The aim of this study was to prospectively analyze the distribution of neck metastases and the outcome in patients surgically treated for tonsillar carcinoma in order to deduce implications for selective neck dissection. The criteria for inclusion in the study were (1) previously untreated, histologically proven, resectable squamous cell carcinoma of the tonsil, (2) curative surgical intent on the primary tumor and neck, (3) no history of prior head and neck cancer, (4) absence of synchronous second primary in the upper aerodigestive tract, lung and esophagus, (5) absence of distant metastases. Fifty-eight previously untreated consecutive patients with tonsillar squamous cell carcinoma were included in this prospective study. Among 22 patients with clinically negative cervical lymph nodes, 4 patients (18.2%) had metastatic lymph nodes on pathologic examination. Occult node metastases were mainly located in ipsilateral II level. No occult metastases occurred at levels I and V. Among 36 patients with clinically positive cervical lymph nodes, 3 patients (8.3%) had an occult pathologic metastatic involvement of cervical lymph nodes of ipsilateral level V. Level I was free of lymph node metastases. Clinical N category >N2a (p=0.003), nodal metastases to levels III (p=0.026) and IV (p=0.009) were significantly related to level V nodal metastases. The 2 and 5 years actuarial disease-free survival was 82.7% (95% CI 71.2-93.5%) and 58.3% (95% CI 36.7-79.9%), respectively. The actuarial recurrence-free survival was 87.9% (95% CI 78.9-96.8%) and 72.2% (95% CI 53.9-90.5%) at 2 and 5 years, respectively. Our findings support the role of a selective lateral neck dissection in the management of clinically N0 necks and in selected N+ necks (N1 and N2a disease located at level II) in patients with tonsillar carcinoma without oral involvement.