Cognitive Daily

In every courtroom drama, the most dramatic scene is always when the star witness points her finger at the villain and proclaims that “he did it!” The confidence with which an eyewitness describes the perpetrator of a crime is often the most convincing evidence in a court battle. But how accurate is eyewitness testimony? Do we really remember everything as accurately as we think we do? How important are other influences on eyewitness testimony? And what if the witness is a child?

Carl Martin Allwood has been working on these issues for years. In his most recent article, co-authored with Anna-Carin Jonsson and Par Anders Granhag and published in Applied Cognitive Psychology, he takes on the issue of child testimony.

Their study tried to replicate as closely as possible the manner of a police investigation of a crime. 12-year-olds were shown a video of a kidnapping filmed from a witness’s eye perspective. Next they filled out a quick questionnaire about what they had seen—the questions gave only two choices of answers, but were very specific in their requests for details about the kidnappers, the crime scene, and the getaway car.

Two weeks later, participants were presented with two sets of answers to the questionnaire: their own, and either a classmate’s or a teacher’s. Based on these two sets of answers, the children had to rate the degree of confidence they had in their original responses. In fact, the second answer set was a fake, carefully constructed by the experimenters to agree or disagree with the child’s own answers. For every correct response, half of the fake answers agreed and half of them disagreed with the child’s answers. The same was done with incorrect responses. The point of this was to see what influence of teachers teachers and peers had on the child’s confidence in his or her responses. The two-week delay was deliberately chosen to simulate the often lengthy police crime investigation process.

The students rated their confidence in their own answers on a scale from 50% to 100%. They were instructed that 50% meant they were guessing—they had no idea of the correct response, and that 100% meant that they were absolutely certain. Thus, these ratings correspond to the actual odds of their response being correct.

An analysis of the data revealed that there was no difference between the teacher- and the peer-responses to the student answers—they were equally influenced by both their peers and their teachers. This surprised Allwood and his colleagues, because previous studies had suggested that children were more influenced by adults than their peers. Previous research had been with younger children, however, and as they enter their teen years, the influence of peers does begin to rise.

However, the respondents were significantly influenced by the corroboration of other witnesses. When the teacher or peer response agreed with the child, confidence was much higher than when the responses disagreed. Take a look at this graphical version of the results:


The diagonal black line indicates where confidence ratings would correctly correspond to actual accuracy. Any results above that line represent underconfidence, because confidence ratings are less than the actual accuracy. In this case, nearly all of the ratings reflect overconfidence, with even 100% confidence ratings corresponding to a maximum of less than 70 percent accuracy. More importantly, in cases where the teacher or peer agreed with the child, confidence increased, even though the “disagree” confidence ratings still reflected overconfidence.

The same team conducted a similar study in 2004 on adults, and when they compared those results to these results with children, they found that children were influenced to a greater degree by people who disagreed with them: while both adults and children are overconfident about memories, children become less confident to a greater degree when faced with disagreement than adults do.

To sum up, based on these results, both children and adults are overconfident about nearly all their memories, but children are more likely to change their minds when faced with disagreement.

So what memories can we trust? Clearly this study emphasizes details such as the escape vehicle or the environment around the kidnapping. But would witnesses actually disagree about whether a kidnapping took place, or how many kidnappers were involved? We may have hunches, but this study doesn’t tell us the answers. Considerably more work will need to be done to learn exactly how much information is necessary for us to form accurate memories.

Allwood, C. M., Jonsson, A.-C., & Granhag, P. A. (2005). The effects of source and type of feedback on child witnesses’ metamemory accuracy. Applied Cognitive Psychology, 19(3), 331-344.


  1. #1 MikeS
    August 2, 2005

    Excellent blog. Hope you don’t regard the description as pejorative. Please keep going, this is fascinating stuff. My pet interest is philosophy, though I have done some psychology courses, and I am often struck that many of the questions philosophers ask, particularly about theory of mind, qualia and dreaming, would be better answered if they were more familiar with cognitive psychology and neuropsychology.

  2. [...]

    The jury’s out: eyewitnesses are unreliable

    or, rather, kids are influenced by what their peers think when they are used as eyewitnesses. and: more about the topic of the relative lack of obj [...]