The authors undertook a prospective and comparative echocardiographic study of 2.5, 3.5 and 5 MHz ultrasonic transducers for the detection of left ventricular mural thrombosis in 53 patients with left ventricular dysfunction. Thirty-three patients had advanced ischaemic heart disease following anterior myocardial infarction and 20 had dilated cardiomyopathy with a left ventricular ejection fraction of < or = 40%. Eighty-two per cent of patients had anticoagulant therapy. The diagnosis of thrombosis was based on Asinger's classification. Eleven thrombi were detected, an incidence of 21%. Using the 5 MHz transducer as a reference, the sensitivity of the 3.5 MHz transducer was 100% and much greater than that of the 2.5 MHz transducer (55%) which was associated with 4 false positive results. The specificities were respectively 97 and 86% for the 3.5 and 2.5 MHz transducers. There was no correlation between the apical Doppler flow velocities and the presence of mural thrombosis. Atrial fibrillation was significantly associated with mural thrombosis (p = 0.04). The increased sensitivity associated with transducers of higher frequency is, however, limited by the echogenicity of patients. The introduction of transducers of variable frequencies should facilitate the diagnosis and improve the sensitivity of echocardiography in detecting left ventricular mural thrombosis.