In this study, we developed a new electronic musical instrument, Cymis, with which notes are played by touching a touch screen and the volume of the note is controlled by expiratory pressure. It can be played with about the same amount of exhalation as a keyboard harmonica. Performance experiments were conducted with healthy adults (11 persons) in their 20s. First, all were able to play solo using Cymis. Next, simulated ensemble performances were conducted. Using a system that can predict the ideal timing for sounding notes from visual information, six novice performers performed a piece (Amazing Grace) after having had from one hour to several weeks of practice. For each note in the ensemble performance, the difference between the actual and ideal times at which the note was produced, td, was measured. In the experiments, |td| was found to be less than 200 ms for more than 85% of the total number of notes played. In an ensemble performance in which ve experienced performers listened to the sound of the performance, this result was more than 80%. This study shows that a novice performer using Cymis can play in an ensemble with about the same degree of temporal accuracy as an experienced performer.