The Cal3D simulation program was used to study the interaction between the belt restraint systems and a Hybrid III dummy in two sled tests. The elastic properties of the dummy joints and thorax were obtained from static tests. The two belt algorithms in the Cal3D program were compared and simulation results indicated that the “harness model”, which utilizes multi-segment representation for the belt is more suitable than the “simple belt model”, which has only one fixed point on the torso. Simulating a frontal impact of a lap belt restrained dummy indicated that the dummy motion consists of two distinct phases; a forward translation followed by a rotation. During forward translation the belt is primarily in contact with the abdomen while during rotation the belt is interacting with the upper legs. For a lap and shoulder belt restrained dummy, considerable head acceleration was induced by chin/chest impact. The flexibility of the neck also has a significant effect on the magnitude of this impact. Exercising the model indicated that shoulder belt alone is not sufficient to prevent submarining and proper anchor point location is important to achieve adequate torso restraint. It was also found that a stiffer belt system would reduce head acceleration and excursion, rib deformation and chest viscous response, but induce higher spine and sternal accelerations.