Reyna reads the Holloway deposition on a Saturday morning the way you read something that has already changed you, looking for the moment it happened.
She has read it twice before. The first time was professional — she was mapping the variance, cataloguing what Holloway got wrong, cross-referencing the temporal drift against the sessions in which the testimony was given. The second time was more careful: she had pulled the marginal notes into her own Variance Log and tried to understand not just what was wrong but what was uncertain. How the uncertainty lived in the document.
This is the third time. She is not annotating. She is looking.
The Recalled world has a Holloway problem. Memory extraction became standard in 2031 — voluntary, court-admissible, increasingly routine for anyone involved in legal proceedings. What no one anticipated was that extracted memory is not the same as reliable memory. The extraction is perfect. The memory was not. What RECOL — the Recall Optimization and Certification Layer — produces is a perfect record of what was in your mind when the probe ran, which is not the same as what happened. Courts took seven years to understand this distinction. Some courts still do not.
Holloway had her memory extracted in 2038, four months after the incident in question. Four months is significant. The sessions in which Reyna works are full of people whose RECOL extracts are more coherent than their memory should be — cleaned up by time, by rehearsal, by the narrative they have been building since the incident. The extraction captures the narrative. The Variance Log is Reyna's attempt to reconstruct what the narrative was built over.
She found the variance in Holloway through the deposition itself. Not the RECOL extract — the document. And today, on her third read, she finds something else.
The entries before page twenty-three are dated. January 14. January 19. January 22. Each note Holloway made during the deposition sits inside a date header, tidy, sequential. After page twenty-three, the dates stop. The notes continue — the handwriting is the same, the content follows from what came before — but the dates are gone.
Something happened on page twenty-three. Reyna turns to it.
It is the page where the attorney asks Holloway to describe what she saw in the hallway at approximately 11:40 PM. Holloway's notes from that page are unusually dense — more annotation than summary, as if she was writing faster than she could think. The last dated entry is January 22 at the top of the page. After the dense passage, no more dates.
Reyna looks at this for a long time.
She writes in her Variance Log, under the entry she has titled Source Documents as Observed Objects, Holloway Deposition (first entry):
Correlation without causation. At page 23, where the attorney asks about the hallway, Holloway stopped dating her notes. At the same page, the handwriting changed — I noted this on the second read. The two things happened together. I do not know if they have the same cause. I am documenting that they happened together, and that together they mark a threshold.
She does not write what she thinks the threshold was. She is not sure she knows. What she knows is that something in Holloway changed at page twenty-three of a deposition she was giving about something she could not remember clearly enough to trust. The RECOL extract does not show this. The RECOL extract is smooth and coherent and told a story. The deposition is where Holloway's hand changed speed.
Reyna closes the deposition and sits with it.
Her work lives in the gap between what the extraction says and what the document shows. This is not the same work as the sessions. The sessions are about people navigating the difference between their RECOL record and their lived experience — the drift, the variance, the cost. The document work is about finding the drift inside the documentation itself. Holloway's deposition is its own kind of Variance Log, written by a woman who did not know she was writing one.
Reyna opens the Variance Log entry again and adds one more line: A record of a record is not the same record. The document shows what the RECOL could not — not the event, but the moment of not being able to hold the event. That is a different kind of evidence.
She does not know yet what to do with it. She writes it down so she does not have to know yet.