A series of full-scale car barrier crashes and an additional series of sled tests were carried out. The tests involved two types of three-point restraint harness, anthropomorphic dummies, and in some tests a human subject. These tests evaluated the effectiveness of the harnesses under various conditions, i.e., different hardness of seat cushion, slack in harness loops, and angled frontal impacts. The loads in anchorage points and the retardation of the dummy as well as that of the vehicle were recorded. The restraint kinematics were further evaluated by means of high-speed motion-picture cameras. The effectiveness of the two harness systems was found to be high and about equal, but, with respect to retardation and general restraint features at the rebound stage, somewhat favorable to the harness system which employed a single sliding strap at the strap junction. Harder seat cushions and slack in harness gave increased retardation and loading figures in some recording points. The harnesses offered good restraint features in angled frontal impacts up to 30°.