As the New York Times reports today, eyewitnesses to the August 9 shooting of Michael Brown by Ferguson, Missouri, police officer Darren Wilson don’t agree on much. While there’s one sequence on which something like a consensus has developed — after Wilson stopped Brown and his friend Dorian Johnson for walking in the street, an altercation ensued in which, as the Times puts it, “Officer Wilson’s firearm went off inside the car, Mr. Brown ran away, the officer got out of his car and began firing toward Mr. Brown, and then Mr. Brown stopped, turned around and faced the officer” — after that, things get murky, with different witnesses giving different accounts.
This shouldn’t surprise anyone: Psychologists have known for a long time have known that eyewitness testimony is very — and sometimes dangerously — unreliable, and they may not offer much clarity for those hoping for a clear final answer to what happened in Ferguson two weekends ago.
“Surveys show that large proportions of people, at least in the United States, think that human memory works like a video tape or a DVD,” said Scott Lilienfeld, an Emory University professor who co-authored a Scientific American article about the shortcomings of eyewitness testimony. “And we know of decades of psychological research that human memory, including eyewitness memory, doesn’t work that way.”
Rather, it's much more flawed and easily malleable than a video recording. As Lilienfeld points out in the article, this misunderstanding can have dire consequences. Referencing an Innocence Project report, he and his co-author write that "73 percent of the 239 convictions overturned through DNA testing [since it was introduced] were based on eyewitness testimony. One third of these overturned cases rested on the testimony of two or more mistaken eyewitnesses."
Memories of events “can be very affected by people’s biases, preconceptions, expectations, and the like,” Lilienfeld said. He explained that neuroscientists believe memory is “really reconstruction rather than a reproduction” — we’re not drawing on some untainted recollection of the event, but “rebuilding” it, in a sense, every time we summon the memory. All sorts of impurities can seep in each time this reconstruction occurs.
For example, eyewitnesses to an event are just as susceptible to media influence as the rest of us. “Once something gets out in the media, it’s really hard to un-ring that bell,” said Lilienfeld. “People have heard what other people have said, people have heard what other people have reported … and then their memory may be tainted, perhaps forever tainted by those recollections.”
That’s part of the well-documented phenomenon of “interference” in memory recall — that is, one memory affecting another one. If an eyewitness sees an event and then discusses it with another eyewitness, that subsequent discussion can retroactively affect how both witnesses remember the event itself.
Then there are more standard biases — does the event in question fit a well-rehearsed script in the witness’s mind? “If there’s unintentional desire to remember things in a particular way,” said Lilienfeld, that can affect recall. So if the witness, say, expects police to treat black people violently, or on the other hand has racist views about black people, they may remember a given event in a manner that better fits the ingrained cognitive script than the actual events that took place.
Finally, the fact that a gun was involved in Brown’s death likely only worsened eyewitnesses’ ability to accurately recall what happened. Lilienfeld explained that there’s a phenomenon called “weapons focus” in which people (understandably) fixate on a weapon when one is present. “That can impair people’s accurate reporting of the event,” he explained. “That’s a well-known problem in eyewitness testimony — they’re focusing so much on the weapon that they might be missing what the other person is doing.”
Many of these problems wouldn’t be so nettlesome if people better gauged the accuracy of eyewitness accounts — both their own and others — rather than wrongly assumed them to be generally infallible. The problem is that on the one hand, individuals appear to be swayed by confident eyewitness accounts, and on the other, individuals aren’t very good at properly gauging their own reliability as an eyewitness.
Lilienfeld said that researchers have been trying to figure out for a while whether there’s a correlation between the confidence of an eyewitness and how accurate their testimony actually is. Some researchers think there’s none, while others, including Lilienfeld, think there’s a slight correlation (that is, confident eyewitnesses are, all things being equal, just a bit more likely to be accurate eyewitnesses), but overall the correlation hovers not too far from zero. So there’s little real debate over whether members of homo sapiens’ general tendency to overrate their abilities extends to eyewitness testimony: It does.
In what cases, then, is eyewitness testimony reliable? “I wish there were a simple answer to this,” said Lilienfeld. “I’d say when there’s consistency around eyewitnesses,” particularly when they haven’t communicated in a way that could lead to them unintentionally (or not) adopting the same storyline. A 2003 literature review (PDF) by Gary Wells and Elizabeth Olson, then both Iowa State University psychologists, offers some other hints, even though it focuses mostly on eyewitnesses’ ability to correctly pick perpetrators out of a lineup rather than visually parse complicated sequences like Brown’s shooting.
All things being equal, according to the review, young children and the elderly are less reliable eyewitnesses than younger adults. People are better at successfully identifying a suspect when they are the same race or ethnicity as that suspect. Anxious people may be less likely to make mistakes. People are more likely to successfully identify suspects with very attractive or unattractive faces as compared to average ones. Even simple disguises worn by suspects seriously degrade the accuracy eyewitness accounts. Telling a witness they did a good job identifying a suspect may retroactively make them more confident in their testimony.
These are all just observations stripped of context, however. Every event is different, and it’s very hard to know which biases or other psychological quirks that can influence eyewitness accounts are at work in a given incident. Ideally, one could quickly sequester a witness from influences that might affect the accuracy of their story, whether other witnesses or media coverage of the event, and then get their story down before the passage of time increases the odds of the memory being altered — but that simply isn’t how things work in the real world.
The one clear takeaway from the voluminous literature on eyewitness accounts is that we need to be very, very careful about relying on them, especially in cases like Ferguson, in which multiple eyewitnesses who haven’t communicated disagree about what they saw.
Asked whether all this research makes him more supportive of using cameras to record interactions between police and the public, Lilienfeld gave his shortest answer of the interview: “Yes.” Unfortunately, there was no camera to record Brown’s death, leaving us with a bunch of conflicting stories and no clear way to untangle signal from cognitive noise.