We compared all frozen section examinations of ovarian tumors during a 6-year period in our institute with the final diagnosis from paraffin sections. In this period, 946 ovarian tumor specimens were removed for histologic assessment; 176 (18.6%) had frozen section examination. Final histological diagnosis was divided into benign (55.1%), borderline malignant (10.3%), and malignant (34.6%). Sensitivity of the frozen section method for malignant or borderline disease was 83.5% and specificity for a benign lesion, 92.8%. Predictive values and 95% confidence intervals were computed: 100% (93-100%) for malignancy, 62% (32-86%) for borderline malignancy, and 92% (85-96%) for a benign disease. Diagnostic problems occurred in large borderline tumors of mucinous cell type. Analysis of the 12 false negative diagnoses revealed that a sampling error was involved in 11 cases. A judgment error was made in the only false positive and in 1 out of 12 false negative frozen section diagnoses. It is concluded that when surgeons and pathologists are aware of the limitations of frozen section diagnosis of ovarian tumors, peroperative histologic examination can be worthwhile and prevent under- and overtreatment of gynecologic patients.