UC San Diego

In recent years, the field has found that the confidence-accuracy relationship for positive identifications (ID) made from a police lineup is often strong while the relationship for lineup rejections is typically much weaker. The reason for this asymmetry remains unclear. Here, we report results from signal-detection-based simulations and models, as well as from mock-crime lineup experiments, to help explain why this is often observed. When a face is identified from a photo lineup, the selected face is presumably the one that generates the strongest memory signal, with confidence presumably being determined by the strength of the signal associated with that face. When a lineup is rejected, the entire set of faces is collectively rejected due to no face generating a sufficiently strong memory signal to be identified. One theory is that confidence for rejections is determined by the average strength of the memory signals instead of the singular memory signal generated by the most familiar face. Averaging could wash out what would otherwise be a strong confidence-accuracy relationship. Chapter 1 investigated whether changing the lineup task such that participants reject only a singular face instead of a set of faces will strengthen the confidence-accuracy relationship for rejections. We found support for this hypothesis. Chapter 2 used multiple data sets for an in-depth modeling paper investigating whether the averaging of signals is the basis for confidence in a rejection. Our model-fitting analysis found that confidence in a lineup rejection is not based on the average signal and is instead based on the most familiar face, just as is the case for positive IDs. Chapter 3 investigated whether response bias, not averaging, may be a determinant of the strength of the confidence-accuracy relationship. Inducing a more conservative response bias should theoretically weaken the relationship for positive IDs while strengthening it for lineup rejections because a conservative criterion shift increases the range of possible memory signals associated with that decision. Our results support this prediction, showing that the degree of range restriction directly corresponds to the strength of the confidence-accuracy relationship for lineup rejections.