In this study we analyze how learners constitute what it means to learn and know a song. This is investigated in the context of four 9-to 10-year-old children in dyads teaching each other to sing a song of their own choosing. How the children take on this task is studied in terms of how they dialogically co-construct pedagogical and musical values throughout the collaborative tasks. The empirical data consist of video observations of the children engaged in dyads. Informed by a sociocultural perspective, with an emphasis on mediational means, scaffolding and appropriation, the study seeks to examine how young people's instructional methods are facilitated and constrained by communicative resources of different kinds. The empirical data is analyzed as interactively unfolding activity. The study shows that the children make a distinction between learning and knowing a song, in terms of tool use. In teaching, learning is communicated as supported by mediational means in the form of external visualization tools, while knowing the song, from the participants' point of view means to be able to sing the song without any such mediational means. From a sociocultural theoretical perspective, this difference is conceptualized as the gradual learning process of moving from a materialized practice, based on external artifacts, to an embodied practice, that is, a change in mediational means rather than developing musical knowing without tools.