On 3 Jun 2003 at 9:23, Christopher BJ Smith wrote: > Back in my undergrad days I had lunch with one of my theory teachers > and we discussed some interesting research that had been done into > finding out how the brain reacts to different types of music. I said > I thought this might have big implications for theorists and > composers, and he asked me if I thought, some day, we might be able > to identify and analyse every aspect of music as we hear it. I > responded (in my youthful enthusiasm) yes! and he replied, "Well, > then, you are a much more dedicated theorist than I am." > > That conversation affected me deeply. Now I have been working on > (Warning! True section ends, joke ahead!) my Grand Unified Music > Theory, which will reduce all aspects of all music to a simple > formula that will fit on a half-page of paper. One will be able to > understand everything about a piece by simply reading the analysis. > This will eliminate the need to listen to music at all! > > I'm almost finished it, just a few kinks to work out...
Well, believe it or not, there was a positivistic belief in the idea of complete analysis back in the 50s and 60s and 70s. There was also a belief that computers would make it possible to analyze a lot of things that were too complex to be analyzed by humans (or just too much data). One such project took as its goal the abstraction of ideas about musical style. The goal was to use the analysis in the service of making determinations of authenticity based on style. First, you'd take a body of demonstrably authentic works by composer X and run them through the handy-dandy-musickalizer, and out would come your data. Then you'd run the demonstrably authentic works of composer Y and Z through the same hand-dandy-musickalizer. Then you'd take the unknown work (or the one with conflicting attributions) and run it through the mechanism and then you'd have a definitive answer as to who wrote it. I am not making this up, folks. It was called "Activity Analysis," and at base, was actually a really excellent idea. It was based on the idea of determining rates of change in surface rhythm, melody and harmony (all of which is quite musically sensible, actually), and then doing some kind of transformation on the result the eventually boiled down to a single number (and this is where things got bogus, of course). One of the first test cases involved some pieces with conflicting attributions to Mozart and Dittersdorf. The problems were manyfold: 1. Two distinct musical styles could threoretically result in the same number, so it wasn't telling you anything definitive. 2. Although they did consider the factors separately as well, this didn't help, because those, too, were boiled down numbers for a whole piece or a whole movement, and many things about "activity" levels a particular piece are going to be drawn from conventions for a particular genre of piece (a minuet is a minuet is a minuet, at a certain level), so you might not actually be examining style of composers so much as aspects of generic configurations that would cause a composer to make particular stylistic choices within that genre. 3. Because it is about rates of change, it tends to privilege rhythm over other things, since it's all about change over time. I'm not as bothered by this objection as by the others, because to me, that's what music is, changes in pitch and rhythm and harmony over time, of varying densities and intensities. But it does tend to place the time element at the forefront as analytic determiner. 4. Your "control sets" (i.e., the demonstrably authentic pieces) have to be very pure. It is the last aspect in which the original practitioners fell down the worst, as they used 4 symphonies that have been attributed to Mozart for which there are absolutely no authentic sources and which have not been really believed to be by Mozart for the last 20 years or so (the research happened in the early 70s, at a time when most scholars had already concluded that these pieces are not by Mozart). While I can shoot down what was wrong with the whole endeavor -- the main problem was the goal of having definitive "factual" answers to questions that are not scientific -- I sure as hell wish I could run MIDI data through an activity analysis program, and get a score that had 3 activity numbers for ever measure in a piece! Indeed, I don't know of much in the way of computerized analytical tools at all. I raised this subject a few years ago on the this list and was pointed in the direction of a handful of programs that seemed to be designed for heavy-duty non-tonal theorists (set theory is easy to program in comparison to analyzing functional harmony). I havent' kept up with the subject at all, and would like to know if anyone is aware of any work like this? Is this, perhaps, a job for a Finale plugin? For instance, what if I could designate the cadence points in a piece and then have the plugin summarize the number of different cadence types? Sounds simple? Well, it's not, because the repertory in which this would be useful is early polyphony, where it is actually a question of some import. So, you'd really be looking at the final configuration and how it relates to the penultimate configuration and then classifying accordingly in terms of voicing and cadential pitch. In any event, is the plugin toolkit advanced enough to do that kind of thing? It would need to have some temporary data storage capability, both for the UI settings and for each piece (you wouldn't want to recalculate certain kinds of information each time), so I have my doubts. Anyone? -- David W. Fenton http://www.bway.net/~dfenton David Fenton Associates http://www.bway.net/~dfassoc _______________________________________________ Finale mailing list [EMAIL PROTECTED] http://mail.shsu.edu/mailman/listinfo/finale
