The EU-project “iTalk” aims at developing artificial embodied agents that will be able to acquire complex behavioral, cognitive, and linguistic skills through individual and social learning. Within this frame, the Bielefeld team investigates social aspects of learning and focuses on a scenario, in which a human tutor presents and explains some task to a learner, who/which observes the action and, in turn, attempts to understand its structure in view of eventually imitating it. We use ‘tutoring’ in parent-infant-interaction as an empirical model to explore the topics of (i) action structuring / acoustic packaging (-> Lars Schillingmann), (ii) tutoring spotter (-> Katrin Lohan), and (iii) variability (-> Karola Pitsch). Karola’s research focuses on the interplay between the tutor’s and learner’s actions. She accesses the topic of ‘variability of tutor’s actions’ from an interactional perspective detailing the sources and consequences of observable actions in the unfolding course of action. This leads to an interactional account of “motionese” behavior and the investigation of feedback strategies in both parent-infant- and human-robot-tutoring. Methodologically, Karola uses Conversation Analysis to reveal interactional patterns and attempts - in interdisciplinary collaboration - to link qualitative research with quantitative approaches and formalization. Results of her empirical research are used as basis for the design for human-robot-experiments, which she evaluates with regard to the systems’ functioning and to understand the ways in which the users’ reactions and expectations are shaped in situ.