The Relationship Between Confidence and Accuracy in Eyewitness Identification Studies: Is the Conclusion
Few claims about eyewitness behavior have generated as much controversy as those concerning the relation between a witness' confidence in an identification decision and the objective accuracy of that decision (Penrod & Cutler, 1995; Sporer, Penrod, Read, & Cutler, 1995; Wells, 1993). On the one hand, beliefs about a positive confidence-accuracy (CA) relation have been expressed in survey research by members of the public (Brigham & Bothwell, 1983; Deffenbacher & Loftus, 1982; Noon & Hollin, 1987; Yarmey & Jones, 1983) and the legal profession (Brigham & Wolfskeil, 1983), as well as the judiciary through court decisions (e.g., Neil v. Biggers, 1972; Penrod & Cutler, 1995; see Brooks, 1983). On the other hand, psycholegal researchers have presented numerous research studies and meta-analyses of them in support of the conclusion that little or no statistical (or practical) CA relation exists (Bothwell, Deffenbacher, & Brigham, 1987; Deffenbacher, 1980; Lieppe, 1980; Wells, 1993; Wells & Lindsay, 1985; Wells & Murray, 1984). For example, Bothwell et al. (1987) obtained an average correlation coefficient of r = .25 across 35 field studies of person identification. This same position has been shown to characterize the majority of eyewitness testimony experts polled by Kassin, Ellsworth, and Smith (1989). Despite the seeming unanimity of these responses by the
research community, it is worth noting that very strong CA relations have occasionally been obtained in forensically valid simulation research (e.g., Brigham, Maass, Snyder, & Spaulding, 1982; Krafka & Penrod, 1985; Malpass & Devine, 1981; Read, 1995) and several investigators have attempted to identify features that characterize situations producing meaningful or valid CA relations (Bothwell et al., 1987; Cutler & Penrod, 1989b; Deffenbacher, 1980; Fleet, Brigham, & Bothwell, 1987; Sporer et al., 1995). The earliest such attempt was by Deffenbacher (1980), who offered the optimality hypothesis as an explanatory principle, according to which the more optimal the encoding and retrieval opportunities for person identification, the greater the likelihood of a positive CA relation.