Seeing Is Believing

(Flickr/Boston Bill)

On a Saturday night in July 1984, Jennifer Thompson, a 22-year-old, straight-A student at Elon College in Elon, North Carolina, returned to her apartment after attending a party. Thompson wasn’t feeling well and went to sleep. Her boyfriend left around 11 P.M. About four hours later, Thompson awoke with a man on top of her. He held a knife to her neck. He smelled of alcohol and cigarettes. Thompson screamed. “Shut up, or I’ll cut you,” the man threatened. 

Before, during, and after the rape, Thompson willed herself to study his features so she could increase the odds of identifying him later. She made mental note of his close-cropped hair, his small almond-shaped eyes, his high, broad cheekbones, his wisp of a mustache.

When the rape was over, Thompson lured the man into the kitchen by promising to pour drinks for both of them. His attention briefly diverted, Thompson ran out the door to a nearby house. The neighbors called 911. Later, Thompson would learn that her assailant broke into a residence less than a mile from her apartment and raped another woman that night.

Thompson worked with Detective Mike Gauldin on a composite sketch of the rapist. “The mouth wasn’t quite right,” she recalled, “and the ears stuck out too much, but it was close to the picture in my mind.” The second rape victim, Mary Reynolds (a pseudonym), wasn’t sure she could identify her assailant. One fact was certain: He was a black male raping white women in a small city with a legacy of shaky race relations. 

Two days after circulating the composite, police asked Thompson to return to the station house. Gauldin and two other detectives sat with her. They placed in front of her an array of six photographs. The detectives specified that the rapist might not be in the array. 

“I assumed they must have a suspect,” Thompson later said. Why else would they have called her to come in? “All I had to do was pick him out. And if I failed to do that, would he go free? Would he find me?” Thompson eliminated four photos easily. Of the two that remained, one brought back “the image of the man performing oral sex on me so violently I thought I would be sick right there.” 

She pointed to a photo. “Yeah. This is the one. I think this is the guy.”

One of the detectives asked, “You think that’s the guy?”

Thompson replied, “It’s him.”

Another detective asked, “You’re sure?”

She said, “Positive.” 

A minute later, Thompson asked, “Did I do OK?”

One of the detectives answered, “You did great, Ms. Thompson.”

The session had taken five minutes.

Eleven days after the rape, Thompson went back to the police station to view a live lineup of seven men. No partition separated her from the suspects. The men in the lineup could see and hear her. The detectives told Thompson she should not feel pressure to identify any of the seven, but she felt pressure nonetheless. “If he was here, now he knew what I looked like in broad daylight,” Thompson recalled thinking. “He knew my name. If he was here, I couldn’t screw this up.” She did not want to make eye contact as each man stepped forward. But she had to. Thompson concluded her rapist could have been No. 4 or No. 5. 

She listened to their voices again. The voice of No. 5 resonated. She forced herself to look at him. His body was “long and lean.” He sported a “light mustache.” His eyes looked “cold.” Thompson chose No. 5. In the hallway, she asked Gauldin how she had done. He replied, “We thought that might be the guy. It’s the same person you picked from the photos.” Thompson felt validated. “We got him,” she thought.  

His name was Ronald Cotton. He ended up in the photo array and then in the live lineup because of the composite drawing. Twenty-two years old, a ninth-grade dropout, employed at a restaurant, he’d been identified by a manager there who had seen the composite. Cotton had been previously incarcerated for breaking and entering with intent to rape, an arrest he insisted derived from a misunderstanding, not from a violent sexual act. 

When Cotton learned that police were searching for him, he went voluntarily to the station. He spoke to detectives without a lawyer present, offering what he believed was a solid alibi, only to realize later he had mixed up his nights. As a result, police were no longer listening to explanations, and he was sitting in the county jail, unable to raise enough money for bail. 

As Thompson prepared to testify at Cotton’s trial, scheduled for January 1985, she learned that Mary Reynolds had not identified Cotton during a lineup. Instead, she had identified a man recruited by police to serve as a filler. Rather than wondering if she had been mistaken, Thompson felt even greater resolve. At trial, when asked if she could identify her rapist, Thompson pointed at Cotton. She noticed his expressionless visage, “as if he didn’t care at all what he had done to me.” 

Cotton’s defense lawyer emphasized the lack of physical evidence. Would the jurors, he asked, convict Cotton solely on the testimony of the victim? After all, Thompson had said nothing about identifying marks such as Cotton’s scars. The defense wanted to present testimony from an expert about the vagaries of memory, especially during and after a traumatic event. The judge said no. The jury deliberated four hours before finding Cotton guilty. He received life imprisonment plus 50 years. 

About three months into his prison sentence, Cotton noticed a new inmate being escorted to a cell. He looked strikingly similar to the composite of the man who had raped Thompson. Approaching the inmate, Cotton asked where he was from. Burlington, the new inmate answered, one town over from Elon. They exchanged names: Ron Cotton. Bobby Poole. Cotton soon learned the crime that had brought Poole to prison was rape. Their physical resemblance caused some of the corrections officers and inmates to confuse the two. 

During his second year in prison, Cotton won a new trial from the North Carolina Supreme Court. The justices ruled that Cotton’s lawyer should have been permitted to tell the jury about Reynolds’s failure to identify the defendant. Cotton’s hopes shattered, however, when, a few months before trial, the prosecutor charged him with raping Reynolds; three years after not identifying Cotton, Reynolds said she had recognized him but had been too frightened at the lineup to speak out. The new trial would consolidate the two crimes.

After Thompson’s and Reynolds’s testimonies, the judge sent the jury away so he could hear arguments about whether the defense would be allowed to question Poole with the jury present. As the jurors waited outside the courtroom, Cotton’s lawyer asked Poole if he had raped Thompson and Reynolds. Poole denied everything. An inmate cast doubt on that denial by testifying that Poole had told him he had raped the women. Cotton’s lawyer, still outside the presence of the jury, pointed out that Poole had blood type A, the same type found at the Reynolds rape scene. Cotton had blood type O. Nonetheless, the judge ruled that the evidence pointing to Poole was too precarious for the jury to hear. 

Thompson saw Cotton and Poole in the courtroom, but when asked if Poole could have been the rapist, she answered no. Later, she recalled being furious that Cotton’s lawyers “would try to point the finger at someone else, using a jailhouse snitch to do it, no less. There was never a doubt in my mind that Ronald Cotton was the man who assaulted me.” The jurors deliberated about an hour before finding Cotton guilty again. 

Thompson did her best to move on with her life. For a long time she found it difficult to trust men. She eventually married, and in 1989 learned she was pregnant with triplets. The rape, though, continued to haunt her. She saw Cotton’s face while trying to sleep. She wished he would die in prison. 

Then, in March 1995, Detective Gauldin and a prosecutor visited Thompson and her husband at their home. Gauldin was delivering news he knew would upset her. A new team of lawyers had taken on Cotton’s case and had arranged a DNA test based on evidence still in the possession of the police. None of the evidence contained Cotton’s DNA, but no DNA showed up from another man, either. That result did not constitute a finding of innocence. The next step, the lawyers had told Cotton, would be to test the rape kit from the hospital where Thompson had been taken right after the assault. But the blood sample from Thompson within the rape kit had deteriorated. 

Gauldin told Thompson that she could give a blood sample voluntarily or perhaps face a court order later. “Why do I have to go through this again?” Thompson remembered asking. “I’m supposed to be the victim.” Hoping to soothe her, Gauldin said, “Look, we all know it’s not going to change anything. It’s just a last attempt. Ronald Cotton is going to stay in prison. But they could drag this thing out.” Thompson decided to comply, thinking that, finally, she would never have to hear Cotton’s name again.

That is not how the script turned out. In June 1995, the next round of testing found insufficient DNA material in Jennifer Thompson’s rape kit but found Bobby Poole’s DNA in Mary Reynolds’s rape kit. Confronted with that evidence, Poole confessed to both rapes. 

Cotton attended a brief court hearing and walked out a free man after 11 years of incarceration. 

Thompson, the seemingly ideal eyewitness, had been mistaken. She had chosen the wrong man. She had failed to identify her assailant. 


Ronald Cotton’s conviction based on Jennifer Thompson’s mistaken identification did not surprise Gary Wells, who worked hundreds of miles away on a university campus in Ames, Iowa. 

By 1995, Wells had already spent 21 years researching the uncertainties of eyewitness identification, explaining why faith in such testimony bedevils the criminal-justice system, and devising inexpensive protocols to reduce the number of wrongful convictions. Many jurors find eyewitness testimony more compelling than any other evidence, using the commonsense reasoning that “she was there, I wasn’t, who am I to question her identification?” Even when defense lawyers cross--examine eyewitnesses and establish inconsistencies between the version in police reports and the version told at trial, jurors’ acceptance of the eye-witness is often unshaken. For Wells, improving the accuracy of eyewitness testimony is a matter of morality but also a law-and-order issue: When an innocent person is convicted, the actual rapists and murderers remain at large, free to rape and murder again and again. 

Although people like to think of themselves as perceptive, Wells’s research has established that there is a Jennifer Thompson inside everyone, a well-intentioned individual who is certain of what she saw yet is sometimes mistaken. When eyewitnesses are confronted by a stressful situation—as a victim of rape, for example, or as a witness to a crime that involves a gun—the odds of a correct identification are no better than flipping a coin.

Over the past 40 years, Wells has become the most visible eyewitness researcher in a field crowded with accomplished researchers. No one else in the field possesses Wells’s combination as a big-picture thinker, laboratory researcher, and effective advocate for reform within the judicial system. Under six feet and of slight build, with thinning, longish hair, he’s not physically imposing. His attire tends toward the informal, jeans and a sport jacket. He could blend in anywhere. 

Wells, who is 61, grew up in Hutchinson, a city of 37,500 in Central Kansas. His dad was a firefighter; his mom sold beauty products. Not expected to attend college, Wells settled on Kansas State University because he could not afford out-of-state tuition, and it offered a small scholarship. By his freshman year, he was married to his high-school sweetheart. By his sophomore year, he was the father of a baby boy. 

Uncertain of a major, Wells enrolled in a psychology course during his freshman year. He liked the field’s newness. “When you looked in the back of the textbooks, everyone was still alive,” he says. In such a relatively young discipline, Wells thought he might make an impact. He also liked the opportunity psychology offered for a student to prove or disprove mistaken beliefs that had morphed into societal norms. “If you don’t measure it, you can’t study it scientifically” became Wells’s foundational principle. 

In 1973, Wells entered Ohio State University’s combined master’s and doctoral program in psychology. During his second year, he heard a man speaking loudly in the hallway of the psychology building, hoping to rouse someone’s attention. The man was a lawyer from Cincinnati, representing a client he believed had been wrongfully arrested because of mistaken eyewitness testimony. Wells approached the lawyer, who was carrying an envelope containing a photograph of the lineup that had led to the client being charged with a crime. Yes, Wells told the lawyer, he and his colleagues knew something about the workings of memory but had not applied that knowledge to the reliability of eyewitness identification.

The lawyer handed the lineup photo to Wells, who tacked it onto the bulletin board in his graduate student office. Within a month, he and a colleague decided to learn more about eyewitness identification. (All these years later, Wells is unsure about the name of the lawyer and whether the client had been misidentified.)

Thomas Ostrom, a professor in the psychology department who specialized in human memory, helped steer Wells to the work of Robert Buckhout. A psychology professor at Brooklyn College, Buckhout was not a traditional academic; has crime rate was a self-described smartass. Early in his adult life he served in the Air Force, rising to captain. He earned his doctorate at Ohio State University and bounced around the academic world before landing at Brooklyn. Publishing in both academic and nonacademic journals, Buckhout argued that psychologists could improve outcomes within the criminal-justice system by sharing their expertise with police, prosecutors, judges, and jurors about the uncertainty of eyewitness testimony. He frequently agreed to assist the defense, criticizing prosecutors who relied on eyewitness identification while ignoring contrary evidence. Law-enforcement officials tended to dismiss Buckhout as a soft-on-crime crank, but his zeal left a mark on a younger generation of psychologists, including Wells.

Soon after encountering the lawyer and immersing himself in Buckhout’s writings, Wells conducted an experiment that yielded a stunning insight. He devised a scenario in which he asked individuals to watch his bag while he left the room. A confederate then grabbed the bag, dropped it, and picked it up again, giving everyone a good look at him before he bolted. When shown a lineup, a majority of the witnesses chose the wrong man. More telling, those choosing the wrong person were just as confident as those who identified the right person. The confidence level of witnesses, a crucial factor for police, prosecutors, judges, and jurors when assessing testimony, suddenly seemed shaky as a gauge of accuracy.

The law-enforcement establishment bridled at such findings. So what? professional associations of prosecutors, police, and judges retorted. Laboratory experiments mean nothing in real life, where victims and other eyewitnesses have been vetted by skilled police detectives and district attorneys. Some people are confident. Others are tentative. That’s human nature. Sure, mistakes occur, but not in most cases, and when they occur, they are well intentioned. What does all this academic research tell us about an individual witness in a specific case? Nothing. 

Wells was determined to show law enforcement that eyewitness testimony plays into the memory center of the brain, a mysterious realm that scientists are still just beginning to understand. Most important, Wells wanted to convey the message that psychologists could offer suggestions for improving the criminal-justice system based on research, suggestions almost certain to reduce the number of wrongful convictions.

“The first thing to understand,” Wells has written, “is that the usual safeguard of cross-examination simply does not work with eyewitnesses. … Cross-examination is designed to distinguish between truth tellers and liars and it is a fairly effective tool for doing so. … In the case of an eyewitness, however, lying versus truth telling is not the issue. … Mistaken eyewitnesses believe what they are saying, their errors are genuine, and those errors of memory tend to be indistinguishable from accurate memories.”


The first psychologist to question the accuracy of eyewitness testimony was Hugo Munsterberg. One of the early giants of psychology, the Prussian-born Munsterberg was invited to teach at Harvard in 1892 by William James. Munsterberg advocated for the application of psychological insights to a variety of fields, including law. His most controversial book was also his most enduring, On the Witness Stand, published in 1908. A collection of articles on everything from jury selection to police interrogation, it argued that police, prosecutors, and judges should be wary of eyewitness testimony. 

Munsterberg did not conduct controlled research, but he offered himself as an example of an unreliable eyewitness after his home had been burglarized. Despite Munsterberg’s close examination of the crime scene and renowned memory (he lectured without consulting notes), his testimony under oath was riddled with errors. In his book, Munsterberg emphasized that such mistakes, while not inevitable, occur as a matter of course: “We never know from the material itself whether we remember, perceive or imagine, and in the borderland regions there must result plenty of confusion which cannot always remain without dangerous consequences in the courtroom.”

Munsterberg lived before the era of modern electronics, but he implicitly understood that human memory is not the equivalent of a tape recorder or camera, popular but mistaken analogies relied upon by police, prosecutors, judges, and jurors. Witnesses testify under oath that they remember a face at a crime scene; even though they are wrong, everyone believes the testimony, as everybody would believe Jennifer Thompson. 

Munsterberg lost the campaign to insert psychological insights into courtrooms. Sixty years later, though, psychologist Elizabeth Loftus would validate his assertions. While earning her doctorate from Stanford University in the late 1960s, Loftus became intrigued with Munsterberg’s work. Teaching at the University of Washington, among other schools, she focused on the malleability of brain centers, especially how memory could be surprisingly unreliable. The process of viewing something and then storing the images in the brain and retrieving those images is extraordinarily complex, affected by myriad factors. False memories, for instance, can be created after an accurate memory has been established. 

An example of Loftus’s memory research with obvious implications for cases of wrongful conviction: She asked students to view a film of an automobile accident. Half the students answered the question “How fast was the white sports car going when it passed the barn while traveling along the country road?” No barn appeared in the film. About 20 percent of the students said they had seen the barn. By showing how her students “saw” something nonexistent, Loftus established how contamination can affect memory. A victim chooses No. 2 from a lineup of six men, for example, because he had already seen No. 2 in the photo array. Contamination has occurred. It was through experiments like this that Loftus changed the entire conversation about eyewitness testimony. The question was no longer “How good are eyewitnesses?” but “What are the factors that affect eyewitness performance?” Buckhout had not made eyewitness research respectable. Loftus did. 

By the time Wells settled in at Iowa State University, in 1988, he had developed an elegant, deceptively simple framework for his research. He differentiated between the elements police and prosecutors could not control (what he called “estimator variables”) and those they could (“system variables”). An example of an estimator variable might include the adequacy of light at the crime scene, the duration of the crime, or the racial and ethnic identities of victims, witnesses, and perpetrators. 

For instance, research has shown that eyewitness unreliability tends to increase when the alleged perpetrators and the witnesses are of different races. Caucasians are less reliable when trying to identify African Americans, and African Americans are less reliable when trying to identify Caucasians. Witnesses viewing an image of a Caucasian male holding a weapon, with an African American male beside him, tend to recall that the African American was brandishing the weapon. Given the disproportionate number of African Americans among the wrongfully convicted, the research findings were important. They could be useful during a trial. But those studies could not improve the accuracy of eyewitnesses, because the elements were beyond the control of police and prosecutors. 

Wells decided to focus on the elements that could be controlled. He embarked on a series of experiments that involved multiple witnesses observing staged crimes in progress and then viewing lineups and photo arrays after the action occurred. The experiments consistently showed that many of the witnesses identified the wrong alleged perpetrator. Clearly, some mechanism within the human brain was triggering the mistakes. But which mechanism? How could it be bypassed?

Clarity began to emerge when Wells experimented with six-person lineups and six-person photo arrays typically used by police. One hundred witnesses to a staged crime viewed a lineup and a photo array that included the perpetrator. Fifty-four accurately picked the perpetrator. Twenty--five failed to. Twenty-one said they did not think the perpetrator had been included. Then a different group of 100 viewed the same staged crime, live lineup, and photo array. This time, however, Wells did not include the perpetrator in the lineup or the array. Thirty-two witnesses correctly chose nobody. Almost all of the remaining witnesses selected the innocent individual who most resembled the perpetrator. 

What had happened? Wells theorized that when witnesses are asked to choose among multiple suspects, those witnesses often make a relative judgment, identifying the suspect who looks the most like the perpetrator, even when the perpetrator is absent. Imperfect memory plus the desire to please police and prosecutors were producing incorrect identifications and leading to wrongful convictions.

Revelation after revelation emerged from the experiments, and those revelations led to recommendations: Law-enforcement officers unaware of a suspect’s identity should administer live lineups and photo arrays, thus eliminating verbal and physical cues, intentional or unintentional. Witnesses and victims should be told that the lineup or photo array might not include the perpetrator, thus reducing the pressure to identify somebody when certainty is absent. During lineups and photo arrays, witnesses should be presented one person or one photo at a time, thus compelling a direct response (yes or no) about each alleged suspect rather than a relative response from witnesses (more likely or less likely). 

Early in his career, Wells published his findings in the Journal of Personality and Social Psychology, the most prestigious publication in the field. “For boldness,” says James M. Doyle, a defense attorney who co-authored a book on eyewitness testimony with Loftus, “this was the equivalent of a high-school English student sending a short story off to The New Yorker. … Typically, articles in scholarly journals are cited once or twice by other academics, then forgotten. Wells’s ‘system variables’ article has been cited hundreds of times. The term has now, like ‘Xerox’ or ‘Kleenex,’ passed out of the realm of proper nouns dressed in capital letters and into the everyday language of the field. It is not uncommon to see the term ‘system variable’ used with no reference to its origins. It is as though it has always been there.” 

Wells’s visibility rose. He lectured across the country. Journalists called on him as a resource. Defense attorneys hired him as a well-paid expert witness, a practice Wells gradually halted because he decided it damaged his credibility with law enforcement. “It should not be the purpose of expert testimony merely to raise doubt about the guilt of a defendant in a given case,” Wells says. “Expert testimony does nothing to address the misidentification problem unless it helps the legal system improve its methods for collecting the evidence.”

Wells’s research led him to an incontrovertible conclusion: Because of the fallibility of eyewitness testimony, the legal system produced avoidable wrongful convictions. Wells hoped what he had learned in the laboratory and the field would win the attention of the ultimate arbiters—the U.S. Supreme Court. But his research, plus that of Loftus and others, clashed with the Court’s controlling decision regarding eyewitness identification, the 1977 ruling in John R. Manson v. Nowell A. Brathwaite

The case centered on the eyewitness testimony of a Connecticut State Police trooper named Jimmy Glover. On May, 5, 1970, Glover made an undercover purchase of heroin from a seller in a Hartford apartment. Glover said that he observed the seller from outside the apartment, with the door open about 12 to 18 inches. It was 7:45 P.M. According to Glover, natural light from an apartment window provided adequate illumination to recognize faces inside. He handed the suspect $20 through the partially opened door. The door closed. Soon it reopened, and the suspect handed two bags of heroin to Glover. 

Glover did not know the identity of the seller. At police headquarters, he described the suspect to two colleagues. One of the officers thought the description fit Nowell Brathwaite and obtained his photograph from the local police. On May 7, Glover found the photograph on his office desk and believed it matched the seller in the apartment. At trial, Glover identified Brathwaite as the seller. The prosecution never explained why Brathwaite had not been placed in a lineup or why the photograph had not been shown to Glover as part of an array. When Brathwaite testified, he denied being at the apartment on the day of the sale but could not offer an unbreakable alibi. The jury convicted Brathwaite, and he began a prison sentence of six to nine years. 

The Connecticut Supreme Court rejected Brathwaite’s appeal, as did the federal district court. A federal appeals court reversed the conviction, however, stating that Glover’s photographic identification should have been excluded from the trial as “unnecessary and suggestive.” The U.S. Supreme Court agreed to hear the case.

It is impossible to know whether Brathwaite was innocent or guilty. This much is certain: The seven justices who ruled on June 16, 1977, that Brathwaite’s conviction should be restored believed Glover’s testimony was untainted. 

Writing for the majority, Justice Harry Blackmun said “reliability” (in this case, of Glover) “is the linchpin in determining the admissibility of identification testimony.” The factors in determining reliability, according to Blackmun, included the opportunity of the witness to view the suspect, the witness’s degree of attention, the accuracy of the description from the start of the case onward, the level of certainty demonstrated by the witness when viewing the suspect after an arrest, and the amount of time passing between the crime and the later viewings. Almost everything Blackmun identified would be contradicted by Wells’s research.

In dissent, Justice Thurgood Marshall (joined by Justice William Brennan) accused the majority of “dismantl[ing] the protections against mistaken eyewitness testimony” and “blindly uphold[ing] the conviction of a defendant who may well be innocent.” Marshall went on: “The use of a single picture (or the display of a single live suspect, for that matter) is a grave error, of course, because it dramatically suggests to the witness that the person shown must be the culprit. Why else would the police choose the person?”


Until the 1990s, anybody claiming that wrongful convictions occur at a high rate invariably encountered skepticism, and often scorn, from the legal establishment: “All inmates claim they are innocent,” went the standard line. As crime rates climbed through the 1970s and 1980s, few people in law enforcement—and even fewer in elected office—wanted to hear that they were sending the wrong people to prison. 

Then two related movements began to occur.

The first involved the use of DNA in forensic laboratories. Attempts to match genetic material to a specific crime suspect had previously suffered because the typing of human blood into a group still left lots of potential candidates—maybe hundreds of thousands in a locale like New York City. DNA molecules, however, are unique to each person. Fingerprints are unique, too. But matching a fingerprint from a crime scene to the print of an individual involves interpretation by a forensic examiner, especially if the print is smudged or otherwise flawed. Testing material containing DNA is more objective. In many rape cases, for example, DNA samples from the victim and the attacker can be matched to allow the presentation of ironclad evidence. 

The first DNA exoneration of a wrongfully convicted suspect occurred in 1989, freeing Gary Dotson from an Illinois prison after a mistaken rape conviction. As DNA testing became increasingly sophisticated, police, prosecutors, defense lawyers, and judges agreed that the results could be considered highly accurate—maybe not flawless but close. As genetic material from crime scenes in current cases showed that the incorrect suspect had been arrested, the conventional wisdom about all guilty prisoners proclaiming innocence began to fray. Then, as testing of preserved genetic material in older cases led to exonerations, the conventional wisdom collapsed altogether. 

The second movement involved the creation of innocence projects around the nation. The largest, the Innocence Project, founded in New York City by lawyers Barry Scheck and Peter Neufeld in 1992, specialized in exonerating wrongfully convicted inmates through the use of DNA testing. Scheck likens DNA testing within the criminal-justice system to the telescope within astronomy, “a way to see things as they really are.” Over the past 30 years, innocence projects—there are now more than 50—have played a significant role in hundreds of exonerations.

With the acceptance of DNA testing and the spread of innocence projects, Wells no longer needed to refute the view that he was a researcher seeking a solution to a “minor” problem or an academic being duped by streetwise cons.

Janet Reno, President Bill Clinton’s attorney general, was the first national law--enforcement figure to express interest in Wells’s work. Alarmed by the number of wrongful convictions coming to her attention in local and federal courts, Reno in 1996 invited Wells to make a presentation to the department. Afterward, Wells told a friend that Reno “asked about blind testing, about pre-lineup instructions, selection of fillers, the use of composites, leading questions, and the use of live versus photo-identification procedures. She got it!”

In 1997, Reno named Wells to the newly created Technical Working Group for Eyewitness Evidence. For the next two years, representatives from police departments, prosecutors’ offices, defense-lawyer groups, and universities argued about what should be included in a manual for law-enforcement agencies. 

Wells and his fellow researchers felt confident that they knew what guidelines to provide police and prosecutors so that eyewitness misidentifications would decrease: Interview victims and other witnesses using neutral words and open-ended questions. Ensure the police officer conducting a live lineup or showing photographs does not know the identity of the suspect. Use the sequential method for mug-shot viewing. Tell the witness that the perpetrator may or may not be present in the lineup. Find fillers for the lineup who resemble the witness’s description of the suspect but who don’t resemble the suspect too closely. Videotape all proceedings. After the witness makes a choice, record the witness’s level of confidence immediately, so it can be compared with later testimony. 

All these techniques seemed obvious to Wells and his allies in drafting sessions, but some law enforcers on the panel felt even the term “guidelines” would box them in. They feared that defense lawyers would cast doubt on investigative techniques that did not comport with the recommendations. Some of the rancor could be explained as a clash of cultures. Police who place their lives in danger every day resisted being told how to do their jobs better by professors. Police have used the simultaneous lineup since the 19th century. Changing to a sequential procedure seemed like going into uncharted territory. Police representatives on the panel were especially resistant to the guideline regarding conduct of a lineup by an officer unaware of the suspect’s identity. Did Wells think cops lacked integrity or that officers were so badly trained they would tip off the witness? 

The prosecutors on the panel proved even more resistant, which surprised the academics. After all, Wells and his colleagues reasoned, they had the imprimatur of Reno, the nation’s top prosecutor. Unlike police, who were required only to have graduated from high school, prosecutors were all highly educated, with a minimum of a law degree. Perhaps Wells should have attributed greater significance to his initial encounter with Melissa Mourges, an assistant district attorney in New York City who served on the panel. “I think you are the devil incarnate,” Mourges said to Wells, who hoped she was joking. 

Despite the wrangling, Wells took heart in the knowledge that Reno wanted a positive result. During the sessions, he sensed the police representatives coming around somewhat. Police tend to favor specific procedures; their conduct manuals tend to be massive, trying to cover every contingency. Wells also grasped that police officers understood that many eyewitnesses were unreliable. Police see misidentifications every day—at lineups and in photo arrays. These are misidentifications that often occur before prosecutors ever become involved in cases. The police in the room worked with Wells to moderate his insistence that officers administering live lineups and photo arrays be unaware of the suspect’s identity. Wells’s proposal was sensible, the police said, but it would take time for it to be adopted. 

In 1999, the Justice Department distributed Eyewitness Evidence: A Guide for Law Enforcement to police agencies in every U.S. jurisdiction. For all of the compromises it contained, the document was a substantial achievement, a victory of science over tradition. Finally, methods to minimize eyewitness error had been memorialized in a widely available, easy-to-read manual. But after the months of arguing and compromising, there remained a huge question: Would police and prosecutors actually change their behavior?


A year and a half after the guide was distributed, New Jersey Attorney General John J. Farmer Jr. issued an order that every law-enforcement jurisdiction in the state had to follow its recommendations. Farmer grasped, in particular, the need for the blind administration of photo and live lineups, the guideline that law enforcement had expressed the most concern about. Knowing the sensitivity of the recommendation, Farmer said that it “is not intended to question the expertise, integrity, or dedication of primary investigators working their cases. Rather, it acknowledges years of research which concludes that even when utilizing precautions to avoid any inadvertent body signals or cues to witnesses, these gestures do occur when the identity of the actual suspect is known to the individual conducting the identification procedure.” 

Farmer’s order was only the initial step. For a statewide criminal-justice system to reduce eyewitness misidentifications, it requires not just the cooperation of police and prosecutors but judges as well. In 2010, New Jersey courts adopted eyewitness reforms after a drawn-out case involving the alleged wrongful conviction of Larry R. Henderson. The state supreme court asked Geoffrey Gaulkin, a retired judge, to serve as a “special master”; he convened ten days of hearings that included testimony from Wells. Gaulkin asked the attorney general, the public defenders representing Henderson, and other parties to cooperate with one another so that the eventual guidelines would be as definitive as possible. Relying heavily on Wells’s research, Gaulkin released an 88-page primer for New Jersey judges that laid out how to reduce impermissible or misleading eyewitness testimony. 

Soon after Farmer’s memo, the chief justice of the North Carolina Supreme Court created a 31-member commission to examine why wrongful convictions were occurring at an alarming rate. It was Wells’s testimony, according to Christine Mumma, executive director of the North Carolina Center on Actual Innocence, that established that reform was possible. “Gary’s presentation,” she says, “convinced the commission members it was necessary and possible to increase the reliability of convictions, and eventually led to unanimous support for eyewitness-identification reform.” The precise number of U.S. police jurisdictions that have changed their eyewitness practices since the guide was published is unknown. But it is estimated that as many as 40 percent of the approximately 18,000 have adopted reforms.

Wells’s research, however, has not held sway in the one arena where it would yield the most impact: the United States Supreme Court. The latest disappointment began to unfold on November 2, 2011, when the nine justices heard oral arguments in Perry v. New Hampshire. The Court had not ruled on the reliability and validity of eyewitness identification since Manson v. Brathwaite in 1977. Wells hoped the consensus about problematic eyewitnesses would lead the justices to revise or overturn Manson v. Brathwaite. But he worried that the specific case leading to a review of eyewitness testimony served as a poor vehicle for reform, because it turned on a narrow, technical legal nicety.

In 2008, Barion Perry had been arrested in Nashua, New Hampshire, after allegedly trying to break into cars parked in an apartment building lot. When police responded to a 3 A.M. telephone call from an apartment resident who had spotted somebody roaming the parking lot, they found Perry holding car stereo amplifiers. A metal bat was on the ground nearby. Perry told police he had spotted the amplifiers in the parking lot and picked them up. One of the police officers entered the apartment building, while another officer stood with Perry in the parking lot. The officer inside asked the witness for a specific description of the suspect. The witness pointed to her kitchen window and said the person she saw break into a car was standing in the parking lot next to the uniformed officer. The prosecutor charged Perry with theft and criminal mischief. 

Before his trial, Perry petitioned to strike the identification, given that it approximated a one-person show-up in the parking lot, pretty much guaranteeing that he would be identified as the culprit. The trial judge denied the request, despite acknowledging reasons to question the accuracy of the eyewitness identification: The parking lot was poorly lit; Perry was the only African American male in the vicinity; he was standing next to a police officer; and the eyewitness later failed to identify Perry from a photo array. 

Perry was convicted of theft but acquitted of the criminal-mischief charge, and the New Hampshire Supreme Court affirmed the lower court’s ruling. The U.S. Supreme Court agreed to hear the case on the narrow question of whether the due-process clause of the Constitution “requires a trial judge to conduct a preliminary assessment of the reliability of an eyewitness identification made under suggestive circumstances not arranged by the police.”

Wells was chief author of an amicus brief submitted on behalf of Perry by the American Psychological Association. Drawing on the breakthroughs he and his colleagues had made in the previous decades, the brief sent the message—clearly but with some subtlety, Wells hoped—that the 1977 ruling was a reasonable proposition then, “but we know much more today.” 

On January 11, 2012, the Supreme Court rejected Perry’s petition by a vote of eight to one. Justice Ruth Bader Ginsburg wrote the majority opinion as if eyewitness research had never been conducted. Sonia Sotomayor, the only justice who is a former prosecutor, cast the dissenting vote. In her opinion, she pointed to Wells’s work, scolding her colleagues for failing to recognize the power of the research that “cast doubt” on the Court’s 1977 precedent. 

 “In my rational mind,” Wells wrote after the decision, “I knew that the Perry case was not likely to yield any substantial solutions to the serious eyewitness identification problem in the legal system, a problem that I have been working on for 35 years. But my emotional side, aroused by the splendor and mystique of the majestic Hall and Courtroom, permitted me a temporary illusion of hope. … [As I exited], I wondered if the Court would wait another 34 years before they again took up the serious problem of eyewitness identification evidence. I hope not … I don’t think I can climb these stairs at the age of 95.”

Among the four amicus briefs filed on behalf of Perry, there was one from Jennifer Thompson. 

You may also like