Objective: Two scoring systems based on trans-vaginal sonographic findings of ovarian tumors were evaluated with respect to the prediction of ovarian malignancy. Study design: In this retrospective study, 64 premenopausal non-pregnant patients with an ovarian tumor underwent trans-vaginal sonography from 1991 to 1993. In each women a numerical score was calculated using two methods. Method A included four variables: inner wall structure, wall thickness, presence of septa and echogenicity. The point scale ranged from 4 to 15 points. Method B included three morphological characteristics: volume, wall structure and septal structure, the point scale ranging from 0 to 12 points. The threshold values for predictability of malignancy were 9 and 5 points for methods A and B, respectively. Results: The mean age of the patients was 32.2 years (range 22-44). The mean gravidity was 2.6 (range 0-5) in 45 women, 19 women being nulliparous. Fifty-five masses were surgically proven to be benign and 9 women had primary malignant tumors. Sensitivity and specificity for prediction of malignancy were 89% and 73%, respectively, for method A, and 89% and 70%, respectively, for method B. Positive predictive value and negative predictive value for malignancy were 35% and 97% for method A and 29% and 64% for method B. The mean morphology scores on benign and malignant masses, calculated by method A, were 6.94 +/- 2.36 S.D. and 12.0 +/- 2.4 S.D. (P < 0.01). The same scores, calculated by method B, were 4.16 +/- 1.16 S.D. and 9.44 +/- 2.96 S.D. (P < 0.01). We found a considerable overlap in the scores of different types of ovarian tumors. Conclusions: Both methods are easy to apply and provided explicit data. Method A was shown to be more effective. The number of false positive results was relatively high. The most important single sonographic characteristic of the malignant masses are the wall structure abnormalities.