Memory demands on facial composite identification

Face composites created with FACES 3.0 composite software were used to identify target photographs under varying conditions of memory load. Composites created while viewing a mugshot were presented concurrently with a five‐photograph lineup of actual criminal suspects, immediately preceding the lineup, or with a filled delay between composite and lineup. Compared to a baseline using the mock witness technique, results demonstrated above‐chance identification of target photographs from their composites in all four experiments. Identification was highest with concurrent composite‐lineup presentation and lower under immediate, 30‐second‐delayed, and 4‐minute‐delayed conditions, with no significant identification differences across the three memory conditions. Confidence ratings in target selection only weakly predicted identification accuracy at best. These findings extend the limited research on computerized facial composite systems while addressing composite software use, efficacy, guidelines and limitations. Copyright © 2006 John Wiley & Sons, Ltd.