User:FloraC/On the canon of music

This text body concerns the canon of music. Not that canon. The other canon. Canon means the source material accepted as officially part of a story. Now the first question is: what does it have to do with music? Actually, music often manifests itself in many forms such as recorded, performed, and written out, which technically all carry different information. The very notion of thinking of those forms as equivalent – as a piece of work – cannot be taken for granted. There are some limitations to the material, and we are here to figure them out.

Digital audio has many advantages that lend itself to fit the role of a canon in contemporary music. By the subject is meant the digitally rendered, losslessly compressed audio file supplied in openly accessible formats. In terms of the signal form, it may be a waveform, a spectrum or a spectrogram. Most essentially, all the sounds can be perfectly reproduced, and never degrade with time – except for the impact from cosmic rays. Such a level of reliability is incredible if you think about it. Indeed, there has never been a moment in history as exciting as this where the magical technology of digital audio can take on the basis of the supreme form of music.

That unfortunately leads to two direct consequences that prompt us to re-evaluate it. First, what about live performance? Why perform if digital audio is the supreme form of music? After all, no performances could be as authentic as the canon. Second, what if the musical quantities fail to fit the digital grid? The lack of musical feature extraction in digital audio is to blame for both problems.

Musical features are what make a human process music distinctly from how an insentient machine does, because human beings are gifted at perception in the pattern of musical features rather than of signal itself. For example, we think of music as made of notes. That knowledge is not recognized by the audio at all.

Contrary to digital audio's absolute lack of musical feature extraction, the musical score may be of more relevance, as it is made up of musical feature extraction. Modern staff notation was so created for the duty of collecting all the information required to reconstruct the musical ideas in a piece, and no doubt, all it carries now is condensated wisdom as it underwent the numerous rounds of iteration in the history.

Note that music is a prototype of nonlinear art to which emergence applies. To clarify, notions such as linearity and emergence are largely alien to the flow of music, so the saying is a gaze on music from logos. But it is necessary. On the positive side, at least it explains why artificial intelligence engineers have always been interested in training their little bots through none else than deep neural networks.

Of course, the score extracts musical features in a way that is mostly linear. Nonlinear features are also found in scores, but they are not central. Rhythm is a strictly linear feature, unfolding along the horizontal axis. Pitch is also a strictly linear feature, unfolding along the vertical axis for each part. Note that using the nominal–accidental chain in place of a plain piano-roll-style pitch space is not nonlinearity – it is called rank-2, which is still linear. Both types of quantities are perfectly accurate as they are recorded in their native grid instead of the foreign-to-score digital grid.

Dynamics are very nonlinear; other expressions, even more so. For what? Dynamics, the musically relevant element, does not rigidly translate to loudness, the digitally relevant element, since dynamics is a musical feeling that defies logic. For the same reason, dynamics does not translate to velocity, and velocity does not translate to loudness, either. One could hear what it is like in many scorewriter programs should dynamics be treated as a linear feature translated straight to velocity. It typically feels like the work is nowhere near completion whenever played in that setting.

Generally speaking, music is less linear than the score makes it out to be. In other words, the score is an underfitted model, even with dynamics and expressions. Because music is nonlinear, it is vague in which way it is nonlinear. It is this underdetermined nonlinearity that opens up the possibility of all the interpretations of the same piece, besides the rather debatable external factors such as social and historical considerations.

Among all the possible interpretations, some are closer than others to the composer's intention. It is a producer's job to implement one which is as close as possible to the composer's intention, and the result is typically rendered in the form of digital audio. So when we say digital audio, we do not merely refer to the form. We refer more to the implementation which supposedly reflect the composer's intention.

Therefore, regarding canon, digital audio and musical score combine to a better reference point than either form alone. The score is the feature set of the audio; the audio acts as the exemplar interpretation of the score.

While live performance is not canon, it is never devoid of meanings: it is the exploration of the interpretation space, where the performer adds their own taste to the space defined by the composer. Thereby a discussion of intention between the composer and the performer solidifies; a "dialog" of music is achieved.

Release Notes

© 2023 Flora Canou

Version Stable 0

This work is licensed under the Creative Commons Attribution-ShareAlike 4.0 International License.