To retrospectively assess the diagnostic accuracy, cytologic features and pitfalls of ThinPrep (TP) (Cytyc Corporation, Marlborough, Massachusetts, U.S.A.) versus conventional (smear) preparation (CP) in salivary gland fine needle aspiration biopsies (FNABs) and second, to evaluate the reproducibility of the cytomorphologic criteria used in the evaluation of FNABs prepared by CP versus TP. All salivary gland fine needle aspiration biopsies (SGFNABs) between January 1996 and June 1999 were retrieved from the cytology files of the University of Michigan Hospital. Histologic correlation was identified when available. Two cytopathologists reevaluated the slides for artifacts, cellular preservation, background material, cellularity, and cytoplasmic and nuclear details. Seventy-four of the 134 (55%) cases identified had histologic follow-up. Fifty (68%) cases were processed by TP and 24 (32%) by CP. FNAB processed by TP and CP correctly identified malignancy in 14 and 9 cases, respectively. There were three (4%) false negative cases. These included two acinic cell carcinomas and one mucoepidermoid carcinoma. There were 37 true negative cases (24 TP and 13 CP) and one false positive case of cellular pleomorphic adenoma (cytologic interpretation, mucoepidermoid carcinoma). All discrepant cases were processed using the TP method. The overall specificity and sensitivity were 98% and 88%, respectively. However, specificity and sensitivity for TP-processed SGFNABs were 96% and 82% as compared to a 100% specificity and sensitivity for CP. Additionally, there were 10 (14%) nondiagnostic cases, 8 of which were processed by TP. Cytologic artifacts associated with TP included diminished/distorted extracellular and stromal elements, cellular shrinkage and tissue fragmentation The diagnostic accuracy of TP-processed SGFNABs approaches that of the CP. However, there are several artifacts that may lead to erroneous diagnoses. Additional studies, that depend on real-life clinical samples processed by TP are suggested to modify current diagnostic criteria.