The main content of the paper is concerned with the way people compensate their driving behaviour for degraded vehicle stability. It also outlines briefly the background of man–machine simulation and the use of driving simulators in the U.S.A. The problem was investigated by means of a fixed base driving simulator at the Davidson Laboratory as part of a larger programme concerned with driver behaviour. The simulator has a two-degree-of-freedom analogue simulation of the vehicle dynamics. The display system is photographic using a real road film, and the vehicle is subjected to random appearing disturbances of the wind gust type. The disturbance, vehicle, and driver behaviour are recorded both as analogue and digital signals for experimental control and later data analysis. The paper describes and gives the derived results from an experiment with nine subjects, driving two vehicles, one understeer and one oversteer, simulated to be travelling at 60 mile/h. While driving, the vehicles were subjected to a random appearing gust spectra having a maximum amplitude equivalent to 10 mile/h. Methods of assessing overall driving performance and what the drivers did in the two vehicles are developed and the results, in terms of a driving error and a control response function, are presented. The driving error results show that, in general, drivers had significantly higher error scores with the over-steering vehicle. The driver control response function gives a measure of the reaction to the various disturbing frequencies of the disturbance. The results show that seven out of nine drivers had a control response function that was higher with the oversteering car. The remaining two subjects had the highest error scores with the understeering vehicle. A driving effort function is derived and it is shown that, with exceptions which are discussed, the driving error may be regarded as a function of driving effort. The results substantiate the hypothesis that the driver will try to compensate for degraded vehicle stability, but show that the compensation is only partial and increased errors result. The implications for real driving situations are that degraded vehicle stability, either of a basic vehicle or occurring during manoeuvring, will be only partially compensated for by the driver, and increased driving errors from the driver–vehicle system are likely to result. The techniques used for the simulator experiment are suitable with only minor modifications for real driving situations.