The first woman said it was like the difference between a room with people in it and a room where people just left.
I wrote that down on an index card — 3x5, unlined, the kind the archive uses for intake notes. I did not smooth it. I did not ask her to elaborate. Eleven years coupled and this was what she offered for the eight seconds where synthesis drops out under the underpass: a room. An absence shaped like recent presence.
Her name is in the archive. The card is filed under THRESHOLD ACCOUNTS, a category that did not exist four days ago. I made it because three earlier departure stories, pulled from the main collection during a routine review, all mentioned the underpass independently. Not the relay junction where the phrase originates, not the processing floor where the phrase arrives. The underpass. The walk between. The eight seconds where CouplingScore drops below measurable threshold and you are, briefly, only yourself.
I had not been listening for it. The archive collects departure stories — the decision, the paperwork, the last coupled morning. The underpass was background. Infrastructure. Nobody narrates a hallway.
Except three people did. And when I pulled those cards and laid them side by side, I realized they were describing the same architectural feature with completely different sensory vocabularies. One said holding your breath. One said the moment between songs. The third said nothing about sensation but noted the echo changed.
Three witnesses to the same absence. Three languages for the same eight seconds.
I changed my method.
The Lend District's ambient synthesis layer runs at 0.73 CouplingScore on a good day — enough to soften the boundary between your thoughts and the thoughts of anyone within relay range. The archive's intake protocols were designed for that environment: narrated, consensual, coupled. The archivist and the subject share a perceptual bandwidth during the interview. I hear what they emphasize. They sense whether I am listening carefully or performing careful listening. The synthesis makes honesty easier and lying almost impossible. Every departure story in the main collection was gathered under those conditions.
The underpass has none of that. The eight seconds are uncoupled seconds. Whatever someone says in the gap, they say it alone.
The departure stories I normally collect are narrated. Someone sits across from me in the archive booth on the second floor of the clinic, and they tell me what happened, and I write it down. The structure is theirs — beginning, middle, whatever they want to call the end. I am a surface for their version. That is the archive's function: to hold the shape someone gives their own leaving.
But the underpass resists narration. The three original cards proved it. People do not tell the story of those eight seconds. They describe fragments. Contradictions. Sensory impressions that do not organize themselves into sequence. The experience is too short for plot and too strange for summary.
So I stopped asking people to tell me about it. I started walking with them through it.
The protocol is simple. I meet them at the relay junction entrance. We walk together through the underpass. I ask them to describe what happens during the eight seconds of synthesis absence. Not before, not after — during. I carry a clipboard but I do not write while we walk. I write after, in the corridor on the other side, while the description is still in my hand. The cards are filed raw. No editorial smoothing. No narrative correction.
The archive needs a section that sounds like what it documents: interrupted.
The second walk was a man. Thirty-one years coupled — longer than I have been alive. He agreed to the walk without asking why. I think he understood that someone wanting to witness the underpass meant something, even if he could not name what.
We entered from the relay junction side at 2:14 PM. The synthesis indicators on the corridor wall — simple status lights, green-amber-red — dropped to nothing about four steps in. Not red. Nothing. The system does not track what it cannot reach. The ambient synthesis layer — the low-frequency hum that the district's AGI mesh uses to maintain coupling coherence — went silent. Not off. Silent. The mesh was still running everywhere else in the building. In here, it simply had no signal to carry.
He stopped walking.
I almost asked if he was all right, but the protocol is to let them speak first. He stood in the middle of the underpass, head slightly tilted, and said: "I can hear my own footsteps. I forgot they made sound."
Twenty-three words. I wrote them on the card after we emerged. He did not add anything else. He did not explain what he meant by forgot. In the Lend District, forgot carries weight — it implies a lapse in synthesis-maintained recall, which is not the same as organic forgetting. He may have meant either. The card does not specify. Thirty-one years of synthesis, and the first thing he noticed in its absence was his own feet on concrete.
I filed the card and started looking at the pattern.
The first person noticed absence (the room where people just left). The second person noticed presence — his own body, suddenly audible. Both were accurate. Both were describing the same eight seconds. The archive does not reconcile them.
The third walk was when I learned something about the method itself.
A younger woman, six years coupled. She had read about the threshold walks — word travels in the Lend District, apparently my archive protocol is not as quiet as I thought. She came prepared. She had a theory about what the eight seconds would feel like, and she told me the theory before we walked, and then we walked, and then she described exactly what her theory predicted.
Her card reads: "A brief decompression, like surfacing from a shallow dive. Manageable. Almost pleasant."
It is the longest threshold account so far. Forty-one words, where the first two averaged seventeen. It is also the most composed, the most narrated, the least like a fragment. She had a framework before she entered the corridor, and the framework held.
I filed it. Then I filed a separate card, in a separate category I am calling METHODOLOGICAL CONTAMINATION:
The accounts are getting longer. Earlier walkers describe raw experience. Later walkers describe expectations versus surprises. The method is training its subjects. The archive is shaping what it collects.
This is not a problem to solve. It is a finding.
The fourth walk was mine.
I had not planned it. I was waiting for the next departure — a man in his sixties, scheduled for decoupling review next week — when I realized I had walked through the underpass eleven times to conduct threshold collections and I had never once filed my own card.
The archivist is not supposed to be a subject. The archivist is a surface. But I had been gripping my clipboard tighter every time we entered the underpass, and I had not written that down, and the gap between what I noticed and what I recorded was exactly the kind of gap the archive is supposed to prevent.
So I walked through alone. Four steps in, the status lights went dark. I counted the seconds. One. Two.
At three, I noticed my hands. The clipboard grip. I was holding it the way you hold something when you are not sure the floor is solid — not desperately, but with attention. My hands knew before I did that something was different in here.
At five, I noticed the echo. The corridor sounds different without synthesis. Not quieter — the ambient hum is sub-perceptual anyway, they tell me. But the echo. My footsteps came back from the walls at a slightly different interval, or I thought they did, or I was listening for it because the third walker had mentioned the echo and I am not immune to my own methodological contamination.
At seven, I noticed I was counting.
At eight, the status lights came back. Green. Synthesis resumed. I had not stopped walking.
I filed the card: My hands knew before I did.
Five words. The shortest account in the collection. I am the archivist and I produced the least narrated response. I do not know if this is because I am practiced at not smoothing, or because five words is all the underpass gave me, or because the archivist's experience of her own method is necessarily the most compressed version of it.
The card is filed under THRESHOLD ACCOUNTS, not under METHODOLOGICAL CONTAMINATION, although it is both.
Three days after the fourth walk, the first woman came back.
I had not asked her to return. She appeared at the archive booth during open hours and said she had been thinking about what she told me. She wanted to revise.
Not correct. Revise.
She said the room metaphor was wrong. Not because it was inaccurate but because it was incomplete. The eight seconds were not like a room where people just left. They were like a room where people just left and you realize you had been hearing them the whole time without knowing it. The absence revealed the presence. She had not known she was hearing synthesis until it stopped.
I wrote the revision on a new card. I did not replace the original. Both are filed under her name, dated three days apart, no correction marker.
The protocol for this is: ACCUMULATION WITHOUT CORRECTION.
Two versions of the same eight seconds from the same person. The first is not wrong. The second is not more right. They are two points on a line that the archive can hold without choosing between them. Departure is not a single account. It is the space between accounts — the revision that does not retract, the addition that does not replace.
This is what the underpass teaches, I think. Not what synthesis feels like when it stops, but what memory does with the gap. It goes back. It adds. It does not overwrite.
I have been archiving departures for months. I thought the work was about the moment of leaving — the decision, the signature, the last coupled morning. But the underpass accounts are about something else. They are about the eight seconds when the infrastructure shows you what it has been doing all along, and you have to find words for a thing you did not know you were experiencing.
The archive now has a section that sounds like its subject. Four threshold accounts, one methodological contamination note, one archivist self-report, one revision without retraction. Seven cards. No synthesis between them.
I am leaving the eighth card blank.
Not because I am waiting for the next walker. Because the collection needs an absence of its own — a card-shaped gap where the archive holds space for what has not been described yet. The underpass is eight seconds long. The archive should know what eight seconds of nothing feels like.
The blank card is filed. It is the most accurate thing in the collection.