Subliminal

eventually did. She opened her eyes and made out the form of someone crouching in the darkness at her side. A moment later the man jumped on her, put a knife to her throat, and threatened to kill her if she resisted. Then, as the intruder raped her, she studied his face, focusing on being able to identify him if she survived.

Thompson eventually tricked the rapist into allowing her to turn on a light and fix him a drink, at which point she escaped, naked, out the back door. She frantically pounded on the door of the next unit. The sleeping occupants didn’t hear her, but the rapist did, and he came after her. Thompson raced across the lawn toward a brick house that had a light on. The rapist gave up and moved on to a nearby building, where he again broke in, and raped another woman. Thompson, meanwhile, was taken to Memorial Hospital, where the police obtained samples of her hair and vaginal fluid. Afterward, they took her to the station, where Thompson recounted her study of the rapist’s face for the police sketch artist.

The next day the tips started pouring in. One pointed to a man named Ronald Cotton, twenty-two, who worked at a restaurant near Thompson’s apartment. Cotton had a record. He had previously pleaded guilty to a charge of breaking and entering and, while a teenager, to sexual assault. Three days after the incident, Detective Mike Gauldin summoned Thompson to headquarters to look at six photos, which he lined up on a table. According to the police report, Thompson studied the photos for five minutes. “I can almost remember feeling like I was at an SAT test,” she said. One of the photos was a shot of Cotton. She picked him out. A few days later, Gauldin presented Thompson with a physical lineup of five men. Each man was asked to step forward, utter a line, then turn and step back. At first unsure whether the rapist was the fourth man or the fifth, Thompson eventually settled on the fifth. Cotton again. According to Thompson, when informed that this was the same man she had identified from the photo lineup, she thought to herself, “Bingo, I did it right.” In court Thompson pointed her finger at Cotton and once more identified him as her rapist. The jury reached a verdict in forty minutes, and the judge sentenced Cotton to life plus fifty years. Thompson said it was the happiest day of her life. She celebrated with champagne.

The first sign that something was amiss, other than the defendant’s denials, came after Cotton, working in the prison kitchen, encountered a man named Bobby Poole. Poole bore a resemblance to Cotton and, therefore, also to the face in the police sketch based on Thompson’s description. What’s more, Poole was in prison for the same crime, rape. Cotton confronted Poole about the Thompson case, but Poole denied any involvement. Luckily for Cotton, Bobby Poole blabbed to another inmate that he had indeed raped Thompson and the other woman. Ronald Cotton had by pure chance run into the actual rapist. As a result of the prison confession, Cotton won a new trial.

At the second trial Jennifer Thompson was asked again if she could identify her rapist. She stood fifteen feet from both Poole and Cotton and looked them over. Then she pointed at Cotton and reaffirmed that he was her rapist. Poole looked something like Cotton, but thanks to the experiences that she had had during the time after the rape—her identifying Cotton in a photo, then in a lineup, then in the courtroom —Cotton’s was the face forever burned into her memory of that night. Instead of becoming a free man, Cotton emerged from his second trial with an even harsher punishment: he got two life sentences.

Seven more years passed. What was left of the evidence from the ten-year-old crime, including a fragment of a single sperm from the perpetrator, languished on a shelf in the Burlington Police Department. Meanwhile, the new technology of DNA testing was making the news, thanks to the double-murder trial of O. J. Simpson. Cotton prodded his attorney to request that the sperm fragment be tested. Eventually, his attorney was able to get the test done. The result proved that Bobby Poole, not Ronald Cotton, had raped Jennifer Thompson.

In the Thompson case, all we know is that the victim misremembered her attacker. We’ll never know how accurately or inaccurately Thompson remembered the other details of her attack because no objective record of the crime exists. But it is difficult to imagine a witness more reliable than Jennifer Thompson. She was bright. She stayed relatively calm during the assault. She studied her attacker’s face. She focused on remembering it. She had no prior knowledge of or bias against Cotton. Yet she fingered the wrong man. That has to be disturbing, for if Jennifer Thompson was mistaken in her identification, perhaps no eyewitness can be trusted to reliably identify an unknown assailant. There’s plenty of evidence to suggest that this is the case—some of it from the very people who organize lineups like the one that resulted in Cotton’s arrest.

About seventy-five thousand police lineups take place each year, and statistics on those show that 20 to 25 percent of the time witnesses make a choice that the police know is incorrect. They can be sure of this because the witnesses have chosen one of the “known innocents” or “fillers” that the police inserted into the lineup simply to fill it out.² These are often police detectives themselves, or inmates plucked from the local jail. Such false identifications don’t get anyone in trouble, but think about the implications: the police know that a fifth to a quarter of the time a witness will identify an individual who they are certain did not commit the crime, yet when a witness fingers the person who is their suspect, the police—and the courts—assume that that identification is reliable. As the above statistics reveal, it’s not. In fact, experimental studies in which people are exposed to mock crimes suggest that when the true culprit is not in the lineup, more than half the time eyewitnesses will do exactly what Jennifer Thompson did: they will choose someone anyway, selecting the person who best matches their memory of the criminal.³ As a result, false eyewitness identification seems to be the leading cause of wrongful conviction. An organization called the Innocence Project, for example, found that of the hundreds of people exonerated on the basis of postconviction DNA testing, 75 percent had been imprisoned because of inaccurate eyewitness identification.⁴

You would think that such findings would result in a massive overhaul of the process and the use of eyewitness identification. Unfortunately, the legal system is resistant to change, especially when the changes are fundamental—and inconvenient. As a result, to this day the magnitude and probability of memory error has gone virtually unnoticed. Certainly the law occasionally pays lip service to the fact that eyewitnesses can be mistaken, but most police departments still rely heavily on lineups, and you can still convict someone in court solely on the eyewitness testimony of a stranger. In fact, judges often prohibit the defense from introducing testimony about the scientific research on the flaws of eyewitness identification. “Judges say it’s either too complicated, abstract, and unconnected for jurors to understand, and other times they say it’s too simplistic,” says Brandon Garrett, the author of a book called Convicting the Innocent.⁵ The courts even discourage jurors who are deliberating from using the trial transcript to aid their memory of the testimony they heard in court. The state of California, for example, recommends that judges inform juries that “their memories should prevail over the written transcript.”⁶ Lawyers will tell you there are practical reasons for that policy—for instance, that deliberations would take too long if jurors pored over the trial transcripts. But to me, that seems outrageous, like saying we should believe someone’s testimony about an incident rather than a film of the incident itself. We’d never settle for such thinking in other areas of life. Imagine the American Medical Association telling doctors not to rely on patients’ charts. “Heart murmur? I don’t remember any heart murmur. Let’s take you off that medication.”

IT’S RARE TO have proof of what actually happened, so in most cases we’ll never know how accurate our memories really are. But there are exceptions. In fact, there is one example in which those who study memory distortion were provided with a record that couldn’t have been surpassed had they orchestrated the incident themselves. I’m referring to the Watergate scandal of the 1970s. That scandal concerned a break-in by Republican operatives at the headquarters of the Democratic National Committee and the subsequent cover-up by the administration of President Richard Nixon. A fellow named John Dean, the White House counsel to Nixon, was deeply involved in orchestrating the cover-up, which eventually led to Nixon’s resignation. Dean was said to have an extraordinary memory, and as millions around the world watched on live television, he testified at hearings held by the United States Senate. In his testimony, Dean recalled incriminating conversations with Nixon and other principals in such great detail that he became known as the “human tape recorder.” What endows Dean’s testimony with scientific importance is the fact that the Senate committee later discovered that there was also a real tape recorder listening in on the president: Nixon was secretly recording his conversations for his own later use. The human tape recorder could be checked against reality.

The psychologist Ulric Neisser did the checking. He painstakingly compared Dean’s testimony to the actual transcripts and cataloged his findings.⁷ John Dean, it turns out, was more like a historical novelist than a tape recorder. He was almost never right in his recollections of the content of the conversations, and he was usually not even close.

For example, on September 15, 1972—before the scandal engulfed the White House—a grand jury concluded