Select Page

This is the final post in a series about memories of eyewitness testimony. The first post described how memories are formed and remembered, while the second talked about how memories decay can be distorted and manipulated. This post will give some suggestions on how to increase the accuracy of information an eyewitness provides.



In the previous two posts on eyewitness testimony, we have seen how fragile and inaccurate memories can be, and how easily misinformation effects can take hold. John Dean’s testimony in the Watergate case was shown to have a large number of errors and inaccuracies after tape recordings of the conversations in question were uncovered. People forget about half of what they learn within one hour of learning it. Subtle changes to questioning can affect the memory (the estimated speed of a car involved in an accident increased by more than 20% when the question use the keyword “smashed” instead of “contacted”). False memories can be created out of nothing. A person’s confidence about the accuracy of their memory is not predictive of the accuracy of the memory. With all of this, one might wonder why legal cases would even bother with eyewitness testimony when it can so easily be corrupted.

Whilst it has been found that DNA and recorded audio/video footage are considered to be more trustworthy by jurors, eyewitness testimony still has its place in the legal system as “the next best thing” after these options. Ultimately, many cases will not have DNA or recordings of the events, so rely on memories of eyewitnesses.  However, in over 70 percent of the more than 360 cases in which a person convicted of a crime was later found innocent through the use of DNA testing, at least one mistaken eyewitness identification was involved, according to the Innocence Project.

As we saw in the previous posts, our brains often “fills in the gaps” of parts of events in memories that we may be uncertain about. For instance, studies have shown that people are more likely to recall seeing broken glass after being asked whether a car “smashed” into another, rather than “contacted”, and John Dean clearly remembered being asked to take a seat by President Nixon in a meeting, when tape recordings showed this never happened.  Additionally, external information can cause a misinformation effect by distorting a memory. For example, an eyewitness might see the events of a crime and store that information in memory, and then read in a newspaper article that another witness said the perpetrator was carrying a firearm. Even though the original witness didn’t perceive any weapon at the time of the events, that information may get implanted into their memory, without being consciously aware that this piece of misinformation did not come from their own experience.

In 2014, the National Academies released a report entitled “Identifying the Culprit: Assessing Eyewitness Identification” that detailed these same limitations with memory within the context of eyewitness testimony.  Additionally, they listed several important procedures that improve the accuracy of memories from eyewitnesses and are based on decades of research in memory, perception, and eyewitness testimony.

First, eyewitness identification has undergone radical changes in the past decades. In the past (and even present in many countries and jurisdictions), a group of individuals are presented simultaneously and the eyewitness is required to pick the relevant individual. This can be problematic in several ways. First, the eyewitness is very, very likely to identify the person that best fits their memory of the relevant individual – even if that person is not in the lineup. That is, the eyewitness is making a relative judgment (“this person looks most like who I remember out of everyone in the lineup”, rather than “this person looks like the person who I remember”). This is especially the case when the lineup is presented simultaneously, as the eyewitness can compare each member of the lineup to each other before coming to a decision. 

Second, it is possible for lineups to be biased: imagine a case where the main suspect of a crime is a tall man with blond hair. If the lineup contains one man somewhat fitting this description and the remainder being short men with dark hair, it’s highly likely that the blond man is going to be picked out as he is the only member of the lineup that even somewhat resembles the description. Third, it is possible that those administering the lineup can consciously or unconsciously bias the eyewitness if there is a particular person they suspect is implicated.  Lastly, in high-stress events like witnessing a crime, memory accuracy will also be affected and our brains will fill in the gaps to cover parts of the event that were not confidently remembered so that the memory is coherent, if not completely accurate. All in all, these practices only serve to distort memories, and therefore decrease the accuracy of eyewitness testimony and identification.

The National Academies report addressed each of these points, based on many years of scientific research. To address relative judgments, which predominantly occur when simultaneous lineups are used, research has found that sequential lineups increase eyewitness accuracy. That is to say when each member of a lineup is presented one by one and the eyewitness is told that they cannot go back to a previous lineup member, eyewitnesses are much more likely to accurately choose the relevant person. Additionally and importantly, accuracy from sequential lineups can be further improved in two more ways: first, police can present a “blank lineup” as an initial screening test, where the suspect is not present in the lineup at all. If an eyewitness picks someone from this blank lineup, then it is apparent that their memory of the event if not reliable and any testimony they provide should be disregarded.  Police can also explicitly instruct eyewitnesses that the lineup may not even contain the person of interest, so they do not have to choose anyone if they do not believe that they can positively identify a person.  Both of these additional procedures have been found to decrease misidentifications and increase accuracy.

To counter the lack of awareness of how memory works, it is recommended that courts bring in experts to advise juries about the ways that eyewitness evidence can fail, such as in the ways that this and the previous posts discuss. The report also urges the use of “double-blind” lineups, where the police officer administering a lineup does not know who the relevant person is, leaving them unable to offer any conscious or unconscious biasing cues to the eyewitness. It recommends that police officers record a witness’s confidence level upon first identifying a suspect — a time when their assessment of their own confidence is more likely to be accurate, and when their memory is less likely to have been contaminated.

Since the National Academies’ report was released in 2014, at least 19 states have passed legislation or adopted rules requiring reforms recommended by the report, such as recording witnesses’ confidence levels and using double-blind lineup procedures. The federal government has also made changes to lineup procedures in federal cases – in 2017 the Department of Justice released new guidelines for how lineups using photographs should be conducted, based on procedures recommended in the report. Even prior to this, the International Association of Chiefs of Police issued a new model policy for conducting lineups, also drawing upon the report.

To further improve eyewitness memory in the future, machine learning models are being employed to estimate the accuracy of a witness.  As a simple example, scientists can quantify the effects of specific distances and/or reduced lighting on human vision, and therefore models could quantify and factor in these characteristics and provide a numerical estimate of how accurate eyewitness testimony is likely to be. Indeed, a similar approach is already being successfully used in medical diagnosis and prognosis. 

This series of blog posts has introduced some basic concepts of memory and explained how fragile it is as it can very easily be distorted and manipulated. It is a very personal and subjective recollection of an event and is certainly not perfect like an audio or visual recording. Additionally, entirely false memories can be implanted simply through the power of suggestion. However, through more rigorous and science-based procedures, eyewitness memories can be highly accurate, which can give police and jurors more confidence that an event really did occur as an eyewitness claims it did.


Frueh, S. (2020). Using Science to Improve Eyewitness Testimony.