The use of a quasi-naturalistic riding method to investigate bicyclists' behaviors when motorists pass.

The present study aimed to investigate how motorized vehicle-related factors, road-related factors, and bicyclist-related factors influenced motorists' decisions about initial passing distances and bicyclists' behaviors after the motorists started to pass. A quasi-naturalistic riding method was used for thirty-four participating bicyclists riding an instrumented bicycle in real traffic. The study included 1380 incidents of left-side passing by motorists and revealed that the factors studied influenced both the motorists' initial passing distance and the bicyclists' position (lateral distance from the passing motorists), wheel angle, and speed control behaviors while the motorists passed. Some factors were related to the motorized vehicles; for example, the initial passing distance and the mean lateral distances were smaller when motorcycles passed than when cars and small trucks passed. The bicyclists demonstrated weaker lateral stability when they were passed by buses. In addition, a longer passing time caused the bicyclists to demonstrate cautious but less stable riding behaviors. For road-related factors, a clear and longitudinal solid line separation helped to maintain a greater lateral distance between motorists and bicyclists and assist the bicyclists in riding out of the motorists' path, maintaining a low speed, and remaining stable. Moreover, when bicyclists avoided road surface hazards, they reduced the initial passing distances that the motorists had chosen. Considering bicyclist-related factors, the motorists selected a greater initial passing distance for female bicyclists. There were also gender differences related to the bicyclists' lateral control, and differences in road use experience may explain this gender difference. Furthermore, the bicyclists' wheel angle, speed, and speed variation affected the motorists' initial passing distance. The present study demonstrated that the quasi-naturalistic riding method is capable of collecting rich data concerning bicyclists' behaviors, which could potentially be utilized in various types of studies. However, this method requires a large sample and considerable time and effort for data processing.