This study examines non-lexical vocalisations accompanied by the German deictic “so” or “so_was” as a multimodal package in music rehearsals. Focusing on the sequences prior to the performance of a piece, where this practice often occurs, I demonstrate that the multimodal package, vocalisations + “so” or “so_was”, is utilised by participants in the service of establishing a joint decision on how to perform the next part of a musical piece. The analysis shows that participants demonstrate the rhythm in which the next piece should be played and, by seeking their interlocutors’ approval in the next turn, orient to their shared deontic authority with regard to decisions about their joint performance.