We prospectively compared the value of MRI and clinical palpation for detecting cervical metastases in patients with primary cancer of the head and neck. Sixty patients with squamous cell carcinoma of the upper aerodigestive tract were evaluated with MRI and clinical palpation before undergoing a total of 81 neck dissections. The results of preoperative clinical palpation and MRI were compared with the histopathologic outcome. The sensitivity and specificity were 75.6% and 97.5%, respectively, for clinical palpation and 73.2% and 95%, respectively, for MRI. The rate of occult cervical metastasis was 24% with clinical palpation and 26.8% with MRI. The use of MRI did not improve the rate of early detection of occult metastasis, nor did it improve the detection of extracapsular spread. Our findings show that we could not depend on palpation or MRI alone to determine the need for elective neck dissection. However, MRI can be used to improve the preoperative grading of cervical lymph nodes. In selected cases, this may direct surgeons to convert the treatment plan to choose a more conservative neck dissection or, after sentinel node sampling and frozen-section control, to convert the treatment to a more radical dissection.