The constant-volume hypothesis regarding the four-chambered heart states that total pericardial volume remains invariant throughout the cardiac cycle. Previous canine studies have indicated that the pericardial volume remains constant within 5%; however, this hypothesis has not been validated in humans using state-of-the-art technology. The constant-volume hypothesis has several predictable functional consequences, including a relationship between atrial ejection fraction and chamber equilibrium volumes. Using cardiac magnetic resonance (MR) imaging (MRI), we measured the extent to which the constant-volume attribute of the heart is valid, and we tested the accuracy of the predicted relationship between atrial ejection fraction and chamber equilibrium volumes. Eleven normal volunteers and one volunteer with congenital absence of the pericardium were imaged using a 1.5-T MR scanner. A short-axis cine-loop stack covering the entire heart was acquired. The cardiac cycle was divided into 20 intervals. For each slice and interval, pericardial volumes were measured. The slices were stacked and summed, and total pericardial volume as a function of time was determined for each subject. In the normal subjects, chamber volumes at ventricular end diastole, end systole, and diastasis were measured. Pericardial volume remained invariant within 5 +/- 1% in normal subjects; maximum variation occurred near end systole. In the subject with congenital absence of the pericardium, total heart volume, defined by the epicardial surface, varied by 12%. The predictions of the relationship between atrial ejection fraction and chamber equilibrium volumes were well fit by MRI data. In normal subjects, the four-chambered heart is a constant-volume pump within 5 +/- 1%, and constant-volume-based modeling accurately predicts previously unreported physiological relationships.