Twenty-nine evaluable patients with metastatic carcinomatosis in whom initial workup failed to reveal the primary site were entered into this trial. Patients with histological evidence of adenocarcinoma (n = 15) received FAM, while patients with anaplastic carcinomas (n = 14) were given AVM. Pretreatment characteristics were similar for the FAM- and AVM-treated patients with regard to age and sex, but 47% of patients on FAM had liver metastases as compared with 36% for the AVM group. Of the 14 patients on AVM, 1 (7%) achieved a complete response lasting 16 months, and 3 patients (22%) achieved a partial response for 10, 12 and 20 months, respectively. No patient on FAM reached a complete response, and only two patients (13%) showed a partial remission for 7+ and 24 months, respectively. The median survival for the AVM patients was 8.5 months, not significantly different from a median of 5 months for the FAM-treated group. AVM caused substantial myelotoxicity, resulting in five hospitalizations for leukopenia and fever; the FAM regimen was better tolerated with no episodes of leukopenic fever. AVM appears to be more effective than FAM in the treatment of carcinomas of unknown origin. A higher response rate was achieved with AVM, despite the fact that patients on this combination had undifferentiated carcinomas and a larger proportion of three or more metastatic sites (36 vs. 13% on FAM), and received a lower percent of the planned dose than did the FAM patients. Further clinical trials to fully establish the role of vinblastine in the treatment of metastatic carcinomatosis of unknown origin seem warranted.