Frozen section is often requested in the intraoperative assessment of patients, presenting with ovarian masses, to provide guidance for appropriate surgical management. To assess the accuracy of frozen section and identify causes of diagnostic error, we reviewed 914 consecutive ovarian frozen sections performed over a 5-year period in 2 laboratories; one of which provides a general surgical pathology service and, the other, a specialist gynecologic pathology service. Cases, in which there were significant diagnostic discrepancies between the intraoperative and the final histological diagnoses, were reviewed. The series included 552 benign lesions (60.4%), 96 borderline (atypical proliferating) epithelial tumors (10.5%), and 266 malignancies (29.1%). The overall accuracy of frozen section diagnosis was 95.3%. There were 43 cases with diagnostic discrepancy; 20 (3.8% cases) of which were reported in the specialist laboratory and 23 (5.9% cases) in the general laboratory. Underdiagnosis of tumor type accounted for 32 of 43 discrepant cases and was most frequent in borderline mucinous tumors. The most common cause of overdiagnosis was the misinterpretation of serous cystadenofibroma as borderline serous tumor. Slide review of the 41 assessable cases indicated that sampling error, pathologist misinterpretation, and suboptimal slide preparations contributed to misdiagnoses in 17, 23, and 9 tumors, respectively (in 9 cases, 2 factors were contributory), whereas no specific error was identified in the remaining case. Technical factors and pathologist misinterpretation were more common in the general pathology laboratory. This study confirms that ovarian frozen section is a generally reliable technique, but there are problematic areas, particularly involving the assessment of borderline tumors.