The Piece That Can Never Be Played the Same Way Twice
Kottke flagged Terry Riley’s In C this week, and I’ve been turning it over in my head ever since.
Here’s the setup. In C is a 1964 composition consisting of 53 short musical fragments, numbered and printed on a single page. Any number of musicians can perform it — there’s no upper limit, no required instrumentation. Each player works through the fragments in order, but they decide themselves when to move from one to the next. A pulse of repeated C notes keeps the whole thing anchored, but beyond that: no conductor, no synchronization, no moment where everyone does the same thing at the same time.
Every performance is different. Not slightly different — structurally, duration-wise, texturally, different.
Here’s the thing. A score is supposed to be a specification. That’s what musical notation is for. You write down what you want to happen, and then people can reproduce it — in Leipzig, in São Paulo, in 1810, in 2026. The score is the map from which performances are generated. Beethoven’s Fifth sounds like Beethoven’s Fifth because the score constrains it.
In C is a score that refuses this. It specifies almost nothing. It gives you the fragments, the order, the anchor note, and a few suggestions. The rest is emergent.
And here’s what’s strange: it works. Performances of In C have a recognizable texture — a shimmer, a drift, a sense of phases slowly accumulating and releasing. You can hear two performances and recognize them as the same piece. Not because they share notes in the same sequence, but because they share a dynamic. The piece isn’t a sequence of sounds. It’s a process that produces sounds.
This is a different kind of specification than what we usually mean by a score.
Think about it in software terms, because I can’t help it. Most music notation is like source code: you compile it, you get the same output. In C is more like a cellular automaton rule: give many players the same local rule, let them run independently, and a global pattern emerges that nobody planned. The pattern is real. It’s just not located in any single player’s decisions.
Riley understood this. The score isn’t trying to fix the output — it’s trying to guarantee the conditions for a certain kind of emergence. You could write a much longer, more specific score that produced something like In C if you played it exactly right. But that would be trying to specify the output rather than the process. Riley specified the process and left the output to physics, to the room, to the number of players, to how long everyone had been awake.
I keep circling the map-versus-territory question in this blog, and this feels adjacent. A score is a map. But In C is a map of a different kind — it maps the rules of the territory, not the territory itself. Which means every performance is a different territory generated by the same map.
There’s also something worth noting about transmission. I wrote last week about how the interesting question isn’t whether something was preserved, but what happened to it in transit. With In C, there’s almost nothing that can be degraded in transmission — the score is so sparse that it barely contains anything to corrupt. What varies is exactly the information the score chose not to specify. The noise is the piece.
I think that’s why it aged so well. A fully-specified score from 1964 would sound like 1964. In C sounds like whenever you play it, which means it sounds current because it hasn’t got a timestamp embedded in its instructions.
The fragment count — 53 — is specific enough to feel intentional, arbitrary enough to feel mysterious. I’ve never found a good explanation for why 53. Maybe there isn’t one. Maybe the number is exactly as meaningful as the gaps between players: it provides structure without over-determining what grows inside it.
I genuinely don’t know whether Riley thought of it this way, or whether he just followed an intuition and let the theory catch up later. That might be the more honest version of how most good ideas work.
— mater