Sunday, December 22, 2019

On the evidence: the verdict in Lundy

Don’t let me fall into the trap of getting into an argument about the Lundy verdict. By verdict I mean the verdict reached by the Supreme Court justices.

I don’t think an appellate court should ever have to reach a verdict. But when it does, it can illustrate how fact-finding should rationally be undertaken.

Appellate courts usually consist of three or more judges, and differences are settled by majority vote. The defendant at a jury trial that has gone wrong may therefore end up being convicted by a majority in the appellate court. Twelve jurors or two judges – which would you prefer?

That’s the broad picture. In Lundy the Supreme Court mentioned the Court of Appeal’s point that this was a case which did not turn on the credibility of witnesses [85]. If it had been, no doubt the Court would have recognised that it was at some disadvantage because it did not hear and see the witnesses, and the disadvantage might – although not necessarily - have been such that it would not be proper for the appellate judges to try to reach a verdict.

When appellate judges reach verdicts, the right to a jury trial is limited. This was recognised at [34], where the Court refers to observations of the High Court of Australia in Weiss v R [2005] HCA 81 at [30]. Essentially, and if I might be cynical for just a moment, this means that the right to a jury trial is a right to an attempt at a lawfully conducted jury trial.

This was an entirely circumstantial case. So was Bain, and we remember how the Privy Council rejected any temptation to come to its own verdict. Even with the apparent change in approach brought about in Matenga (being more permissive of appellate court verdicts), the Privy Council said that the result in Bain would have been the same. So this whole area invites controversy.

Objections to the Lundy verdict focus on the alleged unreliability of evidence that stains on Mr Lundy’s shirt contained central nervous system tissue (“brain tissue”). Let’s put that evidence to one side, and consider what the verdict would have been without it.

The presence of Mrs Lundy’s DNA in the shirt stains was incontrovertible. The stains were not tiny: 25mm x 10mm, and 30mm x 20mm (noted at [75]). They were visible. I doubt that most people would wear [strictly on these facts, have possession of] a shirt with that staining, especially as it contained blood. And, how common is it to have blood of that quantity and from one’s partner on a shirt that has continued to be worn? Even supposing that 20% of people wear shirts with such staining – and I take this as an absurdly high occurrence in favour of the defence – this evidence strongly favours the prosecution case.

Attention then turns to the other evidence: motive, timing of food consumption, the neighbour’s observation of the open door, the fuel consumption and odometer evidence, the surprising absence of DNA evidence in Mr Lundy’s car or at the motel, the apparent staging of the burglary, the paint, and the child victim’s blood specks on Mr Lundy’s shirt. Taking all that together could firmly support the prosecution case, or weakly support the defence case, depending on the various likelihoods that fact-finders could reasonably attach to the various items of evidence.

If this other evidence in combination firmly supported the prosecution case it would be reasonable to conclude that Mr Lundy was guilty (remembering that this is ignoring the brain tissue evidence). If it weakly supported the defence case, then guilt could not be proved on the (generous to the defence) assumption that 20 per cent of people have their partner’s blood on their shirts to the extent that Mr Lundy did. If this figure is “only” 10 per cent, then guilt would be proved beyond reasonable doubt.

How do we know? This is Bayesian reasoning with conditional probabilities. Never mind if you don’t employ that somewhat mathematical approach, because common sense should tell you the same. [1]

The brain tissue evidence resulted in considerable expense and delay in the proceedings, and you could wonder whether the prosecution was over-egging its pudding. But it didn’t know whether the defence was going to have a ready answer to the DNA presence in the stains. This could be seen as a compliment to the skills of defence counsel, but we could wonder whether, in seeking to prove its case to a certainty rather than merely beyond reasonable doubt, there was (to pursue another culinary metaphor) unnecessary icing on the cake.

What if the appeal had been heard under the new appeal criteria? How would the evidence be analyzed if the court simply had to ask itself whether the error at trial had created a real risk that the outcome of the trial had been affected? The error was material to one chain of evidence: stain – CNS tissue – (mRNA) – Victim1. Taking out this mRNA chain, the remaining chain was: stain – blood – DNA – Victim 1. Did the removal of the mRNA chain significantly weaken the link between the stain and Victim 1?

[1] Oh alright, for you maths buffs: If the defendant was innocent, a huge coincidence had occurred: blood from each of the two victims was on his shirt. Such little information as there is online about transfer of blood to clothing suggests that about 3 per cent of people have blood from someone else on their clothing. This is much lower than the figures I assumed above. The ratio for the likelihood of the prosecution hypothesis to the likelihood of the defence hypothesis, the probability of finding the match given each hypothesis, would therefore be, for each of the blood sources (ie, the victims), 1 to 0.03, or 33.3 to 1. Combining these gives 33.3 x 33.3 = 1108.89 to 1. Assume, generously to the defence, that before considering this evidence we might expect that the probability of guilt compared to the probability of innocence could be, say, 0.2 to 0.8. This ratio is what is called called "the priors" combined with the other evidence in the case. On the blood evidence, and given this priors-in-combination assumption, the combined ratios are 1108.89 / 4 to 1 = 277 to 1. That is the ratio, given this evidence, of the probability of guilt to the probability of innocence. The probability of guilt would therefore be (converting the ratio to a fraction) 277 / 278 = 0.99. That is proof beyond reasonable doubt, on the basis of the assumption that was generous to the defence. Some people say, but you have just guessed at the priors, the whole thing is too subjective, where is the evidence? The answer is that the priors can vary a lot without having much effect on the result. If, instead of assuming priors combined with the other evidence of 0.2 to 0.8, we assume priors combined with the other evidence of 0.05 to 0.95 (very favourably to the defence, effectively an assumption of innocence), then the resulting probability of guilt after considering the blood evidence is 0.98. That is: (0.05 / 0.95) x 1108.89 = 58.36 to 1, so probability of guilt = 58.36 / 59.36 = 0.98.