Summary: In the assessment of malignant potential of ovarian tumours, frozen section has been found to be accurate in 97.1% (168 of 173) of cases. The positive predictive value of frozen section in the diagnosis of a malignant lesion was 100% (34 of 34). Errors were mainly made in the diagnosis of borderline tumours with a predictive value of 87.5% (7 of 8). The negative predictive value was 98.4% (127 of 129). Frozen section however, was less accurate in the diagnosis of specific histological type with an accuracy rate of 91.9% (159 of 173). Macroscopic features were found to be useful in the intraoperative prediction of malignant potential. Completely cystic tumours were benign in 96.4% (108 of 167) of cases. Solid/cystic tumours were malignant in 69% (27 of 38) of cases. Completely solid tumours were malignant in 56% (9 of 16) of cases. Frozen section in completely cystic tumours only marginally improved the clinical macroscopic diagnosis of malignancy. The sensitivity and specificity of ultrasound scan in the diagnosis of malignant/borderline tumours were 82% and 86% respectively. The false negative rate of 7% makes laparoscopic excision of unsuspected malignant ovarian cyst a significant possibility. The predictive value of ultrasound scan in the diagnosis of malignant ovarian tumour was 62% (26 of 42). In the preoperative assessment of malignant potential of ovarian tumours, this study shows that ultrasound scan has a high false positive and a significant false negative rate. Careful intraoperative assessment of gross features and the use of frozen section especially in those with solid/cystic and solid tumours will help achieve a high accuracy rate in the assessment of ovarian tumours.