Two hundred and ninety-eight evaluable patients with non-Hodgkin's lymphoma were stratified according to histology, treated with either BCNU, cyclophosphamide, Oncovin (vincristine), and prednisone (BCOP) or cyclophosphamide, Oncovin (vincristine), and prednisone (COP), and evaluated at 3 months. Those with a good partial (PR) or complete response (CR) were then separated and randomized to be treated with either cycle-active therapy (methotrexate, cytosine arabinoside, and 6-thioguanine) or more induction therapy with COP or BCOP. Patients not achieving a good PR at 3 months received cycle-active therapy. The results indicate (a) that there is a significant advantage for good over poor histologies with regard to good PRs at 3 months; (b) that the addition of cycle-active therapy (as administered in this study) is of advantage when the tumor has been significantly reduced only for patients receiving COP induction; and (c) that BCOP has an advantage over COP in diffuse histiocytic lymphoma where the percentage of CRs, their durability, and subsequent survival are superior for patients treated with BCOP. Since this lymphoma accounts for about 25% of all non-Hodgkin's lymphoma patients, this regimen represents a useful tool for the chemotherapist.