A staged crime paradigm was used to evaluate the costs and benefits of 'fair' lineups. Unsuspecting witnesses (n=90) to a staged crime were given the opportunity to identify a criminal from relatively fair or unfair lineups. One fair and one unfair lineup contained a picture of the criminal, while one fair and one unfair lineup contained a picture of an innocent suspect who resembled the criminal. In the criminal justice system only identifications of the criminal or the innocent suspect lead to further proceedings. The results indicate that high-similarity lineups (those in which lineup members resemble the criminal) produced less identifications of the criminal and of the innocent suspect than low-similarity lineups. However, the reduction in identifications of the criminal was much less dramatic than the reduction of identifications of the innocent suspect. A Bayesian analysis of the data led to the further conclusion that identification evidence obtained from relatively fair high-similarity lineups in superior to similar evidence obtained from relatively unfair low-similarity lineups. The cost (in lost convictions of guilty suspects) of using fair lineups (high similarity) appears to be rather small. Tabular data, footnotes, graphs, a note, and 13 references are included. (Author abstract modified)