Assessing crossing and communication behavior of pedestrians at urban streets

•Different data collection methods may cause divergent results in pedestrian behaviour.•In a field study, crossing, gaze and gesture behaviour was measured in urban traffic.•Self-reported and video-observed pedestrian behaviour is similar.•Instructed persons do not cross and communicate differently from regular pedestrians. Pedestrians belong to the most vulnerable road users and in order to reduce accidents and fatality rates, research on pedestrian crossing behavior is necessary. However, findings on pedestrians’ crossing, gaze and gesture behavior are heterogeneous and sometimes contradictory. Different methods that are used to assess pedestrian behavior might provide an explanation. In a field study, we measured pedestrian behavior in real traffic by making use of instructed and non-instructed pedestrians, combining and comparing objective and subjective data. Instructed pedestrians deliberately interacted with a research vehicle in urban traffic and filled in 2687 self-report questionnaires about experienced scenarios from their point of view. Interaction scenarios focused on instructed pedestrians either crossing or not crossing the street when the research vehicle passed by. In addition, 75 h of video-recorded interaction scenarios with instructed and non-instructed pedestrians were gathered using a camera placed on the inside of the research vehicle. Self-report and video data were matched (N = 644) and evaluated regarding instructed pedestrians’ crossing, gaze and gesture behavior. Substantial agreement was found between self-reported and video-observed data for crossing action and gesture occurrence and somewhat lower agreement for gaze occurrence. Furthermore, the crossing, gaze and gesture behavior of instructed pedestrians (N = 1423 situations) and non-instructed pedestrians (N = 369 situations) was compared and no major discordance was reported. Therefore, instructed pedestrians can be an externally valid method to assess crossing behavior while implications are given on how self-reported and observational data can mutually supplement each other.