The relationship between the position of a body's center of gravity (CG) as determined by a whole body kinematic model and that given by 3 CG position estimation techniques using the ground reaction force center of pressure (CP) has been investigated in this study. The CG and CP are related by the Newtonian mechanics equations of motion. Data required to determine directly the center of gravity position, a key variable in posture and locomotion, are usually not obtained in posture or gait trials; rather, force plate data and center of pressure data are often obtained. Consequently, previous studies have developed estimations of CG position history from CP data. The results of 3 CP-based CG estimations methods are here compared with kinematically determined CG positions in humans. The CP position varies about the CG position and has a higher frequency content than the motion of the CG. This observation, based on the authors' data and mechanics theory, provides the basis for the methods considered in this study. All current methods employ a filtering technique to obtain CG position from CP position time histories during standing trials. In most cases the mean square error is less than 0.1 cm/sup 2/. Finite-duration impulse response filters with periods of 1.0 s to 1.50 s gave the best results when compared with the CG position based on kinematic data. A low-pass filter with cutoff frequency of 0.4 Hz to 0.5 Hz provides the best comparison for this approximation method. The accuracy of the methods diminishes as more dynamics are introduced to the trial. The average mean square differences for walking-in-place trials is from 1 to 10 times greater than that for the standing. CG position estimates from CP data must be interpreted cautiously when nonstatic tasks are monitored, especially in unstable patients. >