By Emma Young
A witness to a crime has to describe the offender’s face in as much detail as they can before they work with a police expert to create a visual likeness — a “facial composite”, sometimes called a photo-fit, or e-fit. But the way this is typically handled in police stations could be reducing the accuracy of these images, according to a published in the Journal of Experimental Psychology: Applied.
There have been concerns that the process of describing facial features might create a so-called “verbal overshadowing” that interferes with the visual memories of the offender. Recent work had suggested that waiting half an hour before starting on the composite should allow this predicted over-shadowing to fade away, and so make for a better composite. However, the new research, led by Charity Brown at the University of Leeds, has found that in more real-world situations, a delay actually makes things worse.
In the first of three studies, the team split 96 participants who reported being totally unfamiliar with the UK TV soap EastEnders into eight different experimental conditions. Each participant watched a brief video clip that featured two of a total of 12 actors from the programme. They were asked to focus either on the content of the conversation (to simulate the kind of incidental facial information that an eyewitness might gather when they don’t realise that they are actually witnessing a crime) or to focus on the faces.
Either four to six hours or two days later, these participants were asked to verbally describe one of these faces — to report on the shape and colour of the eyes, nose, forehead, and so on. Then there was either a delay of half an hour, or no delay, before each participant worked with Laura Nelson at Lancashire Constabulary Headquarters on a facial composite. (Nelson had no idea which particular actors each participant had seen).
Another group who were familiar with EastEnders were given the composite images to identify. A separate group also rated how similar each composite was to photographs of the actors.
The results were clear. There was only one experimental condition that produced worse likenesses than the others: when the descriptions were given two days after viewing the clip and there was a 30 minute delay before work started on the composite. Unfortunately, this is also the closest to real-world conditions in police investigations, the team notes. Facial descriptions are often not collected until a few days (at least) after a crime, and it’s not uncommon for witnesses to be offered a break before starting on the composite.
Two days after seeing a face, witnesses’ facial descriptions tend to be less full but more accurate than descriptions given four to six hours afterwards (perhaps because they’ve had more time to process which facial features were most striking, the team suggests). But a half-hour delay immediately after giving the verbal descriptions seems to then impair their access to details of their recalled descriptions that would otherwise contribute to a better facial likeness, the team writes.
This effect held whether Nelson used the “holistic” facial likeness approach, common in the UK, in which an initial face is tweaked until it fits the witness’s memory of the offender, or the system in which a facial composite is built up from selected constituent features. (Whichever method is used, verbal descriptions are always gathered first.)
“The results have real-world but counterintuitive implications for witnesses who construct a face 1 to 2 days after a crime,” the researchers conclude. “After having recalled a face to a practitioner, an appreciable delay (here, 30 min) should be avoided before starting face construction”.