Can the traditional lineup procedure be modified in such a way as to still secure positive identifications of guilty persons while minimizing the risk of misidentifications? Witnesses interacted with a ‘culprit’ who they were later called upon to identify from a 20‐person sequential video lineup. Both culprit‐present and culprit‐absent lineups were employed and witnesses could choose more than one suspect. In 61 culprit‐present lineups 43% chose only the suspect, while in 93 culprit‐absent lineups no‐one did. In culprit‐present lineups 25% chose him along with foils, while 5% did so in culprit‐absent lineups. Larger lineup size, and the ability to make multiple choices, helped lower the probability of choosing only an innocent suspect. Using Baysian analysis, the probability that a chosen defendant was innocent, based on the lineup alone, equalled 0.03, compared to 0.258 in traditional lineups. © 1998 John Wiley & Sons, Ltd.