Objective. Frozen section is an important and helpful adjunct in the intraoperative diagnosis of ovarian tumors. This retrospective study was undertaken to determine the accuracy of frozen section diagnosis of ovarian masses and the reasons of discordance. Methods. From January 1995 to December 2003, 1494 ovarian specimens were received for histopathological evaluation, and 617 of them were submitted for frozen section examination. Results. The final paraffin section diagnoses of these 617 cases were a nonneoplastic lesion in 18.3% of the cases, benign tumor in 56.1%, borderline tumor in 6.2%, and malignant tumor in 19.4%. The overall accuracy was 97%. Twenty-one cases were incorrectly diagnosed by frozen section. All of them were false negatives. There were no deferred cases. The majority of the cases of disagreement were mucinous and borderline tumors. The sensitivity for benign, borderline, and malignant tumors were 100%, 87%, and 87%, respectively. The specificity for benign tumors was 97%; for borderline tumors 98%; and for malignant tumors 100%. Conclusion. Our data confirm that frozen section diagnosis is a reliable method for the surgical management of patients with an ovarian mass. However, diagnostic problems can occur in mucinous and borderline tumors during frozen section examination. The clinicians and pathologists must be aware of the pitfalls of this method; therefore, a good communication established between them is necessary to obtain more accurate results and to minimize the number of deferred cases. © 2005 Elsevier Inc. All rights reserved.