In a series of 100 patients with head and neck carcinoma, the preoperative histopathologic findings of palpation and magnetic resonance imaging were compared with regard to both laterality and lymph node level (I through V). The overall error for palpation in detecting affected sides was 32%. Gadolinium-enhanced magnetic resonance images reliably upgraded 60% of the clinically negative necks, the overall error of magnetic resonance imaging being 16%. However, for both modalities, the sensitivity per level was too low to allow for selective neck dissections in case of only one positive level. These findings show that apart from primary tumor grading, magnetic resonance imaging can improve the preoperative grading of cervical lymph nodes. In selected cases, this may change the treatment plan to a "wait-and-see" policy or a more conservative type of neck dissection.