The purpose of the present experiment was to examine how the upper limb movements (shoulder, elbow, forearm, wrist) of patients with spinal cord injury (C6 level) were affected while using two types (type 1: pronation type, type 2: supination type) of self-helping device. Five subjects were required to eat 5 spoonfuls of yoghurt. We recorded the position of 11 light reflecting markers attached to the subjects' body with three cameras. We divided the eating action into three phases, the scoop phase, reach-to-mouth phase, and reach-to-plate phase. These kinematic landmarks were used to define the dependent variables. We calculated five joint angles (shoulder flexion, shoulder abduction, elbow flexion, forearm supination, wrist extension) with a three-dimensional video-based motion analysis system (APAS System, Ariel Dynamics), and analyzed how they changed at each phase. We compared them while using type 1 and type 2. While using type 1, the scoop phase played a larger part than other phases, and shoulder flexion, shoulder abduction and elbow flexion angles increased, not only in the reach-to-mouth phase but also in the scoop phase, and the supination angle decreased. This result suggests that patients who can supinate their forearm had better use type 2, and also that it is important to consider upper limb movements in the scoop phase when we provide patients with a self-helping device. In this study, however, we focused only on upper limb movements. We also have to analyze head, neck and trunk movements and examine the relationship among upper limb, head, neck and trunk.