Improving Witnesses\u27 Predictive Confidence Judgments by Enhancing Test Domain Familiarity

Recent research on witnesses’ pre-identification confidence (“predictive confidence”) suggests that these judgments are moderately related to identification accuracy when witnesses experience encoding variability and appropriate statistical techniques are used. However, other research shows that under ecologically valid conditions involving a single identification, the relationship between predictive confidence and accuracy deteriorates. One potential explanation for this lack of a meaningful confidence-accuracy relationship is that witnesses are unfamiliar with the lineup task leading them to underestimate its difficulty. Identification difficulty is partly determined by the similarity of lineup fillers to the suspect, which witnesses cannot anticipate when they make a predictive confidence judgment; in light of this, the current study tested whether witnesses’ predictive confidence could be improved by exposing participants to “sample fillers” that matched (or did not match) the similarity of fillers in the actual lineup. The current study also explored whether witnesses’ self-reported memory strength predicted their identification accuracy. Finally, to overcome limitations of using continuous measures (such as memory strength or predictive confidence) to make a dichotomous decision as to whether to show a witness a lineup, the present experiment evaluated whether witnesses’ dichotomous judgments about their ability to make an accurate identification decision could predict their subsequent identification accuracy. Witnesses viewed a mock crime under one of eight encoding conditions, and one week later were shown “good”, “poor”, or no sample fillers prior to reporting their predictive confidence, memory strength, and dichotomous lineup prediction, and attempting to make a lineup identification. Results indicated that viewing good sample fillers did not significantly improve the predictive confidence-accuracy relationship, and that although exposure to either type of sample filler decreased witnesses’ predictive confidence, they were largely overconfident relative to their level of accuracy (thereby harming calibration). Witnesses’ self-reported memory strength and dichotomous prediction also failed to successfully differentiate accurate and inaccurate eyewitnesses. Results suggest that real-world decisions as to whether to present witnesses with a lineup based on their predictive confidence are misguided. Implications of retention interval on the use of predictive measures are discussed