A finite element human model has been developed to simulate occupant behavior and to estimate injuries in real-world car crashes. The model represents an average adult male of the US population in a driving posture. Physical geometry, mechanical characteristics and joint structures were replicated as precise as possible. The total number of nodes and materials is around 67,000 and 1,000 respectively. Each part of the model was not only validated against human test data in the literature but also for realistic loading conditions. Additional tests were newly conducted to reproduce realistic loading to human subjects. A data set obtained in human volunteer tests was used for validating the neck part. The head-neck kinematics and responses in low-speed rear impacts were compared between the measured and calculated results. The validity of the lower extremity part was examined by comparing the tibia force in a foot impact between the test data and simulation results. Preloading was applied to the Achilles tendon in order to simulate the muscular bracing in emergency braking. For the thorax part, a new test method was developed to obtain the viscoelastic responses of the human thorax under realistic loading conditions. In most papers in the literature, thorax tests were conducted by impacting the chest with a rigid cylinder representing the steering hub. In this study, the thorax was compressed not only by a rigid cylinder but also by a diagonal belt and by a wide band, representing the seatbelt and airbag, respectively. This paper describes the difference in the mechanical response of the human thorax under different types of loading and then compares the test data and simulation results.