The purpose of this research is to help a technological improvement by measuring the movement data of the human body, and comparing an expert with an inexperienced person's movement. Solving optimization problems for higher-dimensional multi-body systems requires vast amount of calculation. Accordingly the expert's motion is assumed to be the optimal motion, and is compared with the beginner's motion. To make the beginner's motion approach to the expert's, the kinematical teaching elements are proposed. The kinematical teaching elements are defined as kinematical constraints for correcting the motion of human body. In this paper, the proposed kinematical teaching elements are examined. The human subjects are modeled in fifteen body segments, fourteen joints and twenty-five degrees-of-freedom multi-body systems. The target motion is one backward somersault. The previously measured data, reconstructed by spline interpolation, is used for the data of the expert's movement. To examine the effectiveness, two teaching elements are created and are applied to the beginner's motion. Finally, it is found that the beginner's motion approaches to the expert's.