To assess the completeness of free recall descriptions of criminals, the first study required participants to describe persons in staged crime incidents. Witnesses often provided vague descriptions and left out information such as sex and race that they certainly noticed. When describing criminals, witnesses were most likely to describe clothing. Hair color and height were also frequently described, although these descriptions were often vague as well. The second study tested the superiority of foil selection by match-to-description rather than by similarity-to- suspect as a method of lineup construction. Study participants were tested on their memory of a man who had just spoken to them. The target man was selected least often from the similarity-to- suspect array. The third study involved a staged crime; witnesses completed open-ended descriptions of the criminal, responded to specific requests for further information about his appearance, and were asked to identify the criminal from photo arrays. Correct identifications were made equally often for lineups with the criminal absent. When lineups were presented sequentially, correct rejections increased significantly. False identification rates appeared to be inflated by selecting lineup foils who fit vague descriptions of the criminal but otherwise differed from the suspect on default values. 15 references and 2 tables