This manuscript reports the results of the first cooperative study on primary central nervous system lymphoma (PCNSL) in Japan. Of 196 patients registered, 170 were judged as having PCNSL. No patients were immunocompromised. Of the 170 patients with PCNSL, 93 were males and 77 were females. The mean was 56.7 years. One hundred and nineteen tumors were confirmed histopathologically, and 51 were diagnosed by neuroimaging alone. All the tumors were non-Hodgkin's lymphoma. According to the Working Formulation for Clinical Usage (WF), 96 out of 119 tumors were classifiable: 53 were diffuse large cell type (55.2%), 17 immunoblastic type (17.7%), 9 diffuse small cleaved type (9.4%), 6 diffuse mixed type (6.3%), 5 polymorphous type (5.2%), 5 small lymphocytic type (5.2%) and 1 small non-cleaved type (1.0%). Of 21 tumors studied immunohistochemically, 18 were B-cell type and 3 were T-cell type. Irradiated patients (144) survived significantly longer than non-irradiated patients, (median survival time, MST: 19.2 and 2.7 months, respectively; p < 0.001). There was a remarkable difference in survival among patients of the intermediate lymphomas; MST (18 months) of patients with large cell lymphoma was significantly shorter than MST (over 96 months) of patients with other intermediate grade lymphomas (small cleaved and mixed) (p < 0.001) and had no significant difference from MST (9 months) of patients with high grade lymphomas. If patients were irradiated with more than 40 Gy, higher doses and different modes of irradiation brought no further survival advantage. Chemotherapy was performed in 87 of 144 irradiated patients (60.4%). No regimens were effective in prolonging survival. Of 144 irradiated patients, a complete or partial response to initial treatment was demonstrated in 91 (63.2%) and 43 patients (29.9%), respectively. Improvement in performance status was confirmed in 82 patients (57.0%). Despite a good response to initial treatments, 88 out of 144 evaluatble patients have died of PCNSL (MST: 19 months). Multivariate analysis based on the Cox hazard model revealed that histology of tumor, age at onset, performance status, and radiotherapy were prognostic factors. Neither chemotherapy nor mode of surgery was a beneficial factor.