Simulation models of side impact have been developed as part of the basic research aimed at understanding the mechanisms by which injuries are caused, and the methods by which car design can be improved to minimise injuries. Early simulation models were based on experimental tests in which cars were hit by rigid impactors. These models represented the side structure of the target car by a single mass/spring element. However, analysis of tests, using both mobile deformable barriers and cars as impactors, have shown that this simple model provides a poor representation of reality. When a car side is hit by a non-rigid impactor, there is considerable relative movement of the different structural components, such as the sill, the A and B posts, and the door. The way in which this relative motion occurs, particularly between the door and the sill, can have a considerable effect on occupant injuries. It is essential that these motions are properly represented if a simulation is to be effective. The paper discusses ways in which a simulation can model these complex motions, both in terms of the car structure and of its interaction with the occupant. The paper also describes some of the difficulties in obtaining adequate calibration data for the dynamic model from quasi-static tests. The present stage of development of simulation at TRRL is described, together with results from a set of quasi-static crush tests on one car model. It concludes that much has been learnt from using simulation models as a complement to dynamic tests, but that current models are far from being able to provide a complete substitute for full scale impact tests.