Purpose: Cervical lymph node metastasis is the most important factor of the prognosis and therapeutic planning in head and neck cancer. With increasing interest of minimally invasive neck surgery, more accurate preoperative assessment of cervical lymph node becomes more essential. We evaluated the diagnostic accuracy of $^{18}F$ FDG-PET in the assessment of lymph node metastasis in patients with primary head and neck cancer and compared the results with those of CT/MRI. Materials and Methods: Thirty-two patients (M/F=27/5, $56{pm}10yr$) with biopsy proven head and neck cancer (16 supraglottic cancer, 9 tongue cancer, 7 others) underwent FDG-PET and CT/MRI (25/7) within 1 month before neck dissection. Based on lymph node level, the diagnostic sensitivity and specificity of FDG PET and CT/MRI for the metastasis of cervical lymph node were compared. Results: Of 153 lymph node levels dissected in 32 patients, 32 lymph node levels of 19 patients were positive for metastasis by histopatholologic examination. The overall sensitivity and specificity of FDG-PET were 88% (28/32) and 93% (113/121), whereas those of CT/MRI were 56% (18/32) (p=0.002) and 92% (112/121), respectively. The diagnostic sensitivity and specificity of FDG-PET were different according to location of lymph node levels, and those of ipsilateral level 11 were lower than those of other levels. Conclusion: FDG-PET is more sensitive in detecting metastatic cervical lymph node in head and neck cancer than CT/MRI. FDG-PET might be useful in guiding the extent of neck dissection.