In the summer of 1984, a man broke into the Burlington, North Carolina, apartment of 22-year-old Jennifer Thompson-Cannino and raped her at knifepoint. Police met her at the hospital that same night, where they worked up a composite sketch based on her description of the attacker. Three days later, a detective showed her photos of six men, asking if she saw his face before her.
After a few minutes of studying, Thompson-Cannino managed to narrow down the field to just two. Finally, she settled on the picture of Ronald Cotton, also 22, who worked at a nearby restaurant. “This is the one,” Thompson-Cannino told the detective, according to her 2010 memoir. But she hedged a little, too. “I think this is the guy.”
The detective pressed her. “You ‘think’ that’s the guy?” he replied. “It’s him,” Thompson-Cannino answered. He asked if she was sure; this time, she said she was “positive.” When she asked how she’d done, he offered reassuring praise. “You did great,” he said.
Police lineups and photo arrays are mainstays of shows like “Law and Order,” which portray them as time-honored investigative techniques that confirm the results of diligent detective work. But the story of Ronald Cotton, featured on a 2009 episode of “60 Minutes,” is one of several cautionary tales discussed in a report on the inherent unreliability of eyewitness identification procedures. The report was commissioned in 2016 by the U.S. Court of Appeals for the Third Circuit—which, as its authors note, is the first federal court to formally tackle the subject.
The task force’s findings, which appear in the fall 2019 volume of Temple Law Review, offer a snapshot of the criminal legal system’s chronic over-reliance on eyewitness accounts. As the report notes, mistaken identifications have been involved in nearly 70 percent of post-conviction exonerations based on DNA evidence. In 2012, Supreme Court Justice Sonia Sotomayor called eyewitness evidence “a unique threat to the fairness of trial,” citing research showing that misidentifications are the “single greatest cause” of wrongful convictions in America.
Several days after she identified Cotton as her attacker in the photo array, Thompson-Cannino picked him out of a live lineup, too. Again, though, she hesitated, telling the detective administering the procedure that Cotton “look[ed] the most like” her rapist. Again, he asked if she was certain, and again, she replied that she was. The detective then revealed that Cotton was the same man whose photo she had selected previously, which allowed her to feel, as she later put it to PBS’s “Frontline,” “a huge amount of relief.” She had gotten it right, and would get the justice she deserved.
For this and another assault that took place that same night, Cotton was found guilty and ultimately sentenced to a pair of life sentences plus 54 years in prison. He would serve more than 10 years before modern DNA testing confirmed his innocence—and the guilt of Bobby Poole, who was already incarcerated alongside Cotton for a separate series of rapes in the area.
People who witness a crime may identify the wrong person as the perpetrator for a multitude of reasons, many of which relate to basic psychology, the authors of the report write: In the room with law enforcement, a nervous witness wants to be helpful, and may strain to deliver the answer they believe the authority figure standing next to them wants to hear.
Officers want just as badly to solve the case and may unintentionally provide suggestive clues about the suspect’s identity—a phenomenon known as the “expectancy effect.” Beginning a lineup with comments like “We got the guy and just need you to pick him out,” for example, can prompt an otherwise hesitant witness to hazard a guess.
If a witness has previously seen someone—in an earlier photo lineup, for example, or on social media—they might incorrectly believe they remember that person as the perpetrator. For this reason, the report’s authors recommend against including the same person in successive identification procedures, and urge law enforcement to discourage witnesses from doing their own amateur detective work online.
Police sketches, another staple of criminal procedurals, are another potential source of error. Research shows that people “typically process faces holistically, not feature-by-feature,” which means that a composite cobbled together from a witness’s description of individual features may end up looking nothing like the perpetrator. Yet police, with little else to go on, are likely to start looking for people who look like the composite—straying further away from the witness’s actual description, and perhaps ignoring promising leads in the process.
For the witness, the composite immediately becomes the most tangible representation of someone they may have only briefly glimpsed. During subsequent identification attempts, they may focus on a person’s resemblance to the sketch, as opposed to their recollections of the perpetrator. Because composite or sketch evidence is involved in about a quarter of DNA exonerations, the report authors recommend that law enforcement “only use them rarely and with great caution.”
The process by which Thompson-Cannino identified Cotton some 35 years ago was riddled with these kinds of compounding errors. Cotton’s face appeared in her photo array, and then again in her live lineup several days later. Poole, the man who committed the crime for which Cotton was found guilty, did not appear in the initial photo array or the subsequent lineup.
On both occasions, when she told detectives that she wasn’t especially confident, they pressed her until she offered more-assured responses. And after she implicated Cotton, she received immediate feedback that she had done the right thing.
Courts have long been at least somewhat aware of the shortcomings of these identification procedures. Whether police show a suspect in a photo array or alone, “it is obvious that risks of suggestion attend either form of confrontation,” Supreme Court Justice William Brennan wrote in United States v. Wade.
And as Brennan noted in a different case, Watkins v. Sowders, it is difficult for jurors—impassive observers with no firsthand knowledge of what took place—to be skeptical of an earnest witness who saw a crime with their own eyes and swears to tell the truth in court. Quoting renowned psychology professor and memory expert Elizabeth Loftus, Justice Brennan wrote that “all the evidence points rather strikingly to the conclusion that there is almost nothing more convincing than a live human being who takes the stand, points a finger at the defendant, and says ‘That’s the one!’”
By the time jurors file into the courtroom, though, it is far too late to warn them about all the different mistakes that can occur along the way. At that point, a suspect is an indicted defendant, and after confirming and reconfirming that the defendant is the responsible party, the witness has grown comfortable repeating this narrative to others. As a result, they may have formed a “memory” that never took place.
The physical evidence linking Cotton to the crime was thin, but Thompson-Cannino’s compelling, unflinching testimony proved to be enough to convince the jury of his guilt. After that conviction, Poole’s name surfaced in connection with the attack, and at a retrial obtained by Cotton’s lawyers, Thompson-Cannino even had the opportunity to face Poole in court and identify him as the true perpetrator. But in that proceeding, too, she remained adamant about whom she had seen that night, and whom she had not. “I was absolutely, positively, without-a-doubt certain [Cotton] was the man who raped me when I got on that witness stand,” Thompson-Cannino told ABA Journal in 2001.
“And nobody was going to tell me any different.”
There are simple strategies available to law enforcement for minimizing the risks of these errors, and the report goes through them in detail: For example, whenever possible, police should conduct double-blind lineups and photo arrays, where neither the witness nor the officer administering the procedure knows who the suspect is. If double-blind procedures are impractical—for example, in smaller departments where everyone knows the suspect’s identity—officers can at least use “blinded” techniques, in which the officer cannot see which suspect or suspects the witness is viewing at any given moment.
So-called filler lineup participants—the people asked to stand next to a suspect in a lineup—should match the elements of the witness’s description of the perpetrator, and not merely look similar to the suspect. The authors also caution against giving witnesses books of mugshots to browse, a practice that may result in witnesses over-committing to their initial identification and being less reliable during subsequent attempts. In 2001, New Jersey Attorney General John Farmer ordered law enforcement agencies to do away with “mugshot-searching” altogether and instead present witnesses with sequential, one-at-a-time lineups. The move, prompted by a U.S. Department of Justice report published two years earlier, made New Jersey the first state to embrace such a shift based on the evolving understanding of memory science.
Using standardized instructions can reduce the likelihood of tainting the proceedings, too. The report suggests that police issue a series of caveats and reminders before each identification attempt: that the suspect “may or may not be present,” for example, and that “it is just as important to free innocent people from suspicion as it is to identify the guilty.” In 2007, lawmakers in North Carolina passed the Eyewitness Identification Reform Act in an effort to modernize identification procedures throughout the state. The act prescribes a set of initial instructions to be given to witnesses, and requires that law enforcement either conduct identifications using independent administrators who are not involved with the case, or use blinding techniques that prevent administrators from knowing whose face the witness is looking at.
Once the witness has made an identification, the report’s authors add, police should take one final, critical step: Immediately ask witnesses for a self-assessment of their confidence, and record it without comment. Detectives should also refrain from giving any sort of suggestive feedback, even after the attempt is ostensibly “complete.” No matter what evidence later comes to light, it can be difficult to convince a witness who saw a detective pump his fist in triumph that the answer they gave was anything other than the right one.
Ronald Cotton and Jennifer Thompson-Cannino became close friends after his exoneration in 1995. They co-wrote a best-selling book, and continue to act as advocates for reforming the legal system’s uncritical treatment of eyewitness testimony. Cotton married, had a daughter, and bought a home using the $110,000 in restitution money he received from the North Carolina state government. It at first offered him only $5,000, but Thompson-Cannino lobbied on his behalf, urging lawmakers to ensure that he received a more generous settlement.
None of this, of course, erases what happened to Cotton, who spent a decade of his life behind bars for a crime he did not commit. But cases like his shine a badly needed spotlight on the risks associated with relying on eyewitness testimony—no matter how confident the person offering it may be—in order to lock people up.