Assessing the Effect of Eyewitness Identification Confidence Assessment Method on the Confidence-Accuracy Relationship

• Published in: Psychology, public policy, and law Vol. 28; no. 3; pp. 414 - 432
• Main Authors: Arndorfer, Andrea, Charman, Steve D.
• Format: Journal Article
• Language: English
• Published: American Psychological Association 01-08-2022
• Subjects: Best Practices; Face Perception; Human; Identification; Measurement; Self-Confidence; Witnesses; confidence assessment methods; calibration; eyewitness identification confidence; confidence-accuracy relation; lineups
• ISSN: 1076-8971; 1939-1528
• DOI: 10.1037/law0000348

Historically, best practice recommendations suggested obtaining eyewitnesses' lineup identification confidence reports in their own words. More recently, best practice recommendations call for the collection of confidence reports using scales containing either words or numbers. Clearly, historical and contemporary recommendations are inconsistent. This article provides a review of the existing relevant scientific literature and presents new data to empirically assess the effect of confidence assessment method on the confidence-accuracy relationship. Although small, the extant literature has consistently failed to show any significant difference in the confidence-accuracy relationship as a function of confidence assessment method: Verbal confidence reports are as diagnostic of identification accuracy as numeric confidence reports. We present data from a basic repeated face recognition paradigm in which participants (n = 634) each attempted 16 separate lineup identifications and were randomly assigned to indicate their confidence using either a numeric scale, a verbal scale, or in their own words. Consistent with our review of the scientific literature, both calibration and confidence-accuracy characteristic analyses demonstrated that (a) confidence is predictive of accuracy, including when witness confidence is obtained via open-ended responding, and (b) the confidence-accuracy relationship is not dependent on the method used to obtain confidence estimates. Thus, results support using any of the recommended methods (i.e., scales containing either words or numbers or a verbatim account in the witness's own words) for collecting confidence statements. However, consideration of other criteria (e.g., subjectivity associated with the interpretation of verbal confidence reports) may support recent calls for scale-based confidence collection.