filed for New Scientist, 25 March 1999
People asked to identify unfamiliar faces from high quality video images make mistakes at least one in five times, according to new studies by psychologists. Identification from surveillance cameras, on which the police and courts are increasingly dependent, is unreliable and could lead to miscarriages of justice, they say.
Vicki Bruce from the University of Stirling and Mike Burton from the University of Glasgow tested the ability of 230 Open University students to correctly match pictures of faces grabbed from video with still photographs of ten similar faces. The faces, which were all young, clean-shaven short-haired Caucasian males, were selected from a Home Office database of 200 police trainees.
To their surprise, Bruce and Burton found that even in ideal conditions - using high quality pictures, full-frontal faces and neutral expressions - only 70% of the identifications were correct. When the face grabbed from video was smiling, the proportion of accurate matches dropped to 64%. When it was shown at an angle of 30 degrees, the figure was only 61%.
In a second experiment, 60 students from the University of Stirling watched short high quality video clips of unknown faces and then tried to match them with ten photographs. Even when they were told that one of the ten photographs matched, and were allowed to rewind and pause the video as much as they wished, only 79% of the identifications were accurate.
However, other experiments showed that when faces caught on poor quality video were familiar to students, over 90% of their identifications were correct. Bruce and Burton's studies are to be published later this year in two US journals, Psychological Science and the Journal of Experimental Psychology: Applied.
"Although people are very good at identifying faces they know, they are very bad at comparing faces they don't know," says Bruce. "What we are learning is that two images of the same face can look remarkably different while images of different faces can look remarkably similar." Juries should be formally warned about the hazards of making identifications from video pictures, she suggests.
Britain leads the world on video surveillance, with over a million closed circuit television (CCTV) cameras recording peoples' movements in streets, shopping centres and offices. Their use by the police to identify suspects and by the courts to secure convictions is becoming commonplace. Yet, according to a 'Dispatches' programme on Channel Four this week (25 March), people may have been imprisoned after being wrongly identified from CCTV pictures.
Graham Davies, a psychologist and Home Office adviser from the University of Leicester, has also found high error rates in video identification experiments. This could cause miscarriages of justice, he warns. "I am very concerned about CCTV because it bestows a kind of spurious scientific glitz on identification."