Revisiting the confidence-accuracy relationship in eyewitness identification: a metacognitive perspective.

Abstract

The fallibility of eyewitness memory has been demonstrated in decades of research. This presents a challenge to legal practitioners and triers of fact when evaluating the trustworthiness of eyewitness identification. A witness's initial confidence in their identification decision is a potentially useful indicator of accuracy. While the field generally agrees that there is a meaningful relationship between confidence and accuracy, uncertainty remains about whether high confidence can serve as a direct indicator of accuracy in real-world legal settings. In this review, we integrate findings on the confidence-accuracy (CA) relationship in eyewitness identification research with prevailing metacognitive theories of retrospective confidence and individual differences. We argue for greater consideration of the influence of individual-level factors - particularly metacognitive bias and domain-specific competence, which influence how effectively people monitor their memory - on the CA relationship. Understanding these variables may help answer the key question facing courts: given a witness's expressed confidence, how likely is it that their identification is correct? We discuss how measurement approaches from metacognitive research can be used to clarify when, and for whom, confidence reliably predicts accuracy and suggest future directions of research that may help better understand this question.