The Unreliability of Eyewitness Testimony

Interesting article:

The reliability of witness testimony is a vastly complex subject, but legal scholars and forensic psychologists say it’s possible to extract the truth from contradictory accounts and evolving memories. According to Barbara Tversky, professor emerita of psychology at Stanford University, the bottom line is this: “All other things equal, earlier recountings are more likely to be accurate than later ones. The longer the delay, the more likely that subsequent information will get confused with the target memory.”


Memory is a reconstructive process, says Richard Wise, a forensic psychologist at the University of North Dakota. “When an eyewitness recalls a crime, he or she must reconstruct his or her memory of the crime.” This, he says, is an unconscious process. To reconstruct a memory, the eyewitness draws upon several sources of information, only one being his or her actual recollection.

“To fill in gaps in memory, the eyewitness relies upon his or her expectation, attitudes, prejudices, bias, and prior knowledge. Furthermore, information supplied to an eyewitness after a crime (i.e., post-event information) by the police, prosecutor, other eyewitnesses, media, etc., can alter an eyewitness’s memory of the crime,” Wise said in an email.

That external input is what makes eyewitness testimony so unreliable. Eyewitnesses are generally unaware that their memory has been altered by post-event information, and feel convinced they’re recalling only the incident itself. “Once an eyewitness’s memory of the crime has been altered by post-event information, it is difficult or impossible to restore the eyewitness’s original memory of the crime,” Wise told Life’s Little Mysteries.

Posted on June 4, 2012 at 6:36 AM33 Comments


Kevin an Auditor June 4, 2012 8:00 AM

I’m surprised Bruce is citing “Scientific American” – It’s not what it once was.

I’m an auditor in private industry. I also investigate “incidents of extrodinairy loss” and have a PI license in two states. I conduct a lot of “interviews” (o.k., they’re interrogations, but the witness can always say “I resign”, and walk out – they often do). Something the interviewer must be certain of is that they themselves are not the source of erroneous “memories” (or statements, anyway). I am certain that this a large part of the false confession problem with our police.

Firstly, I don’t lay out the complete set of data that implicates the witness. I offer them the opportunity to explain anomolous behavior/circumstances in the course of narrating the events and procedures under discussion. This has let some probably innocent people off the hook, because, although they were not following procedure, they had a very sensible (from their point of view) reason for doing so. (The supervisor who told them to do it that way, on the other hand…)

Secondly, regarding seemingly trivial points that actually constitute very damning actions, I make a point of letting the witness make their statement, and then deliberately mis-state it back to them (i.e. WITNESS: “so I put all the checks in the envelope for Al to count and he put them in the safe” ME: “You counted the checks and put them in the safe?” I let the witness correct me on points I don’t want ever questioned again. Addionally, the witness is sometimes frustrated that I don’t seem be listening, and, in addition to correcting me, adds to the story (digging themselves deeper). I always try to conduct interviews as soon as possible after incidents, but have no means of locking up people to prevent them from discussing it with other people. On the other hand it allows colluders to get their verbatim stories straight, and the only thing more suspicious than “innocent” people who can give a minute by minute account of what they did three days ago is a group of three people who can do it and agree on every point! (the first video/login/receipt that shows them wrong fires them all – yes I get written statements)

This is my first post on Bruce’s most excellent blog. I only check it twice a week, and seem to miss topics I could add to in a timely manner.

vasiliy pupkin June 4, 2012 8:18 AM

Memory could be contaminated as Prof. Loftus stated as other types of evidence.
There is no 100% reliable type of evidence in the same way as no 100% security. It is always trade off. Unfortunately, the price it too high for errors in criminal cases when innocent person is placed behind the bars for many years (or executed) for crimes were not committed, but just attributed to that person due to currently unreliable mechanism of crimal justice.
I guess future of more reliable proving who is guilty/innocent is not within court procedure at all, but rather through other objective methods of truth verification: e.g. fMRI and other brain studies for evaluation of validity of witness/victim’s testimony.

Petréa Mitchell June 4, 2012 11:18 AM

“This is why, though the camera can lie, at least it doesn’t change it’s mind much over time….”

Bah. A few hours with Mr. Photoshop, and the camera will say anything you want it to.

T June 4, 2012 11:54 AM

@vasily From a little personal experience with studying the brain via fMRI and similar methods that track neurological activation and traffic – these solutions aren’t exactly foolproof.

Some (as yet undetermined, but I believe statistically significant) percentage of the population generate either erratic sequences or total false positives (truth blackouts). That is to say, that when they’re telling the truth, the process is (presently) indistinguishable from that of an intentional deception. These subjects use the same neurological sequences and activations to tell the truth that other people use to lie. That could well be related to memory being a process of perception, recognition, and reconstruction.

It might even be possible to communicably teach the technique, if you have access to the equipment.

Likewise it seems possible for a person to freshly observe an incident, and then describe it radically differently to how it objectively occurred and still pass a neurological test for “truthiness” with flying colours. Memory isn’t only mangled during warehousing and recall, but also during the process of recording and stowage.

The answers aren’t always inside the skull.

Petréa Mitchell June 4, 2012 12:57 PM

“Likewise it seems possible for a person to freshly observe an incident, and then describe it radically differently to how it objectively occurred and still pass a neurological test for ‘truthiness’ with flying colours. Memory isn’t only mangled during warehousing and recall, but also during the process of recording and stowage.”

Yes, indeed. Memory is never a recording to begin with, just a sparse collection of (possibly erroneous) data points from which we interpolate something that feels like a recording.

Furthermore, if the memory changes later, the witness will still believe they are telling the truth when reporting the edited memory.

Jason Wright June 4, 2012 1:35 PM

The article is right on point. What should remain clear however is that “eyewitness tesimony” is roughly equivalent in reliability to phrases like “according to sources close to the event” or “unverified” photos, videos, etc., and that it nonetheless drive news coverage and public opinion. Unless there is concrete proof, verified via more than one reliable channel then it belongs more appropriately on a gossip page than as a basis for any major decisions — be it a life or a war.

If we treat heresay and compelling stories as truth in and of themselves, we deserve the world we pull over our own eyes. Anybody still remember Nayirah (aka al-Ṣabaḥ) and the testimony to congress regarding Iraqi soldiers, babies and incubators?

Trust but verify.

Figureitout June 4, 2012 1:35 PM

“This is why, though the camera can lie, at least it doesn’t change it’s mind much over time….”
Bah. A few hours with Mr. Photoshop, and the camera will say anything you want it to.

Tampering with photos has been going on for a while now it seems =>

–Almost like cryptanalysis, I’m sure “vetting” photos is a mind-numbing process.

Memory is never a recording to begin with, just a sparse collection of (possibly erroneous) data points from which we interpolate something that feels like a recording.

–Though extremely rare..And then there were savants with eidetic memory (Stephen Wiltshire, Kim Peek, Daniel Tammet)

OT: Thanks Bruce for not mentioning the “Trayvon-Zimmerman” aggregating distraction, sensationalizing one story and completely manipulating the case with the national media; AKA “prolefeed”

Erich Schmidt June 4, 2012 1:35 PM

I have observed people erroneously relating events from only minutes or hours prior, seeming unconsciously and/or unintentionally. From my somewhat limited and non-professional research into this you’ve got to watch out for personality issues that can interfere with memory or recall, not sure which. It seems it can be too painful for the person to believe what actually happened. Very weird to be confronted with.

Paul Collins June 4, 2012 1:54 PM

After reading The Invisible Gorilla, I’m skeptical that any memory or perception can be relied on to “extract the truth.” Perception can be significantly wrong, let alone memory of it.

Doug Coulter June 4, 2012 2:36 PM

I’m very aware of this but perhaps from a slightly different angle, as a scientist, philospher, and engineer.

Nearly everyone does not “record” events – I agree with that statement above. What they “record” is their filtered impressions of an event. “He made me mad” vs “He said these words”, for example. I’m sure most can think of more than they’d like me to type in for them to read, this is real common and should be obvious.

This is to save storage space – it’s lossy compression. I call it lazy-brain syndrome, or part of that syndrome. People want “what to think” and are in general too lazy to learn “how to think”.

It’s real work to train yourself to just record objective data – something I had to learn to be an extra-effective technical type. All the post analysis in the world of “impressions” vs “what was really going on” is useless or nearly so. GIGO is another common term for this problem.

This is also why paradigm shifts (20c!) or taking up a new religion are so often very difficult. When all you have is the interpretations of your life-experience stored, some new understanding that invalidates that is like losing that whole part of your life. Whereas, if you’ve stored the raw data, you could always re-analyze it in the context of the new understanding, and not only not lose anything, but gain.

An ancillary point is that this does not appear to be totally fixed in our hardware, but can be learned. Few bother – it’s real work and it’s ongoing work, just like learning how to think vs letting some talking head tell you what to think.

So, the unreliability of eyewitness data is in some measure related to how they approach their own memory discipline – with the huge bulk not recording “raw data” for later reprocessing.

Captain Obvious June 4, 2012 2:38 PM

It’s pretty easy to tell if a witness is lying about the details…they say anything other than “I don’t remember.” Whatever is rehearsed with the attorney is far more memorable than what the witness may have seen.

Bryan W June 4, 2012 3:13 PM

@Captain Obvious

I’m pretty sure that former Attorney General Alberto Gonzalez is not widely seen as a liar for the few memories he bestowed in congressional testimony, but for averring that “I don’t remember” more than 70 times.

Rajesh June 4, 2012 3:29 PM

The man can try to become perfect with time in his acquired skills. But the Time can prove his senses imperfect sometimes. Its all about how we take it. How we have seen is more important
than what we have seen. Here in this case it becomes the most important how someone is being interpreted..

NobodySpecial June 4, 2012 3:30 PM

@Captain Obvious – that’s the problem with jury trials and lawyers.

A perfect, honest and reliable perfect witness would say “as far as I remember” and “I can’t be certain but”…. which would be torn apart on the witness stand.

Chris Lawson June 4, 2012 4:43 PM

@Kevin the Auditor,

I think you’re talking about a very different situation to the average police investigation. From what you said, you seem to be interviewing suspects rather than general witnesses, and you have the advantage of usually having very reliable forensic evidence of wrongdoing in the form of signed checks, money transfers, ledgers, etc.

There’s nothing wrong with exploiting these advantages, but I think you’re generalising them to other scenarios.

DoctorT June 4, 2012 4:47 PM

The facts that eyewitness accounts are moderately unreliable soon after an event and get worse with time have been known for centuries and studied by psychologists for decades. The claim that “legal scholars and forensic psychologists” can “extract the truth from contradictory accounts” is self-promotional bullshit. They apparently hope that experts such as themselves will be brought in every time a trial hinges on the testimonies of eyewitnesses. That will be just as reliable as the efforts of “memory recovery” experts to help people remember (eg: completely invent) past horrors.

Quirkz June 4, 2012 5:03 PM

@ Doug – “All the post analysis in the world of “impressions” vs “what was really going on” is useless or nearly so.”

I see this a lot. Can’t count the number of times I’ve observed what appeared to be a simple disagreement between two people, but in the retelling one side indicates they’re sure their opponent said everything with a sarcastic snarl, a malicious gleam in their eye, and perhaps while waving a knife and kicking a puppy. Something like “they disagreed with me” can turn into “they were mean to me” very quickly.

Of course, I’m not immune. I remember one incident where I retold a funny story back to the person who had told it to me just a week or two before, but I apparently mangled nearly every detail in the process.

Jim B. June 4, 2012 8:26 PM

If we are talking traumatic events, from what I understand, memories are far more accurate a few days later because the immediate stress response significantly impedes recall. The memories then degrade from then.

tensor June 5, 2012 1:09 AM

“It takes years to consolidate a memory. Not minutes, hours, or days but years. What you learn in first grade is not completely formed until your sophomore year in high school.” — Dr. John Medina,

He also claims your brain can process only a tiny amount of the information your senses provide to it during each second. This may be one reason several persons who were eyewitnesses report radically different accounts of the same scene — they actually did see it differently, processing different selections of the same visual information they all received.

Clive Robinson June 5, 2012 3:00 AM

This adds little (other than academic gloss) to that which has been known for ages.

I’ve made the statment before that there is no such thing as truth only perspectives, and there are as many perspectives as there are witnesses (be they participents, bystanders, cameras or other imperfect method of recording).

A few simple differences between humans are their height, visual accuity, colour and sound perception. For instance I might see a dark blue coat you might see dark green or black even under near perfect lighting conditions. And as we know near perfect lighting conditions are almost as unlikely at a crime scene as getting the Euro Millions winning ticket. Sound is a funny thing, because learning to speak generaly effects the way we hear for the rest of our lives… It is known that people who are brought up with a pitch dependent language are more likely to be able to accuratly tune an instrument “by ear” than those whose language is pitch independent. We also know that certain languages that do not contain certain sounds cause the speakers in later life difficulty in pronouncing names (my own having a couple of consonants not universaly spoken). There have been studies carried out that litters of kittens brought up in a visually simple environment with only horizontal lines for the first few weeks of their life when put in another visual environment where there are physical vertical bars the don’t see them and walk into them.

So it can be said that even if a human did have perfect recall of what they saw they possibly could not see it accuratly and many details (such as height) would be reported incorrectly.

But a little test for everybody, just try to recall what the weather was at 10AM last easter sunday, that is if the sun was shining, or if it was raining or both, what the percentage of cloud cover was if any and the clour of the clouds, which way the wind was blowing and how hard…

Now ask yourself “how do I check this” and “was I correct”, if you can remember who you were with and where ask them and see what they remember oh and what they were wearing… Now answer honestly would you put any veracity on there or your statments?

Jonadab June 5, 2012 6:01 AM

Wait, there are people who do not NOTICE when this happens to them? Seriously?

It is certainly true that my memories of an event are frequently obscured by memories of subsequent related information (most commonly, hearing someone else talk about the event), but generally I am aware of this fact. I was aware before I was old enough to go to school that there were some events in my own past that I could only remember mom or dad talking about and could not recall the event itself. By the time I was in fifth grade I knew that there were some events in my past that I could remember telling other people about myself, and even though I’d been able to remember the event itself at that time, now I could no longer do so and could only remember what I’d said about it later.

Now you’re telling me other people don’t realize when this happens to them, and think their altered memory is still in its original state?

That’s absolutely terrifying. I’m not sure I can believe anything anyone says ever again.

Laura June 5, 2012 9:01 AM

It’s worse than that, @Jonadab. Some of your own memories that you think are original are no longer in their original state. You just haven’t hit evidence of it yet. At least the odds are pretty good that’s the case. Baring fancy memory advantages (eidetic, etc), I suppose.

Point being, of course other people are sometimes aware their memories are compromised. And sometimes they’re not aware of it. I’ve experienced both, myself.

vasiliy pupkin June 5, 2012 9:08 AM

Let say professional hypnotizes cooperative witness (with presence of his/her lawyer to guarantee that only information related to particular event could be addressed/extracted and privacy of witness is protected!), establish rapport, then in that state ‘move’ him/her back in time to the point of crimnal event and giving only instructions to concentrate on particular details, ask questions related to what witness is actaully seeing, smelling or hearing. All session is videotaped for the court.
Yeah, nothing is not 100% reliable, but fMRI looks like the first babby step in right direction.

karrde June 5, 2012 9:56 AM

@Doug Coulter, memory of events vs. memory of impressions.

I’ve run into this problem multiple times. Usually when attempting to help someone figure out what is happening on their computer.

They remember their impression of what happened, and I know (from how the computer/device is behaving) that something happened which they do not remember or did not understand.

If the event was something they didn’t understand, then it got filtered into the unimportant-stuff portion of memory, and probably disappeared quickly.

No One June 5, 2012 10:32 AM

Re: Memory malformation — my memories of Australia have the cars driving on the right side and the drivers on the left. Now, that is obviously incorrect but even the memory of me driving a car for a short time has me in the left-side front seat and driving on the right side of the road. WTF, brain?

-B June 5, 2012 11:32 AM

“–Though extremely rare..And then there were savants with eidetic memory (Stephen Wiltshire, Kim Peek, Daniel Tammet)”

and Sheldon Cooper.

Petréa Mitchell June 5, 2012 11:32 AM

vasiliy pupkin:

Unfortunately, studies of hypnotic recall generally find it to be elaborate confabulation. You are, however, right to suggest that the questioning ought to be taped.

Ideally, someday witness testimony will be treated similarly to physical evidence from the crime scene– collected as soon as possible with proof that it has been collected correctly (i.e., the questions recorded as well as the statements) and an understanding that the longer the gap between the event and the collection, the more it will have degraded.

Less ideally, as long as the court system continues to rely on testimony recited at the trial, I’d really like to see some information on the incredible unreliability of eyewitness accounts taken months after the event included in the standard jury instructions. But we’re probably a few years away from that.

Tony June 6, 2012 5:38 PM

I have learned in 30 years of forensic work that both victims and perpetrators recall facts that are convenient to recall at that given moment, truthfulness or accuracy seldom match the reality of what has or has not occurred. People have tackled this memory recall problem for thousands of years and in all likelihood will for thousands more. Sadly the science of the interview is not as accurate as the 6th century technique used to establish guilt or innocence when using the magic donkey, (obviously tongue in cheek).

The 6th century magic donkey test of field interrogations.

When confronted with multiple suspects a magic donkey would be called in to judge a persons guilt. The donkey would be placed inside a dark tent and the suspects lined up in front of the tent. This donkey had the ability to tell guilt from innocence. Prior to being placed in the tent the donkeys tail was secretly covered with black lamp soot which of course was kept secret from those who were about to be tested.

The suspects were told that they simply needed to enter the dark tent alone and pull on the donkeys tail and that if the donkey remained silent they were proven innocent. If the donkey brayed when its tail was pulled the donkey had determined who the guilty party was as its well established that magic donkeys in dark tents only bray when the tail is pulled by guilty men.

Each man was sent into the tent alone for a minute or two and then retrieved from the tents back door. The hands of the suspects were examined and those who had soot on their hands were most likely innocent of the crime as they had nothing to fear from pulling on the magic donkeys tail. The person who exited the tent with clean hands obviously did not pull on the donkeys tail having known it would determine his guilt.

I trust that any criminal who hung out in street gangs at the time was savvy to this little test and probably stroked the donkeys tail softly and hoped for the best. And probably a number of innocent men were hanged because of false positives from one irritated donkey, what the hell it was an early honesty test and about an accurate as anything developed in the forensic study field since.

Now that’s science that might convince a jury.

Dr. Al Bloom June 6, 2012 5:53 PM

This piece illuminates an issue that I and a number of my associates are seeking remedies for. Thanks so much for covering it with so much authority.

Dirk Praet June 6, 2012 6:33 PM

I have been skeptical about the reliability of eyewitness testimony ever since I first saw “12 Angry Men” with Henry Fonda.

One of the important elements in this movie is that stereotypes not only lead to a biased way of interpreting the evidence, but also to a confirmation bias, which is the tendency to only seek information that confirms one’s expectations and ignore disconfirming information.

Many of the jurors initially expected that the boy was guilty and so they only remembered details in the case that supported that expectation. They also seemed to disregard or ignore details in the case that would disconfirm their expectations. For example, the jurors failed to notice the significant details such as the way the old man walked with a limp or that the female eye witness had marks on her eyes that were caused by prescription eye glasses. All of these details would have disconfirmed their expectations, but were overlooked.

Leave a comment


Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via

Sidebar photo of Bruce Schneier by Joe MacInnis.