Wells’s most pivotal and controversial experiments have dealt with the structure of the lineup itself. In conducting simulated lineups over the years, he identified two pathways by which people make decisions. One involves absolute judgment, the kind of instant recognition in which someone might immediately cry out, “That’s him!” The other is a more deliberative process, in which the witness compares one face to another. “People would say things like, ‘I know it can’t be numbers one, two, four, five, or six, so it must be number three,’” Wells says. He calls that process “relative” judgment, since it involves deciding which face resembles the witness’s memory relative to the others.
In order to compare the two kinds of decision making, Wells and his colleague Rod Lindsay at Queen’s University in Ontario designed a new kind of lineup. Rather than showing six photographs together, they presented the photos one at a time. In other words, they replaced the traditional simultaneous photo array with a sequential one. In this way a witness would have to recognize the culprit instantly rather than pick him out from a group (thinking that the suspect had to be among them).
In the first of many trials, Wells and Lindsay showed a staged crime to 240 students and had half the students pick from simultaneous lineups and half from sequential ones. The sequential lineup reduced mistaken IDs by nearly half. Dozens of studies throughout the country have since confirmed that effect.
Wells says the distinction between “relative” and “absolute” judgment has applications beyond the lineup. Composite sketches, long a staple of police shows on TV, are notoriously inaccurate in real life. Unlike the way we recall other images—for instance, a two-story red-brick house with a screened-in porch and green awnings—humans are not programmed to construct faces by components. “The baby recognizes either ‘Mom’ or ‘not Mom,’ ” he says. “Not ‘Mom has that kind of eyebrows.’ ”
To demonstrate the point, Wells sits me down at an office computer and boots up one of the standard software programs that police departments use to create composite sketches. (Computers have almost entirely replaced human sketch artists.) “Imagine the face of someone you really know, like your father,” he says. “Now we’re going to build it.”
I summon a picture of my father in middle age: wavy hair, square jaw, hazel eyes. Wells clicks, and several hundred facial shapes appear on the screen. “Pick one,” he says. Instantly I see the difficulty of the task. Nothing in my mind’s eye corresponds to the featureless shapes on the screen.
“Just pick,” he says. “We can tweak the images as we go along.”
I pick one image that sort of looks like the right facial shape and ask Wells to alter it to my specifications. The next screen displays several dozen disembodied eyebrows. We repeat the process for many other features, including mouths, hairlines, eyes, noses—the program has 3,850 facial elements in all—until I give up in frustration.
“You start to realize you just don’t friggin’ know,” Wells says, laughing as he presses the print button. The sketch that emerges looks more like an ape-man than a person. “This is not how we store faces. We don’t store them in features. We store them intact.”
Wells says that anything that pushes us from absolute to relative judgment makes our testimony less reliable. To identify the tipping point where we go from recognizing a face to comparison-shopping for one, he has embarked on the biggest experiment of his career, testing nearly 1,600 people under 16 different conditions. In some cases, he will show videotaped crimes with blurred or darkened images; in others, lineups with blurred faces. The idea is to push the boundaries of instant recognition, identifying the point where a witness no longer recognizes the suspect but talks himself into committing to a near match.
Wells is convinced that by documenting that transition—by recording decision times and having test subjects vocalize their internal dialogue—he will be able to devise methods that can separate reliable from unreliable testimony.
These days, wells travels about 26 weeks a year as the unofficial head of a loosely affiliated cadre of academics, lawyers, and police officers proselytizing new procedures for judging eyewitness testimony. One goal is persuading detectives to see themselves as scientists. After all, Wells tells them, “You’ve got a researcher, hypothesis, design, procedure, subjects. You execute procedures, get results, and make interpretations.” He urges them to adopt rigorous protocols such as conducting lineups as double-blind studies: “We do this in medical research all the time.”
Thirteen states have mandated new, scientific procedures for lineups. And more than 40% of police departments in the U.S. have adopted some or all of Wells’s suggestions.
Wells’s efforts have begun to pay off. Thirteen states have mandated new, scientific procedures for lineups such as double-blind or sequential lineups, or have assigned commissions to create policies based on the new social science; five have signed on in the past year alone. And more than 40 percent of the nearly 18,000 police departments in the United States—from rural sheriff’s offices to metropolitan police—have adopted some or all of Wells’s suggestions for lineups, according to a survey by the Police Executive Research Forum. Attorney Barry Scheck of the Innocence Project contacted Wells more than 15 years ago and has been working with him to promote science in the legal system ever since. “His approach on system variables is in sync with my own,” Scheck says. “He has the tone and the approach that could change the field.”
Not everyone is a fan. Several prominent psychologists and attorneys say that laboratory experiments reflect neither the stress nor the confusion of real-world situations, and argue that Wells’s techniques have not been tested thoroughly enough to recommend as a legal policy. “I think he claims far more for sequential lineups than can be supported by the science,” says Roy Malpass, who has written several papers criticizing Wells’s conclusions. Malpass notes that in a 2006 field test of the lineup in several police stations, the state of Illinois actually found sequential lineups slightly less accurate than traditional photo arrays.
Wells and his colleagues from several other universities responded that the Illinois study was sloppily conducted. To prove their point, they collaborated on a field study of their own. The researchers selected four police departments—Austin, Tucson, San Diego, and Charlotte, North Carolina—and gave them laptop computers programmed to randomly administer simultaneous or sequential studies. Witnesses then conducted the lineups themselves, clicking through instructions and options on the computers. With no human directing the interaction, the data-gathering was consistent and double-blind. The results, released last fall, showed that sequential lineups produced the same rate of instantaneous identifications as traditional lineups. More important, the sequential lineups led to significantly fewer false identifications of people known to be innocent.
The implications of this research are enormous, and not only for the untold thousands of innocent people who have been convicted. As Wells puts it, every mistaken ID creates a “double injustice”: one for the convicted person but also another for the rest of us when the person who committed the crime remains free.