In recent years, several kinds of seat systems that aim to reduce cervical spinal injuries in rear impacts, so called ‘whiplash injuries’, have been released by some car manufacturers and seat suppliers in the world. Meanwhile, several kinds of dummies have been developed to be representatives of occupants under such conditions. One of these is the BioRID II equipped with a realistic spine constructed of multiple vertebrae similar to that of a human. It is regarded as the most biofidelic dummy for low speed rear impact. Using this dummy, some typical ‘whiplash protective’ seat systems currently available were dynamically tested to see their performance on injury reduction. From the results of these tests, the design direction to lessen the injury level more efficiently was determined. According to this direction, such parameters as the position of the head restraint and the force-deflection characteristic of the seat back were optimized by means of computer simulation with an in-house developed dummy model. These optimizations made on the existing seat system resulted in lower injury levels in the dynamic tests. In this study, injury levels were estimated mainly by means of Neck Injury Criterion (NIC) currently proposed as the only quantitative criterion based on the hypothesis of spinal ganglion injury. In addition, the upward acceleration at the top of the thoracic spine (T1) that might be one index for injury that was based on the hypothesis of ‘synovial fold impingement’ caused by the upward motion of the torso was taken into account.