The therapeutic use of unmodified monoclonal anti-idiotypic antibody for human B cell malignancies has met with limited success. Some factors thwarting antibody attack can be identified, such as the presence of extracellular idiotypic immunoglobulin (Ig), and the escape of the target cell by antigenic modulation. Another possibility is a change in idiotypic determinants due to somatic mutation. For direct attack on tumor cells in vivo it might be necessary to use antibody derivatives: univalent antibodies will avoid modulation, and chimeric univalent antibodies consisting of mouse Fab' gamma linked to host Ig of appropriate subclass can be engineered to mediate particular effector functions while reducing immunogenicity. Another approach is to use toxin or isotope-bearing antibodies. However, the final eradication of tumor might involve the natural non-specific, and specific anti-idiotypic mechanisms of the host which should not be damaged by antibody therapy, and which appear to be involved in control of tumor progression in patients in long-term remission. Rapidly growing animal lymphomas presently available as models cannot mimic more than a small fraction of human lymphoma but they provide useful information for design of passive anti-idiotype therapy. However host anti-idiotypic immunity must be induced in such models by pre-immunization with purified idiotype. Such a procedure can generate effective anti-idiotypic immunity which is highly protective in mouse and guinea pig lymphomas. Analysis of mechanisms involved should give insight into the role of the idiotype networks in the behavior of human disease.