The proposed project is a competing continuation of our National Library of Medicine research grant to study medical knowledge base construction as a discipline (ROl LM04622). Unless we determine the most efficient and useful methods for knowledge base construction and maintenance, the future of medical expert systems will be bleak. Our previous work explored techniques for multi-centric creation of medical knowledge base entries. During the proposed extension, we will determine desiderata for selecting individuals to be knowledge base builders and maintainers, evaluate the variability in the way physicians construct knowledge bases, and determine how semi-automated tools can assist in knowledge base construction and maintenance. Our studies will focus on the QMR KB as a prototype of medical diagnostic derision support programs. It is likely that the American College of Physicians will adopt this KB as the seed for a national medical knowledge base in the near future.