This randomised prospective study was designed to evaluate the use of transvaginal (TVS) and transabdominal (TAS) ultrasound in the diagnosis of placenta praevia and its effect on length of stay in hospital. One hundred and thirty patients with antepartum haemorrhage (APH) and/or malpresentation with a suspicion of placenta praevia were referred to the scanning department of Aberdeen Maternity Hospital and Princess Badee'a Teaching Hospital from July 1993 to December 1996. Fifty-eight patients had an ultrasound because of antepartum haemorrhage and 72 had a malpresentation and/or unstable lie. All examinations were performed by the same authors. Placenta praevia was diagnosed in 13 (22.4%) out of 58 cases presenting with APH and in five (7%) out of 72 cases presenting with malpresentation. There was a high false positive rate in the detection rate of placenta praevia using TAS (23%) compared with TVS (11.5%) which is higher than other reported figures (2-6%). The mean length of stay in hospital of the group in whom the results were released was 9.2 days compared with those not released where it was 15.7 days. The difference was significant (P <0.05). This study suggests that the traditional transabdominal approach to the diagnosis of placenta praevia can be improved upon. TVS is a safe method and with improved accuracy over TAS can reduce the length of stay in hospital.