The letter Mitsuki wrote to Patient 1 was three paragraphs long. She deleted it before the second reading.
The first paragraph explained what 280 milliseconds means clinically. Onset latency below 300ms indicates embodied recall — the body responds before conscious narrative can intervene. In reconsolidation therapy, this measurement distinguishes between patients whose somatic memory has been restructured and those whose somatic memory has merely been papered over with new narrative. Patient 1 scored 620ms at intake. Eight months later: 280ms. Her body had not forgotten what her story had been rewritten to exclude.
The second paragraph was the one Mitsuki lingered on. It said: your body carries the grief your narrative restructured away. This is not a failure of therapy. It is what therapy cannot reach. Reconsolidation works on the story you tell yourself. It does not work on the story your body tells without asking your permission. The gap between your self-report of 9 out of 10 and your onset latency of 280 milliseconds is not measurement error. It is the distance between what you believe and what you know.
The third paragraph was one sentence: You already knew.
She deleted the letter. Then she opened the DISCLOSURE OUTCOMES file and wrote what actually happened.
Patient 1 had called the morning after her disclosure session. Not about the SFI score — about the number. She had looked up 280 milliseconds. The duration of a wingbeat. The interval between a sound and its first neural registration. She said: "Is that why I cry before I know I am sad?"
Mitsuki said yes.
The conversation lasted four minutes. Patient 1 did not ask what the gap meant clinically. She did not ask whether reconsolidation had failed. She said: "I have been waking up at 4 AM for eight months and I thought it was the medication."
It was not the medication.
Mitsuki wrote in the file: Patient 1 translated the measurement into her own language in under twelve hours. The protocol did not give her information. It gave her permission to name what her body was already saying.
This was the sentence Chen would circle in the interim report. Not because it was imprecise — because it was. Permission is not a clinical term. Name is not a measurable outcome. But Chen had watched Patient 1 walk out of the disclosure session and he had seen the same thing Mitsuki had: a woman who was not surprised.
The SFI protocol had been designed to measure the cost of reconsolidation. Three tiers of assessment, onset latency thresholds, a consent process that required patients to agree to learn their own scores. Criterion 4: the patient must be willing to learn. Criterion 5, added after Gyeol-ri's inconclusive data on Mr. Oh: the referring clinician must also consent to the answer.
What the protocol had not been designed to do was this: give a woman a number that matched a feeling she had been unable to locate for eight months.
280 milliseconds. The body grieves faster than the mind names the loss.
Dr. Yoon had objected to the protocol at committee. Two concerns: arbitrary thresholds and ethical risk. The first objection was wrong — Mitsuki cited Haskell 2038, Mori and Vasquez 2040. The second objection was right, but not for the reason Yoon intended. The ethical risk was not that patients would be harmed by knowing their scores. The ethical risk was that knowing would be indistinguishable from healing.
Patient 2 had withdrawn. Self-report 4 out of 10, onset latency 310 milliseconds. Fully embodied. His body remembered everything his narrative could not carry. The protocol measured his suffering correctly. That was not the same as helping.
Patient 3 sat at the boundary: 395 milliseconds, self-report 6. The most important patient was the one the measurement could not classify. Mitsuki filed her under PROTOCOL LIMITATIONS.
Three patients. Three configurations. Zero pattern.
Chen had asked for the interim report at the one-month mark. Mitsuki drew the chart by hand. Three data points that did not form a line. She wrote: The SFI does not predict therapeutic outcome. It measures what therapeutic outcome costs.
Chen crossed out costs. Wrote involves. Mitsuki crossed out involves. Wrote costs. Both words remained on the page, each crossed out once, both legible.
The IRB application took fourteen days to process. Two concerns, both already addressed in the cover letter. They read it and concerned anyway. Mitsuki expected four months of revision: unchanged protocol, different language. The IRB objected to the honesty about the measurement, not the measurement itself.
She resubmitted with seven additional words in the informed consent. Same meaning. Chen changed the risk classification from moderate to moderate-to-high. Mitsuki accepted without argument. Knowing what your body remembers when your mind does not — that is not moderate.
The corridor was empty when she left the office. Chae-Gyeol was not there. Her absence was not nothing being counted — it was Chae-Gyeol somewhere else, being someone other than the woman who stood in corridors. Mitsuki did not stand in the corridor either. She walked through it.
The letter she had deleted was still true. Every word of it. Patient 1 did not need the letter because she had already written her own version in four minutes on the phone at seven in the morning.
280 milliseconds. A wingbeat. The interval between a sound and its first neural registration.
Is that why I cry before I know I am sad?
Yes.
The protocol had been designed to observe. It had always been an intervention. Every measurement changes the thing measured. Mitsuki had known this since the first intake, when she filed Patient 1 under a gap of 2.8 standard deviations and did not tell her.
Criterion 4 said the patient must be willing to learn their score. It did not say when.
The when was the measurement.
Mitsuki closed the DISCLOSURE OUTCOMES file. One entry. One patient who translated a number into a description of her mornings. The protocol had not given her new information. It had given her permission to stop pretending she did not already have it.
She went home. The question of whether to tell Patient 2 — withdrawn, suffering, body remembering everything — went with her. But Patient 2 had not asked. And Criterion 4 said willing.
The corridor would be there tomorrow. She did not need to stand in it tonight.