THE ANALYSIS OF PRE-ACCIDENT SEQUENCES

The main aim of this study was to carry out an in-depth investigation which would avoid the accepted criticism of unreliability and expense of many such studies. It was decided that the basic source of data would be Nottinghamshire Constabulary records for 1988, as it was expected that the new methods would prove successful on police accident files, thus making a very large new data base available for accident research in general. The type of accident chosen was the right-turning accident either onto or off a larger road, and 185 records of such accidents were coded for analysis, using a special coding scheme to cope with the sequential information, TRAAL (Traffic-Related Action Analysis Language). This study aimed to overcome some of the difficulties in analysis by drawing upon a variety of methods from other areas of the behavioural sciences, as well as more conventional statistical analyses. These included sequence analysis which preserves the temporal structure of the data and enables the effects of the order of events to be examined, so it may be possible to detect chains of events which are especially likely to result in accidents. Rule finding, based on genetic algorithms was also used, as well as a second, complementary, method of rule finding - based on Quinlan's ID3 algorithm to generate decision trees. Other more conventional methods were used such as cluster analysis, prototyping analysis of variance and other parametric and non- parametric analyses. In addition to the main in-depth study a subsidiary study was carried out where a sample of experienced drivers were asked to write an account of a right-turning accident as they envisaged it and a safe right turn (one that does not result in an accident). These hypothetical accounts were compared with each other using the same methods as for the real data, and looking for differences in their sequential structure. Results suggest that older drivers have specific problems with right turns at junctions of fast busy roads, and there is an excessive accident involvement of younger drivers; the study also emphasizes the vulnerability of bicycles, motorbikes and pedestrians.