AIMS--To assess the degree of difficulty in diagnosing partial mole by analysing intraobserver and interobserver agreement among a group of pathologists for these diagnoses. METHODS--Fifty mixed cases of partial mole, complete mole, and non-molar pregnancy were submitted to seven histopathologists, two of whom are expert gynaecological pathologists; the other five were district general hospital consultants, one of whom works in Australia. These participants gave each slide a firm diagnosis of either partial mole, complete mole, or non-molar pregnancy. Some 12 months later, the slides were recorded and again submitted for a second diagnostic round to assess intraobserver as well as interobserver agreement. Standard histological criteria for each diagnostic category were circulated with the slides. RESULTS--kappa statistics showed that complete mole could be reliably distinguished from non-molar pregnancy, but neither non-molar pregnancy nor complete mole could be easily differentiated from partial mole. In only 35 out of 50 cases was there agreement between five or more of the seven participants. Agreement between the expert gynaecological pathologists was no better than for others in the group. Interestingly, the intraobserver agreement for each pathologist was good to excellent. CONCLUSIONS--These results imply that the reported histological criteria are either not being applied consistently or that they are lacking in practical use. An atypical growth pattern of trophoblast, rather than the polar accentuation seen in normal first trimester pregnancies, seems to be the important diagnostic histological feature for partial mole. Ploidy studies might also help with problem cases.