The One-Day Error — Bir Günlük Hata

The herbarium of the Ashcombe Botanical Institute occupied three floors of a Victorian building on the edge of Oxford, and Dr Nil Erdem had come to think of it as a kind of library in which every book was a pressed and dried flower. There were more than a million specimens in its steel cabinets, each glued to a sheet of heavy paper and labelled, usually in fading ink, with a name, a place and a date. It was the dates that interested her. Nil was a phenologist, which meant that she spent her working life asking when things happen: when buds break, when bees emerge, when the first bluebell nods open in a particular wood. Her project, funded for three years and already halfway through its second, was to use a hundred and thirty years of herbarium records to show, with the unanswerable authority of large numbers, how much earlier English spring now arrives. It was patient, unglamorous work, and she loved it with a devotion that her friends in Istanbul found faintly alarming.

The trouble began, as trouble in science so often does, with a line on a graph that refused to behave. Plotting flowering dates for the lesser celandine in Oxfordshire, she noticed a cluster of points from the 1880s and 1890s that sat, with almost insolent regularity, exactly one day later than everything around them. One day is nothing to a flower, but to a statistician it is a fingerprint. When she isolated the anomalous specimens, she found that they had all been collected by the same hand: a certain E. Marsh, of whom the Institute's records said nothing except that he had donated four hundred and twelve sheets between 1884 and 1899. Every one of E. Marsh's dates was, by every comparison she could devise, exactly a day late.

At first she felt nothing but irritation. A systematic error is, in some ways, worse than a random one, because it pretends to be information. Four hundred and twelve sheets amounted to nearly a fifth of her nineteenth-century data for the county; if she simply discarded them, her early baseline would grow thin and wobbly, and her reviewers, who were not known for their mercy, would pounce. The obvious solution was to subtract a day from every Marsh record and move on, with a polite footnote explaining the correction. Her supervisor, a cheerful Scot who had seen more datasets than she had seen breakfasts, advised exactly that. "Your man Marsh probably wrote up his notes the evening after," he said. "Half the Victorian gentlemen in that building did. Shave off the day and don't lose any sleep over it."

She did lose sleep over it, however, because the explanation did not fit. A collector who wrote his labels the next evening would sometimes have remembered correctly, sometimes have been two days out, and occasionally, after a long week, three. Human carelessness is untidy; it has a texture. E. Marsh's error had no texture at all. It was as clean and as constant as a ruled line, and the more she looked at it, the less it resembled a mistake and the more it resembled a decision.

So she went looking for E. Marsh, which, as any historian could have told her, is easier said than done. The Institute's correspondence files contained only a single letter, a curt note of thanks from the Keeper of the Herbarium dated 1899, addressed to "Mr E. Marsh, The Rectory, Long Wittenham". The parish records of Long Wittenham, which she spent a rainy Saturday reading in the county archives, listed no rector, curate or churchwarden of that name. What they did contain, buried in a list of baptisms, was an Eleanor Marsh, born in 1861, the daughter of the rector's housekeeper. Nil sat for a long time in the hushed reading room, with the rain sliding down the tall windows, and felt the floor of her assumptions shift slightly beneath her.

The final piece arrived by accident, the way final pieces tend to. In a box of uncatalogued material from the rectory, donated by a great-great-nephew who had been clearing out a loft in Didcot, the archivist found a slim notebook bound in green cloth. It was Eleanor's field diary, and on its first page, in a small, upright, entirely unapologetic hand, she had written a kind of manifesto. "The Keeper will not accept specimens from a woman, and so I send them under my initial, and he thanks Mr Marsh very kindly," she wrote. "But I will not be lost entirely. To every date I add one day, so that whoever comes after may know my plants from all the others, and know that they were gathered with care." "A careful reader will find me. A careless one deserves a little error."

Nil read the passage three times before she understood that she had been, for some months, precisely the careless reader Eleanor had anticipated, and that she had come within a keystroke of proving her right. The correction she had been on the point of making would have rendered the data tidy and the woman invisible. Somewhere in the eight hundred lines of her analysis code there was already a function called fix_marsh, and she felt, looking at it, something very close to shame. It was also, she could not help noticing, a beautifully designed experiment. Eleanor had introduced a known, constant offset into her own data for fifteen years without a single lapse, which made her, in purely technical terms, the most reliable collector in the county. Once the offset was accounted for, her dates were not merely usable; they were the best in the dataset.

Nil's supervisor, to his credit, read the diary in silence and then said only, "Well. I stand corrected, and so, I suspect, does the Keeper." The paper went through two rounds of review. One reviewer called the Marsh section "charming but tangential", and Nil spent a satisfying afternoon explaining, with tables, why it was neither. In the end the editors allowed her a paragraph and a footnote, and she used them well. The Institute, which is not in the habit of apologising to people who have been dead for ninety years, quietly changed the name on four hundred and twelve catalogue entries from "E. Marsh" to "Eleanor Marsh".

There remained one decision, and it was Nil's alone. The published dataset would be used by other researchers for decades, most of whom would never read a footnote; the sensible thing was to release the corrected dates and let the offset disappear for good. She thought about it for a week. Then she published both columns side by side: the corrected date, for the careless, and the original, for anyone with the patience to wonder why it was different. In the documentation, under a heading that her co-authors begged her to change and she did not, she wrote a single line: "Column F is one day late on purpose. Look closely; someone is waiting to be found."

The following spring, Nil took the train to Long Wittenham with the green notebook's coordinates copied into her phone. She found the meadow by the river where Eleanor had gathered her celandines, now hemmed in by a bypass and a cluster of new houses, but still, improbably, yellow at the edges. It was the fourteenth of March, nearly three weeks earlier than Eleanor had ever recorded the first flower, which was, of course, the whole point of Nil's paper and the sadness folded inside it. She knelt in the wet grass, photographed the flowers, and took out a pencil to write the date in her own notebook. Then, after a moment's thought and with the faint sense of being watched approvingly, she wrote the fifteenth.