Skip to main content

Verified by Psychology Today

Eyewitness Imagination: How Our Minds Change Our Memories

Eyewitness memory can be inaccurate. But can it change itself?

Source: Openclipart

In our last Forensic View, we saw that typical eyewitness errors follow an interesting pattern. The most common eyewitness errors are those of suspect appearance, which is not surprising.

However, the second most common errors are those of the imagination. This is more surprising, and a lot scarier. The frequency of imaginative errors was far beyond those of any other error type (errors of suspect race or sex, or of weapons used, or of important elements of the physical environment of a given crime). Far beyond error rates in any of these important areas, people simply made things up. And they had no idea they were doing it.

How is this possible? Well, it has directly to do with how memory works, rather than with how most people think memory works.

As we’ve discussed in previous Forensic Views, people tend to think of memory as immutable, as some kind of essentially accurate recording we make in the brain (accurate except for the bits that drop out completely, bits which we term “forgotten"). But as we’ve also discussed (Bartlett, 1932; Sharps, 2017), psychologists have known that this popular view is incorrect for most of a century. Bartlett showed that memories are not static, but that they reconfigure in three specific ways: they become shorter, they coalesce around the gist of what actually happened, and they change in the direction of personal belief. This is the really scary one: we tend to remember what we believe happened, not what actually happened.

This fact throws the importance of forensic cognitive psychology into sharp relief. What we remember is not only dependent on physical reality. It also depends on what we think happened in a given situation; and, as we will see, the resultant inaccuracies can be further exacerbated by the nature of the memory process itself.

In research in my laboratory, eyewitness performance is frequently awful. For example, under ideal viewing conditions, people were right less than half the time when they tried to identify a gun they saw earlier, and right less than a quarter of the time when they tried to identify a car (see Sharps, 2017, for review). So, given this genuinely horrible level of eyewitness performance, we began to explore possible sources of eyewitness error. We began to ask: how is eyewitness memory actually elicited in a criminal investigation?

In any investigation, witnesses are going to be asked for their stories repeatedly. Officers at initial contact will ask questions. Then will come sergeants, then detectives, then more detectives, then district attorney investigators, then an assistant district attorney, then a public defender investigator, then a public defender, then a prosecutor in court—a given witness may have to think about his or her memories of a criminal situation many times, not even counting the number of times he or she thought about these issues alone, or recounted them to spouse, parents, or friends.

But why would this matter?

We conducted an experiment (Sharps et al., 2012) in which we did the same type of repeated questioning to which the criminal justice system would subject any witness. We used an experimental scene we have used many times, in which a “suspect” appears to be aiming a firearm at a “victim.” We exposed 92 people, under ideal viewing conditions (generally much better than those of a typical eyewitness situation) to this scene, and we asked them what they saw. We did this repeatedly, as occurs in the real world; our first request for information was followed by three more requests, for the ostensible purpose of adding any additional details the witnesses might recall.

Now, recall that Bartlett showed that memories are not static recordings, but are reconfigured in the directions of gist, brevity, and personal belief. Also, please recall that in our research, errors of the imagination, rooted not in the real world but in the activity of the mind itself, were the second most common error type (see Sharps, 2017).

Our witnesses of course gave accounts of the criminal event they saw initially. But what’s important to realize here is that every time eyewitnesses give an account of a crime, their recounting of that crime is also an event, subject to the same laws of Bartlett reconfiguration.

This means that every time you tell somebody about something, your retelling, as an event, may influence future accounts. Your imagination today can influence your imagination tomorrow.

But does it?

In our experiment (Sharps et al., 2012), our first request for information resulted in a true/false ratio of 3.56; in other words, we got about three and a half true statements for every false one. Not great, obviously, but not too bad.

However, our second ratio was a lot worse, with only 1.39 correct facts for every false one. The fourth question gave us similar results to the second; people were coming in with almost as many false statements as true ones. But on question 3, things were a whole lot worse: we got more false statements than true ones!

Repeating the story resulted in more and more errors, exactly as we’d predict by seeing repetition not as a passive act, but as an active contributor to inaccuracy.

Memories are not static recordings; well and good. But it turns out that memories can actually change themselves, simply by the repeated act of being remembered! And as a result, memory itself can alter the later memories recounted by a given eyewitness; memories on which a completely erroneous conviction of an innocent person may be based.

This is yet another demonstration of the great importance of an understanding of psychology, including forensic cognitive psychology, to the effectiveness and fairness of the criminal justice system.


Bartlett, F.C. (1932). Remembering: A Study in Experimental and Social Psychology. Cambridge: Cambridge University Press

Sharps, M.J. (2017). Processing Under Pressure: Stress, Memory, and Decision-Making in Law Enforcement. Flushing, NJ: Looseleaf Law

Sharps, M.J., Herrera, M., Dunn, L., & Alcala, E. (2012). Recognition and Reconfiguration: Demand-Based Confabulation in Initial Eyewitness Memory. Journal of Investigative Psychology and Offender Profiling, 9, 149-160.