Several archival studies of eyewitness identification have been conducted, but the results have been inconsistent and contradictory. We identify some avoidable pitfalls that have been present in previous analyses and present new data that address these pitfalls. We explored associations among various estimator variables and lineup outcomes for 833 "real life" lineups, including 588 lineups in which corroborating evidence of the suspect's guilt existed. Suspect identifications were associated with exposure duration, viewing distance, and the age of the witness. Nonidentifications were associated with the number of perpetrators. We also consider some of the inherent, unavoidable limitations with archival studies and consider what such studies can really tell researchers. We conclude that differences in sampling prohibit sensible comparisons between the results of laboratory and archival studies, and that the informational value of archival studies is actually rather limited.