Why Science Tells Us Not to Rely on Eyewitness Accounts

Eyewitness accounts feel like the strongest evidence there is, yet mistaken witnesses are behind more proven wrongful convictions than any other single cause.1PubMed Central. An Examination of the Causes and Solutions to Eyewitness Error The reason is that memory doesn’t play back a recording. It rebuilds the event each time it is recalled, and errors can slip in along the way.2PubMed Central. The cognitive neuroscience of constructive memory: remembering the past and imagining the future Scientists haven’t concluded that eyewitnesses are useless. Their point is that how reliable a witness is depends heavily on the conditions and on how police collect the memory.

How Often Eyewitnesses Get It Wrong

Among 375 US cases where DNA later proved a convicted person innocent, a witness had picked the wrong person in 69%.3Journal of Law, Information and Science. Human Memory and the Reliability of Eyewitness Testimony: An Empirical Analysis of Misidentification, Memory Distortion, and Evidentiary Reliability In a wider set of 873 overturned convictions, mistaken identifications showed up in 76%.4PubMed Central. An Examination of the Causes and Solutions to Eyewitness Error These are not rare slips.

The errors come from honest people. In 84% of DNA cases with a misidentification, the mistaken witness was a surviving victim.5Journal of Law, Information and Science. Human Memory and the Reliability of Eyewitness Testimony: An Empirical Analysis of Misidentification, Memory Distortion, and Evidentiary Reliability In 32% of those cases, more than one witness picked the same wrong person.6Journal of Law, Information and Science. Human Memory and the Reliability of Eyewitness Testimony: An Empirical Analysis of Misidentification, Memory Distortion, and Evidentiary Reliability

Witnesses in these cases usually weren’t lying. They believed their memory, and nothing in a sincere wrong memory feels different from a true one to the person holding it.

DNA exoneration cases are useful because they remove doubt. The true culprit’s biology shows the convicted person could not have done it, so the witness’s error is a settled fact rather than an opinion. But that also makes them a skewed sample. They exist only where biological evidence was kept and could be tested, so they mostly involve crimes like rapes and murders. That is why the percentages show how often mistaken identification appears among proven errors, not how often witnesses err in everyday cases.

For a juror, the right conclusion is not “disregard witnesses.” It is that an identification with nothing else behind it deserves a harder look. A case with a recovered item, a recording, or a confession that checks out against the facts stands on firmer ground than one resting on a single person’s pick.

Memory Rebuilds Events Instead of Recording Them

Memory for events is a rebuilding process rather than a replay, and it is prone to errors and illusions.7PubMed Central. The cognitive neuroscience of constructive memory: remembering the past and imagining the future Each time a memory is brought back to mind it can be updated, which can strengthen it or distort it.8PubMed Central. Neural mechanisms of reactivation-induced updating that enhance and distort memory A memory recalled many times, in interviews and retellings, has had many chances to change.

Memory isn’t like a video camera, which could only lose details; memory also adds and changes them. A witness may “remember” a red jacket because someone later mentioned red. The added detail feels exactly as real as the original, and the witness can’t tell which parts came from the event and which came later. Repeated retelling can even turn an early guess into what feels like solid memory.

A vivid memory is not proof of an accurate one. A memory can be rich in detail and emotion and still contain pieces that were never part of the event, because the clarity comes from how easily the brain assembles the scene, not from a check against what happened.

Part of the reason is that we take in only some of what is in front of us, so a memory starts out with holes. When we recall it, the brain fills them with what seems likely, drawing on expectations, similar past experiences, and things learned afterward. The result is a smooth, complete-feeling story. That serves us well day to day, but it is riskier when exact details matter in a criminal case.

One leading idea is that memory evolved partly to help us imagine and plan the future by recombining pieces of past experience.9PubMed Central. Memory distortion: an adaptive perspective That same flexibility means the brain can mix imagined or later details into what it remembers.10PubMed Central. Memory distortion: an adaptive perspective Many memory errors are best seen as side effects of processes that make memory work well day to day.11PubMed Central. On the Evolution of a Functional Approach to Memory12Perspectives on Psychological Science. Adaptive Memory: The Evolutionary Significance of Survival Processing

What Can Distort a Witness’s Memory

Witnesses who discuss an event often absorb details from the conversation into their own memory.13Applied Cognitive Psychology. Combating Co‐witness contamination: Attempting to decrease the negative effects of discussion on eyewitness memory Some of the resulting memories blend what the person saw with what they heard afterward.14PubMed. The co-witness misinformation effect: memory blends or memory compliance? Witnesses who already knew each other, as friends or partners, were even more likely to report details they never saw.15PubMed. “With a little help from my friends…”: the role of co-witness relationship in susceptibility to misinformation

When witnesses compare stories, each hears details the other caught and may start to feel they saw them too. A friendly chat, a family group text, or a shared social media thread all count as discussion. A couple who witnessed something together are not an independent check on each other.

Warning people about this afterward did not undo the effect.16Applied Cognitive Psychology. Combating Co‐witness contamination: Attempting to decrease the negative effects of discussion on eyewitness memory Once the sources are blended, the witness can’t separate them, and telling them to be careful doesn’t restore the original. That’s why police now try to separate witnesses quickly, and why the best protection is not discussing the event before giving a statement.

Distance and light set hard limits. Correct identifications drop as distance grows, at every light level.17Cogent Psychology. A stab in the dark: The distance threshold of target identification in low light At 20 meters, people picked the right person about half the time in good light but only about 1 in 10 times in very dim light, like a dark street.18Cogent Psychology. A stab in the dark: The distance threshold of target identification in low light To judge a witness’s view, ask how far away they were, for how long, and how bright it was. A brief glimpse at night across a street warrants caution.

Stress has mixed effects. High stress lowered how often witnesses picked the right person in one large set of experiments,19PubMed. Elevated stress impairs the accuracy of eyewitness memory but not the confidence-accuracy relationship but another study found stress made no difference to lineup performance.20PubMed Central. Stress, stress‐induced cortisol responses, and eyewitness identification performance

The popular “weapon focus” idea, that a gun pulls attention away from the culprit’s face, sounds convincing. It held up poorly when people watched scenes naturally.21PubMed Central. Revisiting the role of attention in the “weapon focus effect”: Do weapons draw gaze away from the perpetrator under naturalistic viewing conditions?

Faces of another race are harder. People are better at recognizing faces of their own race, even in real-world settings such as store clerks recalling customers.22Journal of Applied Social Psychology. Cross-Racial/Ethnic Eyewitness Identification: A Field Study A witness is over 50% more likely to pick the wrong stranger when that person is of a different race.23Seattle University School of Law Digital Commons. Cross-Racial Misidentification: A Call to Action in Washington State and Beyond And 42% of DNA cases with a misidentification involved a witness of a different race.24Journal of Law, Information and Science. Human Memory and the Reliability of Eyewitness Testimony: An Empirical Analysis of Misidentification, Memory Distortion, and Evidentiary Reliability

How Police Lineups Can Steer a Witness

A lineup is a set of photos or people that includes the suspect and several innocent others, called fillers. The witness is asked whether the culprit is among them, and “no, I don’t see the person” is a legitimate answer. The danger is that a witness who isn’t sure may pick whoever looks most like the culprit compared with the rest.

When the officer running the lineup knows who the suspect is, wrong picks rise and witnesses grow more confident in those wrong picks.25PubMed. Blind sequential lineup administration reduces both false identifications and confidence in those false identifications Knowing the suspect did not raise correct picks.26PubMed. Blind sequential lineup administration reduces both false identifications and confidence in those false identifications An officer can cue a witness without meaning to: through tone, a pause, or a nod, or a remark like “good, that’s our guy” that inflates confidence later. Neither side may be aware it happened. That’s why a blind lineup, run by an officer who doesn’t know the suspect, protects innocent people.

The other faces matter too. Fillers that fit the witness’s description but don’t closely copy the suspect’s face help witnesses tell the guilty from the innocent.27PubMed Central. Optimizing the selection of fillers in police lineups A fair set matches the description, such as age, hair, and build, so the suspect doesn’t stand out for reasons unrelated to the crime. If the suspect is the only one with a beard, a witness can pick him by elimination without truly recognizing him.

At the other extreme, lineups of near-identical faces make it harder to pick the right person.28PubMed Central. Lineup fairness: propitious heterogeneity and the diagnostic feature-detection hypothesis One study found little difference between these filler methods, though.29The LAIR at East Texas A&M. The Effect of Filler Selection Methods on Lineup Fairness and Eyewitness Identification

Showing faces one at a time made witnesses more likely to correctly say the culprit wasn’t there.30PubMed. Eyewitness accuracy rates in sequential and simultaneous lineup presentations: a meta-analytic comparison But a real police field study suggested showing them all at once may be as good or better.31PubMed Central. Estimating the reliability of eyewitness identifications from police lineups Lab and field studies point different ways, and experts agree that blind administration and fair fillers matter more than the display method.

When a Confident Witness Can Be Trusted

Under fair, blind lineup conditions, adults’ confidence and accuracy are strongly linked.32PubMed. The Relationship Between Eyewitness Confidence and Identification Accuracy: A New Synthesis Picks made with high confidence are usually right, around 80% or more, while low-confidence picks are right about 40% of the time or less.33American Psychologist. Initial Eyewitness Confidence Reliably Predicts Eyewitness Identification Accuracy Police field data support this for fair lineups.34PubMed Central. Estimating the reliability of eyewitness identifications from police lineups Scientists once thought confidence meant little, but better analysis showed the problem was contaminated, later confidence.

That distinction is the key. A hesitant pick can become “I’m certain” months later. Feedback, retelling, and trial preparation inflate certainty, and jurors usually see only the final, confident version. So the useful question is what the witness said at the first lineup.

The words a witness uses at the moment of picking, such as “I think it’s number three, but I’m not sure” or “that’s him, no doubt,” are the cleanest read of the memory before anything else touches it. If police write that statement down or record it right then, a court has a fixed point to compare with what the witness says at trial. Without that record, jurors rely on recollections of how sure the witness was, shaped by everything in between.

High confidence is a strong sign only when conditions were fair. Stress lowered overall accuracy, but high-confidence picks stayed about as accurate.35PubMed. Elevated stress impairs the accuracy of eyewitness memory but not the confidence-accuracy relationship Confidence overstated accuracy, however, when witnesses picked someone of another race.36Applied Cognitive Psychology. Confidence and Eyewitness Identifications: The Cross‐Race Effect, Decision Time and Accuracy A non-blind officer can also boost confidence in a wrong pick.37PubMed. Blind sequential lineup administration reduces both false identifications and confidence in those false identifications Confidence is a weaker sign, too, when the witness was told they chose well.

What Courts Are Doing About It

New Jersey’s Supreme Court in 2011 required juries to be told about factors that weaken identifications, from lineup methods to viewing conditions.38University of Colorado Law Review. State v. Henderson: A Model for Admitting Eyewitness Identification Testimony A judge reads the instruction before deliberation, so jurors weigh things like distance, lighting, stress, and how the lineup was run instead of relying on the witness’s apparent sincerity.

Instructions alone may not be enough. The original wording appears to work poorly,39Psychology, Crime & Law. Improving the effectiveness of the Henderson instruction safeguard against unreliable eyewitness identification and jurors may hear a warning as abstract advice and still be swayed by a confident witness on the stand.

Experts help. Without expert testimony, mock jurors weighed strong and weak eyewitness evidence about the same.40Behavioral Sciences & the Law. Expert testimony and jury decision making: An empirical analysis Expert testimony helped them tell the difference without making them distrust all witnesses.41Behavioral Sciences & the Law. Expert testimony and jury decision making: An empirical analysis

How to Protect Your Memory If You Witness Something

Write down what you saw as soon as possible, before talking to anyone. Discussion can permanently blend others’ details into your memory. If you must talk to others at the scene, share what you saw without asking what they saw, and write your own version first.

Write in your own words, without guessing. Describe what you actually saw, and mark anything you’re unsure about as uncertain rather than filling it in. Note where you stood, how far you were, how bright it was, and how long you could see the person, so anyone judging your account later can weigh it properly. If you can, date and sign the note or send it to yourself, so there is a record of what you remembered before anything else could change it.

At a lineup, it’s fine to say the person isn’t there. State your confidence in your own words right away, and don’t ask the officer whether you chose correctly.

Afterward, avoid news coverage and photos of the suspect, and resist reassurance that you “got the right guy.” If you were unsure at first, say so honestly in court.