Can within-individual consistency in episodic recall be used to retrospectively estimate eyewitness memory?

Eyewitness testimony is widely used in forensic scenarios and plays a crucial role for jurors in convictions. However, in the absence of ground truth for the crime, sometimes it is difficult for investigators to assess the accuracy of a particular eyewitness' testimony. In this study, we investigated whether it is possible to retrospectively assess an eyewitness' memory accuracy for the crime event (which presumably has no ground truth), with another staged mock event (that has ground truth). This possibility hinges upon the assumption that individuals are consistent in their memory performances across different episodic events. To test this assumption, our participants first witnessed a burglary theft on video, and were asked to recall details of the crime 24 h later. After one week, participants witnessed another burglary theft (i.e., the probe event) that were either visually dissimilar (Experiment 1) or highly similar (Experiment 2) as the target crime, and were tested on their memory accuracy 24 h later. In both experiments, we observed significant correlation between memory accuracies of the crime and the probe event, and such correlation seemed to be mostly driven by participants' memory for details that are central to the crime rather than peripheral details. Importantly, Experiment 2 not only replicated findings from Experiment 1, but also showed stronger correlation, suggesting that highly similar probe event may be preferable for field use. These results demonstrate that individual differences, as well as its consistency across multiple events, can be capitalized for eyewitness screening.