Despite the impression of "moving pulsation" given by bagana songs, they are characterized by a high temporal regularity at a global level. The analysis of the rhythm is made difficult by acoustical and organological specificities of the instrument. The durations of the sounds are not proportional and regularity is not aimed at by the players. An analysis based on the concept of different levels of time segmentation proposed by Martin Clayton allowed a transcription of a fragment of two bagana songs. Those transcriptions were verified by a master of the bagana and original player of one of the fragment. In this, transcription appears not only as a necessary tool for the ethnomusicological analysis, but also as a means of verifying hypotheses.