Confident witness is not always reliable

A confident witness is generally more reliable than a less confident witness. This is the finding of doctoral research by psychologist Geralda Odinot. But, judges should beware: 100% confidence is no guarantee for 100% accuracy.

Geralda Odinot: ‘A witness statement alone should never be accepted on the basis of the confidence of the witness.'

Criminal cases

Odinot's findings are of great significance for legal practice. Odinot: ‘In many criminal cases the suspect is convicted exclusively or almost exclusively on the basis of witness statements. However, there is a lack of objective means of verifying these statements, so the judge looks for other indicators of the accuracy of a statement.' One of these indicators is the confidence with which a witness talks about a remembered event. Research shows that confidence has a highly persuasive effect, both on legal professionals and on an untrained audience. Odinot's research shows that this of often unjustified: confidence is only a reasonably reliable indicator for accuracy in particular circumstances.

Memory processes

These findings do not come completely as a surprise. Previous studies into the relationship between confidence and accuracy in recognising individuals has shown that the assumed link is weak. But little research had been done into whether this conclusion also applies to memories of experiences and events. 'Picking a suspect out of a line-up relies on different memory processes than freely describing an attack,' Odinot explains. 'In the first case, a rapid and subconscious recognition takes place, while in the second case people make an active reconstruction of what they have seen and heard.'

Margin of error

Odinot carried out a number of lab experiments in which test candidates were shown a video of a staged but realistic account of a traffic accident. One or two weeks later the test candidates were asked a number of open questions about details of the film. For each answer they were asked to indicate how confident they were of the accuracy of their response. The results of the experiments were ambiguous. The more confident the test candidates were, the more the number of correct answers increased and the number of incorrect answers decreased. Confidence, therefore, offers some indication of the accuracy of a memory. However, there is always a margin of error: even with maximum confidence, 6 to 16% of the information given is incorrect, depending on the time which has elapsed since watching the film. The longer ago the film was watched, the greater the percentage of errors.

Cover of Geralda Odinot's dissertation. 'Eyewitness confidence: The relation between accuracy and confidence in episodic memory.'

Odinot warns in her dissertation that the confidence of a witness as an indicator for the accuracy of a memory becomes less reliable over time.


With such a margin of error, the forensic usefulness of the link between confidence and accuracy remains limited, Odinot argues. 'Provided you take adequate care, you can use the confidence of a witness as an indicator of the accuracy of a memory or an event, certainly in the early stages of an investigation. But confidence is no guarantee of accuracy. A witness statement should never be accepted as the truth, simply on the basis of confidence.'


Memory is fallible, which is problematic in a legal context. 'Human memory was not designed for the law court,' says Odinot. At the same time she challenges the suggestion that witness statements about details are by definition unreliable. She believes that this suggestion has come about as a result of psychological research in which the emphasis is on the failure of the memory or the implantation of false memories. 'People do indeed make mistakes, but most of the information remembered is correct,' she says. Her research shows that people are able to remember accurately most details of an event, even after several weeks.

Suggestive questions

At least, if the questions are put in the correct way. If the questions are phrased suggestively, little is left of that predominantly accurate memory. Test candidates who were asked by Odinot where the victim was bleeding while the film they were shown contained no blood at all, appear to be more susceptible to suggestive information after some time has elapsed. Many of them 'remembered' things, often confidently, that they had not seen. Odinot's research again confirms the disastrous effect of suggestive questions.

Interview candidates soon

This danger is, according to Odinot, a good reason to interview candidates as soon as possible following a criminal investigation. If the interview takes place later, there is a chance that the witness memories may in the meantime have been influenced by supplementary or inaccurate information which they they may have absorbed via the media. The initial interview should then be carried out carefully, without any suggestive questions. And, Odinot recommends: 'Always record the interviews of important witnesses on video. A written statement is selective and is often not a literal report of what was said. It might be better from a functional point of view given the great amount of information gathered, but the danger is that important information will be lost.'

Odinot investigated the memories of fourteen witnesses of an armed robbery on a supermarket.

Field study

How conclusive are such recommendations? Can research results obtained in the laboratory be generalised to legal practice? Odinot recognises that it is impossible to reproduce the unexpected and distressing nature of real crimes or accidents. She was therefore pleased to have the opportunity to carry out a field study during her PhD training. With the co-operation of the Zuid-Holland-Zuid police force she studied the memories of fourteen witnesses of an armed robbery on a supermarket in Gorinchem. The robbery was recorded on surveillance cameras, which gave Odinot the opportunity to measure the accuracy of the witnesses' memories. The results confirmed what the psychologist had demonstrated in her laboratory experiments: people's memories are not bad, but a confident witness is not by definition a reliable witness. Odinot carried out her PhD research with the Netherlands Insitute for the Study of Crime and Law Enforcement (NSCR). She is currently working as a postdoc research fellow at the University of Aberdeen in Scotland.


  • Prof. W.A. Wagenaar (Leiden University, Utrecht University)
  • Prof. P.J. van Koppen (Maastricht University, Vrije Universiteit Amsterdam, previously NSCR)

Dr G. Wolters (Leiden University)

Last Modified: 19-02-2009