We have studied the three-dimensional (3D) motion of left ventricular (LV) epicardial points by tracking one to three dozen coronary artery bifurcations in eleven human subjects. Wall motion was analyzed using several different coordinate systems: (1) cylindrical centered about the LV long axis, (2) spherical with origin at the LV center-of-gravity (COG), and (3) spherical with origin at the LV center-of-contraction (COC), the best-fit 3D point toward which the wall moves. The coordinate systems were studied both fixed and moving with time. Three-dimensional motions were decomposed into three directional components, with high radial (in and out) percentages being regarded as the figure-of-merit of a given coordinate system. Average percentage radial motions were fixed cylindrical 16%, fixed spherical COG 35%, fixed spherical COC 47%, moving cylindrical 17%, moving spherical COG 30%, moving spherical COC 91%. Spherical systems were generally better than cylindrical systems, with the COC representing a better origin than the COG. Moving systems were appreciably better than fixed only for the COC model, indicating that the COC, which traverses up and down the LV midline, moves significantly while the other systems are more stationary. At each instant in time, almost all (91%) of the 3D motion of the entire heart wall is directed toward a single moving 3D point, the COC. Thus, there exists in principle a near-perfect 3D heart wall motion model. Approximately 25% of 3D wall motion is unseen in conventional monoplane views. Also, any model that represents 3D wall motion only along fixed straight 3D lines (eg, end-diastole to end-systole) necessarily ignores 27% of the true 3D heart wall motion.