The objective of this study was to test if three-dimensional (3-D) ultrasound (US) provides accurate determination of the cardiac volumes and ejection fraction. The 3-D device (Model 1-Volumetrics, ) is a 3-D acquisition system using a 2-Mhz matrix probe that insonates the whole cardiac volume in a 4-chamber view and collects the entire backscattered US echoes from this volume within one cardiac cycle. The complete 3-D US information stored in the memory can then be cut into 2-D views of any arbitrary orientation. For volume determination, the best 4-chamber view was selected into the memory, then 6 transverse views were displayed at different depths along the ventricle long axis, and the contour of the ventricle was drawn on each of these views. The left ventricle volume in diastole (LVDV) and the ejection fraction (EF) obtained by 3-D US were compared with those from x-ray and isotopic angiographies, and 2-D echo-time motion (2-D Echo-TM). The variations in stroke volume (SV) during a stand test, measured by 3-D US, and aortic Doppler were compared. The correlation between EF evaluated from 3-D US and x-ray or isotopic angiographies was found to be good (r = 0.80 p < 0.001; r = 0.86 p < 0.001), but lower with 2-D Echo-TM (r = 0.59 p < 0.001). For LVDV, the correlation was acceptable with x-ray angiography (r = 0.75 p < 0.001), but much lower with isotopic angiography and 2-D Echo-TM (r = 0.47 p < 0.001; r = 0.55 p < 0.001). A good correlation was also found between the SV changes measured by 3-D US and aortic Doppler (r = 0.79 p < 0.001).