A knowledge-based intelligent system being developed as a computer-assisted tutoring tool that can be used by teachers and students of cervical cytology is described. This system uses a combination of frames and rules as knowledge representation. Combination data structures called "prototypes" characterize the typical cytoplasmic and nuclear features of classes of cervical cells by using a specific prototype for each hypothesized cell type. A hypothesis-directed approach is used for problem solving. Rules are associated with each prototype or hypothesis. There are two major kinds of rules: deducing rules, used for deducing certainty factors (expressed as degrees of certainty of the hypotheses), and strategy rules, used for guiding the user in reaching a conclusion about the test cell. With an easy-to-use interactive menu, the user inputs a series of sets of qualitative cell features, which the system utilizes at various inference stages and finally arrives at a decision about the cell type.