The numbers game
10 January 2000
15 April 2013
28 October 2013
27 March 2013
11 April 2013
15 July 2013
Statistical evidence used to help convict Addleshaw Booth & Co lawyer Sally Clark of murdering her two children is being challenged by leading authorities. Tom O'Sullivan reports on the conflict over the use of such evidence.
It is not a trick question. What is the difference between one in 73 million and one in 8,500? For solicitor Sally Clark it could be the difference between life in prison or a return to her Cheshire home.
Clark was convicted of murder in November. Sentenced to life for smothering her two children in a 14-month period between 1996 and 1998. Her defence claimed the children, Christopher aged 11 weeks and Harry aged two months, were both victims of cot death.
But at her trial a prosecution expert witness, paediatrician Sir Roy Meadow, said the chance of both infants dying of cot death was one in 73 million. He explained that the relevant affluence of the parents - both were solicitors at Addleshaw Booth & Co in Manchester - was a contributory factor to the massive odds.
His evidence severely undermined the defence. In a case where there was conflicting forensic evidence to explain injuries to both babies - no definitive cause of death has ever been established - the statistic was a godsend for the prosecution. This was no coincidence. Lightning could not strike twice. It very simply explained why this woman must have murdered her children - at odds of one in 73 million there was no other explanation.
But now two months later it is the accuracy of the statistic and the monitoring of expert witnesses which is coming under scrutiny. Privately, a number of specialists have questioned the use of the one in 73 million statistic. But now for the first time Professor Peter Fleming, a consultant at Bristol Children's Hospital and the co-author of the unpublished report from which the statistic was picked out, says the prosecution used the statistic out of context and inappropriately. He is now willing to appear at Clark's appeal to explain the statistic and why its use was misleading.
He says: "The statistic was used to support a view being promoted by the prosecution - it should not have been used in the way it was used in court. One of my co-authors gave evidence during the trial but was prevented by the barrister from discussing the qualifications and caveats to the statistic by the way the questions were framed. I am very unhappy about the partial disclosure of unpublished data when used in a court of law where nobody else has the opportunity to look at the research and come up with alternative interpretations."
The statistic does appear in the research paper - jointly funded by the Department of Health and the Foundation for the Study of Infant Death - but with several qualifications. "One in 73 million is a theoretical figure," says the Foundation's secretary general Joyce Epstein. "In our experience we get at least one family every year which has had two infant deaths. That is the difference between a theoretical figure and our observation."
The debate over the statistic has been reignited by Stephen Watkins, director of public health at Stockport Health Authority. Writing in this week's British Medical Journal he dismisses the one in 73 million statistic saying that it "seriously misunderstands probability theory". He goes on to say: "It is speculation whether Sally Clark would have been acquitted without this evidence. But with this mathematical error prominent the conviction is unsafe."
He suggests that the probability could fall to as low as one in 8,500 - if probability theory is correctly applied -- and even dismisses this figure as "meaningless" because it fails to take into account the social, genetic and environmental factors specific to the Clark case - just as the original statistic did. Cot death campaigners point out that statistics are meaningless in such infant death cases because no single identifiable cause of cot death has been established.
Watkins highlights research into the rates of recurrence of infant death - that is, if one child has already died of cot death what is the likelihood of a second child born into the same family suffering the same fate. Some research shows there is no likely increase while others suggest the recurrence rate could be as much as five times more likely for the second child.
Speaking to The Lawyer, Watkins said: "It is grossly unfair to say that lightning cannot strike twice. It strikes twice, on average, every 18 months. This is a conviction by mathematical error. The prosecution said that it is highly unlikely [that two children should die in this way] but there are case studies on it and, by definition, if you have case studies it is not such a rare occurrence. This tendency to pluck such probabilities out of the air without going into details is wrong and it has to be stopped."
Paul Blomeley, the branch crown prosecutor in Warrington, the CPS office which prosecuted Clark, says: "As this case is now the matter of an appeal I cannot comment, but I understand the statistic will be one of the grounds of the appeal and it will be dealt with then. One relies on one's experts and the defence's and then the jury makes its decision. I am not aware of any criticism of the way the case was prosecuted."
But Watkins' concerns go beyond the Clark case into the wider misuse of probability theory, the supervision and monitoring of expert witnesses in trials, and also the need for judges to spell out to juries what a particular probability statistic actually means.
A 1996 Court of Appeal hearing (R v Doheny) spelled out a judge's responsibilities, but the growing use and sophistication of forensic evidence - including DNA - means more criminal cases are calling on expert witnesses and a greater understanding of statistics and probability is required. Other than the 45 forensic pathologists voluntarily registered with the Home Office there is no centrally coordinated list of approved expert witnesses in England and Wales. And so no way of monitoring or assessing their ability to give evidence.
"There are two issues. Lawyers need to better understand the mechanics of probability theory and the question of who supervises expert witnesses needs to be addressed," says Watkins. "Experts stand up in court and their evidence is taken as gospel - and quite rightly most of the time. But what if it is wrong headed, or the expert has a particular theory to promote, or what if they are giving evidence in an area where they are not actually skilled?
"It is possible to be an extremely good doctor without being numerate and not every eminent clinician is best placed to give epidemiological evidence. Expert witnesses can hold a substantial part of defendants' lives in their hands and the defendants deserve the same protection as patients."
Dr Trevor Rothwell is secretary of the Home Office Policy Advisory Board for Forensic Pathology which controls the voluntary list. "The system, leaving aside the Home Office forensic pathologists we try to regulate, is a minefield. Sometimes people say that lawyers go for experts who will say what they want to hear - I think in some cases that is true.
"The Home Office has built-in standards but the opportunity is there for 'experts' to step in and say what they want to say - this is true of medics as well - that is the way the system works. There is no effective monitoring system to assess standards of evidence or the ability of witnesses."
The Council for the Registration of Forensic Practitioners (CRFP) was created as a response to the Royal Commission on Criminal Justice following a string of high profile miscarriages where forensic evidence was found to be wrong. It hopes to begin registering pathologists this spring before adding scene of crime officers, fingerprint experts and other specialists who deal in this type of work and may be called upon to give evidence in court.
Robert Forrest, a professor of toxicology at Sheffield University who sits on the CRFP working party developing a disciplinary code for members, says: "At the moment it is the court which has to decide whether you are qualified to give opinion evidence - most witnesses obviously cannot give opinion evidence - it is a discretionary exception granted to expert witnesses. The CRFP should at least ensure a basic competence in those giving expert evidence before the courts."
It will come as a huge surprise to many that such a basic competence is not already guaranteed.