Skip navigation
CLN bookstore
× You have 2 more free articles available this month. Subscribe today.

Ending Eyewitness Memory Contamination

Memory-Expert Psychologists Recommend Stopping All In-Court Identification and Repeated Lineups

by Matt Clarke

The Heisenberg Uncertainty Principle is a scientific concept in quantum physics explaining that the position and velocity of a sub-atomic particle can never be truly known because the very act of measuring those quantities changes them. Now it appears that memory science has its own Uncertainty Principle, at least as applied to eyewitness identification. A group of world-renowned memory experts reported that exposing eyewitnesses to a comparison identification test—such as photos or a lineup—changes their memories making them less reliable in future comparison identification tests.

In a 2021 paper published by Psychological Science in the Public Interest, a scientific journal of the Association for Psychological Science, a group of world-renowned psychologists who study how human memory works recommended that eyewitness memory of crimes be tested only once in attempting to identify a perpetrator. This is because studies have shown that multiple viewings of lineups contaminate the witnesses’ memory leading to false eyewitness identifications. This includes the dramatic use of eyewitnesses to point to the defendant in the courtroom. For that reason, the paper recommends that eyewitnesses be shown a photo-array or in-person lineup only once and the practice of courtroom identifications be ended altogether unless the eyewitness has never previously been involved in identifying the defendant from a photo-array or lineup.

The researchers—John T. Wixted of the University of California, San Diego; Gary L. Wells of Iowa State University; Elizabeth F. Loftus of University of California, Irvine; and Brandon L. Garrett of Duke University—cited over 70 scientific studies of human memory spanning more than six decades and concluded there was a scientific consensus that memory is changed by subsequent events with the changes being incorporated into the original memory and the subject completely unaware that the memory has been altered.

“The first test is the most reliable test,” said Wixted, who has studied memory for more than 30 years. “The first test probes the witness’s memory but also unavoidably contaminates the witness’s memory. All tests beyond that very first one only serve to test contaminated memory and to contaminate it further. And once a memory is contaminated, there is no way to decontaminate it.”

The paper notes that eyewitness misidentification was identified as one of the causes of wrongful conviction in around 70% of the 375 cases tracked by the Innocence Project in which the convicted person was fully exonerated by DNA evidence. It seeks to explain why faulty memory causes eyewitnesses to misidentify innocent people as criminals. To do this, the paper describes the latest developments in memory science, including findings based on elaborative processing, signal detection theory, and source misattribution. The researchers note that memory contamination comes not only from police procedures to identify suspects but also when crime victims use news or social media to view a suspect’s image.

The problem begins when an eyewitness makes a decision about a face in a lineup (signal processing) and compares that face to the face of the perpetrator stored in memory (elaborative processing). The very act of elaborative processing creates its own memory of the face from the lineup so that the eyewitness will remember that face if encountered in future lineups even if the face was rejected in the initial lineup (source misattribution). Worse, the eyewitness will likely not even be aware that the familiarity of the face comes from the initial lineup and not from viewing the crime.

“Memory is malleable,” said Wixted. “And because it’s malleable, we must avoid repeated identification procedures with the same witness and suspect. This recommendation applies not only to additional tests conducted by police investigators but also to the final test conducted in the courtroom.”

Unfortunately, as the paper notes, there is a popular misconception that memory is unchanging and can improve with time. This leads police to subject eyewitnesses to multiple lineups and photo-arrays if the initial test of the eyewitnesses’ memories does not give the result the police desire. Thus, a person the police suspects of committing the crime may be included in an initial lineup or photo-array, which the eyewitness firmly rejects as not containing the suspect and also in later lineups or photo-arrays only to be identified by the eyewitness as the suspect due to memory contamination.

Researchers have shown that something as subtle as the phrasing of a question can change an eyewitness memory.

History of Eyewitness Identification Recommendations

The problems with eyewitness misidentification have long been known to scientists. In l998, the American Psychology-Law Society (“APLS”) commissioned teams of scientists to draw up recommendations for improving eyewitness identification procedures. Their four recommendations included: (1) having the person administering the lineup (whether in-person or by photo-array) be blind to the identity of the suspect; (2) telling the eyewitness that the suspect may not be in the lineup; (3) making sure the suspect does not stand out in the lineup; and (4) obtaining a statement of confidence level from the eyewitness before any feedback is given by police.

In 1999, the National Institute of Justice issued another set of science-based guidelines, which were similar to the 1998 recommendation, but included the recommendation that at least five fillers (non-suspects) be present in a lineup.

In 2013, a committee was appointed by the National Academy of Sciences to update the eyewitness identification test recommendations. Some of the new recommendations included training police in best eyewitness identification practices, conducting pretrial judicial inquiries into the reliability of eyewitness evidence, and video-recording line procedures. An important recommendation recognized the importance of the first identification test, noting that “[i]n-court confidence statements may also be less reliable than confidence judgments made at the time of an initial out-of-court identification; as memory fails and/or confidence grows disproportionately.” The recommendations also noted that eyewitness testimony is subject to contamination. This was the first recommendation to recognize contamination as a serious issue in eyewitness identification.

The final update of eyewitness identification recommendations came in 2020, when the APLS commissioned Wells et al. to incorporate the most recent scientific findings into updated recommendations. This resulted in nine recommendations, which include interviewing the witness before the lineup and issuing a warning not to attempt to identify the culprit using social media or other sources and a recommendation to avoid repeated identification procedures using the same witness and suspect.

Recognition Memory Theory

The current paper is an elaboration on why it is so critical to avoid repeatedly exposing an eyewitness to the same suspect in multiple identification procedures. The essential problem lies in the recognition procedure itself. Whereas witnesses tend to recall accurate information in other circumstances, the same cannot be said of repeated identification procedures involving comparisons of images to images recalled from memory.

The use of memory to recall and recognize something previously encountered can be described by five concepts: (1) encoding specificity, (2) similarity-based matching, (3) elaborative processing, (4) signal-detection theory, and (5) source attribution.

Encoding specificity refers to the cue that triggers memory to be recalled. For instance, an eyewitness should not be asked, “are any of these faces familiar?” That would cue a generalized search of memory for familiar faces, many of which would not be associated with the culprit and possibly one seen in a previous lineup had one occurred. Instead, the eyewitness should be asked, “do you see the person who committed the crime?” That cues memories specific to the witnessed crime and activates the relevant context (i.e., the face of the culprit) in the brain of the eyewitness.

Similarity-based matching refers to the process by which each face in the lineup is compared to the remembered face of the culprit. The more similar the lineup face is to the culprit’s, the stronger the memory-match signal (“MMS”) is. The problem is that eyewitnesses may simply identify the most similar face rather than rejecting it if it is not the culprit.

Signal detection theory postulates that the MMS is not a simple match or no-match (all or nothing) determination because any face in the lineup may share some characteristics with the culprit. Instead, it is theorized that MMS varies according to the strength of the similarity. In a fair target-absent lineup, none of the faces should generate a strong MMS. However, because “memory matching is an inherently noisy process,” sometimes a guilty suspect will generate a weak MMS or an innocent person will generate a strong MMS. “A troubling implication is that, even under ideal conditions involving no memory contamination and pristine testing procedures, and even on the initial test, misidentifications will inevitably happen from time to time.”

Elaborative processing describes the comparison between a face in the lineup and the remembered culprit’s face. This occurs whether an identification is made or not. “Decades of research have established that the more elaboratively a stimulus is processed, the more likely it is to be later remembered.” Thus, elaborative processing creates an incidental memory that is unintentional. This means that an eyewitness later shown a lineup containing the face of a person shown in a previous lineup may generate a strong MMS due to the previous elaborative processing of that face.           

The assignment of context to a memory signal is call source attribution. It is often difficult for a person to determine the source of a strong MMS upon repeated viewing of a face present in a previously-shown lineup. “Testing memory for the first time using a police lineup almost seems tailor-made for inducing a source misattribution when memory is tested a second time.” A second source of contamination occurs because the eyewitness, consciously or not, will know that police suspect a person of committing the crime if that person’s face is the only one to appear in multiple lineups. This renders the subsequent lineups inherently biased against that person.

Empirical Evidence for Misattribution of Sources

Scientists devised an experimental test to determine whether subsequent pattern-recognition memory tests cause source misattribution. They showed test subjects a mock crime and, 90 minutes later, showed them a 15-mugshot lineup, which did not contain the culprit. Two weeks later, they were shown a lineup and asked whether the culprit was in it.

Neither lineup contained a photo of any person involved in the mock crime. The rate of falsely identifying a person as having committed the crime was 8% for lineups that did not contain a mug shotseen in the previous lineup but 20% for those that contained a mugshot that was also in the previous lineup.

In another experimental study, participants who viewed a video of a person entering an office were either shown 50 mugshots of persons generally similar to the person who entered the office or shown no mugshot, then viewed a target-present lineup 48 hours later. Those who viewed the mugshots and chose one of them as the perpetrator were more likely to identify a person from the mugshots than the actual culprit from the video (70% compared to 8%). If the previously-seen photo was not one the eyewitness picked from the mugshots, the participant was still more likely to identify that person as the perpetrator than the actual perpetrator (28% compared to 18%), and if the target was not in the lineup, the misidentification rate was 38%. This shows that the mugshots contaminated the memory of the eyewitnesses at an alarming rate, and they were unaware of the contamination that caused the misidentifications.

The Effect of Memory Contamination on Confidence

A related issue to misidentification is the effect contamination has on confidence. A shaky lineup identification often morphs into rock-solid certainty in a courtroom identification. But why? The most obvious answer is that repeated viewing of the same person in multiple lineups or simply seeing who is sitting next to the defense attorney tells the eyewitness who the police (and/or prosecutors) believe committed the crime, bolstering witness confidence. Sadly, the eyewitness’ reaction at the initial lineup is often not recorded or misrepresented if the police do not like it. For instance, the statement that, “this one looks the most like the culprit,” becomes “the witness positively identified the suspect” when, in fact, the statement was exculpatory, saying none of them were the culprit, but that one was the most similar.

Two Real-Life Examples

When a 74-year-old woman who had been raped was shown a photo-array, she said that she was “almost positive” John Jerome White was the culprit. She had described her attacker as a Black man who was “well built” and had a “round face.” White was not and did not. Nonetheless, the woman picked White out of a subsequent five-person lineup, and at trial, she was certain it was him. White was the only person who was in the photo-array and lineup.

After spending 22 years in prison, White was exonerated by DNA testing in 2007. Astonishingly, the same DNA testing showed that the man who actually committed the rape had been in the five-person lineup (and was, in fact, well-built and round-faced). He was not identified by the victim because the victim’s MMS of White from being in the photo array was stronger than her MMS of the culprit from her memory of the crime.

There was a similar misidentification by a rape victim who, after viewing a lineup, said the photo of Steven Gary Titus was “the closest one.” This indicates low confidence, not high confidence. However, when she identified him in trial in 1981, she expressed high confidence. A few months after Titus was convicted, an investigative journalist with the Seattle Times assisted police in developing evidence pointing to another suspect. When shown the suspect’s photo, the victim broke down crying and said, “Oh, my God, what have I done to Mr. Titus?” But the police and prosecutors were at fault for the wrongful conviction, not the victim. They ignored her low-confidence initial statement and used her contaminated memory to obtain a high-confidence in-trial identification. Titus was finally absolved of the charge but died of heart failure in 1985, days before he was able to be vindicated by a civil suit in which his family won a $2 million settlement.


Contamination of eyewitness memory is literally a matter of life and death, especially when it comes to death-penalty cases. The researchers identified at least one with suspected contaminated eyewitness memory and a possibly innocent person sitting on death row awaiting execution, having exhausted all appeals. This urgency led them to recommend that an eyewitness’ memory be tested only once using the same suspected person. In short, “courts should per se exclude courtroom identifications if there was a prior identification, but they should sometimes exclude out-of-court identifications,” if improper procedures were used.  


As a digital subscriber to Criminal Legal News, you can access full text and downloads for this and other premium content.

Subscribe today

Already a subscriber? Login



Prison Phone Justice Campaign
Advertise Here 4th Ad
The Habeas Citebook: Prosecutorial Misconduct Side