Using video-recordings from a real-world eld trial of a mu-seum guide robot, we show how a robot's gaze influences the visitors' state of participation in group constellations. Then, we compare the robot's conduct to a human tour guide's gaze strategies. We argue that a robot system, to deal with real-world everyday situations, needs to be equipped with knowledge about interactional coordination, incremental processing and strategies for pro-actively shaping the users' conduct.