Eyewitnesses to multiple-culprit crimes are often asked to try to identify the culprits from different lineups during a police investigation. In 2 experiments (N = 557), we show that disconfirming feedback after an identification attempt for 1 culprit can impair identification performance on a subsequent lineup for a different culprit. In each experiment, witnesses viewed a simulated, 2-culprit crime, followed by 2 police lineups: A culprit-absent lineup for 1 culprit and either a culprit-present or culprit-absent lineup for the second culprit. Following the first lineup, witnesses received disconfirming feedback or no feedback. For witnesses who correctly rejected the first lineup, disconfirming feedback impaired identification performance on the subsequent lineup. For witnesses who incorrectly chose someone from the first lineup, disconfirming feedback impaired subsequent performance when the feedback unambiguously implied poor ability to identify the culprit in the first identification test (Experiment 2) but not when it could have been interpreted as implying poor criterion setting (Experiment 1). Across both experiments, disconfirming feedback also reduced the difference in confidence between correct and incorrect identifications. These results add to evidence that postidentification feedback can affect subsequent identification performance by influencing witnesses’ beliefs about their ability to identify a culprit. Current policy recommendations state that postidentification feedback should be withheld from witnesses until confidence has been documented. These should be updated to recommend withholding feedback for longer if the witness may be asked to view additional lineups, and to ensure that lineup administrators are blind to the results of any previous lineups.