麻豆传媒

The evidence at the crime scene is pointing accusingly at you, but you were never even there. Unfortunately, your lack of alibi puts you in a sticky situation. Your hopes are pinned on the witness who saw the crime taking place.

Surely they will exonerate you? But eyewitness testimony is astonishingly unreliable. Take the case of , who was wrongfully convicted of beating and raping a 74-year-old woman in her home in Manchester, Georgia, in 1979. Her testimony saw White spend 22 years in jail, until DNA testing exonerated him, instead pointing the finger at James Edward Parham 鈥� who had been standing only a few feet away in the original police station line-up.

And White is not alone. 鈥淐lose to three-quarters of wrongful convictions have been in part due to faulty human memory,鈥� says Elizabeth Loftus, a psychologist at the University of California, Irvine.

71% of overturned convictions examined by the US聽Innocence Project involved eyewitness misidentification*

One problem is that our memories are easily contaminated by other people鈥檚 ideas and versions of events. 鈥淓ven people who have extraordinary memories are susceptible to contamination,鈥� says Loftus. This can be hard to avoid, occurring when witnesses talk to one another, see media coverage of the event they鈥檙e supposed to be remembering, or are interviewed by an investigator who unwittingly passes on their pet theory.

Then there鈥檚 familiarity. In White鈥檚 case, for instance, before the live line-up, the victim had picked White鈥檚 face from a set of photographs that didn鈥檛 contain Parham. 鈥淭hat feeling of familiarity 鈥� 鈥業鈥檝e seen that face before鈥� 鈥� is often better than our ability to record the source of that familiarity,鈥� says , a psychologist at Goldsmiths, University of London. 鈥淥nce we have seen a photograph, particularly if we have identified the person in it, we become committed to that identification.鈥�

Extreme stress can also blur people鈥檚 memory of an event, says Valentine, an effect compounded by 鈥渨eapon focus鈥�, in which the sight of a deadly weapon dominates our memory at the expense of the assailant鈥檚 identity.

To reduce these risks, the Innocence Project has drawn up a that have so far been adopted by 14 US states. These include limiting exposure of the witness to a suspect before formal identification takes place; blinding the person administering the line-up to who the real suspect is; telling the witness that the perpetrator may or may not be in the line-up, so they don鈥檛 feel pressured to choose someone; and recording their confidence in the identification, should they make one. Identifications should also be videoed where possible, something that experts say is currently standard protocol in the UK alone.

*Out of a total of 342 convictions overturned in the US since 1989, according to the US Innocence Project. Multiple factors were involved in many of these cases.

Bad science can put you behind bars 鈥� read more of our special report to stay free

This article appeared in print under the headline 鈥淥n the witness stand鈥�