At present, the exact injury mechanisms that cause the so-called "whiplash injury" are still unknown. A number of hypotheses on the nature and location of the injury have been proposed by various researchers. Correspondingly, a number of injury criteria have been evaluated that would allow researchers to predict injuries in low-speed rear-end impacts. To our knowledge, none of the hypotheses have been taken through a complete chain of scientific proof that would lead from a real-world injury via an injury mechanism, an acceptable human test surrogate, and a biomechanically reliable test up to a tolerance limit for the injury-inducing loads. In fact, since there are no generally applicable diagnostic tools to exactly locate and appraise the injury, not even the first step can be performed in a fully objective way. On the other hand, in some countries, the amount of reported whiplash injuries is rising at such an alarming rate that improvements in car seats which increase their protection potential against whiplash injuries should be undertaken with no further delay. The criteria needed to quantify such improvements should be based on the knowledge available today, even if this knowledge is in some part incomplete. Some car manufacturers already have implemented such improvements, and statistical indications on the reduction of injuries in the corresponding cars have begun to surface. Since 1998 the Working Group on Accident Mechanics has, in collaboration with the GDV Institute of Vehicle Safety and Autoliv Germany, performed over 40 sled tests using a new standardized test procedure (ISO draft). In 2 × 13 tests, the Hybrid III/TRID neck dummy as well as the BioRID II dummy have been used. We have concluded that a valid comparison between different seats or protection systems is possible only if the systems have been tested under strictly identical conditions. Another conclusion is that the Hybrid III/TRID dummy does not offer sufficient biofidelity in low-speed rear-end impacts to be eligible for future tests. For the performance comparison, we have employed criteria such as NIC, upper neck loads M_y/F_x and N_km, and rebound velocity. We discuss the nature of these criteria and the reasons why we employ them. We conclude that, while some criteria may be replaced with others or abandoned completely in the future, it is of utmost importance to use more than one criterion to quantify the protection potential of a seat system.