The frontal collision is the most frequent collision type observed in real accident investigations. The impact severity depends on the overlap of the vehicle with the obstacle which also determines the average deceleration of the car and subsequentlyt he loading of the car occupant. The injury severity depends upon anthropometric parameters and the mechanical response. To investigate these relationships 46 frontal collisions with cadavers w-ere performed. The impact velocity was 47-55 km/h, the average sIed deceleration lo-20 g. The subjects, 36 males and 10 females, were aged between 19 and 65 years and were protected with 3-point standard belts, driver air bag - knee bolster, 3-point belt combined with air bag. The injuries were defined through medical investigation during the autopsy in situ or on isolated body parts in a more detailed manner later; the injury severity was coded in accordance with the AIS 90.The most injured body part was the thorax, usually rib fractures were found. The number of rib fractures includes uninjured cases up to 17 rib fractures with thoracic injury severity between AIS 0 and AIS 4. The fracture pattern was characteristic for the restraint system used. Logistic regression analysis was used to predict the probability of injury severity for the explanatory variables. To further restrain the subset of predictors Kruskal-Wallis and F-test were applied. Chest accelerations are suitable parameters to predict thoracic injury severity. By modelling the data with logistic regression models the best biomechanical predictor for the thoracic injury severity according to different goodness of fit criteria was the acceleration measured at the 1st thoracic vertebra. These evaluations take into account only the injury severity and the mechanical response, independent of the restraint system used and the impact severity.