During one cardiac cycle, the volume encompassed by the pericardial sack in healthy subjects remains nearly constant, with a transient +/-5% decrease in volume at end systole. This "constant-volume" attribute defines a constraint that the longitudinal versus radial pericardial contour dimension relationship must obey. Using cardiac MRI, we determined the extent to which the constant-volume attribute is valid from four-chamber slices (two-dimensional) compared with three-dimensional volumetric data. We also compared the relative percentage of longitudinal versus radial (short-axis) change in cross-sectional area (dimension) of the pericardial contour, thereby assessing the fate of the +/-5% end-systolic volume decrease. We analyzed images from 10 normal volunteers and 1 subject with congenital absence of the pericardium, obtained using a 1.5-T MR scanner. Short-axis cine loop stacks covering the entire heart were acquired, as were single four-chamber cine loops. In the short-axis and four-chamber slices, relative to midventricular end-diastolic location, end-systolic pericardial (left ventricular epicardial) displacement was observed to be radial and maximized at end systole. Longitudinal (apex to mediastinum) pericardial contour dimension change and pericardial area change on the four-chamber slice were negligible throughout the cardiac cycle. We conclude that the +/-5% end-systolic decrease in the volume encompassed by the pericardial sack is primarily accounted for by a "crescent effect" on short-axis views, manifesting as a nonisotropic radial diminution of the pericardial/epicardial contour of the left ventricle. This systolic drop in cardiac volume occurs primarily at the ventricular level and is made up during the subsequent diastole when blood crosses the pericardium in the pulmonary venous Doppler D wave during early rapid left ventricular filling.