Opus+

Theory

In this section we examine the theoretical basis underlying the Opus+ project, we briefly consider the nature of music from a philosophical perspective, and trace how its ontology relates to the practice of composition, and discover what ways, if any, this impacts the viability of the Opus+ project.

Music as Physical, Structural and Present

The Opus+ project started as an experiment to explore the limits of music composition using computers. The approach takes a completely 'structuralist' view of music; music is a physical process which is processed as information both by the performer and the listener. In this sense music is not a 'thing' it is a physical process that involves a complex of vibrating instrument parts, physical sound vibrations transmitted through the air, and physical brain processes in both the performer and the listener. Each performance is an 'instantiation' of the musical instructions. A musical score is simply a set of ink-marks giving instructions for human performers to act in certain ways. A CD encodes machine-readable instructions whereby a hi-fi can reproduce, with varying degrees of fidelity, the original physical air vibrations of the music. This becomes clearer with the following question; if all the musical scores existing were never performed and all the CDs that ever existed were never played, nor could ever be played, in what sense would these scores and the CDs be music? And further; can music be produced without either a score or a CD?

Computationally, the amount of information in a performance far exceeds that written on a score; and as such a performer can 'interpret' a piece. In short, performance adds information to composition. This is clear if we consider a recording of a performance, on CD for example. This will contain huge amounts of information not present on the score; micro-tonal changes, unintentional squeaks and buzzes, subtle changes in volume and attack which all conspire to 'bring the music alive' for the listener, and enables critics to discern differences between different renditions of the same piece.

So, if we accept that music only exists when it is 'present', for example when it is being played, what objective features, apart from its physical and structural characteristics, does music have? In what sense does music carry or convey 'meaning'?

Objectivity and Semantics in Music

Questions concerning objectivity and semantics, can open a Pandora's Box of philosophical enquiries, which are probably impossible to resolve definitively, and will therefore not be attempted here. However, the following statements are fairly uncontroversial.

• Music has some physical, structural and informational characteristics which are objective, in the sense that the 'tables objectively have legs'. Music is physically pressure waves in the air; a particular musical phrase either has or has not just been repeated.

• The composer, the performer and the listener can all experience 'meaning' in a musical piece. The truth of this statement cannot be proved, but is supported by subjective introspection and the verbal reports of others.

• This 'meaning' must to some extent be 'conveyed' or 'affected' by the objective properties of the music. If not, what other properties could do this? In other words, if the experience of meaning could be affected without requiring the objective properties of music, then the physical presence of music would be entirely unnecessary, which is plainly not the case.

• The 'meaning' of a piece of music can be associated with an 'emotional' or 'affective' response. Again, this can only be supported by subjective introspection and the verbal reports of others.

All fairly straight-forward, but it does beg the question 'what meaning?'. To what extent is meaning being 'conveyed', or is it being 'affected'? Is it necessarily the case that the composer's meaning is correctly understood by the musician, and then conveyed accurately to the listener? Will the listener necessarily understand what the composer 'meant' by simply hearing the music? Or is this better typified as an example of Chinese Whispers?

Subjectivity and Music

There is little doubt that the performer and the listener subjectively respond to the music. However, it is unlikely that there is much congruence in these responses, either with each other or with any semantics intended by the composer. In fact, the thesis posited here is there is no meaning transferred by a composition whatsoever, not in the sense that language carries meaning. It is more accurate to think of meaning as 'affected' in the listener - not 'conveyed'. Any meaning intended by the composer can be elided by the performance, and masked by the listener's pre-existing psychological pre-dispositions.

For example, the composer might be inspired by 'birdsong', and attempt to express something of birdsong directly, or indirectly, in the composition, but it is unlikely that a performer will understand 'birdsong' or anything like birdsong unless given additional hints in everyday language. The same is true of the listener, only more so, because the composer's intentions will be filtered through the performance. Uncontroversially, any meaning imbued by the composer will be 'coloured' or 'filtered' by the time it is heard by the listener; in the extreme case it will be completely erased by the musical process.

The user's subjective state will effect how the music is experienced. Listening to exactly the same piece of music at different times, on different days, and at different points in your life, radically changes how the music is subjectively experienced. Contrast hearing a track by the band 'Rage Against the Machine' at a university student party as opposed to your mother's funeral. The state of the listener strongly determines how the music is experienced.

However, physical properties and structural elements in the music do affect the listener. Different performances make the same composition sound completely different. Take an extreme case where a performance of Chopin's etudes is played at a rate of one quaver per fortnight, or is performed on a bunch of bananas. It is self-evident that the emotional response of the listener will be utterly different when compared to a more usual performance on the piano at the correct speed.

Given that for the listener there are two sources of 'influence' on their musical experience: the physical and structural properties of the music, and their pre-existing psychological pre-disposition, of what importance is the composer? What is left of the artistic process of composition that affects the listener?

Emergent Structural Composition

What is left of the artistic process of composition that affects the listener? In short only the physical and structural elements of the music produced. In fact, it is only those physical or structural elements which can be perceived by the listener as 'contributing factors' or those that are 'noticed as making a difference' which are important. This includes subtle subconscious physical features, which although not consciously acknowledged by the listener, still contribute to the overall musical experience, for example by 'creating a certain atmosphere'.

For example, on a 12-tone equal temperament instrument, such as the piano, the fact that a piece is in a minor key, as opposed to a major key, will influence most listeners - the piece will often be described as 'sad' or 'emotive'. Whereas the fact that the piece is in the key of A-Flat-Minor rather than A-Minor makes little difference to most listeners, except maybe those with perfect pitch, or with unusually developed critical abilities.

Does the actual artistic compositional process matter? Is it important whether Bartok composes his 5th String Quartet using mathematical formulae based around the proportions of the golden section? Would the piece be any more or less profound if this exact same composition had been created directly by some kind of immediate creative revelation? Or even by him tossing a coin repeatedly? Its not a question of how probable these scenarios are, its a question of the status of the resulting work. It matters not a jot how the work is composed, merely that it is as it is - that the resulting composition results in music with physical and structural properties as such.

Much composition is arbitrary. When pushed, as to why a particular phrase was chosen, or the reason a particular harmonic form was used, composers are often at a loss to really explain their process - at least for some sections of their piece. They might describe their muse, or the source of their inspiration, or some byzantine mathematical decision procedure but this only explains their choice historically not aesthetically . It is very difficult, or even impossible, for a composer to aesthetically justify compositional decisions, as against other choices that could have been made .

One purpose of this project is to test this provocative thesis; because the important features of a musical piece are its physical and structural properties; because there is no meaning conveyed by music, it is only affected in the listener; and because the actual compositional process, as an artistic process is arbitrary, constrained by habit and irrelevant to the end result; that composition can be undertaken 'as effectively' by computers as it can by humans; the phrase 'as effectively' means here that the result is heard as musical. It does not have to be the same kind of music as that produced by human composers. We are interested to what extent musical artefacts will emerge from programs designed around 'constrained random generation', and other artificial intelligence techniques.

Composition Space

If we imagine all the possible musical compositions arrayed in space, where each point in the space represents a particular composition, and nearby points represent similar compositions, they would be provably infinite in extent. Ask yourself how many ways can a piece of music move from middle-C, through a series of notes and back to middle-C? All the music so far composed by humanity in its entirety is finite in quantity, and would occupy infinitesimal volumes dotted throughout this composition space. Computer programs can uncover the musical compositions that lie in these interstices, in the places that human composers would never venture, (sometimes for good reasons), precisely by de-subjectifying the composition process.

This brings us to the other purpose of the Opus+ software; to confound habitual compositional practice, or 'boggle the jam' in common parlance. Composers, like any other human will to a greater or lesser extent work from habit, many deeply ingrained, and many hard-earned by experience. Opus+ will generate musical material independently of any prevailing human foibles, and result in unusual musical fragments or novel musical ideas that might prove useful in expanding compositional repertoire.

To have the best chance of discovering interesting musical material, Opus+ must be able in principle to land at any point in this composition space. Every limit put on the compositions generated by Opus+ will mean there are areas in the space that cannot possibly be encountered; and each excluded area will be infinite in extent. For example, if it was not possible for Opus+ to generate compositions containing dissonant intervals, such as a minor 2nd, think how many pieces of music would be excluded a priori. Some very interesting music relies on using such dissonant intervals. Because it is impossible to determine whether a particular point in the space represents an interesting composition before we land on it we cannot exclude the possibility that Opus+ will generate compositional rubbish, along with interesting or even beautiful material, if we are to allow Opus+ the maximum scope within the composition space.

It remains a very interesting project to attempt to characterize this composition space in terms of its structural dimensions, (there is one dimension for each discernable compositional characteristic and would have many, maybe even an infinite, number of dimensions), the similarity relations between nearby points and to characterize the proportion of compositions judged 'beautiful' as compared to 'ghastly'. However, this is beyond the scope of the current research.

Conclusion

The theoretical basis of Opus+ is based in the idea that music is essentially a physical process with informational structure, and these are its objective features. When music is 'heard as present' a large component of the subjective response of the listener is one of 'affection' caused by these objective properties, rather than by the 'communication' of any 'meaning' from the composer. Therefore, it does not matter how a composition comes to arrange these physical and structural properties of the music, just that these particular physical and structural properties are so arranged. Taken to extremes, this can lead to the provocative thesis that 'the human composer is redundant', but thinking more conservatively and more realistically, it means that artificial composition using computers could in principle produce musical artefacts that would be heard by human listeners as musical. This in turn implies that such musical artefacts could prove useful to human composers to augment their own practice.