The improved prognosis for infective endocarditis (IE) seen in the last decade is due partly to more active surgical treatment and partly to improved diagnosis by echocardiography. To evaluate the clinical value of repeated trans-oesophageal echocardiography (TEE) 34 patients with 35 episodes of suspected IE were included in a prospective part of the study (group A). TEE was carried out for diagnosis, at discharge and about 5 months after hospitalization. Endocarditis was classified using Duke's criteria. In a retrospective part of the study 32 other patients with 34 episodes of IE were included (group B). Both groups were analysed regarding mortality, frequency of surgery and classification according to Duke. The diagnosis was regarded as definite in 49 and possible in 20 episodes. Vegetation-size decreased significantly (p < 0.001) during treatment. In contrast, no significant changes in valvular insufficiency were found. In episodes diagnosed as definite, the mortality was 2/49 (4.1%). The low mortality might be explained by the high frequency of surgery (22%), the inclusion of patients with right-sided IE, and 'early diagnosis'. The first TEE was important for correct diagnosis in patients with small vegetations and for those needing surgery. The clinical value of the additional investigations was low in native valve endocarditis.