Analysis of the results of chemotherapy in patients with carcinoma of unknown primary site is complicated by the small sizes of most treatment series and patient heterogeneity. Careful evaluation of clinical and pathologic information may identify patients with a relatively high likelihood of response to systemic therapy. This includes patients in whom immunohistochemical studies or electron microscopy, or both, suggest a likely tumor type responsive to systemic therapy, such as prostate cancer, lymphoma, or a neuroendocrine tumor. Clinical evaluation can also identify potentially responsive patients, particularly those with clinical features in common with the extragonadal germ cell tumor syndrome. For patients who do not fit into these more treatable categories, most combination chemotherapy programs have response rates of less than 30% and median survivals of less than one year. Randomized trials have not established any clearly superior chemotherapy program. Regimens containing both Adriamycin (doxorubicin) and mitomycin-C produce response rates of approximately 25% but are associated with the possibility of severe hematologic toxicity, and rarely a syndrome resembling the hemolytic-uremic syndrome. The choice between chemotherapy and supportive care only must be individualized, and the latter option is appropriate for many patients. More detailed clinical and pathologic analyses in conjunction with clinical trials, particularly employing newer diagnostic techniques, are vital to provide better prospective data from which to identify relevant clinical subsets that allow an estimate of an individual patient's likelihood of response and the suitability of systemic chemotherapy.