The hammer rang once. A spark jumped off the anvil and did not go out. It waited. They followed.

That summer Tara had an internship at the university, which sounded grander than it was. It meant six weeks in Greenhouse Four, a long glass shed behind the biology building that smelled of wet compost and warm tomato leaves, watering trays, writing labels and fetching things for people who were busier than her.

There were three interns. Tara, from the high school. Nadia, who was a year older and wanted to be a doctor. And Sana, who was quiet and checked everything twice, including the labels Tara had already checked.

The person they worked for was Marisol. She was in the fourth year of her PhD, and her whole thesis rested on a fertiliser: a cheap mix of seaweed and crushed shells that she thought might help bean seedlings grow in poor soil. If it worked, farmers who could not afford ordinary fertiliser could make it themselves. She had explained this to the interns on their first morning, with a mug of cold coffee in one hand and a tray of seedlings in the other, and her eyes had gone bright in a way that made Tara like her immediately.

Marisol had planted twenty-four seedlings in May: twelve in poor soil with her fertiliser, twelve in the same poor soil without it. Every week she measured them. Every evening she typed the heights into a spreadsheet on the lab laptop.

Her supervisor was Professor Adeyinka, a tall, kind, overloaded woman who ran six projects, a department and a grant that paid for all three interns' lunches. She came down to the greenhouse on Mondays, asked how everyone was, and looked at the numbers.

On the last Monday in July, she looked at the numbers and stopped.

"Marisol," she said. "Come and see this."

Marisol had already seen it. She had been up half the night seeing it. The final measurements were in, and every single treated seedling was taller than every single untreated one. Not mostly. Every one. The treated plants averaged eighteen centimetres. The untreated ones averaged twelve.

"That's a beautiful result," said Professor Adeyinka softly. "That's a really beautiful result."

There was a conference in the city on Friday, the biggest one of the year in their field. Abstracts closed at midnight on Thursday. Professor Adeyinka put her hand on Marisol's shoulder and said, "Write it up. We're sending it."

Marisol went pink. Nadia clapped. Tara clapped too.

Sana didn't clap. She was looking at the laptop.

That afternoon Nadia found Tara behind the potting bench.

"Can I show you something?" said Nadia, and opened the spreadsheet on her phone. Marisol had shared it with the interns so they could check their own measurements against it. "Look at the heights. Look at the last numbers."

Tara looked. 18.0. 17.5. 18.5. 18.0. 12.0. 11.5. 12.5. Down and down the column. Every one ended in .0 or .5.

"So?" said Tara.

"So I measured four of those plants myself last week," said Nadia. "With the steel ruler. I got 17.3 and 16.8 and things like that. And on her sheet they're 18.0 and 17.5." She lowered her voice. "And there were twenty-four plants. I've counted the rows on here three times. There are sixteen."

Tara felt something cold go down her back.

At the next bench, a postgraduate student from another lab, a boy called Harvey who had been listening with great interest, leaned over.

"She's faked it," he said. "Obviously. Everyone knows she needs this for her thesis. You should report it. There's a research-integrity email. I'll send it if you want."

Nadia went white. "I don't want to get her in trouble. She's nice."

"Being nice doesn't make the data real," said Harvey, quite pleased with himself.

That evening, Ravi came to collect his mother, who was a technician in the next greenhouse. He was in Tara's year at school and had been repeating experiments since he was twelve, when his mother had let him help with her seed counts on Saturdays. He listened to the whole thing sitting on an upturned bucket.

"Don't send anything yet," said Ravi. "Repeat it before you believe it, and before you blame anyone."

"But the numbers are obviously wrong," said Nadia.

"Maybe," said Ravi. "Once is a story. Twice is a start. If you accuse someone and you're wrong, you can't take it back. Her name goes in an email that never goes away." He shrugged. "And if you're right, it'll still be true on Wednesday. Measure it again. Do it again. Then you'll know something."

Harvey rolled his eyes. But he did not send the email.

On Tuesday morning, before Marisol came in, Sana went through the trays.

She did it the way she did everything, which was slowly and completely. She did not look at the spreadsheet first. She looked at the plants. Then she looked at the planting record from May, a sheet of paper taped to the end of the bench, in Marisol's handwriting: twenty-four seeds, twenty-four positions, twelve with the green label, twelve with the white.

"Twenty-four planted," said Sana. "Sixteen on the sheet. Who got left out?"

Tara and Danya, who had come in with Tara because the summer holidays were long and the greenhouse had a resident cat, helped her match labels to rows.

Eight seedlings were missing from the spreadsheet. Two had died. The other six were still there, in their pots, alive and growing. Each one had a small pencilled note on the planting record: damaged leaf, exclude. Bent stem, exclude. Pot cracked, exclude.

"Which ones did we drop, and why?" said Sana, mostly to herself. She wrote them down. Then she measured them.

Of the four treated seedlings that had been dropped, all four were short. Of the four untreated ones that had been dropped, the two still alive were tall.

Danya looked at the list. "So she took out the short good ones and the tall bad ones," he said.

"She took out the ones that looked wrong," said Sana. "That's not the same as cheating. If you believe the fertiliser works, a short treated plant looks like a mistake. A damaged leaf, a bent stem. And it probably did have a damaged leaf. Lots of them have damaged leaves." She looked at the healthy-looking seedling in front of her, which had a perfectly ordinary nibbled leaf, and which was on the list as excluded. "The ones you threw away are data too. Show me the whole sheet, not the tidy one."

"Did she know she was doing it?" said Tara.

Sana thought about it honestly. "I don't think so," she said. "I think she was very tired and very hopeful."

Tara remembered a night two weeks before, when she had stayed late to finish the labels and found Marisol still in the greenhouse at ten o'clock. The lights had gone off on their timer, and Marisol was measuring by the light of her phone, which she held in her teeth. She had a ruler in one hand and a pencil in the other, and forty pots still to do, because she also measured a second trial for another student who was ill. She had looked up, seen Tara, taken the phone out of her mouth and laughed.

"Don't ever do a PhD," she had said. "Unless you really love beans."

Then she had told Tara about her grandmother's farm, where the soil was so thin that the beans came up yellow, and how her grandmother could never afford the fertiliser the shop sold. That was why the seaweed. That was why the crushed shells. Tara had walked home thinking it was the best reason for an experiment she had ever heard.

She thought about it now, looking at the word exclude in pencil.

On Wednesday at lunch Tara showed the numbers to Tomas.

Tomas was a maths student who carried a twenty-sided dice in his pocket for explaining chance to people, and he had been Tara's maths tutor for a term. He loved a strange number the way some people love a strange insect.

He read the column of heights and started to grin, and then stopped grinning.

"How surprising is this?" he said. "Two ways. First the obvious one. How surprising is this if nothing's going on? If the fertiliser did nothing, how likely is it that every treated plant ends up taller than every untreated one? Very unlikely. That's why everyone's excited. That's what the professor saw."

"So it works," said Danya.

"Maybe. But there's a second kind of surprising," said Tomas. "Real things are noisy. Look at last year's trial, the one Marisol ran on the same beans." Tara had found it in the lab folder. "Plants grown the same way varied by four centimetres either way. Some tall, some short, all over the place. That's what beans look like. Now look at these." He tapped the screen. "They sit in two tidy little rows, like soldiers. Too neat is a kind of surprising too."

He took the dice out and rolled it on the table, as if to remind himself.

"And the endings," he said. "If you measure with a millimetre ruler, the last digit could be anything from zero to nine. Only two of those ten are a whole or a half centimetre. So out of sixteen plants, you'd expect about three to land on a .0 or a .5, just by chance. Sixteen times two tenths." He wrote it down: about 3.2. "She's got sixteen out of sixteen."

"So she's lying," said Nadia unhappily.

"Or she's rounding," said Tomas. "People do. Everyone does. If you're measuring forty plants at eleven at night with your eyes half shut, you look at the ruler and you say 'eighteen and a half'. You don't say 'eighteen point three'. Humans love round numbers." He put the dice away. "What would ordinary mess look like? That's the question. This hasn't got any. Something tidied it. Maybe a person, maybe a tired person, maybe just her eyes."

On Wednesday evening, after Marisol had gone home to write her abstract, Ravi came back with a millimetre steel rule, a roll of masking tape and his mother's permission.

"Measure it again, without knowing which is which," he said.

Sana covered every label with masking tape and wrote a random number on each piece. She kept the key in a sealed envelope in her pocket. Then Ravi and Tara measured every surviving seedling, all twenty-two of them, including the six that had been excluded, to the millimetre. Nobody measuring knew which plant had had the fertiliser. Danya held the torch, because the greenhouse lights were on a timer, and the cat sat on the clipboard.

It took two hours. Then Sana opened the envelope.

They worked it out on the back of a seed packet. On Marisol's sheet, the gap between the treated and untreated plants had been six centimetres: eighteen against twelve.

Measured blind, with every plant counted, the treated seedlings averaged twenty centimetres and the untreated ones eighteen. They had all grown since the last measurement; the gap had not. It was two centimetres. There was overlap. Some untreated plants were taller than some treated ones. The numbers ended in every digit from zero to nine.

Nobody said anything for a moment.

"It still holds," said Ravi at last. "Does it still hold the second time? Yes. The fertiliser seems to help."

"Just much less," said Sana.

"A third as much," said Tomas, who had come to watch, holding the dice. "Which might still be worth a lot to a farmer. Two centimetres on a bean seedling isn't nothing."

Nadia sat down on the bucket and put her face in her hands. "I nearly let Harvey send that email," she said.

"I know," said Ravi. "That's why you do it again."

On Thursday afternoon, ten hours before the abstracts closed, Professor Adeyinka came down to Greenhouse Four with Marisol.

Sana had shown Marisol the blind measurements that morning. She had done it kindly, in the potting shed, with the door shut. Marisol had read them twice and then gone very quiet, and then she had said, "The short ones did look damaged. They really did. I thought I was being careful." And she had cried, briefly, into a seed tray.

Professor Adeyinka was not a villain either. She had trusted Marisol for four years, and she was right to; Marisol was careful and honest and very good. She had a grant report due in September, and the grant paid for three salaries and three interns' lunches, and the people who gave grants liked results. A beautiful result at a big conference would help everyone in that greenhouse. And Marisol had been working on this for four years. If the conference was missed, the next one was in a year.

"So," said Professor Adeyinka. She looked at both sheets on the bench: Marisol's tidy one, with its six centimetres and its soldier rows, and the blind one, with its two centimetres and its mess. "We can send the original, which is what Marisol measured, with a note. We can send the new one, which is smaller and true and might not get accepted. Or we can send nothing and do the whole trial again properly in the autumn, and miss Friday."

She looked at Marisol, who looked at the floor. At Sana, with her envelope. At Ravi, with his steel rule. At Tomas, turning the dice over in his fingers. At Nadia, who had noticed first. At Danya, who was holding the cat.

Then she looked at the watering cans.

"Tara," she said. "You've watered these trays every morning since June. You know these plants better than anyone in this room, including me. And you haven't said a word all week." She folded her arms, not unkindly. "What should we send?"

The greenhouse went quiet. A sprinkler ticked somewhere down the row. The cat jumped down off the bench.

Tara thought of Marisol's bright eyes on the first morning, talking about farmers. The pencilled word exclude. Sixteen numbers ending in .0 and .5. Harvey's email, unsent. The masking tape and the sealed envelope. Two centimetres, still there.

Everyone in the room was partly right.

Tara opened her mouth.

 

 

(Write the sentence Tara says.)

 

The spark dimmed to a warm dot and settled back on the anvil. The Green was just the Green again.

Back on the Green

In September Danya grew cress on the windowsill for a school project: one pot in the sun, one in the cupboard. The sunny one was obviously, wonderfully taller.

He measured it with Tara's ruler. "Four centimetres," he said. Then he looked at the number for a long time.

"Actually it's three point seven," he said, and rubbed it out, and wrote that instead.

Afterword. The idea in this case is that a result can be too clean. Real measurements are noisy: living things vary, rulers slip, and people favour round numbers. A result with no mess in it, a sample with the awkward cases quietly removed, or a column of numbers that all end the same way are reasons to check before you believe, and before you blame anyone. Scientists guard against this by deciding in advance what they will exclude, measuring blind, and repeating the work. Usually the answer is not cheating. Usually it is hope. You can play with samples, blind measurement and repeating a test in InquiryForge, and with how surprising a result really is in InferForge.