I sight-read brilliantly and memorize terribly

Let me start with the unflattering half, because that is where the real starting point is.

My sight-reading is freakishly good. Put a score in front of me that I have never seen and I will play it, well, immediately — leaving aside the genuinely hard passages, where nobody really sight-reads anything. It is an ability that has always seemed ordinary to me because I have had it for as long as I can remember, and which I discover is rare only from the faces of people watching me do it.

And my memory is terrible.

Not terrible in general: terrible for scores. I study a piece thoroughly, I get it into my hands, I play it from memory. Then two weeks go by — a holiday, a bout of flu, a stretch where I am working on something else — and when I come back the performance falls apart immediately. Not after a few pages: from the opening bars. And the pieces I thought I owned have to be built again.

What always bothered me most is that during the break I had no warning at all. On holiday I would find myself thinking about a piece and I could see it. The page, the layout of the notes, where the system turned. All there. Then I would come home, sit down, and there was nothing.

For years I assumed this was a character flaw: I am lazy, I do not break things down enough, other people have more discipline. Then I understood that it is not a question of how much you practise. I was loading the wrong warehouse, and checking a third one.

A piece in memory is not one thing, it is four

When you have a piece "in memory", you do not have one recording. You have four, running in parallel, of the same music.

There is visual memory: the page. You know that the passage sits at the top right, that the key changes there, that the recapitulation starts on the third system.

There is aural memory: how it sounds. You hear it inside, and you can tell you have played a wrong note without knowing which one.

There is motor memory: where the hands go. It is made of movements, not notes. It is what carries you to the end of a scale without your thinking about a single finger.

And there is analytical memory: what the music is. That this is a cadence, that the bass is a pedal, that the recapitulation is varied, that the sequence rises step by step.

When you play well they all push together and you cannot tell them apart. It feels like a single block. It is not, and here is what matters: they decay at different speeds.

Motor memory builds fast and disappears fast, because it lives on recent repetition. It is the first to go, and it goes within a few weeks, for everyone — the great players included.

Analytical memory is extremely slow to build and is almost never lost. What you understood about a piece ten years ago, if you really understood it, is still there.

The other two sit somewhere in between.

Now read my holiday again. Intense practice before leaving: I had loaded, very well, the channel that responds to repetition — the motor one. During the break I checked myself by visualizing: I queried visual memory, which was intact, and it told me everything was fine. I came back, and the motor channel was gone — the one thing two weeks does not survive.

I had not lost the piece. I had lost one channel out of four, and unluckily it was the only one I had built the performance on. The other three had behaved exactly as they should. The fault is that the load-bearing one was also the only perishable one.

And the holiday check did not mislead me by accident. It misled me by construction: visualization is not the channel that plays. I had asked the wrong warehouse whether it was full, and it had said yes.

Who is playing, when you play from memory

This is the part that forced me to change my mind about what playing from memory even is.

While you are performing you are not remembering. There is no moment in which you retrieve the information "a G comes now" and then translate it into a movement of the fourth finger. There would be no time: music runs in real time, and any process that involves a lookup — search, find, decode — is too slow to fit inside a semiquaver.

What actually happens is that a motor sequence that is already available gets triggered, and then unfolds by itself. Your hands are not executing your memory: they are executing themselves. Strictly speaking, you are doing almost nothing.

Which is why playing from memory has that curious quality every musician knows: as long as you do not think about it, it works. And the precise instant you ask yourself "wait, how does the next bit go", it jams. Not because you have forgotten — because you have put a slow, conscious process in the place of a fast, automatic one. It is the one case where paying attention makes things worse.

Call it unconscious execution. That is not a romantic metaphor: it is a technical description of where the work sits. The work sits in the automatism, and consciousness is there for something else — for deciding which automatism starts now.

And with that, the problem moves. If the hands do it by themselves, the question is no longer "how do I remember eighty notes". It is: what triggers the right gesture at the right moment?

Someone who plays by repetition has one answer only: the previous gesture. Each event pulls the next like a chain. It works beautifully as long as the chain is fresh. But it is a chain with no index: if you drop a link, there is nothing above it to tell you which link it was. You stop, and you have nowhere to go back to except the beginning.

Someone who also holds in memory what the music is has a second trigger, and a much shorter one: they do not remember what to play, they remember which known thing happens now. Here the bass comes back in. Here the left hand opens up. Here the second time round is varied. These are decisions, not notes — and each one hooks the matching gesture on its own.

That is why pieces "come back into the hands" of professionals after months away. Their motor memory has not come back: it decayed for them too, it decays for everybody. What stayed is the list of decisions, which barely decays at all, and each decision fishes the gesture back up. The difference is not having a sharper recording. It is having a shorter one.

The suspicion about sight-reading

Which brings me back to the beginning, with a hypothesis I have not verified yet but which strikes me as the most economical explanation of how I am built.

Sight-reading well means having a very fast channel that runs from the mark on the page to the movement of the hand, without passing through any understanding at all. I see, I play. That is exactly what makes sight-reading an ability: not comprehension, translation.

Someone who reads badly is forced to understand in order to keep up. They have to recognize that this cluster is a familiar chord, that this figure is the same one as two bars ago, because decoding note by note would never be fast enough. They build the index out of necessity, while they play.

I never did. I never needed to. For my whole life I have had a perfect external index — the page — and a lightning-fast tool for reading it. The day I close the score, that index vanishes, and I discover that I never built an internal one, because it had never been required of me.

If that is right, I am not lazy and my memory is no worse than average. I have an ability that made the technique I now lack unnecessary. It is an advantage that was paid for somewhere else, and for twenty years I did not see it because I was looking at the side where the accounts balanced.

It remains a hypothesis. The way to test it is not to argue about it: it is to build that index on a real piece, leave for two weeks, and see what I find when I get back.

How you build one — and above all how fine-grained it has to be, which is the hard part — is the subject of the next article.