source file: mills2.txt Date: Sun, 24 Nov 1996 08:23:19 -0800 Subject: Lost post from Brian McLaren From: John Chalmers From: mclaren Subject: "experimental" music -- In sharp contrast with the dismal pseudo- science and incoherent reasoning employed by a minority of forum subscribers in a remarkably feeble effort to "disprove" my criticism of the misuse of the term "experimental" in modern music, Steve Curtin made some cogent points about the term. "Experimental music" was used according to its correct dictionary meaning by Lejaren Hiller. If you read Hiller's 1956 text "Experimental Music," you discover that Hiller employed the term carefully. He called the output of his stochastic composition programs "Experiment 1," "Experiment 2," etc. Notably, Hiller was careful *not* to call the output of his computer programs "music." Rather, he viewed these experiments (which formed the movements of the Illiac Quartet) as systematic investigations of mathematical models of human musical cognition. I did some research on Markov analysis and it turns out that it was originally used in chemistry to measure mean free paths of molecules in chemical reactions between collisions. Thus it's obvious why Markov analysis failed miserably as a music composition tool: a hydrogen atom is a hydrogen atom no matter where it occurs, but a JI minor third played in the deep bass is a grinding sensory dissonance while a JI minor third played in the high treble is a smooth sensory consonance. Atoms are interchangeable and identical; musical intervals are not. If you read Hiller's 1956 book carefully, you will realize that Hiller himself understood that his experiments demonstrated the failure of his mathematical model of human musical cognition. Hiller claims that he was not interested in the aesthetic result of his "experiments," but if you read between his lines you can detect the clear disappointment he felt in not producing more human-sounding music (or, to put it another way, the ease with which listeners could tell that the music went nowhere). Little has been remarked on Hiller's subsequent efforts in this area; however, HIller's own writings about his subsquent "Computer Cantata" (and a listen to the work in question) prove enlightening even today. Unlike John Cage, Lejaren Hiller understood and respected the meaning of the term "experimental." Unlike John Cage, Hiller recognized that science demands hard numbers in order for an experiment to be meaningfully called an experiment. Unlike John Cage, Lejaren Hiller formulated a mental model of a physical process (composition as filtered noise subjected to modus ponens logic in the form of a rule-set); unlike Cage, Hiller reduced his mental model to mathematics (encapsulated in a computer program); unlike Cage, HIller generated a testable hypothesis from his mathematical equations (composition can be generated using such a computer program which is both statistically and perceptually indistinguishable from compositions produced by humans); unlike Cage, Hiller tested his hypothesis (set up a quartet to perform the composition, and recorded the reactions of people who listened to it); unlike Cage, Hiller's music could therefore be meaningfully called "experimental." -- By the way, no doubt some of you will object to my use of the term "computer composition" in a recent post critical of that genre. Many of you will claim that my post denies the validity of all "computer compositions," and since many "computer composers" (Larry Polansky, Carter Scholz, John Bischoff, etc.) have produced interesting and worthwhile music, my post is purportedly incorrect. In fact this boils down to a controversy over the meaning of the phrase "computer composition." What is a "computer composition"? Is a string quartet written out by hand and using a computer only to print the score an "computer composition"? Since the composer did not specify the exact number of 1/300ths of an inch border around the staves, these proportions were generated outside of the composer's direct control, by means of algorithms. Is such a quartet score therefore an example of "computer composition"? Clearly not. Let us then consider a rock band whose manager schedules performances using a computer (with scheduling algorithms). Is this "computer composition"? Clearly not. How about a keyboardist who performs from written scores and uses a computer to load different timbres into hi/r synthesizer during the performance? Is this an example of "computer composition"? Clearly not. Let us then move on to consider a composer who uses elaborate mathematics to generate timbres and chords during the course of a composition. Over his shoulder, the composer has an idiot savant watching who observes which notes the composer plays and calculates rapidly in hi/r head the resulting notes and timbres according to the composer's mathematical procedures. (One such prodigy, Johan Zacharias Dase, boasted such astounding powers of mental calculation that he calculated pi in his head to 400 decimal places over the course of 2 months. Thus it is humanly possible, while rare, to perform such elaborate mental calculations.) Is this an "algorithmic computer composition"? I would say not, since it can be performed by a human. To think of it another way, the elaborate mathematics of the composer can surely be boiled down to a set of charts and tables which can be memorized. Now if we observe that the elaborate mathematics of which I speak might well be figured bass notation of the Baroque period (which, if reduced to a computer program, would require a great deal of calculation to interpret correctly in any key signature as the music modulates), we realize that this is a pretty standard human musical operation couched in deliberately arcane terminology. Obviously, it's not "computer composition" if humans were performing it in the 18th century as a matter of routine musical practice. Now suppose we have a computer program like Lejaren Hiller's which performs many millions of calculations to obtain each note. The mathematics recirculate, requiring some of the output to feed back into the input. Many millions of numbers must be multiplied by weights, sieved with thousands of rules, and random numbers must be generated by multiple 32-bit numbers by one another to obtain overflow and then shifting them right or left by some larger number of bits. This seems a clear case of "computer composition" because the process is too elaborate to be performed by any conceivable human in any reasonable amount of time. -- What do all these cases have in common? Clearly, we make distinctions here depending on the proportion of the computer's input to the process of composition. If the proportion is very high, we can speak meaningfully of "computer composition." If the proportion of computer input is low, clearly we are talking only about computer-aided composition. Thus, true "computer composition" occurs only when the process of composition *demands* a computer and *exclusively* a computer. If a human can perform the same operations, this is not "algorithmic computer composition," it's merely ordinary composition *aided by computer.* This leads us to a clear distinction between *computer-aided composition* and *algorithmic computer composition.* "Algorithmic computer composition" only occurs when the overwhelming bulk of the compositional process takes place in the computer. If most of the compositional process is human, and the computer is used merely to keep track of vectors in ratio space, or display morphological shapes, etc., then we are clearly speaking of "computer-aided composition." My comments in my recent post referred *only* to computer composition--that is, to composition in which the composer sets up a computer program, lets it go, and walks away. Computer-aided composition is a whole different kettle of fish, since it is de facto human composition in which the computer is used as a convenient aid. The line between the two is of course blurry; as the proportion of human effort to computer interpretation approaches 1:1, it becomes hard to say whether the computer or the human is composing. My earlier comments should therefore not be taken as an argument against the validity of all composition involving computers, but rather against the validity of compositions produced mainly by unaided computer operations. To put it bluntly, winding up a computer and letting it spit out music on its own didn't work in 1956 and has failed miserably to produce any interesting musical results in the ensuing 40 years. Changing the number of tones per octave does not promise to improve that track record. However, computer-aided microtonal composition is a wide-open field, and promises many fascinating and aesthetically worthwhile results as new approaches are tried and new algorithms developed. In particular, computer *aided* composition might prove very helpful in harmonizing a melody line in an exotic microtonal intonation with too many notes per octave for humans to navigate easily. --mclaren Received: from ns.ezh.nl [137.174.112.59] by vbv40.ezh.nl with SMTP-OpenVMS via TCP/IP; Sun, 24 Nov 1996 18:45 +0100 Received: by ns.ezh.nl; (5.65v3.2/1.3/10May95) id AA06149; Sun, 24 Nov 1996 18:46:42 +0100 Received: from eartha.mills.edu by ns (smtpxd); id XA05292 Received: from by eartha.mills.edu via SMTP (940816.SGI.8.6.9/930416.SGI) for id JAA20801; Sun, 24 Nov 1996 09:46:40 -0800 Date: Sun, 24 Nov 1996 09:46:40 -0800 Message-Id: Errors-To: madole@ella.mills.edu Reply-To: tuning@eartha.mills.edu Originator: tuning@eartha.mills.edu Sender: tuning@eartha.mills.edu