A witness’s identification of a suspect can be trusted, according to the psychologist John Wixted, but only the first time it happens. Wixted, who studies memory and recognition at the University of California, San Diego, compares a witness’s memory to other forensic evidence: it can be contaminated. Tested early, carefully and only once, it holds up. Ask the same witness again later, and each further test can rewrite what they remember.
more
The doubt about eyewitness memory has a body count of its own. The Innocence Project, a nonprofit that works to overturn wrongful convictions in the United States, traced eyewitness misidentification to 262 of the 375 cases in which DNA evidence later cleared someone, 70% of the total.
Why a second look changes the first look
A police lineup is rarely a one-off. A witness often sees the same face, in a photo or in person, more than once as a case builds. Wixted, together with the psychologists Gary Wells and Elizabeth Loftus and the legal scholar Brandon Garrett, argued in a 2021 paper that this repetition is the problem, not the witness. Seeing an innocent person in one lineup can make that face familiar enough that a witness later points to them for the crime, mistaking recognition for memory of what happened. The recommendation, drawn from a white paper in the journal Law and Human Behavior, is blunt: test a witness’s memory of a given suspect only once.
What a test that counts looks like
Wixted sets out what that one test should look like. The lineup should hold a single genuine suspect alongside people the police already know are innocent, chosen for a similar appearance. The witness should be told the culprit might not be among the photos at all. The officer running the lineup should not know which photo shows the suspect, so they cannot steer the witness, even without meaning to. Most current research favours showing every photo at once rather than one after another.
Confidence matters too, but only if it’s caught at the right moment. A witness’s certainty at that first test predicts whether they are right. By the time of a trial, months or years later, it predicts nothing at all. Wixted’s research has already led courts to revisit convictions built on a witness who grew confident only by trial, having shown none at the original identification.
Where the reputation for unreliability came from
The idea that fear scrambles memory is not new. One of the earliest experiments on it, described in a 1908 essay by Münsterberg, was carried out by von Liszt at the University of Berlin, near the start of the twentieth century. It staged a life-threatening situation and found memories of the frightening, unexpected version more distorted and less accurate than memories of a neutral one.
Research since has built the case further: stressful events leave witnesses with worse recall of a perpetrator’s features, their actions and the details of the scene. After more than 40 years of work on memory and crime, that finding has moved out of psychology journals and into how legal proceedings treat a witness’s word. What the research still has not settled is a simpler question: whether an emotional event is remembered better than a calm one, or worse. Studies split roughly evenly both ways.
What the crime scene decides before the courtroom does
Some of the damage happens before a witness ever leaves the scene. What lands in memory depends on the lighting, the distance, how long the event lasted, and how much attention a frightened witness had left to give it. Attention is finite: someone who fixes on a weapon, an escape route or their own safety has that much less spare for the perpetrator’s face.
An unexpected, threatening weapon draws a disproportionate share of that attention, a pattern researchers call the weapon focus effect. It is not automatic. The researchers Fawcett, Peace and Greve found the effect smaller and less consistent outside the laboratory, and dependent on whether the witness expected to see a weapon at all. A 2021 survey of memory researchers found broad agreement that very high stress impairs accuracy, but real disagreement over how, once timing and the kind of memory being tested are factored in.
Why the fix keeps stalling
Wixted raised his case with Jed Rakoff, a federal judge for the Southern District of New York, at an online event run by the Association for Psychological Science. Rakoff was not convinced it would spread quickly. “Judges are very averse to changing their ways,” he said, pointing to the US legal system’s habit of letting past rulings govern new ones. He argued that real change was more likely to come from prosecutors, whom he called “the ones who will often be making the really operative decisions.”
The stakes are high: steep potential penalties help push about 97% of people charged with a felony in the United States to plead guilty, including some who are innocent. Even at trial, a repeated identification is not automatically thrown out. Federal rules of evidence admit anything judged relevant, and a lineup is usually excluded only for a clear procedural fault, such as an administrator steering the witness, not for falling short of what memory science recommends.
A 2025 review by Mickes, Wilson and Wixted reframes the argument: the major problem is often contamination, not an inherent inability to remember. Often the failure lies less in what a witness saw than in how many times, and how carelessly, somebody asked them to prove it.
