We constructed a diagnostic assistance system for long bone lesions using a relational data base management system, in which we categorized radiographic findings of all parameters in various lesions and stored them as degrees of possibility. Possible diagnoses of unknown lesions were listed by max-min calculation between the data base and findings of the lesions. Integration of parameters narrowed the scope of possible diagnoses. Fifty-six cases with long bone lesions were tested. Experts diagnosed correctly up to 64 percent, while the system showed a correct diagnosis rate of 43 percent as the first rank of possibility. However, the system gave 75 percent correct diagnosis when listed up to the third rank, and this increased up to 100 percent at the ninth rank. The rate of true diagnosis by experts did not exceed 73 percent. The diagnostic assistance system based on possibility analysis proved to be useful for suggesting differential diagnoses in diagnosing long bone tumorous lesions from plain radiographs.