Main Banner

Who killed Jill Dando?

April 2019

Did Barry George kill Jill Dando? The idea might seem outrageous, since Mr George was acquitted of the crime and it is almost universally accepted that the murder remains unsolved. If you Google anything about the case you might assume that expert and public opinion is settled: that a man with mental health problems was made a scapegoat on the flimsiest of evidence, and only through the persistence of decent-minded campaigners was a great miscarriage of justice finally overturned. But the question is more reasonable than some suppose, as indeed the High Court has acknowledged: “There was indeed a case upon which a reasonable jury properly directed could have convicted the claimant of murder.” And the fact that Mr George was released from prison in 2008, seven years into his life sentence, was almost certainly due to a simple error of logic which points to a systemic failure in English justice.

The murder of one of Britain’s best-loved broadcasters in April 1999 was much more than a personal tragedy and a trauma for her family, friends and colleagues. Because of her celebrity status it was bound to cause a furore, especially since one of her roles was alongside me in fronting the police appeal series Crimewatch, then one of the highest-rated factual TV shows. The obvious inference was that the killer must have come from the underworld, and conspiracy theories dominated the news for weeks (as assumed in The Mirror’s headline). The attack was said to be sophisticated and meticulously planned, presumably the work of a hired assassin, or a professional gangster with a grudge.

When more than a year later police arrested Barry George, a local man who had epilepsy and a low IQ, many assumed the cops had lost the plot and simply gone out to find a fall guy. Even when an Old Bailey jury convicted Mr George, rumours swirled that this was a grotesque miscarriage of justice. His flamboyant defence QC, Michael Mansfield, proposed that the real killer was a special agent from Serbia sent in revenge for NATO bombing aimed at halting ethnic cleansing. When, eight years later, the case was sent for trial again, the only forensic evidence was discounted, and Barry George was finally acquitted.

Little wonder that many accounts on social media and even mainstream documentaries play up the widely-held view that the police failed utterly, and that George was no more than a sacrificial lamb.

But the evidence suggests a very different story.

The failure of the detectives was not that they found a patsy to take the blame when their inquiry had run into the sand. It was that they failed to heed warnings about Barry George which should have been flagged up from the start.

Indeed, when the facts are set out in a logical sequence, not as a narrative for the prosecution or defence, anyone who is dispassionate would recognise that, whether or not Barry George was the killer, he was nonetheless subject to an uncanny number of coincidences – many of them striking.

The following is uncontroversial:

  • The murder scene was comprehensively compromised by paramedics in their concern to save Jill Dando.
  • Thereafter there was little forensic evidence other than a casing from the bullet, and, later, the bullet itself and trace residue on Jill’s clothing and in her hair.
  • There were witnesses who had seen a lone man waiting, had heard a scream and had seen the lone man walking hurriedly away. There could be little doubt that, whoever he was, he was the culprit.
  • There was no obvious motive.
  • The inquiry was almost immediately overwhelmed by information, much of it seemingly of little or no evidential value but all of which – some 7,000 lines of inquiry – had to be evaluated.

Accordingly, the investigation began with an open mind about whether the killer was a hired assassin, part of a criminal gang, a lone criminal with a grudge, a jilted lover or someone with a mental disorder. At least that was the theory. The senior investigating officer, DCI Hamish Campbell, insisted this was a “360 degree investigation”. However, as he and his colleagues will now concede, they were under intense pressure from the media, and from top brass in the Met, to prioritise potential criminal connections. Every appeal Jill had conducted for Crimewatch was investigated. Extensive contacts were made with criminal contacts. These all drew a blank.

When other conspiracy theories surfaced these also sucked up thousands of hours of detective work. One of the most risible, yet one that came to have such prominence, was the idea of a Serb assassin. The story might not seem improbable the way it has so often been told: that a NATO missile attack on a Serb TV station, causing 16 fatalities, led to reprisals. But it requires that the Serbs were easily distracted from the immediacy of their life and death conflict in the Balkans, that they blamed Britain not the US, that they ignored their own failings (in fact their director of the station was later sentenced to 10 years in prison for failing to evacuate the building), that they had the resources to move with remarkable speed (the TV station bombing happened 48 hours before Jill’s murder 1,300 miles away), that they would never claim responsibility for a remarkably successful tit for tat, and that even after the war was over, western intelligence could find no documents or plausible informants to support the claim. More significantly, though, is the origin of the story. Jill’s agent had been asked by the police if anyone might have had a grudge against her, to which the answer was firmly no. But she did recall a letter criticising an appeal Jill had voiced about the Balkans. Its tone was injured but contained no element of threat, But her boss told the press, and the story catapulted from an unremarkable private letter to an international conspiracy.

The murder team spent months on it, but they and the intelligence services turned up nothing.

Finally, some 10 months into the inquiry, DCI Campbell took stock and went back to review all other possibilities.

There had always been indications that Jill’s killing was not professional at all but bore the hallmarks of an amateur. Of course it is possible that a real pro might lay a false trail, pretending to be slipshod and incompetent, but it would have been exceptionally risky.

Again, the following is uncontroversial:

  • No getaway vehicle, and it seems no getaway plan – almost unprecedented for a contract killing.
  • The wrong address, so one where an attacker would have to wait conspicuously for ages (she lived with her fiancé, rarely went to her own house and only decided to go there at the last minute because of a phone call with her agent while she was driving not far away).
  • No backup personnel who could have tailed Jill (we know that from a thorough check on CCTV images).
  • A risky location for a murder: the man who had been seen hurrying away had to walk down Gowan Avenue, a long straight street with few turn-offs where a would-be assailant could be tracked by neighbours or passers-by for at least 40 seconds – a long time for a professional, or any killer who had given any thought to how he would escape.
  • A perilous and unreliable weapon. The gun was not a real one but a re-activated or re-engineered one like a starting pistol – prone to jamming, misfiring or even backfiring. (Neither the Metropolitan Police nor the FBI could find a precedent in a professional assassination, so this would have been unique.)
  • The bullet was re-filled or at any rate hand crimped, increasing all the above risks and improbabilities.
  • Jill had her key in the lock ready to enter, yet the murderer did not push her inside and shut the door to muffle the sound but shot her outside in public where the attack could be witnessed, the assailant seen holding a gun, and the body quickly found.
  • The killer physically held the weapon in contact with his victim, suggestive of a crime of passion but breaking all the rules of an intelligent or trained assassin. Direct contact is always a forensic risk, and as soon as he pulled the trigger he would have been sprayed with biological material as well as gunshot residue.

The forensic psychologist who analysed the murder quickly concluded that any two or three of these features would suggest the killer was not just an amateur, but was taking alarming risks, and that taken together the clues were conclusive. He formally advised that the murderer would be a classic loner with a severe personality disorder.

To his surprise, the investigating team ignored his professional guidance.

Probability supported his view. In recent history (with one notable exception in 1981) no English judge has been assassinated in reprisal for an arrest or conviction. Nor are revenge attacks a feature of life for prosecution lawyers, detectives or arresting officers. It is simply not part of the British underworld tradition. It is extremely rare even in the US where gangsterism and murders are so much more commonplace. It would be very far-fetched indeed for a criminal fraternity to take it out on a peripheral player such as a TV presenter.

Then there was the question of motive. Once a criminal underworld attack had been discounted (statistical improbability, no whispers after a year despite intensive contacts), other theories like a Serb attack ruled out (practical improbability, no supporting evidence or even hints after international intelligence trawls), and no sign of a fall-out with friends, relatives or former lovers, there was no obvious reason for Jill’s murder, especially for a risky killing in the street.

Which left the most plausible explanation. Those unused to forensic psychology often talk about ‘motiveless crimes’, because there is seemingly no rational purpose for someone to act in that way. Yet thousands of these apparently senseless crimes are committed by people who have a very clear internal logic. These include most of the most notorious killers including Peter Sutcliffe, Ronnie Kray and Charles Bronson. In fact Britain’s prisons are crowded with people who are mentally ill. Broadmoor and Rampton are full of them. But that apart, thousands more crimes are committed by people who are not mentally ill so much as suffering from a personality disorder.

Personality disorders result in cognitive impairment, obsessions and narcissism, mostly resulting from genetic or physical damage, such as frontal lobe pathology, but sometimes from severe abuse. These disorders are not curable and so, until recently, even violent offenders were sent to prisons rather that to secure hospitals. Psychiatrists concede they can do nothing for them.

Mark Chapman, who killed John Lennon, suffered a form of personality disorder, and it’s not hard to see similarities in the way Jill was ambushed. Interestingly Chapman modelled his life on a hero of his (in his case, Holden Caulfield, a figure from The Catcher in the Rye) rather as Barry George had modelled himself on famous people, most recently a figure from the rock band Queen, which is why he had adopted the name Bulsara.

What are the chances of Jill being know to such a personality disordered person? The answer is almost 100%. As Jill’s co-presenter, I wrote to DCI Campbell at the start of the investigation, pointing out that in Jill’s case – as with any public person – one needs to consider motives in a very different light from private individuals because celebrities are known to almost everyone. She was routinely on TV, in people’s homes, in their sitting rooms and in their bedrooms.

Thus, albeit ten months too late, personality disorder became a prime line of inquiry

In February 2000, going back to review evidence from the start, it emerged that one individual had been identified several times as a person of interest but, because he was a Walter Mitty character and used different names, the police system had failed to link them.

It turned out that:

  • He had been seen in the road four hours before Jill’s murder.
  • He had been identified as having been by two separate witnesses near the killing in an agitated state soon after the murder.
  • The witnesses, including a mental health worker, were so concerned about him that in the days after Jill’s murder they had rung the incident room repeatedly.
  • He had returned to both witnesses the following day seeking to persuade them he had been there at different times and in different clothes.
  • He had a history of violence against women, including formal warnings, convictions for sexual assault and a prison term for attempted rape.

Further inquiries revealed that:

  • When under surveillance he routinely stalked women.
  • He had once been found in the grounds of Kensington Palace with a balaclava, a knife and a rope – but was never charged because he was thought to have mental problems.
  • Itsuko Toide, briefly his wife in a marriage of convenience, was so frightened of him that she reported his violence to the police and fled back to Japan. (She later told me she had no doubt he was Jill’s killer).
  • Despite denying he had an interest in Jill Dando or the BBC, undeveloped film recovered from his apartment showed he took photos of women from his TV, and he kept copies of the BBC’s in-house magazine.
  • Although he denied he had ever owned or held a gun, a reel of undeveloped film was processed and revealed him posing with a pistol.
  • The pistol he was holding was of the same type that killed Jill.

…and so it went on.

Some argue that, never mind the layer upon layer of coincidences, Barry George simply could not have carried out a murder because he was too slow-witted. Tests showed he had an IQ of 75. That means low average intellectual capability; or to put it another way, roughly 1:12 of the population has his IQ or lower. It is no more unusual than an IQ of 125. Moreover, his record showed he was perfectly capable of functioning and planning:

  • He had once posed as a police officer, having acquired a genuine warrant card. He was convicted of the offence.
  • He persuaded a newspaper that he was the British Karate champion until another paper exposed the fraud.
  • He had posed as a businessman, having created a fictional company called Xanadu Constructional and Mechanical Engineers.
  • He pretended to be a stuntman and was filmed at a stadium performing a daring motorbike stunt, injuring himself in the process.

Then there was the matter of stalking women, sometimes into their front gardens and up to their front doors, a pattern of behaviour which was as deliberate as it was distinctive. And if he could plan an attack in the grounds of Kensington Palace when the Princess of Wales lived there, why not on Jill Dando in an ordinary suburban street?

Others of George’s supporters say his protestation of innocence has been so consistent and sustained that he simply has to be believed. In fact his testimony was riddled with inconsistencies and lies – including that he’d never heard of Jill Dando. But in any case there is good reason to discount his apparent sincerity. People with Mr George’s personality disorders (in his case bolstered by Asperger’s) can live out their fantasies and are so persuasive that they can often win over others. They lie consistently and convincingly.

In fact Barry George had been hoodwinking people for decades, frequently adopting the names of rock musicians, and carrying through with his conceits for months on end.

One of George’s critics, Stephanie Hall, has been through this before. She cites the classic case of Simon Hall, convicted of a murder in 2001 but who so plausibly maintained his innocence that many people, including ‘innocence’ activists and human rights lawyers, campaigned for years to have him freed. Stephanie shared Simon’s surname because she’d known him at work, became one of his keenest supporters, and actually married him in prison. Five years later he confessed to her, revealing details including his motive for the crime, and soon after committed suicide. Stephanie has subsequently researched many similar cases, and has been vociferous about the similarities between Simon Hall and Barry George.

She was not the only such campaigner to feel betrayed. Most miscarriage of justice activists quietly bury embarrassing cases like that of Simon Hall, but one of Britain’s most prominent champions against wrongful convictions (and a one-time broadcasting colleague of mine), concedes it is all too easy to be duped. David Jessel, a former presenter of campaigning TV shows like Rough Justice and Trial and Error, and a long-time member of the Criminal Cases Review Commission concedes: “You always have to reserve a part of your brain for the possibility that the person you campaign for just might be guilty.”

One of the foremost profilers in the US, Pat Brown, says too many coincidences end up with Barry George to be ignored, and is contemptuous of armchair theorists who think he must have been either too dim or too nice to kill. Such people, she says, plainly know little about personality disorders, or indeed criminal profiling.

The fundamental particle

In short, lots of so-called ‘circumstantial’ arrows pointed compellingly in Mr George’s direction – and there’s nothing wrong with circumstantial evidence, especially when so much of it leads to the same conclusion; through criminal history it has been the bedrock on which almost all verdicts have been reached.

But in addition , a particle of firearms residue was found in the pocket of a coat Barry George had been wearing on the day. What’s more, its analysis revealed a mix of barium, aluminium and lead which matched the type of gunpowder recovered from Jill’s hair and clothing. This was the hard forensic evidence the detectives craved, or so it seemed, and it may even have finally convinced the CPS to advise police to charge Mr George with murder. But it turned out to be a poisoned chalice. We cannot know if it played any part in the jury’s decision to convict (English justice, though supposedly open, has absolute secrecy at its heart), but we know for sure that this tiny scrap of residue – just one small part of the case against Mr George and effectively a postscript which was found only after he had been arrested and charged – was what led to his subsequent acquittal.

No one argues about the fundamental science. Such particles are formed under extreme temperature and pressure and are sprayed forwards and backwards when someone fires a gun. The only uncertainty was how this one got into Barry George’s pocket. Supporters claim it led to an appalling miscarriage of justice in which an obviously innocent man was sent to prison. But what’s in question here is not so much a question of physics as of probability, and – overlooked by almost everyone involved in the media extravaganza over Jill Dando’s murder – the logic of probability points in the opposite direction. The implications are so important for English law in general that the whole trail of events deserves setting out in some detail, not least because, unless anyone can find an error in what follows, it means Mr George should not have been released from prison, and the guilty verdict should still stand.

At his trial in 2001 the forensic evidence was offered as tentative but nonetheless of evidential value. Years later, while Barry George was in prison and supporters were protesting his innocence, the Forensic Science Service had second thoughts. That prompted a review by the Criminal Cases Review Commission, a statutory body set up to investigate miscarriages of justice. (Having said that, its interpretation of miscarriages of justice is very much one-sided. It has no powers to intervene if Barry George was wrongly acquitted.) That in turn led to an appeal – in fact his second, his first having been dismissed in 2002. And at the Royal Courts of Justice in London his lawyers successfully argued the forensic evidence should be challenged, leading to a retrial. When this second trial took place, more than seven years after the murder, Mr George’s barrister, William Clegg, persuaded the judge to rule the forensics inadmissible. So the second jury was never told about the firearms discharge residue; and given that nine years on from the murder most witness memories had faded, a new verdict of not guilty was almost inevitable.

Most people have such faith in English justice that they assume the courts must have got this right. Those who always insisted Barry George was innocent felt vindicated. Most of the media, which had previously catalogued his misogyny and violence, turned on a sixpence to deplore the miscarriage of justice that had led to an innocent man’s imprisonment (and two newspapers which persisted in making allegations against him were successfully pursued for libel). The internet was abuzz with stories of his ‘wrongful conviction’, dozens of books and articles have been published to this effect, and supposedly ‘definitive’ documentaries have left viewers in little doubt that the real killer has never been caught.

But was the Appeal Court right? Was it fair that the forensic evidence was ruled inadmissible and that the second jury was kept in the dark? If not, Barry George should not have been released.

Indeed, statisticians who later reviewed the transcripts think the evidence should not have been discounted, and that, on the contrary, it provides ‘weak’ but ‘probative’ support for Barry George’s guilt. This is not mere conjecture, and nor are its implications trifling: it is literally a textbook case of how courts can get things wrong. Two statisticians who worked through the judgment (Professors Norman Fenton and Martin Neil) together with a lawyer, a psychologist and a behavioural scientist (Daniel Berger, David Lagnado and Anne Hsu) expressed serious concerns about ‘oversimplification’, ‘ill-defined assumptions’, and ‘erroneous reasoning’, and warned that the problems go much deeper than the intuitive logic that expert witnesses had proposed and that the lawyers had all swallowed.

So let’s wind back and examine why the particle was thought to be evidential in the first place, why it was then discredited, and why probability experts think it was wrong to be disallowed.

The Forensic Science Service had always conceded that the speck of gunpowder was by no means conclusive proof. It was so tiny – the smallest amount detectable – that all the scientists agreed it was most unlikely to have been planted. Inserting a single microscopic fragment into a fold of clothing would be difficult and, even if experts had gone to the trouble of framing someone, they would have scattered far more convincing evidence than just one infinitesimal speck. Even so, it could have found its way into Barry George’s pocket by accident, perhaps when the coat was handled by police. So which was mostly likely: accidental contamination, or a direct indication of contact with a firearm? At the time of the first trial in 2001 the boffins thought the latter was more likely. But five years later the Forensic Science Service was going through a period of reassessment and worried there was ‘not sufficient data on the environmental occurrence of FDR (firearms discharge residue) to give a safe interpretation of finding a single particle’. As a consequence, they adopted ‘a cautious approach’. In fact it was so cautious that by the time of the appeal they had concluded that a single speck had no practical value at all. They guessed there might be a 1% chance of finding the particle whether or not Barry George handled the gun, and accordingly the evidence was ‘neutral’. To quote their expert witness Robin Keeley:

‘It was no more likely that the single particle of FDR came from a gun fired at the time of Miss Dando’s murder than that it came from some other source.’

Strong stuff, especially as Mr Keeley is known as the founding father of firearms chemistry. Yet when one of the world’s greatest authorities on probability reviewed the Appeal Court transcript he found it, literally, incredible:

‘I cannot believe Mr Keeley said anything so blatantly incorrect.’

So let’s examine why.

On the face of things, calling the firearms residue ‘neutral’ suggests it added nothing to the question of whether Barry George was guilty or not since the chance of leaving residue would be the same in either case. The layman’s implication is that if you turn out all your pockets there might be burned gunpowder in at least one article of clothing – especially if you have been in contact with the police. There is no evidence whatever to support such an assertion. And although the Mr Keeley and his colleagues sounded convincing to the court, their reasoning seems to have gone awry.

The problem is that while they knew a lot about chemical analysis, it appears they were not experts in statistics, and nor were the judges and lawyers. Probability – the likelihood of anything happening – is far less obvious than it seems. (Indeed, the whole gaming industry might be in trouble if we all understood the counterintuitive implications of probability.)

Courts should have learned this from the injustice done to Sally Clark in 1999. She had been sentenced to life for murdering her two sons who died in infancy a year apart. An eminent paediatrician testified that the chance of two babies from the same family suffering sudden infant cot syndrome was 1 in 73 million. But his calculation was based on a false assumption. As the Royal Statistical Society pointed out, there was no reason to suppose the two cot deaths were independent events. There might have been a genetic weakness or some other overlapping cause.

In the Barry George appeal, the firearms experts made a similar mistake. They fell for what statisticians call the zero-sum fallacy: they posed a false dichotomy. They believed that if the particle got into the coat innocuously it would imply that Mr George had nothing to do with the crime; whereas if the particle came from his gun it would indicate his guilt. Since they estimated there was a one in a hundred chance a fragment would be left if he had fired the gun, and a one in a hundred chance if he hadn’t fired the gun, those probabilities were equal.

In fact, of course, those possibilities are not mutually exclusive. True, the fleck could have got there by accident if he was innocent. Alternatively, if he were guilty it could have come from his own gun. But there was at least one more possibility. If chance contamination can happen to an innocent man it can happen to a guilty one too. In other words, it could have got there by accident whether or not Barry George was Jill Dando’s killer.

The seemingly trivial distinction is important because it changes radically the evidential value of the particle. A simple illustration is to consider what happens if you make a bet. An either-or hypothesis is like a single toss of a coin. It either falls heads or tails. But the odds change when you have three possibilities, not two. It gives you another toss of the coin. Consider the chance of flipping two heads in a row: it’s 1 in 4. This is known as the prior probability and the odds are stacked against you. But if you get a head with the first spin the chance of getting two heads now rises to 1:2. This is known as the posterior probability. Now you are just as likely to win as lose. Even amateur gamblers understand this and would raise the stakes after the first roll.

Thomas Bayes, the father of probability calculations, understood this three and a half centuries before the Appeal Court overlooked it, and Bayes’ Theorem provides an exact mathematical way to calculate the posterior probability. In the Barry George case, rather than trying to flip coins, we need to know how many people touch guns.

After all, if everyone in Britain handled guns, the average likelihood – or prior probability – that Barry George handled a gun would be 100%. Everyone else would have been equally likely to have harboured similar residue, so finding a particle would tell you nothing about whether he might have been Jill’s killer. Yet if very few people have access to firearms then finding the speck would be very significant. Consider checking 100 pockets and finding the only tell-tale speck in Barry George’s one.

Similarly, if most people have a particle of burned gunpowder in their pocket, the chance of recovering one from Mr George’s pocket was overwhelming. But if only one person in ten thousand had residue in their pocket then the chance of finding one would be miniscule. In that case a finding would be significant. It is only a combination of factors that can indicate whether the residue had evidential value.

Prof Norman Fenton and a team of statisticians from Queen Mary University and University College London modelled a range of scenarios on the Barry George case so they could calculate probabilities depending on how these circumstances vary. Whatever reasonable scenario they fed into the program, it tended to indicate that finding the particle increased the chance that Barry George was the shooter.

So what is a reasonable scenario? What is the probability that the trace of residue got there without him firing the gun, perhaps as his defence counsel proposed, that it was contamination caused when the coat was removed from its evidence bag to be photographed by the police? To assess the chance of environmental pollution we need to know how many guns there are in London which could have led to haphazard transfer. More particularly we need to calculate what proportion of these particles could have found their way into police stations in the capital.

First let’s look at the number of guns. No one knows how many people in London have direct contact with firearms, but we can make an educated guess. In doing so we shall always err on the side of Barry George’s defence. Most legal guns in Britain are held by hunters and shooters in the countryside. Nonetheless some 4,600 citizens hold handgun and rifle licences issued by the Metropolitan Police. Shotguns are counted separately and leave chemically different residues, but they too could issue particles like that found on Mr George. There are roughly 26,000 licences for these. That means altogether some 40,000 Londoners have access to licensed weapons which could deposit Barry George-type traces. Then there are the police themselves: around 2,500 armed officers in London, along with staff who might handle the guns. And obviously the great majority of the UK’s armed service personnel will come into contact with live weapons, with up to five thousand of them deployed in London for big ceremonial occasions.

The big unknown is illegal guns, including those brought in by visitors. Estimates vary wildly, although one diligent researcher has amassed enough indirect evidence to suggest there are vastly more unlawful guns in circulation than one might suppose from the 400 or so seizures each years recorded by the Metropolitan Police. On the other hand, he dismisses the idea of vast caches of illegal firearms as ‘wild projections built from flimsy starting points’. He also reminds us that many illicit guns are souvenir military weapons which have been unserviceable for years. Let us exclude them (since unused guns they won’t be spreading discharge residue) and make a conservative guess that at very most there are around 60,000 illegal guns in circulation in the capital at any one time, each capable of leaving a gunpowder trace. I stress 60,000 is probably an overestimate – one illicit and operable gun for every hundred adult Londoners at the time of the murder – but an overestimate favours Mr George’s defence. Bear in mind that the more people have guns the more likely it is that a random person on whom a particle is found will be a gun owner, so the less probative the particle found in Barry George’s coat.

On this tally, at the time of Jill’s murder there might conceivably have been as many as 110,000 people in London with access to the types of gun which dispel the sort of debris found on Mr George’s coat. That’s one adult in every 55.

Next we need to know what proportion of residue particles find their way onto people’s clothing, and how the proportion varies (if at all) between shooters and non-shooters. Again we have no precise data, and findings appear to vary between forensic science labs. Even so we can make a reasonable supposition. A survey involving swab tests from 1,300 people and tested across 32 European laboratories found that the mean probability of finding at least one particle was 0.4% for the general population – a figure in line with previous studies – but it was 42.3% for subjects who had been in possession of a firearm. In other words, residue was 100 times more likely to be found on a shooter than on a non-shooter. These were tests on people’s hands, not clothing, but it is not unreasonable to suppose the general finding may hold true for both.

Of course we must allow for the original defence case: the possibility of contamination during seizure of the coat or at police stations. There are two circumstances in which that would be likely. The first is if incriminating evidence was planted by corrupt officers, although as we have seen that was dismissed by scientists as both improbable and impractical with a single microscopic particle. The second is if the coat was handled by an officer who had just fired a gun, where the possibility of transfer is high. But the coat was collected and preserved by scene of crime officers wearing protective suits and gloves, and although it was taken out of its sterile bag to be photographed, it is unlikely the photographer had recently handled or discharged a weapon.

So what about the risk if there was no corruption and no immediately preceding gunfire either from an arresting officer or one of the exhibits team? No experiments have been conducted in the UK to test for the unintended transfer of single particles to clothes seized by police. But a good place to look for scattered firearms discharge residue is in the United States where there are more guns than people, where one in five adults carries a gun routinely, and where 1,000 suspects a year are shot dead by the police. (Even so, contrary to popular assumption, only a quarter of police officers will ever get to fire their guns while out on duty through their entire careers.) Several experiments have shown that the potential exists for the transfer from police or police property, but scientists concluded that in practice it is ‘not an overwhelming, or major concern’. For example, in Pittsburgh, with a homicide rate by firearms more than forty times higher than that of London, researchers conducted a detailed examination of tables and chairs from interview rooms, and from cells at four police stations, and were surprised at how little transfer actually took place, concluding: the risk ‘does appear to be minor at least in this police district.’ Similar results emerge from other countries where murder rates are lower but where police are routinely armed, such as Australia, or from Switzerland where scientists checked seats on police vehicles for the transfer potential of firearms particles. Moreover, where particles have been discovered they were on open surfaces; they had not found their way inside a pocket.

Obviously these ‘minor’ risks are even more minor in London where fewer than one in eight of sworn officers is armed, and where laws on gun ownership are among the strictest in the world. Indeed, even the lawyer who represented Mr George at the appeal – and who is still convinced of his client’s innocence – conceded in his memoirs (Under the Wig, chapter 21) that ‘the prosecution were able to prove there couldn’t have been any such cross-contamination’.

Nonetheless, a credible risk exists for unintended transfer and, in the absence of precise experiments of particle transmission between Metropolitan Police, it would be safer to overstate the possibility than rule it out altogether. So where should we drop the most credible pin of conjecture for a single discharge particle? We could start with caution, assuming that one in every hundred items of evidence seized by police bears undiscovered traces of firearms residue, and that one in every ten of these results in cross-contamination. (This is theoretically far-fetched, and it would render the collection of forensic exhibits almost futile, since if cross-contamination was so common with firearm residue it would also presumably be so for many other physical contaminants; and if exhibits really were so hopelessly exposed it would mean that any forensic connection with any crime would always be deniable.) Nevertheless, again bending over to be fair to Barry George, let us accept this speculation and propose that one item in every thousand gets contaminated with at least one firearm discharge particle.

That would mean we need to build in a 0.1% probability of accidental police contamination.

Set against that is the fact that the particle was in the right-hand pocket. This is of marginal significance since roughly nine out of every ten people are right-handed. Nonetheless, since a right-handed gunman would normally conceal a weapon in his right-hand pocket, it is perhaps worth noting in passing in the case of Barry George.

The photo recovered by police in which he is holding a gun shows that he professionally cradled the weapon with his left hand, but his right hand is clamped round the pistol grip and his right forefinger is on the trigger. Even so, again let’s favour the defence and set this to one side.

If we put all of this together, and if all these estimates are approximately right – that is assuming one in every 55 citizens might have occasional access to a gun, that people who handle guns are 100 times more likely to have firearms residue than those who don’t, and allowing for a one in 1,000 chance of cross-contamination after the coat was seized by the police – then, according to Bayes theorem, the evidence of Barry George handling a gun is far from neutral. In fact the probability rises from close to zero to 71%. That means there is almost three times more likelihood that Barry George handled a gun than any other explanation, which in turn means that the evidence was correct at the first trial and should not have been dismissed by the Court of Appeal. Instead of the clue being ignored it should have been considered as yet another brick in the wall of evidence against Barry George.

Perhaps what is so extraordinary about all this – the forensic scientists’ testimony to the appeal court, as well as the appeal court’s finding, and the second trial judge’s acquiescence in excluding the forensic evidence – is that they fell into a trap that is well-known in criminal trials. Court rules say explicitly, ‘it is vital to ensure that an expert does not give evidence in relation to matters outside of their expertise.’ Yet, as no less authority than a Supreme Court judge has warned, when it comes to statistical probability the rules are frequently ignored. She notes that despite the fact that, ‘humans are particularly fallible at making use of statistics to take rational decisions’, in many cases, ‘the courts take a rather impressionistic approach to what strikes the judge as helpful with little express analysis of what the data.’

Academics point out that the very adversarial nature of criminal trials tends to encourage such impressionistic thinking. As psychologist Prof David Lagnado points out – using the Barry George appeal as an example of lawyers coming to the wrong conclusion – criminal courts have only a binary option, guilty or not guilty, which promotes binary reasoning rather than logical deduction.

Statisticians themselves have been doing what they can to redress the balance, and the Royal Society has published an accessible primer to educate the courts. But as academics point out, the very adversarial nature of criminal trials tends to encourage ‘impressionistic’ thinking. As psychologist Prof David Lagnado points out – using the Barry George appeal as a classic illustration of lawyers coming to the wrong conclusion – criminal courts have only a binary option.

The doyenne of statisticians, the world’s first professor of public understanding of risk, David Spiegelhalter, has been so exercised about this case and others like it that he lectures judges and has written for the Supreme Court on how easy it is for lawyers to be fooled.

So why is the great body of well-researched and scientifically backed evidence against Barry George so widely dismissed and, by some, so vehemently denounced? Why is the narrative that Barry George is innocent so frequently talked up by commentators and why are documentary-makers scared of questioning whether the appeal system managed to put this case wrong instead of right? This bias is by no means unique to Jill Dando’s murder. And perhaps it’s unsurprising, especially among liberals in academe and media who like to be on the side of an underdog, a role that Barry George can fill with ease. It is regarded as campaigning journalism if the media claim a convicted person might be innocent; while investigators who propose a miscarriage of justice in the other direction are likely to have their programmes damned as trial by television. They are also almost certain to be sued. Indeed, anyone who seeks to change the public view of Mr George should know that for all his supposed lack of mental capacity he has proved energetically litigious.

So even I will never say Barry George is guilty (although I cannot undo the fact that I did precisely that on the steps of the Old Bailey when he was first sent down). I have long come to accept that the legend of his innocence is now almost as unassailable in the public view as it is beyond further reconsideration in criminal law.

Nonetheless, an acquittal in English law is not the same as being ‘innocent’. A verdict of not guilty simply means that the jury could not agree that the burden of proof had been met. Moreover, after his acquittal Mr George was twice denied compensation on grounds that, to quote Lord Justice Beatson and Mr Justice Irwin judgment in the High Court, “There was indeed a case upon which a reasonable jury properly directed could have convicted the claimant of murder.”

So while Barry George is unequivocally not guilty in law, it would be a falsehood to suggest there was never real evidence against him. It would be an insult to the police inquiry, to the CPS, to the Treasury Solicitor who prosecuted the case, to the first jury which convicted him – and to the multitude of coincidences that made him such a compelling suspect.

Postscript

I confess I was always of the opinion that Jill was killed by someone much like Barry George, and I went on Newsnight the evening of of her death to say so. By the time of her murder I had accumulated fifteen years experience of working with police and had come to be wary of investigators’ theories. But I was also a psychologist by background, had visited prisons, and was acutely aware of the link between crime problems with mental health. It was not easy to go on presenting Crimewatch without Jill, yet even more frustrating to be appealing for information on her murder when I thought the detectives were looking doggedly in the wrong direction. Hamish Campbell, the SIO, must have found it equally frustrating that I was always downplaying the Met’s theories about criminal gangs.

Having said that I have always had less reason to be emotionally biased about this case than some have supposed. Nor do I harbour ill-feelings towards Barry George. He is plainly psychologically damaged. Indeed, I am greatly impressed by the loyalty of his sister Michelle Diskin who campaigned so hard to clear his name. I am profoundly proud of Jill’s legacy, the world-class Jill Dando Institute which pioneered the new discipline of Crime Science at UCL and finds new ways of helping police and preventing crime. It has spawned similar work in universities around the globe. And I am pleased to support the Jill Dando News initiative which encourages children to consider following Jill into journalism. But while Jill was a marvellous, hugely likeable and generously professional colleague of mine we were never especially close friends. My interest is less about this single case of murder, awful and horribly close to home as it was, than about how a narrative can take hold which swamps so much of the evidence.