Confidence and accuracy of eyewitness identification from lineups

Recent studies of confidence-accuracy correlations in eyewitness-identification experiments have produced highly variable results (Deffenbacher, 1980; Wells & Murray, 1984) with some correlations near 0 (Malpass & Devine, 1981) and others as high as 0.5 (Brigham et al., 1982). Much of this variability may be accounted for by differences, in the design of the research. Specifically, lineup experiments in which the to-be-identified person (target) is always present tend to produce higher confidence-accuracy correlations than studies also providing blank or target-absent lineups replacing the target with a lookalike. A reanalysis of previously collected data (Lindsay & Wells, 1980; Wallbridge & Lindsay, 1982) revealed that lineup foils (nontargets) were identified with significantly less confidence than targets or lookalikes but that targets and lookalikes were identified with equal confidence. Two experiments were conducted to test the hypothesis that confidence of identification is a function of the similarity of the identified person to the original target. The first experiment using nonstudent subjects (n=53) and multiple targets (n=14) to increase generalizability, demonstrated that the person in the lineup who most closely resembled the target was identified with the highest mean level of confidence. The second experiment examined confidence in identification following a crime staged for 260 introductory psychology students. The results indicated that the similarity of foils to the target (established by independent ratings) was predictive of both the frequency and confidence of identification of faces.