This conference took place at Cambridge University, with the usual pomp and circumstance personified in the long list of sponsors of the event, and the fact that the conference booklet had an index…
The full conference programme can be found here.
I was delighted to see some heroes of mine talk, like Jamshed Bharucha, whose inspiring keynote address, ‘Musical Communication as Alignment of Non-Propositional Brain States’ was the highlight of the conference for me, and very relevant to my work. Another highlight was a paper on the second day by Marcus Pearce and Geraint Wiggins called, ‘Information Dynamics in Music Perception and Cognition.’ They had taken the idea of the analysis of music using information theory to another level, redressing past mistakes and drawing on Wiggins’ background of mathematical modelling to create and extremely convincing argument for the method. On asking Wiggins why they were pursuing the topic, he told me that he believed that the time had come when we actually have the tools and know-how to be able to do what the pioneers in the 1950s and 60s could only make a very small start on. Faster, better computers and far better software now allow the ‘information-musicologist’ (copyright V.Hawes 2007) the power to analyse layer upon layer of music, voiding the usual argument against information theory analysis - which is that it assumes music to be linear, which it is not.
Later in the conference, Wiggins, in a response to another paper talked about models in music theory. He said exactly what my colleague Richard had been talking to me on the train about, which is, the decreased worth of a model for music if it is a description model only, in other words it is designed, on using it to ‘test’ music (supervised learning of a system), to give the answer that is already there so it cannot be wrong. The model is a metaphor of what the experimenter/analyst already knows. Many systems in music theory are like this, including some that had already been talked about at the conference - a little like backwards engineering the structure of the music. A better model is one that does not utilise supervised learning for its development, but learns from what actually happens in rigorous testing, developing on the basis of real results. This creates a model that is far more able to describe a variety of music, rather than just the one piece or oeuvre it was designed to ‘explain.’ This kind of model, Wiggins explains, is explanatory, because the structure of the system goes some way to explaining WHY the music’s structure is the way that it is. In this sense, it is a true model and not just a metaphor for the musical structure. This whole approach is reminiscent of one of my pet topics, the explanatory gap between the humanities and the physical sciences. The supervised learning, top-down approach represents one, and the unsupervised learning, bottom-up approach represents the other. Hopefully, expressed Wiggins, they will someday meet in the middle.
Pearce and Wiggins have a project going at Goldsmiths using their new information-based system and this, along with other projects I have been made aware of in recent years, gives me new hope for the relevance of my work in current musicology, and the possibility of having an interested audience for my finished thesis. The time has finally come, says Wiggins, for us to find the baby, extract it from the bathwater (see Musical Communication (Oxford: Oxford University Press, 2005) dust it down, plug it into a computer and tell it all the new things we have learned.