In a retrospective study to determine the accuracy of frozen section diagnoses in ovarian neoplasms, the results of consecutive frozen section diagnoses of 311 ovarian neoplasms from two institutions, New York University Medical Center and State University of New York Medical Center at Brooklyn, from 1980 through 1989 were compared with the final diagnosis results following extensive sampling on permanent sections. The final diagnosis was assumed to be correct for purposes of this study. Ovarian neoplasms were correctly diagnosed on frozen section as either benign or malignant in 292 patients (accuracy of 93.8%). Frozen section diagnoses were incorrect in 11 patients (3.5%). Frozen section diagnosis was deferred in 8 instances (2.6%). The positive predictive value was 100%. The negative predictive value was 95.3%, specificity 100%, and sensitivity 86%. There were no false positives. Of the 11 false negative frozen section diagnoses, 9 (82%) were due to limited sampling for frozen section. We therefore suggest that careful examination with sampling of any suspicious lesions be carried out at the time of surgery for patients with benign frozen section diagnosis, since this may avoid a second staging laparotomy, if the final diagnosis is malignant.