The Rule Out Procedure: A Signal-Detection-Informed Approach to the Collection of Eyewitness Identification Evidence

Visual recognition memory has a remarkable capacity to discriminate between previously seen and novel items. Yet, research on eyewitness lineups suggests that memory is useful for detecting culprit presence, but less useful for detecting culprit absence. We show that this asymmetry is predicted by the equal-variance signal-detection model. When witnesses reject lineups, they provide a global confidence rating that none of the lineup members is the culprit. These ratings do not scale match-to-memory for the suspect and are low in diagnostic value. Consequently, the equal-variance signal-detection model predicts that a one-person showup will have better discriminability than a six-person lineup. A large-scale experiment (N = 3281) supported that prediction. However, a modified lineup in which participants were asked to follow categorical identification decisions by assigning a confidence rating to each lineup member had better discriminability than both the showup and the standard simultaneous lineup. We call this modified lineup the rule out procedure. Results also revealed a relatively weak confidence-accuracy relation for global rejections of lineups, but a much stronger confidence-accuracy relation for rejections of individual faces. Past failures to detect suspect innocence with lineups should be attributed to flawed design, not to limitations of visual recognition memory.