Courts place great weight on video identification evidence but, as Kate Worsley reports, research suggests that such confidence may well be misplaced

I saw a woman mugged on her way to work last week. The boy who did it passed by me as he ran away and I haven't been able to forget his face since. He actually reminded me of someone I used to work with. So, if he's ever caught I may be able to help.

Well no, actually. It's highly unlikely that I remembered the face of a complete stranger at all accurately, according to Vicky Bruce, professor of psychology at Stirling. And if I ever attempted to identify him from a police photo or video-still I'd be even more at sea.

"These days the courts will say, here's the person accused of a crime, here's his picture on video, and there you go, m'lud, there's your man." Bruce says this flippantly, but her research is intensely serious. She knows that the human cost of relying heavily on video evidence to identify suspects in legal trials is too high. Because of what she knows about the psychology of face recognition Bruce believes that mistakes are often made in court.

³ÉÈËVRÊÓƵ

There is a crucial and largely unexplored difference between the way we recall familiar and unfamiliar faces, says Bruce. It is particularly pronounced when we try to match our memory of an unfamiliar face with an image from a photograph or video still. "It's remarkably difficult to compare a picture with a person when both are unfamiliar, or to match two pictures, (one, perhaps) from police files. This causes problems for witnesses," she says.

Over the past decade the growing presence of closed-circuit television in shops, banks and public spaces has become accepted in the courts as a powerful identification tool to back up, or even replace, eyewitness accounts of incidents. But such confidence may be misplaced. "It's very easy to identify a familiar face, the brain builds a representation of the face from all its encounters with it, a sort of composite that aids future recognition. But the ease with which we can do this fools us into thinking we can do it with strangers too."

³ÉÈËVRÊÓƵ

Bruce addressed the problem of eyewitness memory and the difficulties of identifying people caught on video in a speech to the British Psychological Society yesterday. The validity of eyewitness testimony was first investigated in the 1970s, when Lord Devlin's 1976 inquiry drew attention to the fallibility of witnesses in cases such as that of Laszlo Virag, wrongly convicted of an armed offence because he bore a passing resemblance to the real gunman. Psychologists investigating the problem found that changes in viewpoint, expression, lighting - and of course those perennial criminal favourites hairstyle and glasses - made recollections of a face notoriously unreliable.

CCTV images are often blurred, poorly lit and taken from a sharp and unfamiliar angle - we all know it's hard to recognise someone by their bald patch. But preliminary data from a research project Bruce is conducting with Mike Burton of Glasgow University indicates that even when a person's image is captured full-on and on high-quality video tape, that person is very hard to identify.

In one experiment participants were asked to match photographs of ten men with an image grabbed from a high quality video shot at the same time but with different lighting. They matched full-face images 79 per cent of the time but the success rate dropped to 70 per cent when the head angle differed by just 30 degrees. This variation seems to suggest that we can recognise people by being able to see the shape of their hairline.

But it is the fact that the lighting was different that Bruce thinks is crucial, The title of her BPS speech, "Fleeting Images of Shade", gives a clue as to why this might be. Because we have not had time to build up a composite picture of a stranger, we have to rely on the brain's ability to initially perceive objects only in terms of light and shade, particularly shade, she explains. "So when only the lighting has changed we find it difficult to verify whether it is the same person, even when given limitless time." A computer image works the same way, with pixels representing light and dark areas, and so gives us no more information to work on.

³ÉÈËVRÊÓƵ

The findings have surprised Bruce herself, who has worked in the field of face perception for 25 years. "Lighting is a complicated thing that has been neglected. It is really easy for it to change how we see images and facial features. Both human vision and computer vision find light and shade hard to deal with."

Bruce is investigating other areas too, such as how movement affects image perception. When the 18-month research project ends she plans to take her findings to the ³ÉÈËVRÊÓƵ Office in order to establish better standards for the use of CCTV evidence in the courts. She also hopes to develop other ways to enhance human perception, such as computer algorithms.