Shortly after viewing videotape of a staged assault, some subjects rated a group of 18 photos for attractiveness; whereas, other subjects attempted to identify the assailant publicly or privately in the same (target-absent) sent of 18 'mugshot' photos. The remaining subjects had no intervening task. Two days later all subjects attempted to identify the assailant from a six-person, target-present photo lineup which contained their mugshot choice (if made) was well as the original assailant. Lineup identifications were significantly less accurate for subjects who had viewed the intervening mugshots (33 percent accuracy) than those who had only rated the intervening photos for attractiveness (64 percent accuracy) or had seen no intervening photos (69 percent accuracy). When viewing lineups, experimental subjects tended to remain committed to their earlier action of choosing or not choosing anyone from the mugshots. More witnesses who had publicly stated their mugshot choice reproduced their (incorrect) choice at the lineup (78 percent) than did those who made a private mugshot choice (45 percent); this difference was not statistically reliable, however. Witnesses' confidence in their lineup decisions was not significantly related to overall accuracy. Contrary to predictions from self-perception theory, confidence in one's decision decreased significantly as further identifications were attempted. 2 tables, 38 references. (Author abstract modified)