The purpose of traffic lights is to steer the behaviours of the traffic participants and to make these behaviours apparent. Nonetheless, the participants are capable of making certain choices, which are not always apparent to others and which raise the risk of the occurrence of accidents. The following study shows that the technical installations and equipment used for light signals do not cause "good" or "poor" traffic safety, but rather that the "safe" or "unsafe" nature of these installations is dependent upon combinations with other factors. Of the 80,000 accidents registered by police annually, less than 5 percent occurred at light signals in the past few years. The majority of these (approximately 85%) occurred within city limits. In spite of these low figures, the cap was confronted with the fact that intersections with light signals are often dangerous areas and that the causes of accidents are often not known. The goal of the study was therefore to find out why accidents frequently occurred at certain locations, in spite of light signals, and to list measures and recommendations for the reduction of accidents at these locations. The study was divided into two parts: As a first step (characteristic-specific analysis), the data from the police-registered accidents was analysed and combinations of accident-related characteristics as well as possible causes of accidents at light signals were compiled. As a second step (installation-specific analysis), a comparison was made between site-oriented characteristics of the accident-free and accident-prone light signals. The analysis of 2,035 accidents at light signals within city limits provided the results of the characteristic-specific evaluation. Since the examined characteristic data was nominally scaled, the Chi-square test was used for the matrices. Many of the significant results were plausible, even trivial. This analysis showed the importance of the variables/characteristics "location of accident" and "type of accident". High speed seems to be a relevant factor on straight stretches and at junctions; a poor assessment of conflicting traffic flow is a relevant factor at intersections. Pedestrian accidents are most often caused by unprotected conflicts of traffic flow, whereas rolling traffic collisions are mainly due to a lack of attentiveness, which is in turn caused by the high density of complex information confronting the driver. In the installation-specific analysis, accident-prone and accident-free intersections with light signals were compared. Characteristics which affected the occurrence of accidents negatively or positively were sought. Twenty intersections with light signals were analysed in this study; eleven of these were accident-prone (objects of study) and nine were accident-free (control group). Data was gathered according to three sets of variables: technical installations, technical operation and accident variables. The most attention was given to technical installation-related characteristics such as lights, signals etc. The univariate evaluations of the variables showed that, in terms of the chosen characteristics, the differences between the accident-prone intersections and the control group were minimal. To show the cross-linkage of the characteristics, a prediction model, based upon discriminant analysis (statistical method), was developed. This model measured the contribution as well as the significance of the single characteristics. The characteristics which eventually entered the prediction model can be summed up under the term "complexity". The examination showed that technical installation characteristics of signs and signals raised the complexity. There are also aspects which reduce the complexity, such as installations which help to reduce the density of visual information and assist the driver, e.g. warning lights at permanent green signals. It can be generally stated that intersections with light signals, which are already quite complex, should not be made more complex by the addition of further complicated and attention-robbing installations. The opposite is true: traffic participants should be aided by simple, clearly structured light signal installations and their behaviours and intentions should be clearly perceived by other traffic participants. The prediction model developed in this study is an instrument which makes it possible to categorise existing intersections with light signals as accident-prone and accident-free. In order to be generally valid, the statistical applicability of the model needs to be tested with a larger sample. A follow-up study is planned for this purpose. (A)
Samenvatting