Eyewitness Testimony, Eyewitness Mistakes: What We Get Wrong

What kinds of eyewitness mistakes send innocent people to prison?

Posted Aug 21, 2020

Source: Openclipart

In a previous Forensic View (Sharps, 2020), we discussed the fact that eyewitness memory is not a solid record of the past. Eyewitness memories, like other memories, change—memories become shorter, less detailed, and reconfigured in the direction of what we believe we saw, rather than what we actually observed. 

But what kinds of mistakes do these eyewitness processes create? My students and I have researched this, and a lot of what we found was very surprising (e.g., Sharps, 2017). 

In earlier Forensic View posts, we’ve seen that high states of arousal are bad for our memories, but that it’s usually ethically impossible to reproduce those states in the laboratory. So, we don’t try—we run our studies under ideal conditions of lighting and comfort for our research volunteers, in the knowledge that in the street, under violent conditions and frequently poor visibility, what we’ll see in real-world cases will generally be worse.

But even under ideal conditions, things are not exactly, well, ideal. In collaboration with law enforcement colleagues, we’ve created images of realistic crime scenes. The one we’ve used the most is a scene with a “suspect” apparently aiming a gun at a “victim” (the safety precautions used in these studies, conducted under law enforcement supervision, are epic. The "aiming" effect, while realistic, is accomplished with trick photography). 

Anyway, using this realistic scene, we find that the most common eyewitness error is hardly surprising at all. The average witness most commonly makes errors of suspect appearance, clothing, and physique. Blonds become brunettes, the tall become short, grey t-shirts become blue sweatshirts, and so on. Not surprising—but reflect on the fact that these results are obtained under ideal conditions of comfort and lighting, and with plenty of time to observe the scene. Under these ideal conditions, the average witness still manages to make approximately two errors of perpetrator or suspect appearance. That’s a lot of mistakes. 

Suspects are typically arrested based on their appearance as reported to law enforcement, which means the suspect is usually similar, physically, to the person who actually committed the crime. If a primary witness makes two mistakes, once that suspect is in custody, you can see how a lot of the wrong people could wind up charged with the wrong crimes. So, the innocent would be in custody, and the guilty would still be out there, available to commit more crimes if they felt like it. 

Nobody, including law enforcement, wants this to happen; but we’ve shown that under ideal conditions, witnesses make enough mistakes to create wrongful identifications. You can imagine what this means under conditions of darkness, street clutter, and emotional arousal on the part of the witness; but for the moment, we can see that the most common eyewitness error type is suspect appearance. Granted, on average, we only get about two errors of this type; but that’s more than enough.

In our studies, people made other kinds of errors, of course. Environmental details, suspect race or sex, the types of weapons observed; all of these could have real significance in an investigation or in court. But the real surprise for us was the second most common type of error. The average person, under ideal viewing conditions, made one and one-quarter errors of the imagination. In other words, people just made things up. Without knowing they were doing it.  

Some of these errors were truly amazing. People remembered the intentions or emotional state of the suspect, when there was no clue to either of these in the image. People created little biographies of the suspect and victim, based on, well, nothing. And some of the weirdest errors really made no objective sense at all.

You see, we used a photograph of a crime scene, and we asked our witness volunteers what they remembered. Nobody asked witnesses to predict what might happen, or to tell us a story—we were very clear that they were to report only their memories of the scene. And then we asked them what they remembered happening next.

There’s only one reasonable answer to that question. “Dude, nothing happened next. I was looking at a photograph.”

But people told us. They gave us their accounts of memories that had no bases at all.

They remembered what happened after the action depicted in the photo. They remembered the suspect grabbing the victim and climbing onto a roof. They remembered the suspect escaping in a car. My personal favorite remembered that the victim “turned and faced (the suspect) in a violent manner,” and gave a specific estimate of the distance between the victim and her attacker.

And it didn’t happen at all. 

Our witnesses were not people with impairments. They were normal people, of the type who witness crimes every day, and who must report them to the courts and to law enforcement. And on average, they made one and one-quarter mistakes per witness in which they simply made things up. Things that made no sense whatsoever. 

This is certainly not to belittle the persons involved; frequently, they reported exactly what they believed that they saw.

The only problem is that they were frequently completely wrong. Again, the average witness made something up out of whole cloth, at least in our experimental situation, about one and one-quarter times per account.  

We also do not wish to belittle eyewitness testimony; frequently, it’s the only evidence available. But it’s also very important for law enforcement, court personnel, and jurors to be aware of the limitations of eyewitness evidence, and of its potential to mislead a given investigation or court proceeding. 

And it’s not only the original observation, on the part of a witness, that can produce these problems. In the next Forensic View, we’ll examine how our own minds can contribute to false memories in the eyewitness realm. 


Sharps, M.J. (2017).  Processing Under Pressure: Stress. Memory, and Decision-Making in Law Enforcement (2nd ed.).  Flushing, NY: Looseleaf Law.  

Sharps, M.J. (2020, May 8).  You May Question the Witness; And You Probably Should.  The Forensic View, Psychology Today Blog.