The aim of this study was to assess day-to-day variability of left ventricular (LV) volume and ejection fraction (EF) calculated from 3-dimensional echocardiography (3-DE) and to compare the reproducibility of the measurement with magnetic resonance imaging. Forty-six subjects were examined including 15 normal volunteers (group A) and 31 patients with LV dysfunction (group B). Precordial 3-DE acquisition was performed at 2 degrees rotational intervals and repeated 1 week later. Magnetic resonance imaging was performed at 0.5 T. End-diastolic and end-systolic LV volumes were derived using Simpson's rule by manual endocardial tracing of 8 equidistant parallel LV short-axis slices with 3-DE, whereas 9-mm slices were used with magnetic resonance imaging. The mean +/- SD of end-diastolic and end-systolic LV volumes (ml) and EF (%) from magnetic resonance imaging were 182 +/- 75, 121 +/- 76, and 39 +/- 18, whereas those from 3-DE were 182 +/- 76, 121 +/- 77, and 39 +/- 18 respectively. Day-to-day measurements of end-diastolic and end-systolic LV volumes, and EF on 3-DE were not significantly different as assessed with SEE (2.7, 1.1, and 2.4, respectively). Intra- and interobserver SEE for calculating end-diastolic and end-systolic LV volumes and EF for magnetic resonance imaging were 6.3, 4.7, and 2.1 and 13.6, 11.5, and 4.7, respectively, whereas those for 3-DE were 3.1, 4.4, and 2.2 and 6.2, 3.8, and 3.6, respectively. Day-to-day variability of LV volume and EF calculation on 3-DE were small and not significantly different for normal and dysfunctional left ventricles. Observer variabilities of 3-DE were fewer than those of magnetic resonance imaging. Therefore, 3-DE is recommended for serial assessment of LV volume and EF in normal and abnormally shaped ventricles.