The role of self‐rated ability in the accuracy of confidence judgements in eyewitness memory and general knowledge

SUMMARY It is argued that confidence stems in part from self-rated ability in a domain of knowledge and that in eyewitness memory such perceptions are erroneous. Two experiments tested these hypotheses. In both experiments participants rated their relative ability in the domains of eyewitness memory and general knowledge and subsequently took tests of each, giving confidence ratings for each item attempted. In both studies, self-rated ability predicted performance for general knowledge, but not eyewitness memory. Across participants confidence ratings were significant predictors of accuracy for general knowledge, but not for eyewitness memory. In Experiment 1 self-rated ability was predictive of confidence ratings for both domains, although this effect was weaker in Experiment 2. The argument that the accuracy of confidence judgements in eyewitness memory is undermined by a lack of insight into relative expertise is therefore supported. Copyright # 2004 John Wiley & Sons, Ltd. The confidence with which a witness is able to identify a perpetrator in a lineup, or recall a particular detail from a crime is intuitively believed to be a good predictor of likely accuracy. However, the results of numerous psychological investigations indicate that such intuitions maybe misguided. Over the past 15 years or so, there have been several metaanalyses of work on the confidence-accuracy (C-A) relation in eyewitness memory indicating that the relation is either entirely absent (Wells & Murray, 1984), weak (Bothwell, Deffenbacher, & Brigham, 1987) or at best moderate (Read, Lindsay, & Nicholls, 1998). Each of these meta-analyses has advanced explanations as to why the C-A relation is weaker than is intuitively expected, and it is not the intention to review that work here. Instead I take a different point of departure. In the eyewitness confidence literature there have been relatively few attempts to compare the ability of individuals to judge their likely accuracy in eyewitness memory with another domain of knowledge. One exception is the study by Perfect, Watson, and Wagstaff (1993) who compared performance on tests of general knowledge and eyewitness memory. They were careful to ensure that both mean performance, and variability in performance were matched across domains. There were two important findings with regard to the C-A relation. First, when calculated across items using a Goodman-Kruskal gamma correlation (Nelson, 1984), the C-A relation was highly reliable for both general knowledge and eyewitness memory, and in fact did not differ across domains. Second, and