The reliability of eyewitness accounts is one of the bedrock beliefs of our society. In ancient cultures – and in some modern cultures that still follow ancient laws – some crimes could only be proven by eyewitness testimonies. One of the most infamous examples was Pakistan’s Hudood Ordinances, which mandated that allegations of rape could only be proven by four eyewitness accounts; otherwise, the woman was to be punished for making false accusations. Even in our supposedly more enlightened society, eyewitness testimonies still carry great weight in criminal trials – this despite ample evidence that they are often mistaken, resulting in many wrongful imprisonments.
The willingness of juries, and people in general, to believe eyewitness testimony stems from a faulty view of the nature of memory. Human memory is not like a tape recorder or a video camera, creating a record of events and then playing them back exactly as they were first observed. Rather, human memory is basically reconstructive: in most cases, we remember only the basic outlines of an event, and if we’re called upon to retell what happened, the mind fills in the gaps with whatever details are at hand. This means that details that are fed to a person may subconsciously be incorporated into their memories and presented by that person as an accurate account of history.
One of the classic examples of this is an experiment done by the memory researcher Elizabeth Loftus. In it, Loftus interviewed the relatives of her research subjects to get three true stories of childhood events that had happened to each of them. To those three stories, she added a fourth one, which she confirmed with relatives had not actually happened: a story about the subject being lost in a mall at the age of five, crying and being comforted by an elderly woman, and finally being reunited with their parents. The subjects were each presented with these stories and asked to write down as much additional detail as they could remember about each of them. About 25% of Loftus’ subjects “remembered” the fictitious account and described it as a true story that had happened to them in childhood (source). Similar studies along this line have found that, in follow-up interviews conducted later, more and more subjects remembered the false stories over time, with some even embellishing them with details of their own that weren’t part of the original presentation.
Another study by Loftus found that subjects’ memories of events can be altered by questions that presuppose a particular set of facts. In this experiment, 150 students saw a short film of an accident involving a white sports car and then answered questions about it. One set of students was asked, “How fast was the white sports car going when it passed the barn?” (There was no barn in the film.) The other set was simply asked, “How fast was the white sports car going?” One week later, both sets of students were brought in for a follow-up interview, and both were asked, “Did you see a barn in the film?” Students who had previously been asked about the non-existent barn were far more likely (17% vs. 2%) to incorrectly believe that it had been present the first time.
Even memories of highly emotionally charged events – so-called “flashbulb” memories – are just as likely to suffer this distortion. Just like any other memory, subjects tend to forget true details or add embellishments over time. But what’s worrying is that, despite this error creep, people tend to put more confidence in their flashbulb memories and are more likely to believe they’re accurate, even when the record shows otherwise.
And finally, the most sensational example of how memory can fail us: In 1975, the Australian psychologist Donald Thomson was arrested and charged with rape, and was informed by police that the victim had positively identified him as her attacker. This was a great surprise to Thomson, because he had a seemingly invulnerable alibi: he was on live TV at the time, being interviewed in the presence of a studio audience, cameramen, and an assistant police commissioner. As it turned out, the woman who was raped had been watching that very program just before the attack. She had mixed up Thomson’s face on the TV screen with the face of her attacker. Ironically, Thomson had been on TV to discuss the unreliability of eyewitness testimony. (source; see also)
The human tendency to confabulate details and misattribute sources means that memory, especially in the long term, cannot be counted on as a reliable source of information. This doesn’t mean that eyewitness accounts should be excluded from trials and other decision-making altogether, but they should be considered with greater attention to their fallibility. When we question witnesses about the details of an event, we must avoid leading questions that could implant details into their minds. When victims of crime are asked to pick their assailants out of a lineup, those lineups should be double-blind. And testimonies should be given greater weight when two or more witnesses independently agree on the same details or when they are supported by other evidence.
Other posts in this series: