Left ventricular ejection fraction (EF) was determined in 24 persons by magnetic resonance imaging (MRI) using a 1.5 Tesla whole-body scanner and ECG-gated multislice data acquisition in planes corresponding to the true axis of the left ventricle. EF was calculated using three geometric models: ellipsoid single plane (area-length method), ellipsoid biplane and Simpson's rule. In addition, EF was determined by gated radionuclide ventriculography and compared to the values obtained by MRI. There was a good correlation between EF determined by radionuclide ventriculography and MRI using Simpons's rule (r = 0.83, p 0.001). The correlation was poorer when the ellipsoid biplane model (r = 0.64) and the area-length method was used (r = 0.60). EF values determined by MRI were larger than EF values determined by gated radionuclide ventriculography (0.61 ± 0.14 and 0.54 ± 0.14) (p < 0.01). One reason for this may be that a larger proportion of the aortic outflow tract is included in the radionuclide study. Thus, MRI provides a noninvasive and nonionizing alternative for determination of left ventricular EF. EF values obtained by monoplane and even biplane methods should be regarded with caution.