¹The study of how deviations in timing in a performance (with respect to, for instance, a metronomical or mechanical performance) are used to communicate structure or the specific interpretation of the performer, and how listeners are able to perceive and appreciate these deviations.

[2]That they are related we know from a large body of experimental research in Music Psychology on expressive timing. See for example, Clarke (1987), Palmer (1989), or Repp (1992).

[3]This is especially true in a connectionist or neurode network approach to music where it is relatively easy to obtain a reasonably working system, without having a real understanding why it works.

[4]For example, the application of one rule might delay a note a bit (with respect to its notated onset in the score) while another rule might alter that same note to be earlier than notated, resulting in no change at all, an effect not described in any rule.

[5]The pursuit of it should therefore be rejected.

[6]Knowledge on melodic and harmonic structure seems to be dependent on what is known or described on time (e.g., rhythm interferes greatly with the perception of melody) -the other way around, starting with time and temporal structure (as we decided in our research), allows one to ignore (for the time being) the melodic and harmonic structure.

[7]The question, whether music cognition can be described as such a restricted and isolated domain, is still open.

[8]Making a micro-version of an existing program is an alternative to rational reconstruction, a more established AI evaluation methodology (see for instance, Richie & Hanna, 1990), that encompasses reproducing the essence of an AI program's significant behavior with another program, constructed from descriptions of the important aspects of the original program. In this way, one is able to evaluate the published claims made by those programs.