When high confidence implies high accuracy: lessons learned from a decade of field-simulation studies examining...

In recent years, eyewitness experts have provided clear and consistent direction to the criminal justice community that if "pristine" non-suggestive procedures are used, high confidence suspect-identifications imply accuracy: Not a perfect relationship, but much stronger than previously thought. However, when suggestion is present, high-confidence suspect identifications no longer provide reliable evidence of accuracy. Recently, some experts have promoted a much broader picture of how and when confidence implies accuracy to the legal community, noting that confidence can still reliably predict accuracy even when suggestive procedures like showups are used. Although this observation is statistically correct, its practical meaning may be lost in translation to legal audiences, who may interpret this as support for treating high-confidence showup identifications as strong evidence of witness accuracy. In this paper, we argue for researchers to be cautious in how they disseminate research results to clearly specify the limitations on how new findings can be used to evaluate eyewitness evidence in court. In making this argument, we review findings from numerous experiments employing the field-simulation paradigm, which clearly illustrate a situation where a positive correlation between confidence and accuracy exists, but at the same time, nearly half of all false-identifications are made with 90-100% confidence.