Both transthoracic (TTE) and transoesophageal (TOE) echocardiography are used routinely to investigate ischaemic stroke. We retrospectively assessed the incidence of abnormalities on TTE/TOE and whether an abnormal TTE/TOE result could have been predicted on the basis of ancillary tests and clinical cardiological examination. Data from 428 patients were analysed. The diagnostic yield of TTE was 12%. For TOE there was a diagnostic yield of 40% in tests actually performed. Overall, echocardiography altered management in 5% of patients. A significant correlation was found between clinical cardiac disease, stroke subtype and the diagnostic yield of TTE. We conclude that the vast majority of abnormal findings occur in patients who already have clinical evidence of cardiac disease. This suggests that the use of these tests should not be "routine", but determined on an individual patient basis.