Two regimens of neoadjuvant chemotherapy for previously untreated patients with locally advanced head and neck cancer were compared with the goal of identifying a regimen with a greater than 50% complete response (CR) rate. Patients with a performance status of 0 to 2 and normal end-organ function were randomized to receive either four cycles of neoadjuvant methotrexate, cisplatin, and continuous infusion 5-fluorouracil (5-FU) (MPF) (arm A), or four cycles of bleomycin, cisplatin, and methotrexate (PBM) alternating with cisplatin and 5-FU (PF) (arm B). Patients with a performance status of greater than 2 or a carbon monoxide diffusion capacity of less than 50% of the predicted value were assigned to the arm A regimen but were analyzed separately (arm C). Local therapy consisted of surgery (for patients with resectable disease) or radiation therapy followed by two cycles of adjuvant chemotherapy with the regimen that was administered initially. Of the 42 patients who were evaluated, 16 were randomized to arm A, 13 to arm B, and 13 to arm C. The clinical CR rate was 19% on arm A (95% confidence interval, 0% to 38%), 39% on arm B (95% confidence interval, 12% to 66%) (P = 0.41), and 54% on arm C (95% confidence interval, 27% to 81%). At a median follow-up time of 35 months, the 2-year actuarial survival rate was 61% on arm A, 69% on arm B (the P value was not significant), and 38% on arm C. The 2-year survival rate for all 42 patients who were treated was 57% and the median survival time was 31 months. Toxicities of neoadjuvant chemotherapy on all arms consisted of mild to moderate myelosuppression and renal toxicity. The incidence of moderate to severe mucositis was significantly higher on arm A than arm B (P = 0.02). Two cycles of adjuvant chemotherapy were administered to only 11 of 42 patients due to patient refusal or cumulative toxicity. In conclusion, both neoadjuvant chemotherapy regimens resulted in similar response and survival rates, but mucositis was more severe with arm A. However, since neither regimen was likely to cause a CR rate of greater than 50%, this study was closed to further patient accrual.