A music generation system includes a segment manager and a plurality of segment objects. Each segment object has one or more track objects. Each track object containing a time sequence of musical performance data. Some of the track objects are dependent track objects and some of the track objects are controlling track objects. The dependent track objects contain dependent musical performance data that is interpreted based on controlling musical performance data contained in one or more of the controlling track objects. Each of the controlling track objects contains a predefined type of controlling musical performance data, wherein a plurality of different control type identifiers correspond respectively to different predefined types of controlling musical performance data. When a dependent track object needs controlling musical performance data, the track object calls the segment manager with an argument comprising a control type identifier. The segment manager responds by calling one or more segment objects with arguments that include the received control type identifier. In response a called segment object identifies one of its track objects that contains the requested type of controlling musical performance data and returns the musical performance data from the identified track object to the segment manager.