The restoration software had a progress indicator that Elianor had learned, over years of use, to distrust.

It showed eighty-three percent completion on the current filtration pass, which meant approximately nothing in terms of what would actually emerge from the process. The percentage measured computational effort, not acoustic result. She had recovered perfectly intelligible speech from files that clocked in at twelve percent complete, and she had watched a ninety-seven percent pass produce nothing but a cleaner version of silence. The number was not the point. The point was what you could hear when it finished.

She clicked through to the output when the bar reached one hundred.

She listened.

Better. Marginally. The phonemic structure in the middle section of the fragment was slightly more defined — she could distinguish consonant shapes more clearly now, the outline of syllables beginning to emerge from the noise. But the aurora interference in the latter half remained dense and unresolvable with her current approach. The magnetic field data she had from the neighboring stations was too low-resolution, captured by equipment that had not been designed for primary monitoring. It gave her context but not precision.

She needed the primary data.

She had been turning this around for two days now — since she had sent the email, since he had replied in nine words, since she had given him a date and he had simply said he would be there. She had been professional about it, which was the only mode she had available, but the professionalism was sitting over something else, something that she was aware of but was not examining closely because examining it was not currently useful.

She pulled up the research database and ran a search on the Rook Weather Tower monitoring records. What she found confirmed what she had already suspected: the tower had been the only primary aurora monitoring station within range of Lysmere Lake on the night of the incident. The data from that night, if it still existed, would give her a full magnetic field profile — the kind of resolution that would let her strip out the remaining interference layer by layer rather than in the blunt-instrument passes she was currently limited to.

She ran a calculation on how much the resolution would improve with primary data.

The number was significant. The kind of significant that meant the difference between *something that might be speech* and *something she could transcribe*.

She sat with that for a moment.

She ran the broken passage through the current software one more time, concentrating on the section after the name. The waveform showed activity — there was speech there, clearly, the amplitude pattern consistent with multiple words, possibly a full sentence. She could see it. She could not hear it. Not yet.

She thought about what it might say.

She had been trying not to think about what it might say.

---

At ten in the morning, she went down to the archive's physical records room and requested the secondary documentation file for Frostline 19 — not the incident report, which she knew by heart, but the supplementary material that had been collected in the months following the accident. Passenger manifests, communications logs, the technical inspection report that had been commissioned by the transit authority and completed eighteen months after the crash.

She had read all of these before. She read them again.

The technical inspection report was the most interesting, as it always was. It had been produced by an independent firm commissioned by the transit authority itself, which meant it had a particular character: thorough in its documentation of the storm conditions and the track failure, precise about the sequence of mechanical events, and carefully silent about anything that might constitute an admission of preventable cause. The language was the language of events that occurred, rather than the language of decisions that were made. The train had left the tracks due to ice accumulation on the eastern curve. The emergency response had been initiated. The search and rescue operation had followed established protocols.

What the report did not address was why the train had been running at all on a night when the aurora warning level had been elevated to orange.

The Frostline route ran along the shore of Lysmere Lake on its eastern approach. The lake was subject to magnetic interference from strong Aurora Veil events — a documented phenomenon, present in the technical literature going back thirty years. During orange-level events, the track navigation system's magnetic positioning components could be affected. This was also documented. There were internal communications from the transit authority's own engineering department, obtained through a public records request by a journalist two years after the accident, that raised precisely this concern about the eastern Lysmere approach.

Those communications were dated four months before the accident.

Veyrline Transit had not modified the route. They had not updated the operational protocols. They had run the train on the night of a strong Aurora Veil event and attributed the resulting accident to weather conditions.

Elianor had been compiling this documented record for six years. She had it organized in a folder she kept separately from the restoration work, because the restoration work was about finding Niven and this was about something else — something she had not known what to do with until now, because she had no leverage, no platform, no technical basis for a formal challenge.

What she needed — what she had always needed — was a technical case that went beyond the public records. Something that showed not just that the transit authority had been warned, but that the conditions on that specific night had been precisely what had been predicted, and that someone had decided to run the train anyway.

The aurora data from the Rook Weather Tower would show exactly that.

A full magnetic field profile from the night of the incident, from the primary monitoring station closest to the eastern Lysmere approach, would demonstrate with scientific precision what the field conditions had been at the time the train left the tracks. It would prove that the orange-level warning had been accurate, that the interference had been real, and that the decision to run the route had been made in full knowledge of those conditions.

She had not, until now, had a reason to believe that data still existed.

She put the supplementary file back. She returned to her workstation. She sat for a moment with her hands in her lap, processing the particular feeling of something she had been working toward for a long time becoming suddenly, unexpectedly close.

Then she opened a new email draft.

She wrote it in four sentences. She sent it. She went to get coffee.

---

The reply came while she was still in the break room.

She read it standing at the counter with her coffee cooling beside her. *I have the data. Tell me when.*

She read it again.

He had kept it. Seven years. He had left Aurevik without processing it, without publishing it, without submitting it to the institutional archive as he would have been expected to do with research data from a monitoring project. He had taken it with him and kept it intact, which was either professional negligence or something else entirely, and she did not know which and was not going to decide right now.

What she knew was that the data existed. What she knew was that it was complete. What she knew was that within days she would have access to the resolution she needed to hear what was in the forty-one second fragment, and to build the technical case she needed to challenge the Veyrline operational clearance.

She named a date. She sent it. She finished her coffee.

---

In the afternoon she ran two more filtration passes on the fragment, using slightly different parameters. The second pass produced an unexpected result in the opening section — a short burst of clearer audio preceding the name that she had previously attributed to interference artifact.

She isolated it. She boosted it carefully.

It was not artifact. It was a voice. Not Niven's — the frequency characteristics were different, the fundamental frequency lower, the age register clearly that of an adult rather than a teenager.

She could not make out words. But she could make out the shape of speech — a few syllables, low and urgent, something being said quickly and quietly in the way people spoke when they were trying to be heard over noise without raising their voice above a certain level.

She leaned closer to the screen. As if proximity to the waveform would help. It never did. She knew this.

She ran the boost again at maximum safe gain.

There. A sound. Not fully resolvable, not transcribable, but present. Someone speaking, just before Niven said the name.

Two people in the recording. Of course there had always been two people — the incident report confirmed that both Niven and Einar had been in the last carriage. She had known this. But knowing it as a documented fact and hearing it as a human voice in a forty-one second fragment of sound were different things, and she found herself sitting back in her chair with the particular feeling of something shifting.

She had been listening to this file for seven years and she had been listening for Niven. She had been listening for her brother. She had not been listening for the other voice.

She was listening for it now.

She ran three more passes with variations on the filtration parameters. By the end of the afternoon she had isolated enough of the second voice to confirm that it was present in the recording for the entire recoverable portion of the fragment — not occasional, not peripheral, but continuous. Two people talking. Before the name. After the name. Underneath the interference, in the spaces between what she could and could not hear, there was a conversation.

She sat at her workstation in the evening quiet of the archive and thought about what a twenty-one-year-old and a fifteen-year-old might have said to each other in a sinking train carriage, and she did not let herself finish the thought, because the data for finishing it was on a hard drive in a hotel room near Frostline Station, and she had a date, and she would wait.

Waiting was something she was practiced at.

She saved her work. She logged off. She put on her coat. She walked home through the Aurevik evening — the harbor smell, the cold, the amber streetlamps, the particular quality of the northern dark that she had grown up inside and that felt, despite everything, like home.

She stopped at the end of White Lantern Street. She did not look at the corner where the figure had stood three nights ago. She looked straight ahead.

She went upstairs. She lit the lantern. She made dinner.

She was going to need all the patience she had, because what she was approaching was not just the fragment and not just the aurora data. She was approaching the moment when she would know, finally, what had happened in the last carriage of Frostline 19 on the night of the fourteenth of January, seven years ago.

She had waited seven years for that knowledge.

She could wait four more days.

---

The next morning she arrived at the archive early, as she always did, but instead of going to her restoration booth she went first to the equipment room and requisitioned the archive's highest-resolution audio interface — the one used for professional broadcast restoration, which required a sign-out form and a reason. She wrote *Frostline 19 priority project* in the reason field, which was both true and sufficient.

She set the interface up in the booth. She ran a test calibration. She ran it again to confirm. The equipment was in good condition.

Then she sat at her workstation and began preparing the technical documentation she would need when the aurora data arrived.

This was the part of the restoration work that most people did not see — the framework-building that had to happen before the data became usable. She needed to map the expected magnetic interference profile for the night in question, using the secondary station data as a baseline, so that when she had the primary data she could identify immediately where the improvements would be greatest and concentrate her filtration approach accordingly. It was like drawing a map of where the noise was before you had the tool to remove it — you could see the shape of the problem even before you had the means to solve it.

She worked through the morning. She generated four comparison charts, a preliminary timeline of the aurora event's intensity curve based on the secondary records, and a list of seventeen specific sections of the forty-one second fragment where she expected the primary data to produce the most significant resolution improvement.

Section eleven on her list was the part after the name.

She put a star next to it and kept working.

At noon, her colleague Dagna knocked on the booth door and asked if she wanted to come to the break room for lunch. Dagna was forty-two, had worked at the archive for fifteen years, and had the professional tact to never directly ask about the Frostline 19 project but the human decency to keep asking Elianor to take lunch breaks.

Elianor went. She ate soup and half a piece of bread and listened to Dagna and the equipment technician Petter talk about the new monitoring array that was being installed on the northern ridge, and she was present in the way she was always present — fully attentive, engaged enough to respond, absent from herself in the particular way she had learned to be absent when the other part of her mind needed to keep running.

After lunch she went back to the booth and worked for another three hours. The afternoon light faded from the narrow window. The archive staff went home one by one. The overnight monitoring systems came on.

She was still there at seven-thirty when her phone screen lit up with a message from a number she had added to her contacts the day before with no name, just the initials E.R., which was as much as she was prepared to commit to.

*The transit authority announcement came out this evening. I assume you saw it.*

She had. It had come across the newsfeed at four-thirty and she had read it in the break room while the soup went cold and her hands sat on her thighs and she breathed in the counting-and-releasing way she had learned from her own therapist, which she had never told anyone about, because some coping mechanisms were private.

She typed back: *Yes.*

A pause. Then: *We should meet sooner than Monday. Before the announcement gains momentum.*

She looked at the screen for a moment.

She typed: *Tonight. Archive. Eight o'clock.*

She sent it and put the phone face-down on the desk and did not look at it again for ten minutes, which was how long it took her to finish the current documentation section and save it and close the file and go to the break room to make coffee, because if she was going to meet Einar Rook in thirty minutes in the archive building after everyone else had left, she was at least going to do it with coffee.

The reply had come through while she was in the break room. She read it on her way back to the booth.

*I'll be there.*

Three words. The same precision he had used in the email. She wondered if he was always like this — if the economy of his language was a choice or a condition, something he had developed over time or something he had always been. She found that she was curious about this, and then found that she was curious about the fact that she was curious, and then put both of those observations somewhere she did not need to look at right now.

She went back to the booth. She made room on the work table for a second person. She got a second chair from the reading room. She positioned it slightly to the side, which she did not think about until after she had done it.

Then she sat down and waited for eight o'clock.

She spent the remaining time reviewing her documentation. She was precise and focused and she did not let her attention wander, because attention wandering in the direction of what was coming was not productive. She noted three additional sections of the fragment she wanted to prioritize. She reorganized the comparison charts into a cleaner order. She checked the high-resolution interface calibration one more time.

At seven fifty-eight she heard footsteps in the corridor.

She had expected the buzz of the door — she had given him the after-hours entrance code in her message, because the alternative was going down to let him in and she did not want to meet him in the lobby, which was too exposed, too much like a stage. She had wanted the booth, where she was on her own ground.

He came to the doorway of the reading room and she looked up from her workstation and there he was again — the same height, the same dark coat, the same quality of stillness that she had been turning over in her mind since the night before last.

He was carrying a hard drive case.

She said, "You brought it."

He said, "I assumed we'd need it tonight."

She had not told him to bring it. He had simply assumed, correctly, that if the situation had changed enough to require an unscheduled meeting, it had probably changed enough to require the data. She noted this without commenting on it.

"Sit down," she said.

He sat. The chair she had placed slightly to the side. He looked at it for half a second before he sat in it, and she thought she saw the edge of a recognition — that she had placed it there intentionally, that she had thought about it, that the geometry of the room had been arranged with some consideration for him — and then he sat and it was just a chair.

She turned her screen toward him. "I've been preparing the technical framework for the filtration work," she said. "These are the sections of the fragment where I expect the resolution improvement to be greatest with the primary data." She walked him through the charts without preamble, the way she worked with all technical collaborators — directly, without softening, assuming competence.

He followed everything. He asked two clarifying questions, both precise, both about the specific filtration methodology she was using. She answered them. He nodded.

"The interface resolution on my secondary data is approximately a third of what the primary station would have produced," she said. "The difference will be significant in the high-frequency interference bands."

"The aurora event that night peaked at approximately three times average intensity," he said. "The interference in the primary data will be extensive, but it will also be precise — I calibrated the recording parameters for exactly that kind of event." A pause. "I was there for that. It was what I was monitoring."

She looked at him. He said it without emphasis, but the information was significant: he had been in the tower, calibrating equipment, watching the aurora intensify, when the train had gone into the ice. He had been doing science while Niven was dying. She had known this, technically, from the records. She had not known it in those terms before.

She filed it away. "The documentation I have prepared should allow us to move quickly once we have the primary data loaded. I can have the first enhanced pass running within an hour of calibration."

He opened the hard drive case. He set the drive on the table between them.

She looked at it.

It was an ordinary external drive — the kind used for archival research data, heavy-duty, designed for long-term storage. A label on the side in handwriting she did not recognize: *Aurora Monitoring — Rook Tower — Season records, Year 3.* Below it, in different ink, added later: *Jan 14 — full event.*

"The full event record starts at seventeen hundred hours on the fourteenth and runs through oh-four-hundred on the fifteenth," he said. "Peak intensity was between twenty-two hundred and oh-one hundred. That was when Frostline 19 left the tracks."

She picked up the drive. It was heavier than she expected, which was not a rational observation but which she made anyway. Seven years of being unavailable, and it weighed what it weighed.

She set it down. "We'll calibrate the interface to the source format tomorrow," she said. "Tonight — " She paused. "Tonight I want to show you what I have so far."

He looked at her. Something in his expression shifted very slightly — not surprise, but a kind of readiness, the particular settling of someone who has been preparing for something for a long time.

"All right," he said.

She put on her headphones. She held out the second pair, which she had placed on the table before he arrived. He took them without comment and put them on.

She loaded the fragment. She looked at the waveform, forty-one seconds stretched across the screen. Then she pressed play.

The static came first. The aurora interference, the ocean of magnetic noise that she had been filtering for years. And then, in the middle of it, the voice — young, familiar, saying the name that had changed everything the first time she heard it.

She watched him while he listened.

She was not sure what she expected. She had spent a long time imagining how Einar Rook would respond to various things. She had imagined coldness, deflection, the performance of someone managing their public presentation carefully. She had not imagined this: he sat very still, with the headphones on and his eyes fixed on some middle distance that was not the screen and not the room, and something moved across his face that was so unguarded that she felt, briefly, like she was seeing something she had not been meant to see.

The fragment ended. Static.

He took the headphones off. He set them on the table.

He said nothing for a moment.

Then: "There's more in it than that."

"Yes," she said. "I believe so. Which is why I need the data."

He looked at the drive on the table between them.

"Tomorrow," he said.

"Tomorrow," she said.

They sat in the archive in the quiet hum of the monitoring systems, with the hard drive between them and the fragment still loading in her queue, and outside Aurevik was dark and cold and entirely itself, and neither of them said anything else for a while, because some things did not need more words than had already been used.