Objective: To compare the efficacy and treatment outcomes in patients with tonsillar fossa cancer using surgery or radiation as a single modality therapy. Methods: From 1971 to 1991 239 patients with oral pharyngeal cancer were treated at Roswell Park Cancer Institute. Of these patients 90 had tonsillar carcinoma. Seventy-six of these patients received either surgery (SA) (n = 56) or radiation therapy (RA) (n = 20) as single-modality therapy and are the subject of this review. All patients in the radiation arm of this review were surgical candidates who declined primary surgical therapy. Results: Sixty-three percent of the SA and 80% of the RA treatment groups presented with either stage III or stage IV disease (P ⩽ .05). Forty-seven percent of the SA group and 52% of the RA patients had clinically positive regional disease at initial presentation. There was a predictable pattern of nodal presentation, with level II the most frequently involved region. The rate of occult metastasis was 27% and was evenly distributed between T1 and T4 disease. The overall local control rate in the SA group was 75%, compared with 60% in the RA group (P value was not significant). The disease-specific survival (all stages) was 61% in the SA group and 37% in the RA group (P ⩽ .05). The disease-free survival for stage III and stage IV disease in the SA group was 47% and in the RA group 27% (P ⩽ .05). Survival measured against clinical response to radiation therapy, in complete responders (all stages) was 83%; by contrast there were no survivors past 24 months in the partial response group (P ⩽ .001). Conclusion: The results from this study suggest that for early disease (stage I/II), surgery or radiation therapy as single-modality treatment is equally effective. For advanced disease radiation therapy is inferior to surgery as a single-modality treatment, as measured by ultimate survival and the local control of disease. There is, however, a subset of patients with advanced disease who respond to radiation therapy and whose survival is equivalent to our surgical cohort of patients.