Sunday, August 8, 2021

Two crude models of how we make music [thinking out loud]

This is from an email I just sent to my friend Bob Krull. We were colleagues at the Rensselaer Polytechnic Institute (RPI) back in the Jurassic Era. He's now retired and summers in Saratoga Springs in upstate New York and winters Fort Meyers (I believe) in Florida. He works at playing pop, RnB, and jazz piano in both places. That's what we talk about, how to make music.

We spend a great deal of time thinking about improvisation. Both of us are interested in and read extensively in psychology and neuropsychology. I play trumpet and am somewhat more experienced than Bob is. He plays piano. This makes for interesting discussions: we share a high level of intellectual sophistication across differing levels of skill and experience.

And now for something like a model. Actually a pair of them.

One we’ll call mechread, for mechanical reading. The other we can call parset, for parameter setting.

Mechread: How do we play music from a written score? Unless one is very experienced and skilled, one uses mechread. The score gives you a representation of a series of pitches and note values. You call on memory and translate them into a sequence of movements that activates your instrument. The result is a sequence of pitches each of the proper duration. But it doesn’t sound much like music. Too mechanical.

Parset: How is it that I can hear a tune – not any tune to be sure, but many – once or twice an be ready to blow on it fluently? I may even be able to play the melody ‘from memory’? I call up an appropriate ‘engine’, one attuned to the style of tune I’m listening to, set a half-dozen parameters, and I’m reading to go. Picking up those half-dozen parameters is much easier than actually memorizing (via a mechread procedure) the tune. It requires much less information. Of course, the model whose parameters I’m setting, that contains a great deal of (pre-compacted) information.

The thing is, in order to give a musical performance of a written score, you’re going to have to do something like parset. Sure, you can lather on some expression, some of which is likely marked in the score (crescendo, accel., rubato, etc.), but that’s just another layer of mechread. It smooths over the rough edges, but doesn’t breath life into the sound. For that you need parset. Thus when performing effectively from a score, you’re treating the score as a source of prompts to parset.

So now we’ve got mechread and parset going at the same time. And I’m sure mechread is going when I’m listening to a tune in a jam session so I can blow on it three choruses later. You don’t have to have a written score to use mechread. I figure we use a version of mechread to ‘transcribe’ the sounds we hear into ‘inner music’ which we can then ‘pass’ to parset [if we have and have developed the knack].

Now consider this improvisation I did on a Japanese tune, "Kojo no tsuki."

I worked on it a bit before I recorded it, but never wrote anything down and never deliberately set anything to memory other than the melody. I play the melody, in tempo, fairly straight time. Then I improvise a chorus, freeing up the time here and there. Then I go into a medium swing tempo an continue improvising, going through a couple of choruses until I reach a climax and than, after a very short break, collapse back into the melody, in time, and finish. I’m interested in that first improvised chorus. I can hear it in my mind’s ear as I type. I can conjure it up at will. If I were to pick up my horn and play that tune, I’d have no trouble playing that melody. In fact, I’d probably have to exert a deliberate act of will to play something else. Let’s assume I let well enough alone and then switch into a medium swing like I did in the recording. I’m pretty sure I’d do some/many of the things I did on that recording. But I don’t know how close I’d come to what I’d recorded. Don’t even know if I’d play the same number of choruses. But I don’t think I’d deliberately avoid playing many of those riffs. That would be too difficult, and a bit pointless.

That’s more or less how unnotated composition happens. I think.

This post discusses unnotated composition: Some varieties of improvisation.

No comments:

Post a Comment