Experimental studies in the 1960s and 1970s demonstrated the high sensitivity of Doppler ultrasound in detecting gaseous bubbles. More recent studies have shown that microscopic air bubbles, as well as glass microspheres as small as 5 mu to 20 mu, cause characteristic high-intensity signals. Recently it has been demonstrated that less echogenic embolic materials such as thrombus, platelet aggregates, and atheroma can also be detected with a high sensitivity. Such "solid," or formed-element, emboli as small as 200 mu to 400 mu can be detected; the lower size limit of detection was due to an inability to make smaller embolic particles rather than to the sensitivity of the detection process itself. Analysis of the Doppler signals provides some information about embolus size and composition, but accurate characterization in clinical practice is not possible using current technology. Studies in experimental models have allowed the detailed description of embolic signals; they appear as a short-duration, frequency-focused increase in intensity, predominantly unidirectional in the direction of flow, and usually contained within the spectral envelope. In contrast, artifacts appear as a bidirectional, high-intensity increase with maximum intensity at low frequencies. These differences have been exploited to develop automatic embolus detection programs, and an off-line version has been successfully validated in an experimental model.