Efforts to improve the Photofit system have focused on evaluating its products and the range of component features rather than operators' training. In this study, 48 volunteer witnesses worked alone or in pairs with Photofit operators to produce composites of 4 different targets who were all clean shaven young men between 18 and 22 years old. A 2x2 factorial design was used with experienced versus novice operator and individual versus pair witness as major variables. Various measures were taken during the composite production to identify operator skills, and the resulting Photofits were evaluated for best likeness and composite quality. The experienced operator took longer than the novice for individuals, while the effect was reversed for pairs. Experienced operators also tended to elicit more elaborate descriptions of the target face. When working with individuals, the experienced operator was approximately 20 percent better than the novice in sorting accuracy. This supports the view that a skill element exists in composite production which should be identified and developed in police operators. No differences were observed in quality between individuals working alone or in pairs. Additional tests indicated that the experienced operator was superior in the initial selection of component features. Training should cover diverse witness conditions and focus on subskills needed in composite production. Tables and 14 references are supplied.