The great variety of ills that man is heir to and the tremendous number of symptoms associated with such diseases make the task of medical diagnosis more speculative than scientific. There is an expanding interest in the possibilities of using computers to help make medical diagnoses more quickly and accurately, but here again, the vastness of the task complicates the establishment of a general quantitative diagnostic procedure. Suggested here are models which simplify this procedure into a sequence of decision processes of manageable size for a statistical approach. Perhaps automatic data processing can make an initial inroad into the area of medical diagnosis through the accumulation and continuous revision of base rates appropriate to the various levels in the diagnostic decision process.