Seventy years ago, in a book called Convicting the Innocent, the Yale Law School professor Edwin Borchard produced a classic study of how the wrong person gets sent to prison or to death. The hapless innocents Borchard profiled included a coal miner and a doctor, Central European immigrants and American blacks, an unemployed religious visionary and an Algerian john named Frenchy. In those days exoneration was almost always a matter of luck—occasionally, for example, a supposed murder victim would turn up "hale and hearty" sometime after the alleged murderer landed in the penitentiary. Today, thanks to DNA evidence (when it is available), wrongful convictions can be reversed more confidently than ever before. And that confidence allows us to analyze the reasons for such convictions with greater certainty than Borchard or his contemporaries could.
Yet what is striking about the death-penalty convictions overturned recently (a hundred have been reversed in the past thirty years) and about other cases in which DNA evidence belatedly showed the accused to be innocent is how clearly the convictions rested on the same flawed foundations that Borchard identified. As in Borchard's day, what tends to do in the wrongly convicted is the kind of evidence that seems clinching, that often is clinching—namely, eyewitness identifications and confessions. But the human memory is not a video recorder; eyewitness testimony is notoriously flawed. And although most of those who confess are guilty, people can and do confess to crimes they did not commit—sometimes because they are coerced; sometimes (more often) because they are bewildered, frightened, or exhausted; sometimes because they are children, or adults with the mental capacity of children; sometimes because their interrogators have presented them with plausible scenarios in which they might have committed the crimes unknowingly—while blacked out, for instance, or while in the grip of another personality.
Children in interrogation rooms will sometimes confess to crimes they did not commit on the assumptions that they will then be allowed to go home, that they are doing what an adult wants of them, and that they can tell their parents the truth and fix everything later. Remember the two boys, aged seven and eight, who falsely confessed to killing eleven-year-old Ryan Harris in Chicago four years ago? The detectives who interrogated them, without their parents present, are said to have served one of the boys a Happy Meal and held his hands, saying, "We are all friends." The mentally retarded, too, will sometimes falsely confess, and for the same sorts of reasons: eagerness to please, naiveté about the legal weight of a confession, a yearning to be back home or to see their mothers. Jerry Frank Townsend, a twenty-seven-year-old retarded man, admitted in 1979 to six murders and a rape and served twenty-two years in a Florida prison before DNA evidence helped to clear him, last year. Charlie King, a retarded seventeen-year-old who was working as a school janitor, went to jail in 1992 for a crime he did not commit but did confess to—the murder of a nine-year-old girl in East Saint Louis—while the real killer went on to murder two more girls. "Even without the use of formal third-degree methods," as Borchard put it, "the influence of a stronger mind upon a weaker often produces, by persuasion or suggestion, the desired result."
Even able-minded adults, subjected to the right combination of coercion, sleeplessness, and grief, can falsely confess. In 1999 Keith Longtin, a forty-four-year-old welder from suburban Maryland, whose case was documented in a Washington Post series last year, allegedly made self-incriminating statements to the police about his wife's murder. Longtin had been held for thirty-eight hours of questioning, during which he slept, the police log says, for a total of fifty minutes. (While he served eight months in prison, the real killer, whose identity was later established by DNA evidence, sexually assaulted five women at knifepoint, one in front of her young child.) In 1988 Christopher Ochoa confessed to raping and murdering a young woman in Austin, Texas; he was later definitively cleared. Now that Ochoa is out of prison, he says that it's hard to explain to people who have never been in his position why he confessed. In interviews and speeches he offers some variation on the following themes: he was terrified, he was only twenty-two years old and had never been in trouble with the law, his interrogation lasted twelve hours, and the detectives in charge showed him autopsy photos of the murdered girl and threatened him with the death penalty.
A 1996 Justice Department report titled Convicted by Juries, Exonerated by Science detailed twenty-eight cases of wrongful conviction. Eyewitness identifications, usually by the victims, were the decisive factor in most of them. Like a confession, the testimony of an eyewitness, particularly a victim, is powerful stuff, sometimes seen as the gold standard of evidence. But in fact—as shown by a generation's worth of careful research on memory and suggestibility by psychologists including Elizabeth Loftus and Gary Wells—eyewitness accounts can be fragmented and changeable and subject to the deep desire to see somebody, anybody, punished for a terrible crime.
Experts have come up with two very good ideas for making wrongful convictions less likely. One is to improve the standard police lineup by letting witnesses see only one purported suspect at a time, so that they can make an absolute judgment about each one. When witnesses see six people at once, they make relative judgments, comparing the six and picking whoever looks most like the person they remember from the crime scene, rather than evaluating each individually. Conducting lineups sequentially seems like a minor change, but research by Wells and others has shown that it reduces the number of mistaken identifications—by as much as one half—without significantly reducing the number of correct ones. Ensuring that the detective running the lineup does not know who the real suspect is, and so does not make leading comments (Don't you want to look at number three again?), helps too, for the same reason that good clinical research is double-blind: otherwise it's easy to contaminate the results with intentional or unintentional bias.