The development of effective autologous cell transfer therapies for the treatment of patients with cancer has been difficult, in part because the cells used to treat each patient are different, as are the patient's tumor and immune status. Much can thus be learned by sequential treatments of the same patient with the same cells, making single modifications in the treatments to determine which factors are critical. The authors have treated a single patient with five sequential administrations of the same cells with minor modifications in the mode of administration and the immune status of the patient. The treatment of this patient strongly suggested that 1) the highly avid recognition of tumor antigens in vitro by a transferred lymphocyte population does not necessarily predict in vivo antitumor activity; 2) the administration of highly avid antitumor autologous lymphocyte populations can be far more active in mediating tumor regression in vivo when administered after nonmyeloablative chemotherapy than when administered without this prior chemotherapy; 3) intra-arterial administration of highly avid antitumor autologous lymphocytes into the blood supply of the tumor can be more effective in mediating tumor regression than the intravenous administration of these same tumor infiltrating lymphocytes; 4) one mechanism of tumor escape from immunotherapy is loss of class I MHC antigen expression by the tumor due to mutation of the beta-2 microglobulin gene.