From a large series of patients who had a "commando" operation for mouth cancer, two groups were selected that were similar in all but one respect. One had histologic evidence of cervical node metastasis that was unsuspected clinically, and the other had both clinical and laboratory evidence of such spread. In those patients who died with uncontrolled mouth cancer, the incidence of recurrence in the neck was higher in the patients who had a therapeutic neck dissection for palpable metastases. Moreover, there was a trend toward a higher five-year cure rate in those who had elective neck dissection before cervical metastases were palpable. This is presumptive, but inconclusive, evidence in favor of elective, radical neck dissection.