The tool consists of a large number of examples of five basic features: hair and forehead, eyes and eyebrows, nose, mouth, and chin. The witness is encouraged to choose those features which most resemble the suspect's in order to produce a composite face which may then be modified by the police artist until a satisfactory likeness is developed. Marked discrepancies exist between levels of face recognition and accuracy of face recall registered by Photofit. One of the reasons for this may be that Photofit demands that subjects initially identify features in isolation, a strategy incompatible with normal coding processes. To test this theory, two experiments were conducted. The first experiment, which involved 36 women volunteers, compared the normal feature-based approach to face building with a condition which permitted subjects to work from an established, average looking face. No superiority for the latter condition could be established. The second experiment, involving 16 undergraduate students, contrasted the standard building method with one in which subjects were offered a variety of facial types as starting points. Again, accuracy was no higher under the novel procedure. It was concluded that results did not justify altering current Photofit methods which emphasize feature selection prior to development of the total facial composite. Three tables, 4 photographs, and 15 references are provided. (Author abstract modified)