A statistical analysis was performed on 40 patients with squamous cell carcinoma of the tongue and mouth floor, which could be followed for 6 months or more after initial treatment in the Department of Otorhinolaryngology, School of Medicine, Keio University during the 14 years from 1983 to 1996. The 5-year survival rate determined by the Kaplan-Meier method for each stage was 100% for Stage I, 77.8% for Stage II, 60.0% for Stage III and 44.4% for Stage IV. Thirteen suffered a relapse after initial treatment and patients with relapses among them have all survived after the subsequent salvage surgery. In contrast, in nine patients with cervical relapse, however, the 5-year survival rate was 11.1% with an unfavorable prognosis. This confirmed that suppressing cervical relapses is important for treating tongue and floor mouth cancers. The treatment strategy in our department is characteristic of positive enforcement of prophylactic neck dissection in the surgery and introduction of neoadjuvant chemotherapy (NAC) in the chemotherapy. Prophylactic neck dissection was performed in the 17 patients and no relapse was observed on the side of prophylactic neck dissection. NAC was performed on 26 patients in consideration of suppressed minute metastases and preserved function and 24 determinable cases were statistically analyzed. Among patients who had received NAC, the oral function was successfully preserved without surgical intervention in six patients both patients who showed complete response (CR) and four out of 14 patients who had a partial response (PR) following NAC. This may indicate that the oral function could be preserved in those patients who exhibited CR following NAC, but that preservation could be difficult in patients who exhibited PR. In addition, concerning the accumulated 5-year survival rate in relation to the effect of NAC, responders (CR + PR) accounted for 90.9% and non-responders (no change + progressive disease following NAC) for 15.0% with a very good outcome noted in the responder group. These figures suggest that responders may have a significantly good prognosis in the multivariant analysis including additional background factors before treatment as well. Accordingly, the present therapeutic measures for non-responders must be reexamined and performed more carefully and accurately as compared with those for responders.