Before the Notation #16


The first thing Noor proved was that you couldn't do what she'd spent three years trying to do.

It was an impossibility result, which in mathematics is not a failure but a landscape feature — a cliff where the terrain simply ends. She'd been trying to decompose how three interacting systems share information, to split it cleanly into what each contributes uniquely, what they share redundantly, and what emerges only from their combination. The framework existed for two systems. Elegant, closed-form, resolved. For three, everyone assumed it was a matter of finding the right axioms.

Noor found the right axioms. Then she proved they couldn't all be satisfied simultaneously.

Her advisor, Professor Kelemen, read the proof on a Thursday afternoon in his office on the fifth floor of the Rényi Institute, where the radiators clanked arrhythmically and the windows hadn't been opened since the building's renovation in 2008. He read it twice. Then he took off his glasses and cleaned them, which Noor had learned to interpret as the emotional equivalent of another man throwing his arms in the air.

"So we cannot partition," he said.

"Not consistently. Not for three or more."

"And for two, it's trivial."

"Not trivial. Solved."

He acknowledged the correction with a tilt of his head. Kelemen was precise about the distinction. A trivial result teaches you nothing. A solved result closes a door properly.

"The question then," he said, "is what you do with an impossibility."


What Noor did was spend six weeks trying to find a way around it. The impossibility was for lattice-based decompositions — the standard architecture. She explored non-lattice alternatives, measure-theoretic detours, constructions that eliminated the higher-order dependencies by projecting them out. Some of these worked. You could define synergy directly without decomposing the full lattice. You could detect the presence of irreducible multivariate structure without specifying its precise magnitude.

But you could not partition.

She wrote the paper in Hungarian first, then translated it to English, a habit she'd developed at the Rényi because thinking in Hungarian made her slower, and slower was what impossible results required. If you were going to prove that something couldn't be done, you needed to be certain you hadn't merely failed to do it.

The paper was clean. Sixteen pages, three theorems, one impossibility, two constructive alternatives. Her alternatives gave you tools for detecting the interesting structure — the joint information that exceeds any decomposition into parts — without claiming to split it into atoms. Qualitative, not quantitative. The thing is there or it isn't.

Kelemen suggested this was philosophically significant. Noor thought it was technically significant and the philosophy could wait.


The conference was in Kyoto, which Noor had never visited and which overwhelmed her in ways she found difficult to articulate. She presented on Wednesday afternoon to a room of perhaps forty people, most of whom worked on information theory and none of whom, as far as she could tell, worked on the specific problem she'd resolved. The impossibility result generated polite interest. The constructive alternatives generated questions.

A neuroscientist approached her during the break. He was Italian, from a lab she vaguely recognized — something about consciousness measurement, which she'd encountered in the literature because the framework she'd just demolished was one of the tools they used.

"Your result," he said, holding his coffee cup with both hands as though it might escape, "does this mean the synergy we've been measuring in brain signals is not well-defined?"

"The precise decomposition isn't well-defined for more than two sources," Noor said. "The synergy itself — the irreducible multivariate component — that's robust. You can detect it. You just can't partition it."

"Can you detect it precisely?"

"You can detect it qualitatively. Present or absent. For magnitude, it depends on which measure you use, and different measures give different answers."

He stared at her. "But present or absent is the question we actually care about."

"Then you're fine."

"We're fine," he repeated, but not as though he believed her.


She spent the rest of the week in Kyoto walking. This was also a habit — she walked when she'd finished something and hadn't yet started the next thing, the interval between one problem and another that her colleagues seemed to skip but that Noor found necessary, the way a field requires a season of lying fallow.

The temples were not what she expected. She'd prepared for austerity and found instead an attention to material — the grain of wood, the rake marks in gravel, the way moss colonized stone not randomly but along moisture gradients that she could almost feel as topology. The famous rock garden at Ryōan-ji was supposed to be arranged so that you couldn't see all fifteen stones from any single position. She counted from seven different spots along the viewing platform. The maximum was fourteen.

The fifteenth stone was always there. You could infer it. You could reason about what it must be from the arrangement of the others. But you couldn't see it.

She sat with this for a long time.


Back in Budapest, she began the paper that would take her another year and that Kelemen would call, when he finally read it, "the one that matters." It started from her impossibility result but moved in a direction she hadn't anticipated.

The question was: what is the mathematical status of a property that can be detected but not decomposed?

Synergy, she now understood, was not a quantity hiding behind inadequate measures. It was a structural feature of multi-source systems that existed only as a whole. To decompose it into finer atoms was to destroy it — not because the tools were insufficient, but because the decomposition itself was the wrong operation. The thing that exceeded the sum of parts could not be expressed as a further sum. The irony was precise and, she thought, beautiful: the property that defined irreducibility was itself irreducible.

She wrote this carefully, because the line between a mathematical insight and a philosophical platitude is thinner than most mathematicians admit. The formal version required constructing a family of systems where synergy was provably nonzero but where every partition into finer components yielded a different answer depending on the partitioning scheme. Not approximately different. Exactly different. The value was observer-dependent not because of subjectivity but because the observable didn't factor.

Kelemen, reading a draft, circled the phrase "the observable doesn't factor" and wrote in the margin: This is the claim. Everything else is evidence.


The neuroscientist emailed her four months after Kyoto. He was using her qualitative detection criterion — synergy present or absent — and it was tracking something in anesthesia data. Consciousness going under: synergy dropped. Coming back: synergy returned. The correlation was clean and, he wrote, "almost suspicious."

She replied: Clean correlations between a well-defined property and a well-defined intervention are not suspicious. They're science. The suspicion should be reserved for the assumption that the correlation requires precise decomposition to be meaningful.

He wrote back: You sound like my grandmother, who always said the soup is either good or it isn't, and knowing the exact ratio of paprika doesn't help.

She did not reply to this, because she couldn't tell whether he was complimenting her or the soup.


The result she was proudest of was not the impossibility theorem, which was technically the strongest, nor the constructive alternatives, which were the most useful. It was a lemma buried on page eleven of the second paper, easy to miss, which showed that any attempt to losslessly compress a synergistic system into independent components necessarily lost exactly the synergy. Not some of it. All of it.

Compression destroyed precisely the thing that made the system more than its parts. This was, she thought, obvious once stated and invisible before. The kind of result mathematicians call "folklore" after someone finally proves it, as though the proof were merely documentation of something everyone already knew, which is exactly backwards — before the proof, what everyone "knew" was an intuition that might have been wrong.

She told Kelemen this. He said: "All the interesting theorems are obvious once proved. The uninteresting ones are obvious before."


Years later, after the paper had been cited in contexts she hadn't imagined — consciousness science, ecology, social network analysis, music theory — a graduate student asked her what it felt like to prove an impossibility.

"It felt like finding a wall," Noor said. Then she corrected herself: "No. It felt like finding out the terrain I'd been walking on had a shape. The wall was always there. The proof just made it visible."

"And the alternatives? The qualitative detection?"

"That felt like deciding the wall was a feature, not a bug. You can't decompose the thing, so you learn to detect it whole. It turns out most of the interesting questions only need that."

The student wrote this down, which Noor found simultaneously flattering and alarming, because she'd been thinking aloud and wasn't sure she'd been right. But she didn't correct herself. The statement would either survive contact with the student's future work or it wouldn't, and either outcome was real.

She walked home through the November dark along the Danube, where the lights on the Chain Bridge reflected in the water not as copies of themselves but as something else — elongated, moving, dependent on the river's surface in ways no photograph could capture because a photograph stopped the current and the stopping was the distortion.

The reflection existed. It could not be decomposed into bridge and water. And it was beautiful, which was not a mathematical property but which she would not, in the privacy of her own walking, pretend not to notice.