Does Eyewitness Confidence Mean the Memory Is Accurate?

Not on its own. How sure a witness feels is not a direct readout of how accurate their memory is, because confidence can be shaped by things that have nothing to do with what they saw. The research is more specific than a simple yes or no. A large 2017 synthesis by John Wixted and Gary Wells concluded that confidence can be informative, but only at the first identification and only when the procedure is clean: one suspect in a lineup, fair fillers, an administrator who does not know who the suspect is, and no feedback. Under those "pristine" conditions, a high-confidence first identification tends to be fairly accurate. Once a witness has been told they did well, talked to others, or testified repeatedly, confidence can rise while accuracy does not. So confidence matters, but when and how it was recorded matters just as much.

Key takeaways

• Confidence and accuracy are related, but the link depends heavily on when the confidence was expressed and under what conditions.
• Wixted and Wells (2017) found that a high-confidence identification made at the first, uncontaminated test, under pristine conditions, tends to be accurate.
• Confidence is malleable. Feedback such as "Good, you identified the suspect" can inflate it without improving accuracy.
• Low confidence at the first identification is a warning sign, whatever the conditions.
• Confidence measured in the courtroom, long after the event, is far less informative than confidence recorded at the moment of identification.

Why is confidence treated with suspicion?

For years, a running theme in eyewitness research was that confidence and accuracy barely relate. Jurors tend to find a confident witness persuasive, yet case after case showed confident witnesses who were wrong. The concern was reinforced by DNA exoneration data: mistaken identification has been the most common contributing factor in wrongful convictions later overturned by DNA, and many of those witnesses had been sure.

The 2014 National Research Council report Identifying the Culprit set out the underlying reason. Memory is malleable at every stage, from seeing an event to storing it to recalling it, and witnesses are not aware of the distortions. A feeling of certainty is generated by the same fallible system that produced the memory, so it is not an independent check on whether the memory is right.

What did the 2017 synthesis change?

Wixted and Wells argued that the pessimism had gone too far, and that much of it came from measuring confidence at the wrong moment. Their synthesis of the evidence reached a more structured conclusion.

• The timing of the confidence statement is critical. Confidence is most informative at the very first identification, before anything can contaminate it.
• Under pristine procedures, a high-confidence first identification is associated with high accuracy.
• Low confidence at that first test signals a high risk of error, regardless of how good the conditions were.
• Confidence expressed later, after feedback or repeated questioning, is much weaker as a guide.

In other words, confidence is a usable signal, but a perishable one. It tells you the most the first time it is measured, and less and less after that.

What are "pristine conditions"?

The value of a first identification depends on the procedure around it. Wixted and Wells, echoing long-standing recommendations, pointed to several safeguards:

• A lineup should contain only one suspect, with the others being known-innocent fillers.
• The fillers should fit the witness's description, so the suspect does not stand out.
• The person running the lineup should not know who the suspect is, which prevents unintentional steering. This is the idea behind a double-blind police lineup.
• The witness should give a confidence statement in their own words immediately, before any feedback.
• The lineup should be conducted once, cleanly, without repeated exposure.

These points overlap with what makes a lineup fair in general, covered in what makes a police lineup fair?.

How does confidence get inflated?

The clearest demonstration of confidence drifting away from accuracy comes from research on post-identification feedback. In a 1998 study, Gary Wells and Amy Bradfield had participants identify a person from a lineup that did not contain the actual culprit, so every identification was wrong. Some were then told "Good, you identified the suspect"; others received no feedback.

The confirming feedback had a marked effect. Witnesses who were told they were right later recalled having been more confident all along, having had a better view, and having paid more attention, even though they had all picked the wrong person. A later meta-analysis of this post-identification feedback effect, pooling many experiments, found it to be robust across measures of retrospective certainty.

This is why a witness who seems absolutely certain on the stand is not necessarily telling you much about accuracy. Their certainty may have been built up by everything that happened after the identification. A confident account can still be mistaken for these reasons, which is the focus of why can a confident witness still be wrong?.

Does high confidence ever mean high accuracy?

Yes, within limits. The Wixted and Wells synthesis is not a claim that confidence is worthless; it is a claim that confidence is informative under the right conditions and at the right time. A high-confidence identification taken at the first clean test, with proper procedures, carries real weight. The mistake is to treat later, courtroom confidence as if it had the same value, or to assume that because someone is sure, they must be right. Confidence and accuracy are both products of memory, and both can be distorted together.

Is a confident witness who turns out wrong a liar?

No. A witness can be completely sincere, completely confident, and still mistaken. Confidence inflation happens without the witness realising it, and a false memory feels exactly like a true one from the inside. Treating a confident but wrong witness as dishonest confuses an error of memory with a lie. The research consensus is that honest certainty is not proof of accuracy, and it is certainly not proof of bad faith.

How does Murder Book handle certainty?

Murder Book: The Art of Interrogation is an upcoming narrative interrogation game from Gungrounds, an independent studio in Zagreb, Croatia. It is built around unreliable accounts, where how sure someone sounds is one more thing to weigh rather than a verdict in itself. A statement delivered with total conviction is not automatically true, and a hesitant one is not automatically a lie.

The game frames behaviour and confidence as prompts to look closer and to seek independent corroboration, which matches the research: certainty is a signal to be interpreted in context, not a shortcut to the truth. Cases are designed for focused 5–10-minute sessions, and Murder Book is in development for PC (Steam) and mobile, with release currently planned for the end of Q2 2027. For the wider question, see how reliable is eyewitness testimony?.

Sources

1. Wixted and Wells (2017), The relationship between eyewitness confidence and identification accuracy: a new synthesis. Psychological Science in the Public Interest (DOI)
2. Wells and Bradfield (1998), "Good, you identified the suspect": feedback to eyewitnesses distorts their reports of the witnessing experience. Journal of Applied Psychology (DOI)
3. National Research Council (2014), Identifying the Culprit: Assessing Eyewitness Identification. National Academies Press
4. Innocence Project, DNA Exonerations in the United States (1989–2020)
5. Gungrounds, Murder Book fact sheet