Abstract
When a witness selects a face from a police lineup as being familiar (a positive ID), confidence is theoretically based on the face that generates the strongest memory signal (the MAX face). When a lineup in rejected, however, no face is singled out. An interesting theoretical question that has yet to be answered is whether confidence in the case of rejections still corresponds to the (unchosen) MAX face or if confidence instead derives from an aggregate memory signal associated with the set of rejected faces (such as the average memory signal). The possibility that qualitatively different memory signals might be used for confidence in positive IDs vs. lineup rejections is suggested by a curious mismatch between the strong confidence-accuracy relationship typically observed for positive IDs vs. the weak confidence-accuracy relationship typically observed for lineup rejections. In our modeling study, we fit different versions of two standard signal detection models for lineups (the Independent Observations model and the Ensemble model) to data collected from simultaneous lineups that were presented as part of several unrelated projects (combined n = 7,874). One version of each model assumed that confidence in positive IDs and lineups rejections were both based on the MAX memory signal. A second version of both models assumed that while confidence in positive IDs was based on the MAX signal, confidence in lineup rejections was instead based on an average memory signal. Across all data sets, we found that a model assuming an averaging rule for confidence for rejections performed worse than the corresponding model assuming a MAX rule for rejections. These results suggest that confidence in lineup rejections (like confidence in positive IDs) is based on the MAX face. Thus, an explanation for the asymmetric confidence-accuracy relationship for lineup rejections compared to positive IDs remains elusive.