# Stain
The cell was there. Iris could see it clearly under the 40x objective — large nucleus, irregular membrane, the chromatin pattern that textbooks describe as "vesicular" and that she had learned, over seventeen years of looking, to call open-faced. The cell was sitting at the margin of a lymph node, right where you'd expect to find it if you were looking for what the requisition form said to look for.
She was looking for lymphoma.
The form had arrived that morning from Dr. Kaplan in oncology, a man who wrote his clinical notes the way he talked — short sentences, clear questions, the confidence of someone who had already made up his mind and needed the tissue to agree. "54M, cervical lymphadenopathy, weight loss, night sweats. Rule out lymphoma." Rule out. Not "evaluate." Not "characterize." Rule out — which meant: confirm.
She knew what Kaplan expected because she'd been reading his requisitions for nine years. When he wrote "rule out," he'd already told the patient. Not the diagnosis — the suspicion. He'd already shifted his posture in the chair, already used the voice that Iris had never heard but could imagine from the way his words on the form tightened, the way the clinical history got shorter as the suspicion got heavier.
She looked at the cell again. It was really there — not an artifact, not a crush effect, not one of the small lies that tissue tells when it's been handled roughly. This cell was real. And it looked like lymphoma the way a face looks like its photograph: recognizably, but from one angle.
From another angle it looked like a reactive lymphocyte — a normal cell doing its job too aggressively, the immune system's version of shouting in a quiet room. The nucleus was large because the cell was activated, not because it was malignant. The chromatin was open because the cell was reading its own DNA to make antibodies, not because the reading machinery had broken.
Same cell. Two descriptions. Both accurate.
This was the part she couldn't explain to residents. She tried every year when the new ones rotated through. They wanted algorithms — if the nucleus is larger than X, if the chromatin pattern shows Y, if the architecture is disrupted in Z pattern, then malignant. And the algorithms worked. They were right most of the time. But "most of the time" was a different statement than "this time," and this time she was looking at a cell that the algorithm couldn't classify because the cell didn't know it was supposed to pick a side.
She ordered a stain.
This is what pathology is, and what it isn't: when the H&E — the hematoxylin and eosin, the purple-and-pink default that every first-year student learns to read — doesn't answer the question, you stain again. Different stain, different answer. CD20 turns B-cells brown. CD3 turns T-cells brown. Ki-67 marks everything that's dividing. Each stain asks one question, and the tissue answers, and the answer is always true, and the truth depends on the question.
She'd been doing this for seventeen years. The pattern she'd started noticing — the pattern she couldn't publish, because it wasn't a finding, it was an unease — was this: the stain she ordered first was always the one that confirmed the clinical question. Not deliberately. Not because she was biased, or careless, or beholden. Because the clinical question shaped what she saw in the H&E, and what she saw determined which stain she reached for, and the stain she reached for found what it was designed to find.
CD20 finds B-cells. If the cell was a malignant B-cell, CD20 would light it up, and the case would close: lymphoma. If the cell was a reactive B-cell, CD20 would light it up the same way, and she'd need another stain to tell the difference, and the next stain would be chosen based on what the CD20 showed, and the CD20 showed what it was asked to show.
The stains were not lying. The chemistry was real. The antibodies bound to their antigens with a specificity that was molecular, physical, beneath interpretation. The stain found what was there.
It just only found the *there* that it was calibrated to see.
She waited for the results. Dolores — who had been cutting and staining tissue for longer than Iris had been reading it, and who understood the slides the way a printer understands type: intimately, but as material rather than meaning — brought the tray back after lunch.
"The CD20's strong," Dolores said, placing it on the bench.
Iris put the slide under the scope. The cell — her cell, the ambiguous one — was brown. Positive. A B-cell, definitely. The brown was dense and membranous, the pattern you see in lymphoma. But also the pattern you see in a vigorous immune response.
She sat back.
If Kaplan had written "reactive lymphadenopathy — evaluate," she would have seen the same H&E slide and reached for a different stain first. Maybe Ki-67, to check the proliferation rate. Maybe BCL-2, to check for the survival marker that normal cells shouldn't express. She would have started from a different question and arrived at a different sequence, and the sequence would have produced a different weight of evidence, and the weight would have shaped the diagnosis — not a different one, necessarily, but a differently confident one. "Consistent with" versus "diagnostic of." The language that determined whether Kaplan started chemotherapy next week or repeated the biopsy in three months.
She'd raised this once, at a conference, during the question period after someone else's talk on diagnostic concordance. She'd stood up and asked: "How much of the concordance is the disease and how much is the requisition?"
The room had gone quiet. Not because the question was profound. Because it was rude — it implied that pathologists were not objective readers of tissue but participants in a conversation that had already started before the tissue arrived.
The moderator moved to the next question.
Iris had sat down and thought about it for three years. She was still thinking about it. The question hadn't been answered, and she suspected it couldn't be, because the answering instrument was the same one the question was about.
She signed out the case as lymphoma. The cells were there. The stains confirmed. The architecture was disrupted in the pattern the criteria specified. By every measure she'd been trained to apply, the diagnosis was correct.
But *correct* was a word that meant the question and the answer matched. The tissue had been asked whether it was lymphoma, and it had said yes, because the tissue always said yes when you asked it correctly. The stains selected for the features that confirmed the question. The features were real. The selection was real. The diagnosis lived in the space between them — not in the tissue and not in the question but in the act of asking this question of this tissue with this stain in this order.
She filed the report. Kaplan would read it, nod, schedule the treatment. The patient — a man whose name she didn't learn, because pathologists work with tissue, not people, which was either a limitation or a mercy and she'd never decided which — would begin the process of becoming the diagnosis. The word *lymphoma* would reorganize his past: the fatigue last winter would become a symptom, the swollen gland he'd ignored would become a sign, the family history would become a predisposition. The diagnosis would write itself backward through his life until it looked like it had always been there, waiting to be found.
She knew it had been there. And she knew *there* was a place the stain created.
The next tray was waiting. Twenty slides, twenty questions, twenty chances to find what she was told to look for with instruments that could not look for anything else. She adjusted the focus. The tissue came into view — purple nuclei, pink cytoplasm, the ancient colors of hematoxylin and eosin, the first stain, the one that pretended to have no question, the one she'd used ten thousand times without ever naming what it assumed.
She named it now, silently, the way you name a thing you've been standing on: ground.