It is generally agreed that proper pre-lineup instructions can reduce the rate of mistaken identifications of innocent suspects. However, the exact nature of these instructions has not been empirically established. We compared the effects of the detailed pre-lineup instructions recommended by the US Department of Justice to a simple instruction that the perpetrator may or may not be present. Both forms of unbiased instructions significantly reduced mistaken identifications when compared with biased instructions. Comparison of performance using the utility-based analysis deviation from perfect performance demonstrated that unbiased instructions are to be preferred under assumptions of low base rates or high costs of mistaken identifications and do as well as biased instructions under assumptions of high base rates or low cost of misidentification.