The three programs -- WinID2, WinID3, and CAPMI4 -- were tested in the identification of victims in a computer-simulated disaster. One independent variable was the number of victims; 100, 200, and 300 simulated victims were analyzed separately. A second independent variable was search direction, which involved searching postmortem-to-antemortem and antemortem-to-postmortem dental evidence separately. The third independent variable was software program, with each program analyzed separately. Software program performance was measured by mean rank of correct match and total correct matches. Postmortem-to-antemortem search was clearly superior in results to antemortem-to-postmortem search. The number of victims made a significant difference only when there was little or no dental pathology present. WinID2 and WinID3 performed similarly. The CAPMI4 and one dataset of the WinID (without distinguishing between versions), i.e., "Most Dental Hits," tied for best overall performance. 4 tables, 5 figures, and appended discussion of "The Problem of Tied Ranks"