An interobserver analysis of the reproducibility of malignancy grade and of histological type classified according to World Health Organization criteria was performed in 261 ovarian tumours, stages IIB to IV. All available slides per patient including metastatic tumours (16%) were reviewed independently by two pathologists. Overall observed agreement regarding type was 68%. The chance-corrected agreement expressed by the kappa coefficient was 0.55 (95% confidence interval 0.49-0.61). Overall observed agreement regarding grade was 41% compared with a kappa coefficient of 0.19 (95% confidence interval 0.12-0.26). Kappa coefficients, ranging from 0.18 to -1.00 for associations between type categories and from 0.20 to -0.32 for grade categories, indicated that any suggested pattern of association could be caused by chance occurrences. A more objective approach to histopathology is needed in ovarian carcinoma before its prognostic impact can be assessed.