Music is sometimes compared to language as a system of communication, however this comparison is usually at a generic formal, cultural or social level. This paper explores this analogy at the detailed level of interaction: to what extent can musical contributions act as conversational turns? We explore this question through an ethnographic study of music lessons. We describe a new transcription notation designed to capture the interactional details of musical contributions. Using this notation we show that although the ultimate objective of a lesson is development of musical performance, the detailed structure of the musical contributions depends on their interactional organisation. We show that musical contributions display interactional structure at the turn and sub-turn level and are closely integrated with other verbal and non-verbal cues as part of the unfolding conversation.