To assess the diagnostic value of fine needle aspiration biopsy in individual lesions of the salivary glands. During a 16-year period (1979-1995), 841 salivary gland lesions were investigated by fine needle aspiration with cytologic examination. The lesions affected the parotid gland area in all cases except 85 and 19, respectively, in which the submandibular gland and oral cavity (more often the palate) were involved. Approximately 3% (25/841) of the aspirates were inadequate, and in 97% of cases they represented the result of samplings performed outside our institution. The 816 diagnostic aspirates comprised 245 cases negative for tumor cells (benign cysts, inflammatory, malformed, degenerative lesions), 571 positive for tumor cells (benign; malignant; not otherwise specified [NOS]); metastatic to the salivary gland tissue; or to the intraparotid or periparotid lymph nodes). Among the 245 cases negative for tumor cells, 36 underwent surgery, and the absence of a neoplasm was histologically confirmed in all cases but one (a low grade mucoepidermoid carcinoma, underdiagnosed as a retention cyst). Among the 571 patients having a cytologic diagnosis positive for tumor cells, in 518 the lesion was removed and the tumor nature confirmed. The cytologic diagnosis of a neoplastic lesion demonstrated a good correlation with histology except in an oncocytic adenoma diagnosed as a Warthin's tumor, in a basal cell adenoma diagnosed as pleomorphic adenoma, in a low grade mucoepidermoid carcinoma of the palate diagnosed as a polymorphous low grade adenocarcinoma and in a vegetative intravascular hemangioendothelioma diagnosed as a possible malignant tumor, NOS. The accuracy was 97%, the sensitivity to the presence of a tumor 98% and the specificity for absence of a neoplasm 98%. Despite the relative rarity of salivary gland tumors, if established diagnostic criteria are present and strictly observed, the great majority of the common variants of the nonneoplastic and both benign and malignant salivary gland tumors can be diagnosed with a high level of accuracy. There remains a proportion of "problem cases" due to the rarity of the lesions, and in these circumstances the uncertainty must be conveyed to the surgeon openly, leaving the diagnosis open, with a few suggested differential diagnoses.