Evaluating the Claim That High Confidence Implies High Accuracy in Eyewitness Identification

Two provocative claims about eyewitness confidence have recently been advanced in the eyewitness-identification literature: (a) suspect identifications made with high confidence are highly accurate and (b) high-confidence suspect-identification accuracy is unaffected by variations in memory strength. Several recent publications have reiterated these claims in spite of providing weak or no support. We present four criteria that can be used to evaluate the empirical support for these claims. Regarding the claim that high confidence implies high accuracy, it is necessary to consider whether (a) high-confidence suspect identifications are in fact highly accurate and (b) high-confidence suspect-identification accuracy is dependent on the assumption of a perfectly fair lineup. Results of a base-rate analysis show that lowering the threshold for the claim that high confidence implies high accuracy undermines the claim substantially. Likewise, relaxing the assumption of a perfectly fair lineup may also undermine the claim. In regard to the claim that high-confidence suspect-identification accuracy is unaffected by variations in memory strength, it is necessary to demonstrate that (a) the conditions under comparison actually differ in memory strength, and (b) there is evidence that high-confidence suspect-identification accuracy is equivalent across conditions. We conclude with discussion of whether laboratory experiments have the capacity to provide valid estimates of high-confidence suspect-identification accuracy in real cases. We believe laboratory experiments have the capacity to estimate high-confidence suspect identification accuracy at the time of an identification procedure, but due to a prominent selection bias, we are skeptical of their capacity to estimate high-confidence suspect-identification accuracy at trial.