Part of a sustained program that has involved the design of personally tailored information systems responsive to the needs of scientists performing common research and teaching tasks, this project focuses on the procedural and content requirements for accomplishing need diagnosis and presents these requirements as specifications for an interactive, computer-based, knowledge delivery system for scientific researchers. A discussion of selected concepts from cognitive science and artificial intelligence describes their use as models for characterizing the research process or to specify procedural requirements for the system. Applications of these concepts in two manual "simulations" of the diagnostic process are described, including the research project as a procedural script (illustrated with scripts in microbiology and industrial psychology), the researcher as a problem solver, the diagnostic system as a problem solver, and the researcher's knowledge needs as schemata/frames. A 7-step process is described which is designed to successively narrow and define the researcher's current problem and his knowledge surrounding that problem, thus providing a basis for developing a search strategy, selecting relevant knowledge, and developing information products to fill knowledge gaps. Fifteen references are listed, and appendices provide examples of knowledge schemata. (RAA)