This paper reports an experiment which was part of a project to investigate how drivers estimate risk. Thirty drivers, all of whom had held their driving licence for a minimum of 4 years (maximum 38 years), drove an instrumented car along a set 21.4 mile route, in and around Cambridge city. Drivers were grouped into three categories, according to annual mileage, but groups were equated for average age, number of years licensed, and sex. Each driver drove the route in the same direction, during day-time. The test route was chosen to reflect a range of different types of driving situations, traffic flow conditions and accident histories. At 40 junctions along this route the driver was required to indicate their current feeling of personal risk and (b), after the junction had been driven through, to estimate how many accidents (serious enough to be reported to the police) had occurred at that junction. An auditory tone signalled when the 'feeling of risk' estimate was to be made. Only these two estimation tasks were carried out during the test run. Other measures, EGA video record of the entire drive, self-report ratings on how competently a range of driving manoeuvres could be carried out by the subject, and a record of which junctions were previously known by the driver, were also obtained for each driver, but will not be reported on here. An alternative methodology was introduced here, which has demonstrated that not only are drivers' feelings of risk and estimates of accident frequency sensitive to differing levels of objective risk, but that drivers' speed through junctions is also related to accident likelihood at that site. However, in spite of these encouraging findings, it is apparent that the 'objective risks' are not 'known' in any real sense, since the amount of unexplained variance is very large.
Abstract