The gastrointestinal tract is a common primary extranodal site for non-Hodgkin's lymphoma. There is however no uniform consensus on its pathological classification, clinical staging system and management. This paper reports the experience in the management of 425 Chinese patients with primary gastrointestinal lymphoma in Hong Kong from January 1975 to June 1993. There were 230 (54 per cent) males and 195 (46 per cent) females. Their median age was 53 years. The primary sites were: the esophagus in three (1 per cent), stomach in 238 (56 per cent), small intestine in 131 (31 per cent) and large intestine in 53 (12 per cent). According to the Working Formulation, there were 20 (4.7 per cent) small lymphocytic, 10 (2.4 per cent) follicular small cleaved cell, 15 (3.5 per cent) follicular mixed, five (1.2 per cent) follicular large cell, 40 (9.4 per cent) diffuse small cleaved cell, 50 (12 per cent) diffuse mixed, 181 (43 per cent) diffuse large cell, 30 (7.1 per cent) immunoblastic, five (1.2 per cent) lymphoblastic, 10 (2.4 per cent) diffuse small non-cleaved cell and 50 (14 per cent) unclassifiable lymphoma. Immunophenotyping was performed in 199 (47 per cent) patients: 90 per cent B-cell, 7 per cent T-cell and 3 per cent uncertain. According to a Manchester system, 81 (19 per cent) patients had stage I disease, 44 (10 per cent) stage II, 85 (20 per cent) stage III and 215 (51 per cent) stage IV. B symptoms were present in 275 (65 per cent) patients and bulky disease in 104 (25 per cent). Surgery followed by chemotherapy was the mainstay of treatment. Of the 408 patients treated, 63 per cent had a complete remission with relapse rate of 42 per cent. For those with complete remission, 47 per cent were free from disease at 5 years. The overall median survival of all patients was 45 per cent at 5 years. Multivariate analysis revealed that significant independent prognostic factors predicting better survival were young age of < 60 years, low grade histology, stage I and II disease and absence of bulky tumour. For gastric lymphoma, aggressive surgery did not significantly improve their outcome. Chemotherapy appears to play an important role in the management of gastrointestinal lymphoma. Better classification of the primary gastrointestinal lymphoma and more refined stratification of the patients according to the prognostic variables may allow individualization of treatment. Prospective randomized studies are essential to define the relative roles of surgery, chemotherapy and radiotherapy.