The witness who was sure

Confidence and accuracy are different things, and courts have spent two centuries learning how different.

A witness who says 'I am certain' is believed more readily than one who hesitates. The instinct is understandable: in ordinary life, certainty usually tracks knowledge. In identification evidence it tracks something else. Studies of real cases where the conviction was later overturned by genetic evidence found that mistaken witnesses were, on average, no less confident than accurate ones. Some were more so.

The mechanism is not dishonesty. Memory for faces is reconstructive, and each retelling rebuilds the memory using whatever is available, including the photograph the witness was shown last week and the reassurance of an officer who said they had identified the right person. Confidence rises with rehearsal while accuracy does not. By the time of the trial, the witness is describing a memory that has been revised a dozen times, and the revisions are invisible from the inside.

The practical response has been procedural rather than psychological. Confidence is now recorded at the first identification and the recorded figure, not the courtroom performance, is what counts. Line-ups are run by an officer who does not know which person is the suspect, so no unconscious encouragement can leak. Witnesses are told the suspect may not be present at all.

None of this makes identification reliable. It makes the unreliability measurable, which is a smaller claim and a more useful one. A system that knows how often it is wrong can decide how much weight the evidence should carry. A system that trusts confidence has no way of finding out.