Although many questions remain unanswered, recent clinical and pathologic studies have shed considerable light on the subject of carcinoma of unknown primary site. It is now clear that some patients in this group have extragonadal germ cell tumors. This is suggested by the superior treatment results in patients with clinical features of extragonadal germ cell tumor and is confirmed by the finding of the diagnostic chromosome abnormality in tumor cells of some patients. These patients have tumors that are unrecognizable using all available pathologic techniques other than molecular genetic analysis; most patients also do not have elevated serum tumor marker levels. Young men with poorly differentiated carcinoma located predominantly in the mediastinum or retroperitoneum should be strongly suspected of having germ cell tumors; chromosomal analysis should be obtained if possible, and these patients should be treated as for germ cell tumor. It is clear that some responsive patients with poorly differentiated carcinoma do not have extragonadal germ cell tumors. A few patients initially thought to have poorly differentiated carcinoma actually have non-Hodgkin's lymphoma. With the widespread availability of immunoperoxidase staining for LCA, this diagnostic error should be minimized. Other responsive patients have poorly differentiated neuroendocrine tumors. The nature and spectrum of neuroendocrine tumors is still being defined; however, our initial documentation of cisplatin responsiveness has been confirmed, even in poorly differentiated neuroendocrine tumors with a known primary site. It is likely that additional responsive subgroups also exist but have not yet been identified. With the availability of a diagnostic chromosomal marker, the answers to other questions regarding the relationship of poorly differentiated carcinoma and germ cell tumors will soon be forthcoming.(ABSTRACT TRUNCATED AT 250 WORDS)