The estimation of segment inertial parameters (SIPs) is an important source of error in inverse dynamic analysis. In most individual cases SIPs are derived from extrapolation of known SIPs of a certain population through regression equations (proportional models). Another well-known method is the use of mathematical approximation of the shape of human body segments combined with estimations of segment densities (geometric models). In the current study five males and five females performed four different lifting movements in the sagittal plane. A full body linked segment model was applied twice to the same data set, once using a proportional and once using a geometric anthropometric model. As a full body linked segment model is an overdetermined system of equations, four equations could be formed to test the summed effect of SIP errors on the inverse dynamic analysis. The overall performance in terms of coefficients of correlation was better for the geometric model as compared to the proportional model. When a back lifting movement was performed, the equations indicated systematic errors in the proportional model. However, when a leg lifting movement was performed, the equations indicated systematic errors in the geometric model. Therefore, analyzing only one kind of movement does not suffice to draw conclusions with respect to the reliability of an anthropometric model.