Each track is made up of a linkage of regions, and each of the regions is made up of a linkage of nodes or clusters. Sound data are stored on a cluster-by-cluster basis, and reproduced by tracing such linkages. Each time editing is performed on a virtual track of the track, a track data of a corresponding leading region in the edited virtual track is stored in memory as a history record of the track. When an undoing instruction is given, desired record data is selected from the history records of the leading region. Where the sound data in the trailing-end cluster of a preceding region and leading-end cluster of a succeeding region amount to less than 50% of a predetermined total data quantity of sound data to be contained in a proper cluster, the sound data represented by a trailing-end offset and the sound data represented by a leading-end offset are combined together and then written into a reproducing cluster. In reproduction, the sound data written in the reproducing cluster are reproduced for a connecting portion between the preceding and succeeding regions.