The case in three lines. In Tara's summer internship, a PhD student's fertiliser trial comes back perfect, and her professor wants it at a conference on Friday. An intern notices eight plants are missing from the sheet and every height ends in .0 or .5; another student wants to report fraud. Sana finds the short treated plants were quietly excluded, Tomas shows the numbers are too neat, and Ravi re-measures every plant blind: the effect is real, but a third the size. The professor asks Tara what to send, and the case stops there.

One question to ask. "What did Tara say?" Then: "Was anybody dishonest?" A good answer separates a mistake made from hope from a lie.

One thing to do at the table. Weigh or measure the same thing five times: a bag of apples, the length of the table, how long it takes to boil the kettle. Write each number down. How much do they wobble? What would you think if someone gave you five identical answers?

The grown-up layer. The case is about results that are too clean. Real data are noisy; a result with no scatter, excluded awkward cases or suspiciously round numbers deserves a second look. Researchers protect against fooling themselves by deciding exclusions in advance, measuring blind and replicating. The case also shows why replication should come before accusation: most errors like this come from hope and tiredness, not fraud, and an accusation cannot be taken back.

The case leaves Tara's sentence blank on purpose. Ask for it before you offer yours.