An acute abdominal pain program run on a microcomputer is described and the experiences tabulated for a series of 194 patients seen in a general hospital setting. The initial diagnostic accuracy of the program compared favorably with that of attending physicians and house staff, and suggests that improvements in the program can lead to a more effective and more accurate abdominal pain program. The possibilities for developing other programs, particularly chest pain, fever, and other global categories, are obvious. We are presently working on a chest pain program and a diagnostic strategy program. Based on the results reported here, we believe that several factors could improve program accuracy. For example, with additional clinical studies and refinement of the program structure, with more expert knowledge, and with further algorithmic development, the program could be made to outperform the average clinician and possibly approach the level of true clinical experts in abdominal pain diagnosis by mimicking their analysis. Indeed, since a program of this type is capable of incorporating the expertise of many different clinicians, it has the potential of outperforming any given expert in specific cases.