Cardiac wall motion has been studied extensively. It is usually determined by indirect two-dimensional measurements for the true three-dimensional (3D) motion with its specific speed and direction. Errors are also introduced by using internally fixed reference systems and by the inability to identify precise points on the heart wall during the cardiac cycle. Because of these limitations, the endocardial and epicardial wall motion and their relationship are still unclear. This study was designed to assess endocardial and epicardial wall motion by measuring the direction and speed of implanted markers in an externally fixed 3D coordinate system. Fifty-seven pairs of endocardial and epicardial metallic markers were placed at anterior, lateral, posterior, basal, and apical regions of the left ventricles of 14 normal mongrel dogs. Biplane cineradiographs were performed at 50 frames/sec, and the 3D motions of the markers were analyzed using a specially designed computer system. It was found that the speeds, directions, displacements, and phases of the movements of corresponding endocardial and epicardial points were highly correlated. The correlation coefficients were 0.77 to 0.95 for the mean directions, 0.61 to 0.96 for the mean speeds, and 0.59 to 0.96 for the mean displacements at various regions of the heart, and the periodic movements of the endocardium and epicardium were always in phase. The mean epicardial speeds and displacements are fixed proportions (approximately 70%) of the mean endocardial speeds and displacements despite the differences in absolute values between regions in the same dog and the same regions in different dogs. The correlation coefficients for endocardial and epicardial instantaneous speeds, directions, and velocities ranged from 0.68 to 0.83, 0.81 to 0.88, and 0.77 to 0.86, respectively, for different regions of the heart. The correlation coefficients were significant for both the mean values and the instantaneous values. Thus, when only fixed epicardial points are accessible for wall motion measurements in clinical situations, it is possible to infer the endocardial motion from the epicardial motion.