Abstract: In today ́s traffic, drivers communicate with pedestrians explicitly through a variety of means, including gestures and eye contact or implicitly through the vehicle itself. Road users infer the driver ́s intention from these communication cues. However, as the level of automation increases, the driver will no longer be in control of the vehicle behavior and will instead be engaged in non-driving related tasks. Nevertheless, automated vehicles (AVs) will still need to coordinate with other road users to ensure road safety and traffic flow, without relying on today ́s driver-centric communication cues. In the recent years, many studies have focused on the design of vehicle behavior and additional external human-machine interfaces (eHMIs) to ensure the communication between AVs and pedestrians. In the vast majority of studies, interactions with AVs equipped with a communication interface were found to be more efficient and perceived as safer and more satisfying. To measure the effectiveness of communication, these studies asked participants to explicitly indicate their decision to cross the road. Research on the actual behavior of pedestrians has been neglected so far, probably due to the lack of precise data collection and analysis tools. This dissertation addresses the current state of research and has two aims. First, by presenting five empirical studies, the dissertation takes up the prevalent lack of actual behavior in previous research. It takes a deep dive into exploring pedestrian behavior when it comes to AV-pedestrian interaction, both in laboratory and real-world settings. Secondly, the dissertation deals with the research question of how AVs can communicate with pedestrians. Given that much of the current interaction between road users is based on the motion patterns of the vehicle, we propose to further investigate the role of its "body language" as a communication modality. In addition to eHMIs, we aimed to explore an implicit communication through an artificial vehicle motion, namely active pitch motion. Five empirical studies were conducted to pursue these objectives. All studies analyzed the effect of AV communication on actual pedestrian behavior. Study 1 investigated eHMIs and their behavioral counterparts in terms of actual pedestrian crossing behavior. A motion-based evaluation approach for virtual reality (VR) experiments was developed to operationalize the effectiveness of different eHMIs (Study 1). Derived from the implicit communication literature, Study 2 evaluated a recently discussed artificial vehicle pitch motion in a VR pedestrian simulator. An active pitch motion was pitted against an eHMI and a combination of the two to understand how the motion dynamics of the vehicle affect the comprehension of the AV ́s intention and interacts with the effect of an eHMI. Overall, any form of communication reduced the time taken by pedestrians to make a decision to cross. The results suggest that AV communication is most effective when implicit (active pitch motion) and explicit communication (eHMI) are combined. Study 3 introduced the role of vehicle distance in effective AV communication. The study discussed the potential benefits of a phased communication providing implicit and explicit communication based on distance between the AV and the pedestrian, rather than communicating at a single point in time. Our virtual reality studies suggest that the use of novel eHMIs or pitch motions could bridge the communication gap between AVs and pedestrians. This recommendation needs to be thoroughly validated, and it is of interest how explicit and implicit communication cues influence pedestrian crossing behavior in real-world conditions. In a controlled Wizard of Oz experiment (Study 4), we contrasted an eHMI and an acoustic signal with three pitch motions of the vehicle. Two optical motion tracking systems, a light barrier and a SMARTTRACK3/IF, were used to capture pedestrian behavior in real-world conditions. Consistent with previous work, the eHMI and the acoustic signal had a positive effect on pedestrian crossing behavior, whereas the pitch motion had no effect. Finally, Study 5 compared the applicability of the two optical tracking systems used and concluded with a recommendation on how to capture pedestrian crossing behavior in a real-world setting. Overall, the results of this dissertation demonstrate that the use of behavioral data adds value to the study of AV-pedestrian interaction. In particular, the integration of pedestrian behavior can enhance the understanding of how an AV communication affects traffic flow. Based on this, the dissertation provides an outlook on how behavioral data could be collected in the future in both virtual reality and Wizard of Oz studies to evaluate AV communication concepts. However, the empirical studies also highlighted the current limitations of the research approach, ranging from high inter-individual variation in crossing behavior to the limited generalizability of the findings due to the simplified and artificial experimental setup