We are able to find everything in our memory, which is like a dispensary or chemical laboratory in which chance steers our hand sometimes to a soothing drug and sometimes to a dangerous poison.—Marcel Proust
Describe the last person who served you a coffee. What if I helped refresh your memory? Showed you some photos of local baristas? Pulled together a helpful lineup? Cheered exuberantly when you picked the "right" one? Now imagine that instead of identifying the person who made your venti latte last week, we had just worked together to nail a robber or a rapist. Imagine how good we would feel. Now imagine what would happen if we were wrong.
Last month, a Texas judge cleared Timothy Cole of the aggravated sexual assault conviction that sent him to prison in 1986. Although his victim positively identified him three times—twice in police lineups and again at trial—Cole was ultimately exonerated by DNA testing. The real rapist, Jerry Wayne Johnson, had been confessing to the crime since 1995. Unfortunately, Cole died in prison in 1999, long before his name was cleared.
Our eyes deceive us. Social scientists have insisted for decades that our eyewitness identification process is unreliable at best and can be the cause of grievous injustice. A study published last month by Gary Wells and Deah Quinlivan in Law and Human Behavior, the journal of the American Psychology-Law Society, reveals just how often those injustices occur: Of the more than 230 people in the United States who were wrongfully convicted and later exonerated by DNA evidence, approximately 77 percent involved cases of mistaken eyewitness identification, more than any other single factor.
Wells has been studying mistaken identifications for decades, and his objection to the eyewitness identification system is not that people make mistakes. In an interview he explains that eyewitness evidence is important but should be treated—like blood, fingerprints, and fiber evidence—as trace evidence, subject to contamination, deterioration, and corruption. Our current criminal justice system—blessed by a 30-year-old Supreme Court precedent—allows juries to hear eyewitness identification evidence shaped by suggestive police procedures. In a 1977 case, Manson v. Brathwaite, the Supreme Court held that evidence that was a product of suggestive identification procedures need not be excluded if the identification was nevertheless deemed "reliable." Five criteria for determining whether that identification could be reliable were laid out—including how much opportunity the witness had to view the perpetrator and how certain she was of her identification. In the intervening years, social scientists have called into question much of the science underlying these five factors. Today we know, for instance, that you can have a good long look, be certain you have the right guy, and also be wrong. But Manson is still considered good law.
Jennifer Thompson was 22 the night she was raped in 1984. Throughout the ordeal, she scrupulously studied her attacker, determined to memorize every detail of his face and voice so that, if she survived, she could help the police catch him. Thompson soon identified Ronald Cotton in a photo lineup. When she—after some hesitation—again picked Cotton out of a physical lineup a few days later, a detective told her she'd picked the same person in the photo lineup. As Thompson told Leslie Stahl on CBS last weekend, that assurance led her to think: "Bingo. I did it right. I did it right."
But in this case Thompson got it wrong, although Cotton served 10 and a half years before DNA evidence exonerated him and decisively implicated another man, Bobby Poole. The curious part of the story is that despite Thompson's determination to memorize every detail, when she first saw Bobby Poole in court she was certain she had never seen him before. Indeed, according to Wells and Quinlivan, "Even after DNA had exonerated Cotton and Thompson herself had accepted the fact that Poole was her attacker, she had no memory of Poole's face and, when thinking back to the attack she says, 'I still see Ronald Cotton.' "
How did our eyewitness identification system manage to paint a detailed picture of the wrong face in Jennifer Thompson's mind while somehow completely erasing the right one? Wells and Quinlivan's paper suggests a host of tricks the mind can play, ranging from incorporating innocent "feedback" from police investigators, to increasing certainty in one's shaky memories that become reinforced over time.
Add to that Thompson's determination to regain control over her life, and her need to believe that the justice system was just, and it would have been doubly hard for her to look at a police lineup that, as it happened, did not include an image of the real rapist and walk away. To hear Thompson and other victims tell it, being part of a system that identified and ultimately convicted the wrong man became another form of victimization, and for that reason alone the system needs to be reformed.
The problems with the eyewitness identification system cannot be laid at the feet of crime victims any more than they can be blamed on police investigators. Wells' argument for reforming our eyewitness identification system is that the incentive for the police to subtly nudge our memories goes not only uncorrected by the justice system, but sometimes is rewarded by it. Wells wants the Supreme Court to revisit the scientific assumptions underpinning Manson v. Brathwaite, which allows such identifications to come into a courtroom as long as the identification is "reliable."
Whether or not the John Roberts court wishes to take up the issue of innocent prisoners—there is, for instance, a case now percolating through the New Jersey courts testing the scientific premises of Manson—a few states and cities have used innocent exoneration scandals to rethink their eyewitness identification practices in ways that would begin to restore the credibility of such evidence. Proposed changes include showing victims photos sequentially, explaining to the victim that the perpetrator may not be included in the lineup, and ensuring that whoever conducts the lineup has no knowledge of which person is the actual suspect.
This is not an issue that tracks the usual pro-prosecution, pro-defense divide. Mostly, police departments don't change their eyewitness identification procedures simply because there is no big loud constituency demanding that guys in lineups be treated more fairly. But some of the most zealous reformers of the current eyewitness identification process are lifelong conservatives who recognize that the credibility of the whole justice system is on the line each time an innocent man goes to jail. That's because when that happens, a guilty man often walks free.
A version of this article appears in this week's issue of Newsweek.