: Credibility, Propensity Evidence, and “Pool” Analysis

Case: A, B & C (Children: Perpetrator Finding) [2026] EWCA Civ 219

Court: England and Wales Court of Appeal (Civil Division)

Date: 6 March 2026

1) Introduction

This appeal arose from care proceedings concerning three young brothers (A, B and C). The central issue on appeal was a fact-finding determination: whether the youngest child, C (about 15 weeks old at the time), sustained an incomplete fracture of the proximal left humerus non-accidentally, and if so, which parent inflicted it.

The medical evidence was unchallenged and decisive on mechanism: the fracture required significant force, had no organic cause, and would not occur through ordinary handling, rough play or inexperienced parenting. The parents accepted that only one of them could be responsible. Each denied causing the injury.

The Family Court judge (Judge Magee) identified the mother as perpetrator on the balance of probabilities, while being unable to determine whether infliction occurred before the children left for contact with the father and paternal grandmother (“PGM”) on 16 November 2024 or after their return on 17 November 2024. The mother appealed, principally arguing that the judge mishandled: (i) witness inconsistencies (especially of the maternal grandmother, “MGM”); (ii) the PGM’s minimisation and credibility; (iii) the “wider canvas,” including the father’s propensity to anger/violence; and (iv) overall evaluative reasoning.

The Court of Appeal (Baker LJ, with Arnold LJ and Dove LJ concurring) dismissed the appeal. The decision is an important reaffirmation of appellate restraint in fact-sensitive perpetrator findings in child injury cases, clarifying how “propensity” and “similar fact” arguments operate in family proceedings, and confirming when “pool of perpetrators” reasoning is (and is not) required.

2) Summary of the Judgment

The Court of Appeal upheld the first-instance finding that the mother inflicted the non-accidental fracture. The appellate court held that:

The trial judge’s findings were rooted in permissible credibility assessments and evaluation of inconsistencies—an area where appellate intervention is exceptional.
The judge was entitled to treat the father’s and PGM’s evidence about the weekend contact as reliable enough to discount injury occurring during that period, including the judge’s conclusion that PGM would likely have disclosed wrongdoing by the father.
Arguments about symptom variability (the fracture’s “symptomatology”) did not undermine the judge’s focus: the mother’s shifting accounts of when she first noticed abnormality and her failures to take expected steps (contacting the father; informing the social worker; acting on LP’s advice).
“Propensity” arguments did not reveal any legal misdirection. The judge had the relevant incidents “in mind,” weighed them, and still reached a different conclusion on the totality.
Having identified the perpetrator on the balance of probabilities, the judge was not required to conduct a further “pool of perpetrators” analysis.
The reasons given met the standard for a first-instance judgment; there was no obligation to address every submission exhaustively.

Arnold LJ added an observation (not affecting outcome) that a chronological analysis of events, locations and messages might have helped pinpoint both perpetrator and timing, which are inter-related.

3) Analysis

3.1) Precedents Cited

Fage UK Ltd v Chobani UK Ltd [2014] EWCA Civ 5

This authority was relied upon for two connected principles:

Appellate restraint on findings of fact: an appeal court does not re-try the case; it interferes only with “very clear justification.”
Adequacy of reasons: the oft-cited passage (Lewison LJ) emphasises that a judge must provide reasons sufficient to show the principles applied and why the decision was reached, but need not address every argument or produce an elaborate “jury-style” summing up.

Baker LJ used Fage to frame the appeal: because the case was a “paradigm example” turning on witness impression and reliability, the bar for overturning the trial judge was high. The judgment also used Fage to dismiss the complaint that the judge did not expressly deal with every point (e.g., detailed symptomatology), holding that the reasoning given was sufficient and compliant.

Volpi v Volpi and another [2022] EWCA Civ 464

Volpi is a modern restatement of appellate caution, particularly where fact-finding depends on evaluation of oral evidence. By invoking Volpi, the Court reinforced that even if an appellate judge might have assessed certain pieces of evidence differently, that is not a ground for interference unless the conclusion is outside the bounds of reasonable evaluation.

R v Mitchell [2016] UKSC 55, [2017] AC 571

The mother relied on Mitchell to argue that the trial judge mishandled propensity / similar fact reasoning—suggesting the father’s past anger and aggression should have weighed more heavily in the perpetrator evaluation.

The Court of Appeal distinguished the relevance: Mitchell primarily concerns admissibility principles for similar fact evidence in criminal trials, whereas family proceedings operate within a more flexible evidential environment.

R v P (Children: Similar Fact Evidence) [2020] EWCA Civ 1088, [2021] 1 FLR 652

R v P (Children: Similar Fact Evidence) applies similar fact concepts within the family context. Baker LJ held that the “similar fact” discussion did not materially assist here because admissibility was not the problem: the judge had the father’s prior incidents before her and treated them as relevant. The real question was weight, and the trial judge’s weighting decision was not shown to be wrong in law or perverse.

The appellate court’s treatment of these authorities clarifies a practical point: citing “similar fact” case law will rarely succeed where the first-instance judge has admitted and considered the material; the battleground becomes evaluative weight, where appeals face the Fage/Volpi restraint.

3.2) Legal Reasoning

(a) The “wider canvas” and credibility as the engine of perpetrator identification

The trial judge accepted the unchallenged medical evidence establishing non-accidental injury and a broad window of infliction. With only two possible perpetrators, the case turned on the “wider evidential canvas”: the reliability of each parent’s account and the inferences drawn from their conduct.

The Court of Appeal endorsed the judge’s structured approach to witness reliability:

The mother: the judge’s view evolved from initially honest to ultimately “poor witness” with “numerous inconsistencies,” including shifting accounts about when she observed abnormality in C’s arm.
LP (mother’s friend): treated as clear and consistent; crucially, LP said the mother flagged concern about a “floppy” arm early in the day and was advised to get it checked.
MGM: cautious and frequently “I don’t know/can’t remember”; relevant insofar as it supported the judge’s assessment that the mother had not disclosed LP’s advice and had underplayed early concerns.
Father: emotional but reflective; increased concessions in oral evidence viewed as “evolution of insight” rather than inconsistency.
PGM: minimised some matters (notably the fight), but previously frank with police; the judge concluded she would have spoken up if the father had harmed C during the contact weekend.

On appeal, the mother attempted to re-argue this matrix: that MGM uncertainty was over-weighted, and PGM minimisation was under-weighted. Baker LJ’s reasoning is orthodox: the apportionment of weight to witness features is a core first-instance function, especially where the judge has given “considered and nuanced” credibility reasons.

(b) Symptomatology and the appeal’s “missing expert evidence” complaint

The mother argued that the judgment did not explicitly include expert nuance: after the moment of infliction, symptom presentation could be variable and non-specific. The Court of Appeal rejected the premise that the judge overlooked this, noting she expressly factored in: (i) LP’s observation that C could still use the arm; and (ii) the hospital clinicians not recognising fracture until X-ray the next day.

More importantly, the appellate court reframed the point: the judge’s concern was not that symptoms must have been obvious at all times, but that the mother gave different accounts at different times about when she first noticed something wrong, coupled with arguably counter-intuitive omissions:

not contacting the father to ask if anything happened during contact;
not calling the social worker, unlike when a chest bruise occurred days earlier;
not telling MGM that LP advised medical review.

In this way, the judgment illustrates an evidential distinction that often decides perpetrator disputes: variable clinical signs may explain delayed recognition, but they do not necessarily explain inconsistent narratives and conduct-based inferences (who was contacted, what was said, what was omitted).

(c) Propensity / “similar fact” material: relevance is not determinative

The mother’s main oral focus was that the father had a propensity to anger and aggression, evidenced by incidents involving A (slap), damage to PGM’s car, confrontation with a traffic warden, and a fight with his brother.

The Court of Appeal accepted the obvious: such evidence was “capable of supporting” a finding against the father. But it held that:

The judge did not ignore it; she “analysed it in some detail” and incorporated it into the overall evaluation.
The judge was entitled to conclude that, despite propensity to anger/aggression, the totality pointed to the mother as perpetrator— particularly given the credibility findings and the judge’s acceptance of the father’s acknowledgements and remorse as supporting his general honesty.

The reasoning is significant in practice: “propensity” is seldom a trump card. In family fact-finding, it is one strand of the canvas, and can be outweighed by credibility, opportunity evidence, third-party presence, and conduct after the event.

(d) “Pool of perpetrators” reasoning once a single perpetrator is found

The trial judge included (from the parties’ agreed law note) a summary: courts should identify the perpetrator on the balance of probabilities if possible, and only if unable to do so should they consider a pool and whether each is within a “real possibility.”

Baker LJ held the judge was not required to go further with pool principles because she did identify the perpetrator on the balance of probabilities. This confirms a procedural point: where a court can properly make a single-perpetrator finding to the civil standard, it does not need additionally to set out a “pool” analysis as a fallback or alternative.

(e) Reasons and methodology: credibility-first versus chronology-first

Arnold LJ’s addendum is a practical judicial critique rather than a ground of appeal: it may have been “advantageous” to analyse the evidence chronologically to assist with both perpetrator and timing (inter-related). This is notable because the judge found the mother was perpetrator but could not pinpoint whether injury was inflicted before 16 November contact or after return on 17 November.

The Court of Appeal nonetheless held the credibility-based approach was “fully open” to the judge. The take-away is twofold:

Methodological flexibility: appellate courts will not impose a single “correct” structure for fact-finding judgments where the analysis is coherent and reasons are adequate.
But chronology remains best practice: especially in narrow windows of opportunity, a timeline can sharpen findings on timing and opportunity, which may in turn reinforce (or undermine) perpetrator conclusions.

3.3) Impact

(a) Appeals in perpetrator-finding cases: the bar is high

The case strengthens the message that where the trial judge has: (i) summarised witness evidence, (ii) explained credibility conclusions, and (iii) linked those conclusions to the perpetrator determination, an appeal will rarely succeed without demonstrating a clear legal misdirection, material omission, or irrationality. “Rebalancing the evidence” is not the appellate function (Fage; Volpi).

(b) Propensity evidence: admissible, relevant, but often only one factor

The judgment discourages over-reliance on criminal-law “similar fact” framing where the family judge has admitted and considered prior incidents. The real fight will be over weight—an evaluative domain protected by appellate restraint. Practitioners should therefore focus less on abstract “propensity doctrine” and more on how past incidents concretely bear on: opportunity, mechanism, supervision, escalation patterns, and credibility.

(c) Conduct after suspected injury as a credibility marker

The Court of Appeal’s endorsement of the judge’s focus on omissions (not calling the father; not calling the social worker; not following advice; shifting accounts) underscores that in NAI cases, post-event behaviour can be as probative as the medical window when only two perpetrators are possible.

(d) Judgment-writing practice in fact-finding hearings

Arnold LJ’s comment may influence first-instance practice: judges may increasingly be encouraged to include a clear chronology (people, places, messages) alongside credibility findings—particularly where timing remains uncertain, to reduce ambiguity and enhance transparency.

4) Complex Concepts Simplified

Balance of probabilities: The civil standard of proof: the court decides what is more likely than not (i.e., >50%). It is not “beyond reasonable doubt.”
Non-accidental injury (NAI): An injury the court finds was inflicted rather than caused by accident or normal handling.
Symptomatology: The pattern of symptoms. Here, the argument was that an incomplete fracture might show variable signs after infliction, potentially explaining delayed detection.
Pool of perpetrators: If the court cannot identify who inflicted an injury on the civil standard, it may identify a “pool” of possible perpetrators (those with opportunity) and decide whether each person falls within a “real possibility.” This case confirms that if the court can identify a perpetrator on the balance of probabilities, it need not proceed to pool analysis.
Similar fact / propensity evidence: Evidence of past conduct used to suggest someone is more likely to have acted similarly again. In family cases, the key question is usually not strict admissibility but the weight the court should give such evidence in the overall evaluation.
Interim care order (ICO): A temporary order placing a child under the local authority’s care during proceedings, pending final decisions.
Threshold document: The local authority’s pleaded facts said to satisfy the legal threshold for state intervention (risk of significant harm attributable to care given or likely to be given).

5) Conclusion

A, B & C (Children: Perpetrator Finding) [2026] EWCA Civ 219 is a reaffirmation of three core propositions in child injury fact-finding:

Appellate courts rarely disturb credibility-driven perpetrator findings absent clear error (Fage UK Ltd v Chobani UK Ltd; Volpi v Volpi and another).
Propensity evidence is relevant but not determinative; “similar fact” authorities do not convert a weight dispute into an error of law where the judge considered the material (R v Mitchell; R v P (Children: Similar Fact Evidence)).
Once a perpetrator is identified on the balance of probabilities, pool analysis is unnecessary; the court’s duty is to give adequate, not exhaustive, reasons for its choice (Fage).

The judgment also offers a practical lesson for future cases: chronological mapping of events, alongside credibility analysis, may help courts resolve not only who inflicted an injury but when—thereby strengthening the transparency and resilience of findings in the most contested fact-finding disputes.