OBJECTIVE There are few data on the occurrence of microembolic signals (MES) in the acute phase of ischaemic stroke. The objective of our work was to systematically study the frequency of MES in non-selected patients with a first-ever hemispheric transient ischemic attack (TIA) or acute cerebral infarction, and to evaluate the clinical usefulness of MES detection. METHODS 182 consecutive patients with hemispheric TIA or acute cerebral infarction, and 54-age-matched healthy controls were studied. Bilateral transcranial Doppler ultrasound (TCD) monitoring was performed for at least 30 min with a mean time from stroke onset to TCD of 69 h. Stroke severity on admission, early recurrent stroke and dependency on discharge were investigated. RESULTS MES were detected in 20.5% of patients with arterial sources of embolism, 17. 1% of patients with potential sources of cardioembolism and 5% of patients with cryptogenic stroke. They were not registered, however, in lacunar infarctions (p < 0.001). Stroke severity on admission of patients with MES was greater than that of patients without MES (47. 1 vs. 19.4% with the Canadian Stroke Scale < or =6.5; p = 0.009). Early recurrent stroke was more frequent in patients with MES (11.8%) than in those without MES (4.2%) although the difference was not statistically significant. Multiple logistic regression analysis showed that MES increased the risk of dependency on discharge (odds ratio, 4.2; 95% CI, 1.2-14.9; p = 0.01) independently of age, stroke severity on admission and presence of an arterial or cardiac embolic source. CONCLUSIONS There is a strong association of MES in the acute phase of stroke with known potential arterial and cardiac embolic sources. MES have an independent predictive value of poor outcome.