Ten votes to seven. One fundamental question still unanswered.
I pleaded Lhermitte v. Belgium before the Grand Chamber of the European Court of Human Rights.
At the heart of the case was one decisive issue: criminal responsibility.
Three psychiatric experts, confronted with new evidence, revised their position and concluded that Geneviève Lhermitte was not criminally responsible at the time of the acts.
The jury rejected that conclusion.
But it was not required to explain why.
That remains the fundamental problem in Lhermitte.
How can three court-appointed experts reconsider the evidence and conclude that the accused was not criminally responsible, while the jury may reject that conclusion without answering the central issue raised by the defence?
The Grand Chamber nevertheless found, by ten votes to seven, that Article 6 had not been violated.
Ten years later, having revisited the judgment and the Court’s own case-law in detail, I remain convinced that the Court’s own legal logic exposes an unresolved contradiction:
If criminal responsibility is the decisive issue, the accused must be able to understand why the expert conclusion that she lacked that responsibility was rejected.
That answer was never given.
Ten years on, the question remains.
Behind the legal debate lies an immense human tragedy. That should never be forgotten.
In memoriam — Me Xavier Magnée, with whom I had the privilege of pleading this exceptional case.
And with respect for the profound human tragedy at the heart of the case of Geneviève Lhermitte.
A critical reassessment of one of Strasbourg’s most closely divided judgments on reasoned verdicts, jury trials and the right to a fair trial
Abstract. In Lhermitte v. Belgium (29 November 2016) the Grand Chamber held, by ten votes to seven, that a defendant convicted on an unreasoned jury verdict had been able to understand why she was found criminally responsible, although the three court-appointed psychiatrists had unanimously concluded at the end of the trial, on the basis of new evidence, that she had been incapable of controlling her actions. This article argues that the majority conflated two distinct questions: whether the jury was free to reject the experts, and whether the accused could know why it did so. None of the safeguards relied on by the majority, alone or together, answered the one question that was decisive. Read against the evolution of the Belgian assize procedure, in which the legislature itself has required since 2010 that the principal reasons for a jury’s decision be stated, the joint dissent of seven judges offered the analysis more faithful to Taxquet v. Belgium.
Keywords: Article 6 § 1 ECHR; reasoned judgments; jury trial; Belgian Assize Court reform; criminal responsibility; psychiatric expert evidence; Taxquet; arbitrariness.
Disclosure. The author pleaded the case for the applicant before the Grand Chamber, together with the late Me Xavier Magnée. The views expressed are personal.
Almost ten years have passed since the Grand Chamber delivered its judgment in Lhermitte v. Belgium.1 By ten votes to seven, the Court found no violation of Article 6 § 1 of the Convention. The majority consisted of Judges López Guerra, Nußberger, De Gaetano, Lemmens, Jäderblom, Pejchal, Vehabović, Mourou-Vikström, Ravarani and Kucsko-Stadlmayer. The joint dissent was signed by Judges Raimondi, then President of the Court, Lazarova Trajkovska, Laffranque, Sicilianos, Lubarda, Grozev and Harutyunyan.2 That the President of the Court sat with the minority is itself an indication that the case raised a genuine question of principle.
The question is not whether a jury must give reasons; the Court has consistently held that it need not. The question is narrower and more demanding, and its answer turned on two votes. When new evidence emerging at trial leads the same psychiatric experts unanimously to reverse their assessment on the decisive issue of criminal responsibility, can Article 6 accept a verdict to the opposite effect without any procedural mechanism allowing the accused to know why? I shall argue that it cannot, and that the majority reached the contrary conclusion only by treating the fact of the decision as if it were its explanation.
The starting point is Taxquet v. Belgium.3 The Grand Chamber did not require lay jurors to draft reasoned judgments, and the Convention does not prescribe any particular model of criminal procedure. But where a jury gives no reasons, the proceedings as a whole must offer sufficient safeguards against arbitrariness and enable the accused to understand why he or she was convicted. Among those safeguards the Court identified directions or guidance given to the jurors, and precise, unequivocal questions capable of forming a framework on which the verdict is based or of compensating adequately for the absence of reasons.
The requirement is functional, not aesthetic. Reasons, or their procedural equivalent, make the exercise of judicial power intelligible, allow its review and show the accused that decisive arguments were actually considered. This links Taxquet to the Court’s general case-law: a court need not answer every argument, but it must address those capable of determining the outcome.4 The question in Lhermitte was therefore never “did the jury give reasons?” but “did the proceedings enable the applicant to understand why she was held criminally responsible?”.
The Grand Chamber framed the issue precisely. Neither the material facts, which the applicant admitted, nor their legal characterisation, nor the sentence was in dispute. The only question was whether she could understand why the jurors held her responsible for her acts at the time of their commission, despite the psychiatric experts’ change of opinion at the end of the trial.5
That framing matters. A finding that a person committed certain acts and a finding that she is criminally responsible for them are distinct propositions. Responsibility is the bridge between conduct and punishment. Where that bridge is contested on psychiatric grounds, its answer determines whether the ordinary consequences of the criminal law may follow at all. The psychiatric evidence was therefore not one item among many: it went to the only issue the jury actually had to resolve.
The architecture of the case cannot be understood without the chronology of the expert evidence. During the investigation, a panel of three psychiatrists examined the applicant and reported on 30 October 2007; its conclusion did not exclude her responsibility. The trial before the Assize Court of Walloon Brabant ran from 8 to 19 December 2008. In the course of the hearing, the applicant’s treating psychiatrist, Dr D.V., disclosed for the first time the existence of two letters she had addressed to him, one of which had been delivered to his practice on the day before the events.6
The President of the Assize Court considered this material significant enough to instruct the same panel to report again. The three experts then concluded, unanimously, that at the time of the events the applicant had been suffering “from a severe mental disturbance making her incapable of controlling her actions”.7 The sequence is thus: new evidence, renewed assessment, unanimous reversal by the experts, and an opposite finding by the jury.
The jury was entitled to disagree. That has never been in issue. The Convention question is different: what, in the proceedings, allowed the applicant to know why it disagreed?
Experts do not decide criminal cases; judges and juries do. An expert opinion, even a unanimous one, cannot bind the decision-maker on the ultimate question of responsibility, and the Court has long treated psychiatric testimony at an assize trial as one element among others submitted to the jury. The majority was right to insist on that freedom. But it does not resolve the Article 6 issue, because two questions must be kept apart. Was the jury entitled to reject the experts? Yes. Could the accused understand why it did so? That is another matter. The first concerns the independence of fact-finding; the second concerns the intelligibility and non-arbitrariness of judicial power. The more decisive the evidence rejected, the greater the need for a safeguard revealing why it did not prevail.
The majority added that the experts had themselves qualified the scope of their conclusions, describing them as reflecting their inner conviction and as “an informed opinion, not an absolute scientific truth”.8 With respect, this proves too little. Every expert opinion is an informed opinion rather than a scientific certainty. If that caveat sufficed to relieve a verdict of any need for intelligibility, the principle would dissolve in every case involving expert evidence. The caveat explains why the jury could depart from the experts. It says nothing about why it did.
The majority attached decisive weight to the questions put to the jury. The jurors answered “yes” to the first two questions, on guilt of premeditated intentional homicide, and “no” to the final question, which concerned the applicant’s current mental state.9 No separate question asked the jury to resolve her capacity to control her actions at the time of the acts, which was the only contested issue.
The majority reasoned that, since the first question concerned guilt, an affirmative answer necessarily implied that the jurors considered her responsible at the material time. That is true, but it is circular. Everyone knew from the verdict that the jury had found her responsible; what she could not know was why. A binary answer discloses the operative part of a decision, not its grounds. The same circularity recurs in the majority’s remark that responsibility was “a logical conclusion” in view of the jury’s answers.10 The conclusion of a verdict cannot at the same time be its explanation.
The majority also noted that the applicant’s counsel had raised no objection to the questions. The point calls for a precise answer. Under Taxquet, precise and individualised questions are among the safeguards the State must provide; they are not a burden transferred to the defence. Moreover, the dissent observed that the question on mental state made no reference to any concrete circumstance, as Taxquet required.11 The absence of an objection cannot supply reasoning that the questions, by their structure, were incapable of conveying.
The indictment, some fifty pages long, described the facts, the investigation and the applicant’s history and psychology. But it preceded the decisive development. It could describe the accusation and the evidence then available; it could not explain why the jury would later reject an expert conclusion that did not yet exist. The majority acknowledged as much: it accepted that the indictment was of limited assistance, since Article 6 requires the accused to understand not why the investigating courts sent the case to trial but what convinced the jurors after the hearing, and it declined to speculate on whether the indictment had influenced the deliberation.12
That concession is correct, and it exposes the difficulty at the heart of the judgment. If the Court itself cannot know what persuaded the jury, how can material that predates the decisive evidence establish what persuaded it?
The Government’s own submissions brought the problem into sharp focus. Its representative suggested that one could “easily guess” that the experts’ first report of 30 October 2007 had forged the jurors’ deep conviction and that their minds were already made up when the experts changed their opinion.13 The dissenting judges seized on the word. To guess, they observed in substance, is almost the opposite of reasoning and understanding.
The observation deserves to be taken seriously as a proposition of Article 6 doctrine. An accused has no Convention right to a verdict for which lawyers can later devise a plausible explanation. There is a difference between a reason that can be imagined, a reason that can be reconstructed and a reason that can objectively be attributed to the decision-maker. Only the last performs the function that Taxquet assigns to reasoning.
The Government’s hypothesis raises a further problem. If the jury’s conviction had indeed crystallised before the experts reported again, what became of the new evidence? It had been important enough for the President to order a fresh assessment and for three experts to change their view unanimously. Was it weighed and rejected, and if so on what basis? Or had the jury ceased to be open to it? No one can answer these questions. The Government’s explanation did not close the gap; it demonstrated that the gap existed.
The majority relied on the sentencing judgment, which referred to the applicant’s determination and to the cold-blooded manner in which she acted, and on the Court of Cassation, which read those passages as giving the Assize Court’s reason for finding her responsible. It answered the objection that the professional judges had not attended the jury’s deliberation on guilt by noting that they had heard the jurors during the deliberation on sentence and had followed the whole trial.14
These points do not bridge the gap. Guilt and sentence were distinct stages, decided by different bodies: under the procedure then in force, the jury deliberated on guilt alone and in secret. The jurors’ observations on sentence, made after the verdict, cannot reveal the grounds on which the jury had resolved the prior question of responsibility, and nothing in the sentencing judgment purported to record them. What it offered was an ex post rationalisation by judges who had not taken the decision it was said to explain. A reviewing court, likewise, may interpret reasons, test their adequacy and draw their consequences; it cannot know the unexpressed factual reasoning of a body whose deliberations it did not attend. Review can assess reasoning. It should not have to supply it.
Lhermitte cannot be assessed in isolation from the transformation of the Belgian assize procedure that took place around it. Five stages can be distinguished.
(a) The traditional model. Until 2010, the jury deliberated alone on guilt and answered the questions put by the President by “yes” or “no”. The jurors were reminded, in the instruction formerly set out in Article 342 of the Code of Criminal Procedure, that the law did not ask them to account for the means by which they had been convinced. Their verdict was, by design, unreasoned.15 Ms Lhermitte was tried under that model in December 2008.
(b) Taxquet in Strasbourg. On 13 January 2009 a Chamber of the Court found that a Belgian assize verdict answering unreasoned and non-individualised questions violated Article 6 § 1.16 The Court of Cassation ruled on Ms Lhermitte’s appeal in the months that followed, while the principle was already known.17
(c) The legislative response. The Law of 21 December 2009 reformed the Assize Court with effect from 21 January 2010. Its central innovation is Article 334 of the Code of Criminal Procedure: once the jury has returned its verdict, the court and the members of the jury retire together and, “without having to address all the submissions filed, they shall formulate the principal reasons for their decision”. In the words of the Taxquet Chamber judgment, such reasons should highlight the considerations that persuaded the jury and indicate why each question was answered in the affirmative or the negative.18 The Belgian legislature thus adopted exactly the standard at stake in Lhermitte: not an answer to every argument, but the principal reasons on the essential questions.
(d) Taxquet in the Grand Chamber. On 16 November 2010 the Grand Chamber confirmed the violation in Taxquet, while taking note of the Belgian reform.19 From that moment, the question was no longer whether safeguards were required, but how much they had to explain.
(e) Later developments. The Law of 5 February 2016 (“Pot-pourri II”) generalised the correctionalisation of crimes and threatened to marginalise the Assize Court; the Constitutional Court annulled that part of the reform on 21 December 2017.20 The jury survived, but it survived as a reasoned institution.
Set against this chronology, the paradox of Lhermitte is plain. By 2016, when the Grand Chamber ruled, Belgian law itself had long required that an assize verdict state the principal reasons on each essential question. Had Ms Lhermitte been tried thirteen months later, the court would have been obliged to state why the jury rejected the defence of mental disturbance supported by the experts’ final report. The Grand Chamber nevertheless accepted, for her, a procedure that the respondent State had already judged insufficient for everyone else.
The Court’s case-law on French assize courts points the same way. It found violations where nothing in the proceedings allowed the accused to understand the reasons for conviction, and later accepted a written statement of reasons as sufficient.21 Lhermitte was therefore not about the existence of the Taxquet principle, which was settled, but about its depth.
The majority’s reasoning is ultimately cumulative. No single safeguard supplied the explanation; taken together, the indictment, the questions, the adversarial hearing, the sentencing judgment and the review by the Court of Cassation were held sufficient. The method is not illegitimate in itself, since fairness is assessed on the proceedings as a whole. But “as a whole” cannot mean that several elements, none of which answers the decisive question, become an answer when added together. An accumulation of context is not reasoning on the issue that determined the case. Where the missing link concerns the very basis of criminal responsibility, the global assessment should not dilute the requirement that the essential reasoning remain identifiable. Otherwise a specific and decisive explanatory deficit simply disappears into the general fairness of the trial. That, in my view, is what happened in Lhermitte.
Drawing these threads together, the elements that should have led to a finding of a violation of Article 6 § 1 can be stated as five cumulative propositions.
(1) The issue was essential. Criminal responsibility was not one question among many; it was the only contested issue and the legal condition for any punishment.
(2) Decisive evidence pointed the other way. New evidence emerged during the trial and led the three court-appointed experts unanimously to conclude that the accused had been incapable of controlling her actions.
(3) No safeguard addressed that issue. The questions disclosed the result, not its basis; the only question on mental state concerned her present condition and referred to no concrete circumstance. The judgment identifies no direction to the jury on how to approach the experts’ final report.
(4) The substitutes were anterior, posterior or conjectural. The indictment preceded the decisive evidence; the sentencing judgment and the Court of Cassation came after a deliberation their authors did not attend; the Government’s explanation was, in its own word, a guess.
(5) The standard was already recognised by the respondent State. Since 2010, Belgian law requires the principal reasons for the jury’s decision on each essential question, without an answer to every submission. That is precisely the level of explanation that was missing.
None of these propositions requires jurors to write judgments or to answer every argument. Together, they describe a verdict whose essential element, the rejection of the defence of mental disturbance, remained unexplained. That is the core of the Taxquet requirement, and it is why the applicant could not know why she had been held responsible.
Scholarly criticism requires the majority’s position to be stated at its strongest. Four arguments support it. First, fairness is assessed on the proceedings as a whole, and the trial was fully adversarial. Secondly, the jury’s freedom to assess expert evidence is a legitimate feature of a system of lay adjudication, and the experts themselves qualified their conclusions. Thirdly, the applicant’s counsel did not object to the questions. Fourthly, the Court acts subsidiarily and should hesitate to second-guess a national procedure that the respondent State had already reformed.
Each argument, however, answers a different question from the one the case posed. The adversarial character of the trial shows that the issue was debated, not how it was decided. The jury’s freedom explains why it could reject the experts, not why it did. The absence of an objection cannot transfer to the defence a safeguard that Taxquet places on the State. And subsidiarity is weakest where the State itself has already recognised, by legislation, that the procedure in question did not allow a verdict to be understood. The minority, by contrast, asked the one question that mattered and gave it a principled answer. On the Court’s own standard, the minority was right.
There is a further dimension that a purely doctrinal reading tends to leave aside. The Grand Chamber decided by ten votes to seven; the Chamber had decided by four votes to three. Had two judges of the Grand Chamber voted otherwise, the outcome would have been reversed; in the Chamber, a single vote would have sufficed. Across the two formations, the case was lost by the narrowest margins the Court’s procedure allows.
This is entirely lawful. The Court decides by a majority of the judges present, and each judge is entitled to deliver a separate opinion.22 Nor is a narrow majority less binding than a unanimous one. But lawfulness does not exhaust the question of legitimacy. What was at stake for the applicant was not a point of procedure in the abstract: it was the difference between being treated as criminally responsible, with a life sentence served in prison, and being treated as a person who was not in control of her actions. When that difference turns on two votes, the persuasive burden carried by the majority’s reasoning is at its heaviest.
Three consequences follow. First, a narrow majority cannot rely on the authority of numbers; it must persuade. Where, as here, the majority’s reasoning answers the question of whether the jury could reject the experts rather than whether the accused could know why it did, the narrowness of the vote magnifies rather than conceals that gap. Secondly, in such cases the dissent is not a footnote to the judgment but part of its meaning. A seven-judge dissent signed by the President of the Court is a signal that the standard applied remains open, and the Court’s own history shows that minority positions can later become the law. Thirdly, the case exposes a paradox. The Court explained its own division: the applicant could read why ten judges ruled against her and why seven would have ruled for her. What she could never read was why the jury had rejected the experts. The institution that reviewed her trial demonstrated, by its own practice, the very safeguard that her trial lacked.
One need not argue for a qualified majority in Strasbourg to draw the lesson. Where the protection of a fundamental guarantee depends on so narrow a margin, the requirement that the decisive reasoning be identifiable, at every level, is not a formality. It is what allows the person concerned, and the public, to accept a result that could so easily have been the opposite.
The dissent did not require lay jurors to write an academic judgment. It did not make experts sovereign, nor did it question the jury’s freedom to assess the evidence. Its point was narrower and more principled: whether the safeguards actually enabled the applicant to understand why she had been held responsible despite the decisive development that occurred during her own trial. The indictment could not answer that question, because it preceded the development. The questions could not, because they disclosed the conclusion rather than its basis. The sentencing judgment could not reveal what had happened in an earlier deliberation from which its authors were absent. The Court of Cassation could not know what the jury had not expressed. What remained was inference, reconstruction and, in the Government’s own word, guesswork.
Lhermitte raises a question that extends well beyond Belgian jury trials. Every legal system recognises that judges and juries must be free to assess evidence. But that freedom cannot mean that a decisive contradiction between the evidence and the outcome becomes immune from intelligible explanation. Reasoning is one of the means by which the rule of law draws the line between freedom of assessment and unexplained decision-making, and the more consequential the finding, the more that safeguard matters. It does not require a long judgment, nor an answer to every item of evidence. It requires something more elementary: on an issue essential to the legal basis of a conviction, the accused must be able to identify why the decision went against her. In Lhermitte, criminal responsibility was such an issue. I remain unconvinced that the proceedings supplied that answer.
A Grand Chamber judgment settles the case. It does not close the legal debate, and a ten-to-seven division, with the President of the Court in the minority, shows that the debate was genuine at the highest level of the Court.
When the judgment was delivered on 29 November 2016, I did not experience it simply as a lost case. Losing belongs to litigation, and any lawyer who appears in Strasbourg must accept it. This felt different, because I was convinced that the case concerned one of the elementary functions of Article 6: preventing judicial power from becoming arbitrary or unintelligible. The gravity of the facts could not alter that principle; fundamental safeguards acquire their full meaning precisely in the hardest cases. The Convention does not ask whether the accused attracts sympathy. It asks whether the State respected the guarantees owed to every accused person.
Ten years on, I remain convinced of something on which the Article 6 argument does not depend, but which I owe it to say. After many conversations with the late Geneviève Lhermitte in prison, and in the light of what I observed afterwards, I remain convinced that at the time of the events she was not in control of her actions, exactly as the three experts concluded at the end of her trial, and that she never understood why she had been held responsible. The content of those conversations remains covered by professional secrecy. She was granted conditional release in April 2019 and chose euthanasia on 28 February 2023, sixteen years to the day after the events.23 The legal question addressed in this article would arise even if the jury had been right. My conviction is simply the reason why, for me, that question never closed.
I pleaded the case alongside the late Me Xavier Magnée. This article is dedicated to his memory.
These notes are general information on Convention law; they do not constitute legal advice and do not create a lawyer-client relationship.
After the final decision of the highest national court, a case may be submitted to us directly, by the client or by their lawyer.