Our paper concerns the grammar of clause combining in Finnish and Japanese conversation. We consider the patterns of clause combining in our data and focus on the verbal and non-verbal cues which allow participants to determine whether, after the end of a clause-sized unit, the turn will end or continue with another clause-sized unit, resulting in a clause combination. We conclude that morphosyntax alone cannot account for the patterns found in our data, but that the participants orient to, at least, prosodic and nonverbal cues in determining the boundaries of clauses and projecting continuation in the form of another clause. Also important for projection are fixed expressions or 'prefabs'. In addition, semantic and pragmatic factors play a role. In that sense, we explore the question of where the limits of grammar for interaction, understood as the knowledge which speakers share and which forms the basis for the creation and processing of novel utterances, should be drawn, and whether grammar should include, beyond morphosyntax, not only prosodic, pragmatic and semantic features but also bodily behavior.