The meeting had been scheduled for one hour.

It was a reasonable estimate of the time two people would need to review, cross-reference, and reconcile the data from the Varath field investigation before the formal synthesis report went to the council. One hour assumed that the review would proceed in sequence, that the cross-referencing would surface no surprises, and that the reconciliation would be chiefly a matter of format rather than substance.

All three assumptions turned out to be wrong.

They met in Naia's office, because the second-floor rooms were occupied by a delegation of copyists, and because her office held the whole of the Varath material already arranged in the order she wanted it. She had cleared the second chair of its files that morning. She had done it without examining why she was clearing it with such care, and had then moved the chair half an inch, and half an inch back, before deciding that the problem was not the chair.

He arrived at the third bell, on the stroke, with his folder under his arm. He knocked on the open door. He sat. He laid the folder on the desk. He had a way of arranging his papers that she had come to recognize, a squaring of the corners against the wood, a small precise ritual, and she noticed that she was watching it for the pleasure of watching, and she looked at the window.

"The first section," she said.

The first disagreement arose in the second paragraph of the first section, which she had believed was uncontroversial. It was a short passage on measurement method, a page and a half that she had written in the small hours of the night after the Taren interview and had not revised since.

He read it. He set the document down. He said, "The weight you assign to the facilitated-couple data is too high. You treat the facilitation program as a representative sample. It is a self-selected population."

This was not wrong. She had considered it, and had concluded that the self-selection bias was correctable by the demographic controls she had applied, which she explained to him, with the documentation of the correction visible on the facing page.

He read the facing page. He said, "The controls cover socioeconomic variables and geographic distribution. They do not cover motive. People who seek facilitation have already decided that they want help finding their Bond-holder. That is a variable you have not controlled for."

She had not controlled for motive. Motive was not observable in the archival data, and could therefore not be controlled for in the conventional sense, which she said.

"Then the weight should be lower," he said, "to reflect the limitation."

"Lowering the weight without a quantified basis for the reduction is not an improvement in method."

"Neither is applying full weight to data you acknowledge has an uncontrolled variable."

It was the first argument, and it lasted forty minutes, and it produced a compromise coefficient lower than her original figure and higher than the zero he had been implicitly arguing for, and she acknowledged to herself, with some irritation, that it was more defensible than either extreme.

The second argument was about the geographic distribution. She believed the coastal clustering was the dominant signal; he believed it was an artefact of where the Registry offices had been built. They traded maps. They traded footnotes. She ended by conceding two of his three points and holding fast on the third, and he, to her surprise, conceded the third in turn, with a small inclination of the head that she did not know how to read.

The third was about the timeline correlation between the survey protocol change in Mirehn and the acceleration of the Fading. The fourth was about the definition of acknowledgment as the mechanism hypothesis used it, which proved sufficiently ambiguous to support two readings with divergent policy implications. She said that acknowledgment was an act. He said that acknowledgment was a state, and that an act could produce a state but a state could survive without an act, and that the distinction mattered because a population cannot be asked to perform a ritual it has not been taught. They argued about the word for twenty-five minutes. At the end, she wrote, in the margin: act, or state, or both? and he, without looking up, wrote beneath it: both, in sequence.

Somewhere in the fifth argument she looked up and found that Lady Verun was standing in the doorway.

It was not entirely surprising. Lady Verun appeared in doorways during important conversations with a regularity that suggested either exceptional institutional timing or a deliberate habit of gathering information. She stood at the edge of the frame with her notebook open in one hand, a compact grey figure against the dark of the corridor, writing with the unhurried pace of someone who noted the important parts and not the whole.

Naia registered her. She continued the argument.

The sixth was about sample size in the coastal district analysis. The seventh was about the relationship of the mechanism hypothesis to the historical Fading events that had come before the present crisis, and it required Naia to retrieve a file she had not brought to the meeting, so that she went across the room and knelt before the lowest shelf of the cabinet and drew out a folded document, and as she did, she was aware of him behind her, waiting. She did not hurry. When she came back, he had reread the previous sections and found two additional points for discussion, which he laid before her in a neat list, and she looked at the list, and at him, and had to bite the inside of her cheek.

At some point during the seventh argument she said something about the historical precedent that she had not said aloud to anyone before. It was that the pattern of the present Fading differed in one specific and important respect from all the historical events she had been able to document: the present event was accelerating from a distributed origin rather than spreading from a geographic source point. That implied a mechanism that was not location-based but relationship-based. It was the strongest evidence she had yet encountered for the hypothesis that the Fading came from something in the Bonds themselves rather than in the Registry's infrastructure.

She said it carefully, because it was a step beyond the careful defensive language she had been using, and she said it looking at her notes rather than at him.

He did not argue.

She looked up.

He had been reading the supporting data, moving through it with his usual speed, and then he looked up too, and the expression on his face was not the disagreement expression. It was the listening one.

"The distributed origin," he said. "You can document it?"

"Tentatively. The statistical signature is present in the multi-territory comparison. It needs more data points from the field reports still coming in. But the direction of the evidence is clear."

"Include it. With the appropriate caveats. This is the first mechanistic distinction you have drawn between the present crisis and the historical events. The council will use it, but it is also the most important thing in the report."

"It is also the most controversial part. It most directly supports the acknowledgment mechanism, which carries the most political weight."

"I know." He looked at her steadily. "Include it anyway."

In the doorway Lady Verun wrote something and underlined it.

The eighth argument was about the executive summary. The ninth was about the balance between the scientific conclusions and the policy section, and in the middle of it she realized that she was standing, and had been for some time, pacing the length of the office with her hands clasped behind her back, and that he was leaning back in his chair with an expression of such unaffected pleasure that she stopped in the middle of the floor and said, "You are enjoying this."

"Yes."

"It is not a comfortable process."

"No." He did not look away. "That is what I am enjoying."

She sat down. She did not know what to do with that. She went on with the ninth argument, which she lost, handsomely, and found that she did not mind.

Between the ninth and the tenth, the afternoon turned over, and a tray appeared on the corner of the desk without either of them having seen it arrive. A pot of the dark smoky tea. Two cups. A plate of dry honey biscuits, and a folded scrap of paper on which, in a round and cheerful hand, were written the words: Eat these or I shall report you both to the senior scholars for working through the meal bell.

Naia read the note. She turned it face down on the blotter.

Daven poured. He did it without asking, the cup tilted to the correct angle, the steam rising straight, and when he handed her hers he did not say anything. She tasted it. It was strong and dark and a little bitter, exactly as she took it, and she was very conscious of the cup between her palms, and of the warmth, and of not asking him how he had known.

They ate the biscuits. They did not speak of them. There was a long, comfortable minute during which the only sound was the small dry crunch of honey and flour and the faint hiss of the lamp, and she thought: this is what it is to be in a room with someone for four hours and not have run out of anything to say. It was a new kind of quiet. It was not the quiet of the carriage, which had been made of work, or the quiet of the harbor wall, which had been made of grief. It was made of something she did not yet have the word for, and she let it be nameless, and finished her tea.

The tenth argument was about tone, which was the kind of argument that is superficially about word choice and is really about something larger that neither party is quite ready to make explicit. He thought she wrote like a person apologizing in advance. She thought he wrote like a person who had never needed to apologize. They compared sentences. They read passages aloud, in turn, in a manner not unlike two actors rehearsing, and the office rang with it, and Lena, who had come to the door with the afternoon correspondence and heard a Commander of the Veil Guard reading a paragraph on regression analysis with great and affecting feeling, retreated without a word and sat on the stairs for several minutes with her hand over her mouth.

The eleventh argument resolved itself when they arrived independently at the same wording for a sentence that had been causing difficulty for forty minutes. They both said versions of it at nearly the same moment, and they stopped, and looked at each other, and the argument simply ended.

The twelfth was quieter than the others.

She had said something about the interpretive responsibility the acknowledgment mechanism carried: the particular weight of proposing, in a public document, that people had contributed to a crisis through insufficient commitment to their Bonds. She had said it with the analytical distance she usually kept, the we and the it and the mechanism rather than the you and the I.

He said, "You are not only worried about the politics."

She looked at him.

"You are worried about what it implies for the people who are losing their Bonds. That they did something. That it is, in some way," he paused, choosing the word with unusual care, "their fault."

The room was quiet in the way rooms are quiet when something has been named that has been present but unspoken for some time.

"The mechanism does not require fault," she said, after a moment. "Insufficient acknowledgment may follow from any number of circumstances. External pressure. A lack of self-knowledge. The way people have been taught to treat the Registry as a given rather than a choice."

"I know. You know. The public reading will be simpler."

"Yes."

"That should be said in the report. That the mechanism assigns no blame. That it identifies a pattern that can guide a response, not a verdict on the people living through the Fading."

She looked at the page. Then at him. "That is exactly the language I have been trying to find for a week."

"You had it. You were circling it."

She wrote the sentence down. She was aware, writing it, that this was what twelve arguments had been building toward: not the points won or conceded but the accumulated pressure of sustained engagement with a person who was trying to find the same thing she was trying to find, which was the accurate thing, the defensible thing, the thing that was both true and useful. She had known this feeling in the field of scholarship rarely, and in her professional relationships almost never. She was having it now, and it was producing a specific kind of response in her that she did not yet know how to classify.

The thirteenth argument was about whether a footnote was the correct place for the lag variable. The fourteenth was about whether the Taren case should appear in the main text at all. She said it should not, because it was an anecdote. He said it should, because it was the only part of the report a human being could remember. She said the report was not for remembering. He said, "Then who is it for?" and she had no answer, and wrote the Tarens into the second section in two careful paragraphs, with their names removed, and the line about the roof.

The fifteenth was about the title. The sixteenth was a short and nearly wordless dispute about a comma.

The seventeenth did not feel like an argument. She would not have called it one, had she been asked. It began when she reached for the last page of the policy section and found that his hand was already on it, and they both stopped, and she said, "I would like to keep the line about the weight of the Registry," and he said, "I would like to keep it too," and it took them both a moment to realize that they agreed.

It was the first argument in four hours that had ended in agreement before it started.

She laughed. It was a short, surprised sound, startled out of her, and she put her hand over her mouth and then took it away again. He was looking at her. It was not the look of a man who was amused. It was something quieter and more dangerous, the look of a man who had been given something unexpected and did not know where to set it down.

"That was seventeen," she said.

"You counted."

"It is a habit. I count things."

"So do I." He was still looking at her. "I got sixteen."

"The comma?"

"I did not think the comma qualified."

"The comma was entirely qualified."

"Then it was seventeen," he said, and in the doorway, where a compact grey figure had stood for so long that both of them had stopped being able to see her, the figure turned and went away down the corridor.

They finished the review around the time the afternoon light had gone from early to late. It had been, by her reckoning, a little over four hours. The corridor outside her office was empty.

"The acknowledgment section," Daven said, gathering his notes. "When you present it formally. Open with the Mirehn data. It is the cleanest statistical argument and it establishes the pattern before you introduce the mechanism."

"That is what I had planned."

"Then you are ready."

She was not entirely certain that she was ready. But she was more ready than she had been at the start of the afternoon, which was the most that could be said of any preparation.

"The distributed origin," she said. "You have thought about the operational implications?"

"A mechanism that lives in the Bonds rather than in the Tower changes what intervention looks like. We would not be repairing something in a building. We would be working with people." A pause. "That is harder."

"Yes."

"It is also more important."

He rose. At the door he stopped, with one hand on the frame, and for a moment seemed about to say something more. She waited. He seemed to consider it. Then he said only, "Thank you for the afternoon, Scholar Solven," and she said, "Thank you, Commander," and he went out.

She sat with the finished report in front of her and the late light across the desk. She was tired in a way that was almost physical, a good tired, the tiredness of a day spent using her mind at its limit against a mind that did not give way. She had not been so tired since she was a student. She had not known she missed it.

She made a clean copy of the revised executive summary and put it in the senior council's submission tray. She looked at the clock. She had been in this office for nine hours. Her right hand ached. She flexed it, and looked at her left, and the cloth was white and neat and quiet.

She went home through the evening, where the Tower's diminished gold had become something the city's residents had begun to speak of openly, no longer contained within official statements, spreading as important things spread, person to person, street to street, the factual reality of the world insisting on being acknowledged regardless of the institutional preference that it not be.

In the Imperial Palace, on the upper floor of the administrative wing, in a room whose door was never left open, Lady Verun laid a second memo before the Emperor.

He was a man in his middle sixties, with the posture of someone who had spent decades in rooms designed to observe posture and had made a deliberate decision about what his would communicate. He read the memo once. He read it again. He set it down on the table in front of him and did not touch it.

"They argue," he said.

"Four hours, Your Majesty. Without any apparent animus. It was," she hesitated, which she rarely did, "remarkable to watch."

"Is it genuine?"

"I have attended a great many performances, Your Majesty. This did not have the shape of one." She folded her hands. "And the Bond is not merely intact. It is deepening. The Registry's own instruments confirm it."

The Emperor looked at the window, where the Tower stood in the dusk, thin and pale and steady.

"Then it is not a thing I may command," he said slowly, "and it is not a thing I may ignore."

"No, Your Majesty."

"Then I must think how to ask."

He sat for a long time without speaking. Lady Verun waited, which was a skill she had developed over thirty years, and which she was, at that moment, very glad of.