The generalisability of confidence-accuracy studies in eyewitnessing

This paper discusses the evidence on the confidence-accuracy relationship in eyewitness research. It is pointed out that the conclusion often drawn on the basis of such research, that there is little or no relationship between eyewitness confidence and accuracy, is an unwarranted generalisation based on the use of experimental paradigms that are limited in terms of their generalisability to courtroom situations. In particular, almost all studies involve between-subject rather than within-subject designs, thereby limiting the generalisability of findings. A within-subjects analysis examines whether, within an individual, more confident responses are associated with greater accuracy than are less confident responses. A between-subjects analysis examines whether a more confident individual is likely to be more accurate than a less confident individual. A further limitation on the generalisability to real life situations of studies conducted to date is that experiments must involve making errors in identification in order to allow correlational analysis to take place. This means that findings cannot be generalised to those real life situations where all subjects are likely to be completely accurate and confident.