Why Eyewitness Testimony Is Less Reliable Than Jurors Think

On July 29, 1984, a man broke into Jennifer Thompson's apartment in Burlington, North Carolina, and sexually assaulted her. She identified Ronald Cotton in a photo lineup, again in a live lineup, and again in court. She was sincere, and she was wrong. In 1995, DNA testing excluded Cotton and matched another man, Bobbie Poole, who then pleaded guilty. Cotton had served more than 10 years in prison.

Cases like Cotton's are the most common pattern in the DNA exoneration record. The Innocence Project found eyewitness misidentification in 69 percent of the 375 DNA exonerations in the United States through 2020. Here is what memory research shows about why confident witnesses get it wrong, and what courts and police have changed.

How the Cotton identification was built

The details of Thompson's identification show how a memory can firm up after the fact. According to the Innocence Project, when she first looked at the photo lineup she said "it is between number four and number five." After looking again, she chose number five, Cotton. After his arrest she picked him out of a live lineup. A second woman attacked in the same neighborhood that night viewed the same lineups and did not identify anyone. At Cotton's second trial in 1987, that woman did identify him, saying she had been too afraid to do so earlier. When Poole was brought into the courtroom outside the jury's presence, both women said he was not their attacker.

After Cotton's release, he and Thompson met and became friends. In 2009 they published a book together, Picking Cotton, and they have spoken widely about changing how police conduct identifications.

Memory can be rewritten

Memory is not a recording. In a 1974 study, psychologists Elizabeth Loftus and John Palmer showed students a film of a car accident and asked how fast the cars were going. The verb in the question changed the answer. Students asked how fast the cars were going when they "smashed" into each other estimated 40.8 miles per hour on average. Those asked about cars that "contacted" each other estimated 31.8.

In a second experiment, a week after watching a crash film, students were asked whether they had seen any broken glass. There was none in the film. Sixteen of the 50 who had been asked the "smashed" question said yes, compared with 7 of the 50 asked the "hit" question. One word in an earlier question had added a detail to what they remembered.

Confidence grows after the fact

Witnesses often become more certain over time, not less. Law professor Brandon Garrett, writing in Judicature in 2020, noted that eyewitnesses in wrongful conviction cases often testified with complete certainty at trial after showing a lack of confidence when they first made an identification. By the time of trial, he writes, an eyewitness's artificially inflated confidence complicates the jury's job of judging credibility.

That is why research has shifted attention to the first identification. In 2014, the National Academy of Sciences published Identifying the Culprit, a report on eyewitness evidence. It concluded that science has provided an increasingly clear picture of the limits of human perception and memory that can lead to identification errors. It recommended standardized lineup procedures, recording the witness's confidence at the time of the identification, video recording the procedure, and better jury instructions and expert testimony about memory.

The report also called for double-blind lineups, run by an officer who does not know which person is the suspect, so the officer cannot signal the answer even by accident. Garrett describes blinding as central to the scientific method because it reduces the risk of unintentionally biasing the result.

Faces of another race

People are generally less accurate at recognizing faces of a race other than their own. Researchers call it the own-race or cross-race effect. A 2001 meta-analysis by Christian Meissner and John Brigham, published in Psychology, Public Policy, and Law, found that own-race faces were more often correctly recognized and less often falsely identified than other-race faces. In the Innocence Project's DNA exonerations, 42 percent of the misidentification cases involved a cross-racial identification. In the Cotton case, Thompson is white and Cotton is Black.

How courts have responded

For decades, the main federal test has come from Manson v. Brathwaite in 1977. It asks whether a police identification procedure was unnecessarily suggestive and, if so, whether the identification was still reliable under five factors. In 2011, the New Jersey Supreme Court, in State v. Henderson, went further and endorsed detailed jury instructions explaining how memory works. Garrett notes that later studies found such instructions made jurors more skeptical of eyewitnesses but did not make them better at telling accurate identifications from mistaken ones.

When the first look fails

Even in a case where the right man was eventually identified, the first look can fail. On July 14, 1974, people at Lake Sammamish State Park in Washington saw a young man with his arm in a sling who introduced himself as Ted and asked for help with a sailboat near a tan or bronze Volkswagen Beetle. Yet the Lake Sammamish witness detectives considered most reliable failed to identify Ted Bundy from a photo lineup later that year.

In Utah, after Bundy's arrest in August 1975, Carol DaRonch picked him out of an October 1975 lineup as the man who had called himself Officer Roseland and tried to handcuff her. A judge convicted Bundy of kidnapping her in a 1976 bench trial. The same kind of evidence that failed in Washington held up in Utah.

An eyewitness who is certain is usually not lying. That is what makes misidentification so hard to catch. The witness believes it, the jury believes the witness, and the procedure that produced the certainty is often invisible by the time of trial. The reforms that hold up in the research all focus on the first identification: run blind, recorded, and taken down before anything else can shape it.

Sources

1. Ronald Cotton (Innocence Project case page)
2. DNA Exonerations in the United States (1989-2020) (Innocence Project)
3. Identifying the Culprit: Assessing Eyewitness Identification (National Academies, 2014)
4. Judging Eyewitness Evidence (Brandon Garrett, Judicature, Vol. 104 No. 1, 2020)
5. Loftus and Palmer (1974): Reconstruction of automobile destruction (WJEC study summary)
6. Ted Bundy (Wikipedia)

Found an error? Tell us. Corrections are dated and listed on the corrections page.