Decompiling your own automatism

At this point the series has a problem, and the problem is me.

I have argued that a procedure is not teachable because it has been compiled into an automatism, and that the teacher does not transmit it because they are in a different line of work. But a few weeks ago I published a memorization procedure: numbered steps, timings, a calibration criterion. If I were right, that text should not exist.

It does exist, and how I got there is the most useful thing in the whole series.

You get there from a breakdown, not from introspection

I did not arrive at it by looking inwards. Introspecting on an automatism produces nothing, or worse, produces a plausible rationalization, which is the most elegant way of being wrong.

I got there because something broke. I would come back after an interruption and the repertoire was gone. The breakdown forced me to ask what exactly had disappeared, and since not everything had — I could still see the pages, I could still hear the music — I had to admit that what I called "knowing a piece" was made of separate parts with different lifespans.

This is the general point, and it goes well beyond the piano: an automatism can only be read when it jams. As long as it works it is opaque by construction, because its opacity is what makes it fast. The moment it fails, the parts come apart and for an instant they are visible.

From which follows something counterintuitive about who can write procedures. Not the master at his peak — nothing breaks for him, so he has no material. And not the beginner either, who has nothing yet to decompile. What is needed is an expert for whom it went wrong, and who is stubborn enough not to write the breakdown off as a personal failing.

For years I read my own breakdown that way: I am lazy, I do not practise enough. With that explanation in hand I would never have written the method — because a moral explanation closes the inquiry, while a mechanical one opens it.

The strongest objection, and it is Taleb's

Anyone who has read Taleb has the objection ready, so I will make it myself, because it is the right one.

In Antifragile there is a formula that has stayed with me: lecturing birds how to fly. The idea is that the real direction runs from practice to theory and not the other way round — practitioners find things, and then somebody comes along, formalizes, and takes the credit, with the theory appearing to explain something that was already there and working perfectly well without it.

Applied to me it sounds like this: real knowledge is tacit, it comes from doing, and writing procedures is precisely the academic vice of formalizing what needed no formalizing. The birds were already flying.

It is a serious objection and it cannot be waved away. But it does not hit what I wrote, and the reason is precise.

My procedure does not teach anyone to play, and does not claim to generate competence. If you do not already have the hands, the index is of no use to you: it is a list of sentences about music you could not perform anyway. It does not replace the thousands of hours, and it does not shorten them.

It is useful at one point only: when the automatism is interrupted. It is a procedure for coming back, not for learning. It teaches no bird to fly — it tells a bird that stopped flying for two weeks how to start again without redoing everything from scratch.

And that, I think, is the correct location of any well-written procedure. It does not stand before practice, it stands inside its interruptions. It is a piece of scaffolding you put up when the building has subsided, not an alternative design for the building.

The counterweight, which has to be said

There are two ways this argument can do damage, and I write them down because otherwise the series becomes an advertisement for an idea of mine.

A badly written procedure is worse than no procedure. It gives the illusion of possession: you have the list, so you believe you have the thing. It is the same trap I fell into visualizing the score on holiday — I was checking the wrong warehouse and it was answering yes. A procedure with no way of noticing that it is wrong is precisely the kind of object that costs you years.

Which is why, in the memorization method, the part I care most about is not the steps but the measuring device: the empty telling, with its two symptoms and the rule by subtraction. Without that it would have been a list of good advice like any other.

And some things do not verbalize at all. Not all competence is decompilable: certain parts really are only exposure and time, and trying to write them produces texts that look precise and contain nothing. Knowing which ones they are takes the same honesty as knowing when a procedure works.

Where the series ends

It starts with a complaint — nobody taught me anything — and ends with something closer to a task.

If what we really know is compiled, if our masters do not have it accessible and it is not their fault, and if procedures are born from breakdowns, then breakdowns are the only source we have. And most people who work keep theirs to themselves, because admitting them looks like admitting you are not up to it.

I kept mine to myself for twenty years, and in twenty years that silence protected me from nothing. The only thing it produced was doing the same work again every summer.