This paper addresses the general problems of selecting an impact test method. A test method must produce results that (a) are repeatable, (b) can distinguish between tested items that are of different design, and (c) predict real-world performance of the tested item.The paradigm employed here is based on the notion that to identify the most appropriate test method among a set of candidate methods, one must compare the results of laboratory testing over a wide range of products with the results of actual field use of the various products.This is difficult in a case where a new test procedure is to be evaluated and actual field data and laboratory data may not yet exist. This paper provides a methodology for assessing alternative test device concepts by means of simulation and applies it to the side impact test situation. This is done by (1) simulation of field accident experience of vehicles involved in side impacts, and (2) the simulation of test devices and procedures. Simulated test results can then be compared with simulated field accident data. This study shows some interim results for several impact test device design variations.The principal tasks of this study were to: develop a vehicle occupant dynamics model suitable for estimating field results in side impacts; develop a test device dynamic model: select a range of alternative hardware configurations for both the simulated vehicle and the simulated test device: and develop random variable exposure distributions of accidents, vehicles, and occupants so the simulated “field” accidents cover a broad spectrum of exposure conditions. Approximately 90,000 simulations are run, using a Monte Carlo procedure, to estimate the distribution of outcome over the range of exposure conditions.