The N=1 analysis differs from a typical analysis of variance in that there is no within-cell error term. Thus interaction terms are used as estimates of error variance. If the interaction term in question represents a significant interaction, the F tests will be conservative. Tukey's test for nonadditivity will detect a common form of interaction.…

This paper describes NEOMYCIN, a computer program that models one physician's diagnostic reasoning within a limited area of medicine. NEOMYCIN's knowledge base and reasoning procedure constitute a model of how human knowledge is organized and how it is used in diagnosis. The hypothesis is tested that such a procedure can be used to simulate both…