Something very interesting happened late in the Lindsay Clancy trial, and it may tell us considerably more about the jury than anyone in the courtroom was prepared to say openly. After six days of deliberations, the jury foreperson informed Judge William Sullivan that one juror was allegedly refusing to follow the court's instructions concerning reasonable doubt. Defence attorney Kevin Reddington then did something remarkable: he asked the judge to remove that juror. Sullivan refused, observing that it was not his role to choose between "one juror or the 11," but Reddington's unsuccessful motion may nevertheless have revealed something important. We do not know how the jury is divided on the ultimate question of Lindsay Clancy's criminal responsibility, and Reddington presumably does not know with certainty either. Yet his decision to seek removal of the lone dissenter constitutes evidence about what he believes is happening inside the jury room.
The problem can be approached as an exercise in Bayesian probability reasoning. Bayesian reasoning does not require us to know something with certainty; it asks how new evidence should alter the probability we assign to competing hypotheses. In this case there are two simplified hypotheses. Under H1, the eleven-person majority favours an outcome relatively favourable to Clancy while the lone holdout favours conviction or a more serious degree of criminal responsibility. Under H2, the situation is reversed: the eleven favour conviction while the holdout is standing between Clancy and a guilty verdict. We cannot see inside the jury room, but we can ask which hypothesis better explains the evidence emerging outside it.
Before Reddington made his motion, we might have been highly uncertain between those possibilities. The jury had repeatedly reported itself unable to reach unanimity, but there had been no official announcement of which way the majority was leaning. The jury also had several possible outcomes available, ranging from murder and manslaughter through ordinary acquittal to not guilty by reason of lack of criminal responsibility. A deadlock therefore told us surprisingly little by itself. Eleven jurors might have been trying to convict against one Clancy holdout, or eleven might have rejected the prosecution's preferred outcome against one juror insisting upon it.
Then came the foreperson's note. According to courtroom reporting, the note complained that one juror was not following the judge's instructions regarding reasonable doubt. Sullivan questioned the jurors individually, brought them back together and reinstructed them on reasonable doubt. Reddington then demanded removal of the troublesome juror, arguing that there was one juror who "refuses to listen to the law" the court had given on reasonable doubt. The crucial point is that Reddington's action itself becomes evidence because an experienced defence lawyer would presumably assess the probable consequences before asking a judge to remove a deliberating juror.
Ask the counterfactual question. Suppose Reddington believed there was a substantial probability that eleven jurors wanted to convict Clancy while one heroic holdout wanted to acquit her or find her not criminally responsible. Would a competent defence lawyer ask the judge to remove that twelfth juror? It would be an extraordinary gamble. That juror would be the only thing preventing conviction. Removing the holdout and replacing that person with an alternate could transform a hung jury into the very verdict the defence had spent six weeks trying to prevent. A mistrial, by contrast, would leave Clancy unconvicted and force the prosecution to decide whether to undertake the enormous expense and controversy of trying the entire case again.
Yet Reddington did not merely tolerate the holdout; he actively tried to get rid of that juror. This does not prove that the lone juror favours conviction, but it changes the probabilities. Consider the probability that Reddington would make his removal motion if he believed the lone juror was favourable to Clancy, P(Motion | pro-Clancy holdout). That probability should be relatively low. Now consider the probability that he would make precisely the same motion if he believed the lone juror was hostile to Clancy and that the other eleven had reached a result acceptable to the defence, Prob(Motion | anti-Clancy holdout). That probability should be considerably higher. Once the motion occurred, Bayesian reasoning therefore tells us to update towards H1.
There is an important qualification because Reddington could not necessarily know how the jurors were voting. Jury deliberations are deliberately secret, and Sullivan himself was careful not to reveal or endorse the substance of their disagreement. Nor should we assume that Reddington possessed some improper source of information. His evidence consisted of what everyone in court could observe, together with the foreperson's communication, the history of the deliberations and whatever impressions an experienced trial lawyer had formed from watching this particular jury throughout the trial. None of that amounts to certainty, but that is precisely why Bayesian reasoning is useful.
In situations of incomplete information, the decisions of informed actors themselves become evidence. Imagine seeing a veteran poker player suddenly push a large stack of chips into the middle of the table. You cannot see his cards and the bet does not prove that he has a strong hand, because he could be bluffing. Nevertheless, if you have reason to believe he is rational and experienced, the fact that he was willing to make the bet changes your estimate of what his hidden cards are likely to be. Reddington effectively pushed some chips into the middle of the table when he asked Sullivan to remove the juror, accepting a risk that would make remarkably little sense if he thought this person was Clancy's last defence against conviction.
His language makes the inference stronger. Reddington complained that if the court shortly ended up with a mistrial because a juror had disregarded the court's instructions, "it's a shame." That is a curious thing for a defence lawyer to say if the lone juror is preventing eleven others from convicting his client. In that scenario the obstinate juror is not a shame from the defence perspective but potentially salvation, because the holdout produces a mistrial rather than a conviction. If, however, eleven jurors favour an outcome acceptable to Clancy and one juror is blocking it, Reddington's frustration becomes entirely intelligible. He may believe that he has come within one vote of winning the central battle of the case only to see the result frustrated by a juror whom the foreperson alleges is not properly applying the law.
There is now another piece of evidence to add to the Bayesian calculation, and it comes from an entirely different direction. Press reporting has described two female members of the jury as appearing in distinctive pink clothing. That matters because pink has become publicly associated with the movement supporting Clancy and seeking a not-guilty outcome. Supporters outside the courthouse have prominently used pink, making the colour part of the visual symbolism surrounding the trial. A juror appearing in pink could of course be coincidence. Two jurors doing so becomes more interesting, particularly in a case where anyone sitting through weeks of proceedings could hardly be unaware of the symbolism surrounding the courthouse.
Reddington himself has reportedly dismissed the coincidence explanation as unlikely, and there is a reasonable point behind that observation. We should not make the opposite mistake and treat clothing as equivalent to a ballot paper. A woman can own pink clothing without making a political statement, and even a juror sympathetic to Clancy as a person could still conclude that the law requires a guilty verdict. Nevertheless, Bayesian reasoning does not require evidence to be conclusive before it becomes relevant. The question is whether observing two jurors wearing a colour conspicuously associated with the pro-Clancy movement is more probable under a jury containing multiple Clancy-sympathetic members than under a jury in which only one of twelve jurors favours Clancy.
That creates an especially interesting mathematical constraint once the reported 11–1 division is added. If two jurors really were signalling sympathy with the pro-Clancy position and neither subsequently changed her view, they cannot both constitute a one-person minority. At least one must necessarily belong to the group of eleven. If both remained sympathetic to Clancy, both must belong to the eleven. The hypothesis that eleven jurors favour conviction while one lone juror favours Clancy therefore becomes increasingly difficult to reconcile with the observable evidence. It remains logically possible because clothing does not establish a vote and jurors can change their minds, but its probability falls.
The importance of the pink evidence is greater still because it is at least partly independent of Reddington's courtroom gamble. Our first inference comes from the behaviour of the defence lawyer: he wanted the lone holdout removed, which suggests he believed that juror was hurting rather than helping his client. The second comes from observable behaviour by members of the jury themselves. When two substantially independent streams of imperfect evidence point towards the same hypothesis, their combined evidential force can be considerably greater than either viewed separately. We are no longer merely interpreting Reddington's strategic behaviour; we have another potential clue concerning the composition of the eleven.
The accumulating evidence can therefore be understood sequentially. Initially we know only that the jury is deadlocked, which tells us little about the direction of the split. We then learn that the problem has apparently concentrated around one juror, making something close to an 11–1 division substantially more plausible. We observe Reddington trying to remove that one juror, which raises the probability that the lone juror is hostile to Clancy. We hear Reddington describe a mistrial caused by that juror as a "shame," which pushes the probability further in the same direction. We then add the reported presence of two female jurors wearing a colour strongly identified with the Clancy-support movement. No individual observation establishes the verdict preference of the eleven, but the evidence is not arriving randomly. It repeatedly pushes the probability in the same direction.
This still does not tell us precisely what the eleven want. It would be going too far to announce that eleven jurors favour a finding of lack of criminal responsibility. The majority might favour manslaughter rather than murder, second-degree rather than first-degree murder, or lack of criminal responsibility. There may also be complexities within the majority that cannot be reconstructed from the information released in court. The Bayesian inference is narrower but increasingly significant: the available evidence raises the probability that the lone holdout occupies a position less favourable to Clancy than the position occupied by the other eleven.
There is another reason the removal motion was strategically attractive. Reddington did not control whether Sullivan would grant it, and judicial removal of a deliberating juror is an extremely delicate matter. A juror cannot simply be expelled because eleven colleagues dislike the juror's interpretation of the evidence. If that were permitted, the unanimity requirement would become dangerously hollow because eleven jurors could convert an 11–1 disagreement into unanimity merely by characterising the dissenter as unreasonable or unwilling to follow the law. Sullivan therefore had powerful reasons to refuse the motion where the alleged failure to follow instructions was intertwined with the substance of deliberations.
That meant Reddington's downside from making the motion was limited but certainly not zero. If Sullivan refused, as he did, the jury remained intact and the defence had preserved its objection. If Sullivan granted the motion and Reddington's assessment of the jury was correct, the defence might have removed the single obstacle to a favourable verdict. If Reddington was wrong, however, he could have persuaded the judge to remove the very juror preventing his client's conviction. His willingness to take even that possibility seriously enough to make the application is therefore informative. An experienced defence lawyer does not ordinarily volunteer to remove the person standing between his client and a murder conviction.
If the jury ultimately hangs, meanwhile, Reddington has helped construct a powerful account of why it happened. The defence could argue that this was not simply a hopelessly fragmented jury scattered across five possible verdicts but a jury in which eleven people had apparently reached one position while a single juror prevented unanimity amid allegations from the foreperson that this juror was not properly applying the reasonable-doubt instruction. That narrative could matter enormously in deciding whether Clancy should ever be subjected to another trial.
A retrial would require another jury to travel through essentially the same enormous factual and psychiatric battlefield. If it eventually became known that the first jury divided 11–1 in Clancy's favour on the critical issue, the prosecutorial calculation would change dramatically. A prosecutor contemplating retrial would no longer merely be looking at a jury that "could not agree." He would be looking at a prosecution that, on this hypothesis, persuaded only one of twelve jurors to take the position necessary for the result the Commonwealth sought. Even if the eleven favoured a lesser offence rather than complete exoneration, such a division would still say something important about the prosecution's ability to establish its preferred account of Clancy's criminal responsibility.
This is why Reddington's motion, combined with the other clues, deserves more attention than it has received. We are accustomed to treating a jury room as a black box. Evidence goes in and eventually a verdict, or no verdict, comes out, while almost nothing about the internal process is supposed to escape. But black boxes can sometimes be studied indirectly. If an informed participant takes an action whose rationality depends heavily upon what he thinks is happening inside the box, that action contains information, and if separate observable evidence points in the same direction, our confidence in the inference can rationally increase.
This is the same logic used throughout science, intelligence analysis, economics and ordinary life. We infer invisible causes from visible consequences and update probabilities when actors with better information put something at risk. Here there is even a second-order Bayesian process at work. Reddington observes the trial, the jury and the foreperson's note and updates his assessment of what is occurring inside the jury room. He then acts upon that assessment by trying to remove the holdout. We observe Reddington's action and update our own probabilities about the jury. The reported pink clothing potentially provides an additional evidential stream that does not depend entirely upon interpreting the defence lawyer's behaviour.
None of this establishes the Clancy jury's actual division. Perhaps Reddington has badly misread the situation, perhaps the foreperson has mischaracterised the dissenter, perhaps the pink clothing really was coincidence, or perhaps jurors who once appeared sympathetic to Clancy changed their minds after deliberating over the evidence. Bayesian reasoning is not a magic device for manufacturing certainty from missing evidence. Its virtue is precisely the opposite: it gives us a disciplined way of reasoning when certainty is unavailable, allowing uncertain evidence to alter our conclusions without pretending that probability has somehow become proof.
The strongest clue may ultimately remain the simplest, but it is now reinforced by the others. If one juror were all that stood between Lindsay Clancy and a murder conviction, her defence lawyer would ordinarily want that juror protected like Fort Knox. Kevin Reddington instead asked the judge to remove that person, while two other members of the jury had reportedly appeared in a colour that had become conspicuously associated with the movement supporting Clancy. Neither fact tells us what cards are hidden inside the jury room, but together they make one reconstruction of those hidden cards increasingly more probable than the other: eleven jurors appear more likely to favour an outcome relatively favourable to Clancy, while the lone holdout appears increasingly likely to be the juror insisting upon a more serious finding of criminal responsibility.