The ability to target malignant cells for cytotoxicity while sparing normal host tissues has proven to be limited. These limitations have resulted in unacceptable toxicity or insufficiently effective therapy. Continuing investigation of new, potentially useful cytotoxic agents must continue. An alternative approach, also worthy of study, is the selective protection of normal tissues. This approach, used in conjunction with available therapeutic agents, may open the therapeutic window and incrementally enhance the effectiveness of cytotoxic therapy. A variety of methods have been used to protect normal tissues selectively. Regional protection can be used for certain organ systems, such as the oral mucosa. Selective protection on a systemic level is more difficult but agents that seem to protect normal but not malignant tissues selectively are being developed. Among these is amifostine, which was originally selected by the U.S. defense department for study as a radioprotectant. Pre-clinical studies have suggested that amifostine is differentially concentrated in normal tissues but not in malignant tissues. Tissue-specific differences in the activity of alkaline phosphatase, which dephosphorylates amifostine to its active metabolite WR-1065, and in pH are thought to be involved in this relative specificity. Clinical studies indicate that amifostine can reduce the myelosuppression produced by cyclosphosphamide, the combination of cyclophosphamide and cisplatin, and, perhaps, carboplatin. The protective effects of amifostine on nonhematopoietic toxicities are being investigated. Future trials will investigate the integration of amifostine with cytokine-based supportive care in order to define the role of this potentially clinically useful cytoprotectant agent.