The most trusted evidence, and the least reliable
Eyewitness testimony is persuasive to juries and unreliable in fact. Because memory reconstructs, a witness’s account can be reshaped by leading questions, lineup procedure, stress, and post-event information — and their confidence can grow over time even as accuracy does not. Mistaken identification has been a leading contributing factor in wrongful convictions later overturned by DNA evidence.
Why the confident witness is so dangerous
Juries treat confidence as a proxy for accuracy. It is a poor one — and worse, confidence is malleable. Confirming feedback after an identification ("good, that’s our suspect") reliably inflates a witness’s recalled certainty, their memory of how good their view was, and how quickly they say they decided. By the trial, a tentative pick can present as an unshakeable one.
What corrupts an identification
- Leading questions and post-event information (see the misinformation effect).
- Suggestive lineups — where the suspect stands out, or the administrator knows who the suspect is.
- Weapon focus — attention narrows onto a weapon, degrading memory for the face.
- Cross-race identification — accuracy is on average lower for faces of another race.
- Stress and time — extreme stress does not sharpen memory into a reliable record.
The science here has produced concrete, adopted fixes: double-blind lineups (the administrator does not know the suspect), unbiased instructions ("the suspect may not be present"), fair fillers, and — critically — recording the witness’s confidence at the moment of first identification, before any feedback can inflate it. Confidence taken then, under clean conditions, is far more informative than confidence in the courtroom.
None of this means witnesses lie, or that testimony is worthless. Honest, well-intentioned people can be completely certain and completely wrong — that is precisely the finding, and it is why procedure matters more than sincerity. This page is educational and is not legal advice.
Keep going
More from the vault: all 8 pages · related: test your own memory in the Laboratory, déjà vu, deliberate practice.
Frequently asked
Is eyewitness testimony reliable?
Does a confident witness mean an accurate one?
What makes eyewitness identification more accurate?
Sources & further reading
- Wixted & Wells (2017), Psychological Science in the Public Interest: a synthesis arguing that eyewitness confidence, recorded on an initial uncontaminated test under pristine lineup conditions, is a far better indicator of accuracy than the field long assumed. Wixted & Wells, 2017 ↗
- Gary Wells (Iowa State) has studied eyewitness ID since the 1970s; his work drove lineup reforms (double-blind administration, proper fillers, sequential presentation). Wells / APA ↗
- Confirming feedback after an ID (“Good, that’s our suspect”) inflates a witness’s confidence and memory of how good their view was — confidence is contaminated by what happens after the identification. The nuance matters: Sporer et al.’s meta-analytic correlation of .41 for witnesses who pick someone is moderate-to-large, not negligible — and Wixted & Wells (2017) show that under “pristine” conditions (fair lineup, one suspect, unbiased instructions, double-blind administration, confidence recorded immediately) a high-confidence initial ID is in fact highly accurate. It is confidence recorded later, after feedback, that becomes worthless. Wells / APA ↗
- Per the Innocence Project, mistaken eyewitness identification contributed to roughly 70%+ of wrongful convictions later overturned by DNA — the single most common factor. Innocence Project ↗
- The cross-race effect: people recognise own-race faces more accurately than other-race faces — a robust bias across cultures (Meissner & Brigham, 2001 meta-analysis). Meissner & Brigham, 2001 ↗
Educational content, not medical, psychological, or legal advice. Last reviewed July 2026.