This paper describes the results of a three year field study into the crash performance of three basic designs of steering assembly, fitted to British cars. These systems have evolved to comply with current European and United States safety standards. While one design, involving a large-area, self-aligning steering wheel mounted on a conventional column, appeared from the field data to be highly effective in preventing serious chest and abdominal injury, the two systems utilizing axial-collapse steering columns proved to be essentially ineffective in practice. This finding is based on a field accident sample of 108 cases, so sampled as to be representative of the severe end of the collision speed and injury spectra. A test program was undertaken to examine the procedures currently used to evaluate steering assemblies. It was found that tests carried out in compliance with FMVSS 203 failed to differentiate between the safe and unsafe systems described above. Additionally, it was noted that the mode of damage to the steering assemblies produced by normal testing was quite unlike anything seen in the field. Modifications to the test procedures were made that enabled accident damage to be accurately reproduced, but even under these conditions, the peak load injury criterion failed to show any differences between the designs tested. Only when both peak load and effective loaded area were taken into account could the major differences observed in the field be demonstrated under test. In conclusion, suggestions are made for alterations to present steering assembly testing techniques that will allow the impact equipment currently used to predict usefully the field performance of steering assemblies.