A question of identity

The conviction of Barry George for the murder of the TV presenter Jill Dando was one of the biggest stories of last year. But did he really do it? On the eve of George's appeal, Brian Cathcart shows how flimsy the evidence against him was

When the judge summed up in the trial of Barry George last year, one instruction he gave the jury was that if they were not sure – and he stressed that word "sure" – about the identification evidence, then they must return a verdict of not guilty. In other words, no matter what they felt about anything else they had heard in court, if they were not satisfied beyond reasonable doubt that George was at the murder scene that morning, they had to acquit him.

Given such a clear instruction, it might seem fair to conclude that this part of the evidence at least must have been sound, that we can be confident that the man now serving life imprisonment for murdering Jill Dando was there or thereabouts, roughly at the time of the deed. Unhappily for Barry George and – unless something changes when he appeals against his conviction next week – unhappily for defendants in other trials yet to come, that is very far from being the case.

For although police traced dozens of people who used the west London street where Dando died on that morning, only one of them ever categorically identified George. And that witness, a local woman called Susan Mayes, said that she had seen him four and a half hours before the murder was committed. The sighting was at 7am, while the murder occurred at about 11.30am. And the circumstances are as worrying as the timing. The man that Ms Mayes saw across the road was wiping the windscreen of a double-parked, maroon-coloured car and wore a black suit with a white shirt open at the neck – all details difficult to reconcile with Barry George in the role of murderer.

Since a double-parked car in Gowan Avenue would inevitably block the way, we can be sure it would have attracted attention if it had stayed there long, and yet no other witness saw it. The likelihood, therefore, is that it drove off soon after this sighting. By the same token, none of the other people out and about in the half hour after 7am saw this man. The odds, therefore, surely favour him getting into the car (which he had been wiping) and driving away. Yet Barry George can't drive and has never owned a car, maroon or otherwise.

As for Ms Mayes's description of the man's clothing, that raises an even more fundamental problem, for we know what the killer was wearing and it was different. Two other witnesses – neighbours of Jill Dando – actually saw the gunman moments after he pulled the trigger, as he left her gateway and hurried up the street. Both described a man wearing not a suit but a coat, probably something like a Barbour that came down to thigh level. And neither of those two neighbours, who both saw the killer's face, picked George out when they saw him in identification parades. Wrong clothes, wrong behaviour and very much the wrong time; it is no criticism of Ms Mayes, who was testifying in good faith, to say that her evidence is not quite a smoking gun. George, incidentally, denied visiting Gowan Avenue that day and insisted he was at home at 7am.

If this was all the jury had to go on, it is unthinkable that they could have concluded George was near the murder scene – but it was not all. The prosecution presented other identification evidence which was – to say the least – dubious, and before looking at it we should first glance at the troubled past of identification evidence in British courts.

Common sense tells us that when one person sees another committing a crime, and later recognises that person and identifies him or her to police, that is important, even compelling evidence. And as so often with common sense, it doesn't tell half the story. The case of Adolf Beck, a Norwegian who was living in London at the end of the 19th century, is pivotal in British law. A businessman down on his luck, he found himself charged with a serious crime. Over a period of more than a year, a man had been approaching relatively wealthy women in London, gaining their confidence and then, on some pretext, persuading them to part with items of jewellery. Typically, he would say that he knew someone who could remount or repair the jewels. He would then disappear.

Beck was recognised as this man and police brought him in. He was subsequently identified, face to face, by no fewer than 15 of the victims and their close friends, most of them people who had met the thief more than once and for periods of above an hour, and most of them (for what it is worth) educated and informed. Beck was convicted and jailed – but he was the wrong man. The true thief was subsequently identified beyond doubt, and the unfortunate Norwegian eventually released. The case led ultimately to the creation of the Court of Appeal, but it also prompted a review of the problems of identification evidence – a review which has been continuing, in spasms, ever since.

People make mistakes and it is a plain fact, established by thousands of cases before the courts, that the recognition of faces is one area where they are especially fallible. That is problem enough for the law, but it is compounded by a further difficulty, because on those "common-sense" grounds juries tend not to make sufficient allowance for these mistakes. Identification evidence, in short, often carries more weight in court – and sometimes far more weight – than it should, given its unhappy record.

The response of the law, in line with its duty to protect the innocent, has been to tighten procedures, lay down guidelines and issue warnings. Long ago the emotive practice of dock identifications – "That's the man, your lordship!" – was banned. Identification parades rather than simple, one-on-one confrontations became the norm, and parade procedures have been progressively refined. Judges also started vetting such evidence before it was presented to juries in an attempt to ensure that it met certain standards. These standards, currently set out in what are called the Turnbull guidelines, are exacting, though still not so exacting that they sift out all the mistakes. They state that an identification is good if it has been "made after a long period of observation, or in satisfactory conditions by a relative, a neighbour, a close friend, a workmate and the like". An identification is poor if it "depends solely on a fleeting glance or on a longer observation made in difficult conditions". And when an identification is poor there is no room for manoeuvre: the judge must throw it out.

Before we go further, pause for a moment and apply the test to the Susan Mayes identification. She certainly wasn't a neighbour or close friend of the person she identified. By her own account she was holding up an umbrella at the time against the drizzle and was walking as she looked, conditions that might be described as "difficult".

And was it a "fleeting glance"? At first it was said in court that she had the man in her sights for about a minute, but then came this exchange:

Q: "How many times did you look at his face?"

A: "I believe I looked at his face about three times."

Q: "And over what period of time, roughly?"

A: "Five or six seconds."

Q: "Every time, or in total?"

A: "No, I would say in total."

The judge, Sir William Gage, was sufficiently surprised by this reply to ask Ms Mayes to confirm it, which she did. Five or six seconds from under an umbrella in the rain sounds awfully like a "fleeting glance", and yet her testimony stood.

The other identification evidence against Barry George arose from the parades that took place more than a year after the murder. After his arrest in May 2000, George agreed to take part in a conventional line-up before all five witnesses, and that passed off uneventfully, with no one picking him out. Richard Hughes, one of the people to see the killer leave the scene, was among those to draw a blank. It should be said – and the jury was told – that by now George had a beard, which he had not had the previous spring, and which he appeared to have grown in the time since he had fallen under police suspicion. It was not, however, the first time that he had grown a beard. That summer of 2000 the police wanted to hold further parades, but George exercised his right to refuse. Instead, police constructed a video parade, in which a tape showing head-and-shoulders footage of George was run in sequence with similar tapes of seven other men, all also with beards. This video was shown to a succession of witnesses in the presence of a defence solicitor.

One was Teresa "Terry" Normanton, another resident of Gowan Avenue, who informed police a year after the murder that she had seen a man in the street at 9.50am on the day Dando was shot. Shown the video parade, she lingered on George's face and remarked: "I'm sure, but I'm not quite sure. I don't remember a moustache." Charlotte de Rosnay was another resident of the street who saw a man, this time at about 9.30am. Shown the tape, she saw two faces she wanted to see again, numbers two and eight (George was at two). Asked whether she could make a positive identification, she replied: "I don't think so." Charlotte's mother-in-law, Stella, had been staying with her at the time of the murder, and she also saw the man at 9.30am. She, too, dithered between numbers two and eight and declared: "My gut feeling is two. It's frustrating. It's the colour of the skin, too." And then she also declined to make a positive identification.

Three women and three near-misses, you might say. Imagine being in their position and imagine (though Charlotte de Rosnay did not even get this far) that you have a hunch that it was a particular man. This is a case of murder: you would want to be sure. You would want to be very sure. In the interests of justice, you should be very sure. None of these three was very sure.

Here, however, the prosecution adopted an extraordinary approach. Rather than accepting that these were not identifications, it decided to present all three as exactly that. The defence, led by Michael Mansfield QC, responded in horror, asking the judge with force to rule them out before they got to the jury. Such evidence, Mansfield argued, would drive a coach and horses through a century of safeguards and precautions that had been designed to protect the innocent. Not only did the evidence of both de Rosnays depend upon a classic "fleeting glance" – they watched the man rush past the house from an upstairs window – but, crucially, even the strongest of these witnesses, Normanton, had been able to make only what Mansfield described as a "qualified identification".

In reply, the prosecution argued, in essence, that the three non-identifications stood in support of the Mayes identification because there was an "underlying unity" in the descriptions given by all four women. Where one had felt confident enough to pick out George, the others had almost picked him out and after all, that was probably understandable, since a year had passed and there was the business of the beard. The evidence was admitted.

To say the least, this was surprising. Mansfield's argument seems to have failed on the point that neither the law nor the guidelines expressly forbade this use of non-identifications (although that was surely because their authors never conceived that such a ban would be necessary). Moreover, the argument that three witnesses were appearing in support of a fourth might make sense to lawyers and to a judge, but it was unlikely to linger in the minds of jurors. They would see four women in the witness box, and they would be unlikely to keep three of them in a separate mental compartment where their words could be accorded lower value. Juries don't think like that. As for the business of a year having passed since the crime, that was hardly Barry George's fault and yet somehow, for that reason, the prosecution wanted the jury and the court to apply less rigour to the identifications. This was absurd. If any allowances were to be made for confusion and doubt among witnesses, for safety's sake they ought to have been made in favour of the defence.

What happened next compounded all that had gone before. By the time the trial came round in April 2001, Normanton and the de Rosnays were well aware that the man they had dithered over in the line-up nine months before was the man believed by police and prosecution to be the murderer of Jill Dando. In principle this knowledge made no difference, for their evidence was to be restricted to what they had seen in Gowan Avenue on the day of the murder, and what had transpired at the line-up. It was agreed among the lawyers that nothing they heard or thought after the line-up was to be aired.

In practice – and this is hardly surprising – the words used by these women in the witness box clearly reflected the influence of their later knowledge. Each in turn hardened up her evidence against George. Charlotte de Rosnay, when she had not known which of the faces in the line-up was the suspect, had been unable to decide between numbers two and eight. In court, however, she announced boldly that number two – the defendant – had been the "predominant contender" in her mind. The words were out before judge or counsel could stop her. In the same vein, the court heard that Stella de Rosnay, who had not felt confident enough to make an identification, was now "95 per cent sure" that number two was the man.

What was happening here, in effect, was that identification by line-up had turned into something resembling the old-fashioned – and discredited – process of identification by confrontation, or face-to-face. These witnesses, having previously been in doubt, now knew number two was the defendant and were saying: "I knew it all along." This escape backwards to the days before Adolf Beck was completed by a further incident in court, which the judge was unable to prevent and the jury heard loud and clear. Susan Mayes was being led by the prosecution through her account of what happened on the morning of the murder as she set off down Gowan Avenue. Did she see anybody? she was asked. "I saw the defendant," she replied. Counsel was no doubt blameless, but this was a dock identification perfectly equivalent to the "That's the man!" of old. It is not supposed to happen.

One last point: the prosecution's argument rested on the underlying unity of the descriptions, and yet no such unity existed. The two de Rosnays, for example, described suits of different colours, while Terry Normanton was adamant that the man she saw looked just like an E-fit picture released by police just after the murder – an E-fit later accepted as not being of the killer.

One strain of unity did emerge, however. A remarkable number of witnesses, though by no means all, spoke of the man they saw in or around Gowan Avenue as having a "Mediterranean" appearance. In some cases this meant that he was dark, in others that he seemed tanned or had olive skin. Did this match Barry George in the spring of 1999? No evidence was produced to suggest it did, and the man in court in 2001 was certainly more pasty-faced than tanned, though this was presumably inevitable after a year in custody. Moreover, of the two witnesses who saw the gunman, neither used the word "Mediterranean" or any similar word in their descriptions.

There were other grave flaws in the case against Barry George, in forensic evidence as well as in the absence of motive and of a weapon, and in the existence of alibi witnesses. But it is the identification that must arouse greatest concern. To begin with, the evidence of three and possibly four of those key witnesses was at least borderline "fleeting glance", as described in the Turnbull guidelines. The latest of these sightings took place an hour and 40 minutes before the murder, an interval in which many other witnesses traversed the scene without seeing George. None saw the man in the clothes the murderer wore. Most of all, the fact that the three non-identifications were allowed to go before the jury, though it may in some way be within the letter of the law, surely stands in flagrant contradiction of its spirit.

Do three "I'm not sure"s now equal one "That's the man"? If so, what are the other units in the currency? Perhaps two "I think so"s add up to one "I'm sure". And maybe four or five times "I dunno, but I suppose it could be him or him" will be enough to convict – especially if the witnesses have made up their mind that it was the defendant by the time they get to court.

We have known for a century that identification evidence carries far more weight with juries than it should, and that it needs to be treated with the greatest care. For reasons that are difficult to understand, however, in the case of Barry George all sorts of risks were taken and dangers invited. Assuming that the jurors took note of the judge's summing up, we have to accept that they were "sure" that Mayes, Normanton and the de Rosnays saw Barry George hanging around in Gowan Avenue that morning, waiting for Dando to appear. Even on what was said in court that seems a strange conclusion, but in any case what was said in court included far, far too much that was emotive and unreliable.