Memory Researchers Push for Better Police Procedures to Protect Eyewitness Testimony

The 70 percent has a denominator worth naming. It is a share of convictions that DNA later overturned, not a share of identifications that witnesses made ^[1]. Read literally, it says mistaken identification was present in most provable wrongful convictions between 1989 and 2019, and that fewer than 30 percent of those cases turned on something else ^[2]. It does not estimate how often a witness who picks a face out of a fair lineup picks the wrong one. Identifications that led nowhere, and convictions with no biological material left to test, never enter that sample.

Each time a memory comes out of long-term storage it goes through reconsolidation, and new information seeps in and alters it ^[3]. The order of police procedures therefore does more damage than their content. John Wixted, a memory researcher at the University of California, San Diego, says a poorly done lineup is one of the easiest ways to spoil a witness's memory ^[8]. The sequence he describes is a showup, in which a single handcuffed suspect is presented and the witness says he might be the man, followed days or a week later by a six-person photo array containing that same face. By the time the array arrives, what the witness recognises is the showup ^[9]. "The memory has changed, and the witness doesn't know it," Wixted said ^[10].

Laura Mickes, a memory researcher at the University of Bristol, told Scientific American that memory is "like any other piece of forensic evidence." "If it's contaminated, it's not going to be good," she said ^[7].

Mickes is describing handling, and that sits awkwardly beside the same article's opening line, which calls human memory an unreliable witness at its core ^[17]. The two positions point at different failures. One indicts the instrument; the other indicts the chain of custody. The procedural reading depends on Wixted's confidence result: certainty expressed in court predicts accuracy poorly, while certainty expressed on a first fair photo lineup tends to predict it ^[11][12]. Scientific American reports the direction of that relationship without an accuracy rate for either condition.

People get about 40 percent of the details wrong about where they were on September 11, 2001 ^[5], which leaves roughly 60 percent right ^[6], and the report does not separate central details from peripheral ones.

Some of what degrades an identification is fixed before any officer arrives: distance from the crime, how well the scene was lit, whether witness and perpetrator were the same race ^[16]. Procedure is the part anyone can write down and audit, and memory researchers have produced evidence-based recommendations for police meant to avoid inadvertent contamination, though the best way to question a witness remains open ^[18]. Intoxication is one of the newer questions, according to Lilian Kloft-Heller, a forensic psychologist at Maastricht University ^[13]; some evidence indicates that witnesses who were drunk or high report fewer details while remaining accurate in the details they do report ^[14]. "What is most important is that they are questioned well," she said ^[15].