In the past, many studies have been dedicated to the comparison of dummies and human body behavior in different impact conditions. However, the complex boundary conditions generated by a complete restraint system render it difficult to compare both human surrogates in a car environment. Furthermore, the great dispersion among car occupants is an additional difficulty which is difficult to overcome with experimental studies, Computer simulation, as far as a validated human body model is available, gives a unique possibility to assess the influence of some restraint parameters, whilst all remaining parameters are unchanged.To this end, a 3D finite element human body model validated in many different impact configurations against a large number of biomechanical corridors was used. In order to compare responses, models of Hybrid III and Eurosid 1 dummies were also used. Three typical impact configurations were chosen, both frontal and lateral, using the currently most advanced restraint systems with belt load limiter, pretensioner, knee bolster, frontal and lateral air-bags and pelvis door padding. A comparative study was then carried out by varying some of the restraint system parameters.This comparative study highlighted significant differences between both human body and dummy models responses in a vehicle crash environment. The results showed that the human body model is more sensitive than the dummy models to the modification of some restraint system parameters such as the load limiter level or the air-bag pressure. Furthermore, as it was done previously by different authors, it was noticed that the deflection criterion was questionable in some situations. However, the dummy models were able to give reasonable trends in most situations and verified that the currently most advanced restraint systems are well optimized with regard to the current biomechanical knowledge.