Digital devices such as social robots are increasingly being developed as artificially intelligent learning tools that could support and expand early childhood education by providing new ways to engage children in social interaction. Given this potential, research in child-robot interaction has begun to investigate which aspects of a robot’s behavior provide advantages within the interaction. However, a perspective that addresses the child’s communicative behavior within a child-robot interaction is seldom explored. In this chapter, the results of a long-term child-robot study with preschool children are presented, in which children’s multimodal response behavior during a word learning task with a social robot was in focus. The results reveal that children not only used different communicative multimodal signals such as gestures or delay markers when interacting with the robot, but also changed their behavior over the course of the sessions. The findings suggest that children shape their responses to a robot in a manifold multimodal way beyond verbal lexical utterances. Finally, implications for future child-robot interaction research, as well as identifying issues that will need to be resolved for social robots to be helpful and responsive in interaction with young learners, are discussed.