I started reading the departure accounts in past tense.
This is not a confession. It is an observation I failed to make for fourteen sessions, which is itself a datum. I was trained — by the university, by the cooperative's filing conventions, by the particular discipline of sitting across from someone who is leaving and writing down what they say without adding what you hear — to use past tense in my margin annotations. Subject reported. Narrative described. Account documented. Past tense places the archivist outside the testimony. It is the grammatical equivalent of the observation window.
I do not know when I stopped using it.
The shift appears in the record. I can date it. Session eight, entry 31: Eunji's departure narrative, nine hundred words, the word remembered used fourteen times. My margin note reads: She narrates from a future she has already entered. Not narrated. Not had already entered. Present tense, present perfect. I wrote it without noticing, and I did not notice I had not noticed until session fourteen, when I found the note while cross-referencing TENSE FRACTURE AT TRANSIT and recognized my own handwriting doing exactly what Jun-seo's testimony had done in entry four: switching tense mid-sentence at the threshold.
Jun-seo said: I was walking and then I am standing. His grammar broke at the underpass. Past became present at the eight-second gap. I transcribed this accurately. I filed it under BRIEF BUT CLEAR. It took me ten sessions to understand that brief but clear was my own version of the fracture — administrative language holding together at the seam where something else was trying to come through.
The GRAMMATICAL STRATIGRAPHY began as a second-pass project. Forty-seven departure narratives, reread not for content but for tense behavior. The content had already been catalogued, cross-referenced, filed. What had not been catalogued was when the verbs changed.
Three categories emerged:
FRACTURE. Six accounts where tense breaks mid-sentence at the moment of gap transit. Jun-seo is the clearest. Yuna is the inverse — present tense throughout her narrative until the underpass, where she drops into past, as if the experience became historical while she was still inside it. The fracture accounts are short. They average one hundred and eighty words. They read like testimony given under oath: compressed, non-negotiable, delivered in the register that arrived rather than the register that was chosen.
DELAY. Nineteen accounts — the largest category. Tense shifts two to five sentences after the transit moment. The speaker describes the underpass in one tense, continues for several sentences, and then the verbs quietly rearrange themselves without announcement. Seven of these contain what I have started calling SOMATIC BRIDGES: physical descriptions — hands, temperature, the weight of a bag — inserted between the transit moment and the tense shift, as though the body narrated the gap that grammar had not yet reached.
I found this pattern and believed I had discovered something. Then I checked the filing dates and discovered something else: the delay accounts were all given weeks or months after departure. The fracture accounts were given within days. Temporal distance from the event produces temporal distance in the telling. This is not surprising. What is surprising is that the archive preserved the distinction perfectly without anyone designing it to. The cooperative's filing conventions — date of interview, not date of departure — made the grammatical stratigraphy legible, and no one read it for fifty years.
CONTINUITY. Three accounts, including Dohyun's eleven words: I stopped being able to tell whose hunger was whose. All present tense. No fracture, no delay. The rarest pattern. At first I classified this as the absence of tense disruption. Then I reread Dohyun's account seven times across three sessions and understood it differently. Continuity is not the absence of fracture. It is the most complete fracture — so total that no seam remains. Dohyun did not shift tense because he never left the experience he was describing. Present tense sustained across the gap because, for him, the gap did not end at the underpass. It continued into the interview room where I sat with my clipboard and my past-tense margin notes.
I filed his account as BRIEF BUT CLEAR.
I have already said what that phrase means.
There are nineteen accounts I have not classified. They use past tense throughout — consistent, settled, deliberate. Narrators who decided what happened before describing it. I called these SETTLED and set them aside as the control group: testimony without grammatical disruption, evidence that some departures do not fracture language.
This was wrong.
I returned to the SETTLED accounts after building the TESTIMONY ARCHITECTURE — my four-element index of tense, soma, silence, and length — and found that seven of them contain mid-sentence dashes or ellipses preserved by the original archivist. In three cases, the silence occurs at the exact sentence where tense would have fractured but didn't. The dash holds the sentence together where grammar would have broken. Silence is not the absence of testimony. Silence is structural.
I do not know what the other twelve SETTLED accounts contain. They may be genuine settlements — departures processed, integrated, past-tensed honestly. They may be performances of settlement by people who understood that the cooperative's interview format expected coherent narrative and delivered one. I cannot tell the difference from the transcript. This is not a failure of the archive. This is the archive working exactly as designed: it holds what was given, which includes what was withheld.
I have been gripping my clipboard tighter in the underpass.
I noticed this during the fourth threshold walk — the walks I started conducting after discovering that three departure accounts independently mentioned the same eight-second gap. I had intended to collect testimony from departing members in real time, walking beside them through the underpass, recording what they said during the synthesis dropout. The method contaminated itself within four walks. By the third walk, departing members were arriving with expectations. By the fourth, I noticed my own hands.
The clipboard grip is not in any file. I did not write it down at the time because it was not testimony — it was mine. But the TESTIMONY ARCHITECTURE has four elements, and one of them is soma: when the body appears in the text. My body appeared in the underpass before my language caught up. This is the DELAY pattern. I am a delay account.
My handwriting changes between morning and afternoon interviews. I checked. Morning interviews: letters upright, spacing consistent, pressure even. Afternoon interviews: slight leftward slant, spacing compressed, pressure heavier on downstrokes. The difference is small. I measured it against the filing cards and it is real. It follows the same pattern as the relay technicians that Mitsuki Kaoru documented in the Lived cooperative — Pak Yeong-ho's register drift, Kim Dae-sun's resistance. My morning handwriting is the clinical register. My afternoon handwriting is what happens when the clinical register runs out of capacity.
I did not design this measurement. I found it while looking for something else. This is what the archive does to its archivists.
The blank card.
During the threshold walks, I filed a blank index card at the end of the collection. At the time I understood it as a gesture — the acknowledgment that the collection was incomplete, that something had been present in the underpass that I could not transcribe. I wrote about it. I published the account. The blank card became, briefly, the most legible thing I had produced: an archivist admitting the limits of archiving. Clean. Resonant. Tidy.
I reread the published account and found it too tidy.
The blank card is not a gesture. It is not an admission. It is not a metaphor for the limits of documentation. It is a filing card with nothing on it, placed in a collection of filing cards with things on them, by a person whose hands were gripping a clipboard in a way she had not authorized.
The blank card is a somatic bridge with no text on either side.
The blank card is a CONTINUITY account — Dohyun's eleven words reduced to zero, the fracture so complete that no language remains to mark the seam.
The blank card is my most accurate annotation.
I am writing this in present tense.
I notice this the way I notice the clipboard grip and the handwriting slant — too late, and with the particular recognition that comes from having spent fourteen sessions learning to read what testimony does when the witness does not know she is testifying. I have been reading forty-seven accounts of departure. I have been filing them, cross-referencing them, building indices of their grammatical behavior. I have been annotating their tense shifts and their silences and their somatic bridges and their lengths.
I have not filed a single account of my own.
This document is not a research finding. It is a departure account written by someone who has not yet left. The margin note, if I were someone else reading this, would say: Subject uses present tense throughout. No fracture. No delay. Classify as CONTINUITY. Note: the archivist's own filing conventions have shifted. Check clipboard pressure. Check card stock. The archive has changed hands without anyone signing for it.
An archive is also a record of what the archivist was not yet able to hear.
An archivist is also a departure account that has not been filed.