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

The July issue of the Green Gazette was the last one before the summer, and it came out in the middle of exams, when half the Green was revising and the other half was being asked to be quiet. Tara had wanted something cheerful in the middle pages. Femi in Year 11 had given her exactly that.

BLUEBERRIES BOOST MEMORY, SAYS STUDY

Femi had found it on a university's website, in a press release with a photograph of a bowl of blueberries and a smiling woman in a white coat. New research suggests blueberries may help memory, it said, and then it had a quote from the scientist: These are exciting early results. There was a picture of a brain, made of blueberries, which the Gazette had also used.

Femi had written it up well. He had explained that blueberries had things in them called flavonoids, and he had put in a box at the bottom called EXAM TIP. It said: Revising? Try a handful of blueberries every morning!

People loved it. Mr Patel, who ran the greengrocer's at the corner of the Green, put the page in his window. By Tuesday he had sold out of blueberries. By Thursday he had ordered three times as many, and sold out of those too.

On Friday there was a letter in the Gazette box. It was typed, on paper with the university's crest at the top.

Dear Editor, it said. Thank you for featuring our research. I am the scientist who led the study, and I am writing because I am a little worried. Several parents have now emailed me to ask how many blueberries their children should eat before an exam. I would like to make sure your readers know what our study did and did not show. With best wishes, Dr S. Akhtar.

Tara pinned it to the board above the back table at the Whetstone café. She read it twice.

She had been pleased with the middle pages. Now she was uneasy. She did not know why yet.

"I was going to say a university did the study," Tara said.

She had brought the letter to the back table. Danya was there, eating, as it happened, a bowl of blueberries, which he had bought from Mr Patel's second delivery and was eating very slowly to make them last. Across from him sat Source, who read the footnotes first and the article second, and who was a little stubborn about dates.

"I was going to say it like this," said Tara. "Femi didn't make it up. It's a real university. He quoted their own press release, word for word. The scientist is even the one who wrote to us. I can write back to Dr Akhtar and say we only printed what her university said."

She had the sentence half written. Thank you for your letter. Our feature was based on your university's own announcement.

It felt good. It felt safe, the way it always feels safe to stand behind someone with a title.

Source read the feature, and then the letter, and then the feature again.

"Who would know this best?" said Source.

"The university," said Tara. "It's their study."

"Who said it first, and when?" said Source. "A famous name is not a source. A press release is the university's announcement about a study. It isn't the study. Find the book, not the poster."

"The press release is the university," said Tara.

"The press release is the university's press office," said Source. "Somebody whose job is to get the study into newspapers. The study is what the scientists wrote. They're not always the same thing."

"Says who they're different?" said Danya, through blueberries.

"Dr Akhtar," said Source, pointing at the letter. "That's what she's saying. Politely."

Tara looked at the letter. What our study did and did not show.

"How do we know what the study actually says?" Danya asked.

Nobody knew. That was the problem.

Deed found the study that evening.

Deed read long documents for fun and kept them on her phone, and she found the study in a journal, behind a link at the bottom of the press release that nobody had clicked. It was eleven pages long, with tables, and a summary at the top.

"Go to the original document," said Deed. "Not the press release. Not the feature. This. Read it where it was first written down."

She read the last sentence of the summary out loud, slowly.

"Daily blueberry supplementation may be associated with a small improvement in word recall in this sample. Further research is needed."

"May be associated with a small improvement," said Tara.

"In this sample," said Deed. "Further research is needed. That's what the scientists wrote."

Then Deed found the press release and read its first line. "New research suggests blueberries may help memory."

Then she held up the Gazette. "BLUEBERRIES BOOST MEMORY."

She laid all three side by side on the table.

"A summary of a summary drifts," said Deed. "Look. The scientists said may be associated with a small improvement. The press office said may help. We said boost. It grew twice on the way to our middle pages. Not because anyone lied. Because each time somebody wrote it shorter, they left out the careful part."

Tara looked at the three sentences.

She felt something shift, the way it had in June with the poll, and in May with the crossings. It was the feeling of a true sentence turning out to be standing on the wrong ground.

Tomas carried a twenty-sided dice for explaining chance, and he took it out as he sat down, as he always did. He read the study's tables while everyone waited.

"How surprising is this if nothing's going on?" said Tomas. "That's my question. So let's see how big it was."

He read from the table.

"Forty people," said Tomas. "Older adults, not students, by the way. Twenty got blueberry powder in a drink every day. Twenty got a powder that looked and tasted the same but had no blueberry in it. Nobody knew which they had. That's good. For twelve weeks. Then they all did a memory test. They were read a list of thirty words, and had to say back as many as they could."

"And?" said Tara.

"The blueberry group remembered eleven words, on average," said Tomas. "The other group remembered ten."

"One word," said Danya.

"One word out of thirty," said Tomas. "About three points in a hundred. Real things are noisy. With only twenty people in each group, some people are just having a good day, or slept well, or are better at lists. A difference of one word could easily be chance. The scientists say so themselves, in the small print on page eight."

"So it's nothing," said Tara.

"It's something," said Tomas. "Just not everything. It's a small difference in a small study of forty older people. That's a reason to do a bigger study. It isn't a reason for every Year 11 on the Green to eat a handful every morning."

He rolled the dice. It came up three.

"Too neat is a kind of surprising too," said Tomas. "So is too exciting."

Ravi was a lab technician's son and had been repeating experiments since he was twelve. He listened to all of it without saying anything, and then asked one question.

"Has anyone done it again?"

Deed looked. She searched for other studies with the same test, the same powder, the same kind of people.

"No," she said. "This is the first."

"Once is a story," said Ravi. "Twice is a start. Do it again."

"But it was done properly," said Tara. "Nobody knew which powder they had."

"It might be exactly right," said Ravi cheerfully. "I hope it is. I like blueberries. But lots of small studies look exciting the first time and then come out smaller the second time, or not at all. Not because anyone cheated. Because the first time, by chance, the groups were a bit different, or it was a lucky week. You only find out by doing it again." He shrugged. "Repeat it before you believe it, and before you blame anyone."

"So nobody's wrong," said Danya. "Femi's not wrong. The scientist's not wrong."

"Nobody's wrong," said Ravi. "Somebody's in too much of a hurry."

Attest came last, with the Gazette's red pen. Attest subbed every story before it went to print and drew a red circle round any claim without a name next to it. The July feature had come back with one circle, round the word BOOST. Next to it Attest had written: their word or ours?

"Who says so? I wrote that," said Attest. "Then Femi showed me the press release, and it had a university crest on it, and I let it go."

"I saw it," said Tara. "I thought you meant the headline was too big."

"I meant: did they say boost, or did we," said Attest. "But I didn't check. Nobody did, until Source asked who said it first."

"Let me listen first," Tara said, before anyone could tell her what to write.

She took her pad and wrote three headings: what we printed, what the study said, what nobody knows yet.

Under the first she wrote blueberries boost memory; exam tip, a handful every morning. Under the second, forty older people; one word in thirty; "may be associated with a small improvement". The third column: whether it's real, then whether it works for students, then whether anyone can repeat it.

"Hm. So what does a clarification like this have to carry?" Tara asked.

"Every claim names its source," said Attest. "The study, by name, not just the press release. What it said, in its own words. How many people, how old, for how long. How small the difference was. That it hasn't been repeated. And that the exam tip was ours, not theirs."

"Do we say blueberries are bad?"

"No," said Attest. "We never said that, and they're not. Put a name next to it. Say what we got wrong, which is one word in the headline and one box at the bottom."

"Hm. That's a lot for one box," said Tara.

"Five sentences," said Attest. "One of them can be about fruit."

Tara looked at Mr Patel's window, which she could see from the back table, across the Green. The Gazette page was still in it, a little faded now, with the brain made of blueberries.

"I think we missed someone," Tara said slowly. "The people revising. We told them a handful every morning. Some of them will have done that instead of something else, because it's easier than revising. We didn't just make the headline too big. We gave them a tip that wasn't in the study."

Nobody answered at once.

"Who you say it to is your call," said Attest. "That's your bit."

Mrs Hallam, who advised the Gazette and signed off every issue, came in at four. Femi came with her. Mr Patel came at quarter past, in his apron. Mrs Doyle came in with her son Liam, who was in Year 11 and very embarrassed to be there. Dr Akhtar came last, on a video call on Mrs Hallam's laptop, propped against the sugar bowl.

Femi spoke first.

"I quoted the university," he said. "Word for word. May help memory. That's their sentence, on their website, with their crest. I didn't make anything up. I thought boost meant the same as help. If I'd left the study out, I'd have been hiding real science from people."

It was the strongest version of his argument, and every part of it was true.

Mr Patel spoke next.

"Can I just say," he said, "blueberries are good for you. Whatever the study says. Fruit, vitamins, fibre. I've sold more fruit to teenagers this month than in the whole of last year. Where's the harm in that? If the paper says sorry, they'll go back to crisps."

That was true too.

Then Mrs Doyle spoke, with her hand on Liam's shoulder.

"Liam gets very anxious before exams," she said. "Really anxious. He couldn't sleep. Since he read your page, he eats a bowl of blueberries every morning, and he says it makes him feel calmer, like he's doing something. I don't care whether it's the flavonoids. He's calmer."

Liam went red and said nothing. But he did not disagree.

Then Dr Akhtar spoke, from the laptop.

"I wrote may on purpose," she said. "And small. And further research is needed. Those are not the boring words. Those are the most important words in the whole paper. I'm glad people are eating fruit. But I've had six emails from parents asking how many blueberries to give a child the morning of an exam, and the honest answer is that we have no idea. We studied forty people over sixty-five. We didn't study a single teenager."

Nobody said anything for a while.

"Who checked that?" said Danya, very quietly.

Mrs Hallam took the corrections box out of her bag. It was the small grey panel at the bottom of page two, with a border round it and room for a few lines. She had used it ten times that year. She set it on the table in front of Tara, with the July feature beside it.

On the table in front of Tara was the feature, with its brain made of blueberries and its EXAM TIP box. There were Deed's three sentences side by side: may be associated with a small improvement, may help, boost. There was Tomas's note, one word in thirty, forty people, all over sixty-five. There was Ravi's question, has anyone done it again?, and Attest's red circle, their word or ours? There was Mr Patel in his apron, and Liam, calmer, and a scientist on a laptop leaning against the sugar bowl.

"You're the editor," said Mrs Hallam. "What does it say?"

She handed Tara the pen.

Tara looked at the empty box.

 

 

(Write the clarification Tara prints under the July feature.)

 

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

Back on the Green

That afternoon Lan, who was seven, told everyone at the park that eating carrots lets you see in the dark.

Danya asked her who said so.

Lan said her friend's big sister. Her friend's big sister had said it was a science fact.

They looked it up together on the library computer. It said carrots were good for your eyes, but they did not let you see in the dark, and the story was very old.

Lan said she would still eat the carrots. She said she would stop telling people about the dark.

Afterword. The idea in this case is how a claim grows on its way from a study to a headline. Scientists write carefully: "may be associated with a small improvement in this sample". A press release, written to get attention, shortens it to "may help". A headline, shorter still, says "boosts". Each step can be made by an honest person in a hurry, and each one drops a careful word. Before you trust a health headline, find the original study and read what it actually says; check how many people it involved and who they were; see how big the difference was; and ask whether anyone has done it again, because many small studies come out smaller, or not at all, the second time. You can practise reading a claim back to its source in NewsForge, and judging how surprising a result is in InferForge.