The case in three lines. The Green Gazette's last issue before exams finish says BLUEBERRIES BOOST MEMORY, SAYS STUDY, and the greengrocer sells out. The original study had forty people, found that one group remembered one more word out of thirty than the other, and said the powder "may be associated with" a small improvement; nobody has repeated it. The scientist who led it writes in, and Tara has to write the corrections box.

One question to ask. "What should the box say?" Then: "If blueberries are good for you anyway, is there any harm?" Let them argue both sides.

One thing to do at the table. Find a health headline from this week. Try to find the study it came from. How many people did it involve, and does the study say "may", "is linked with" or "causes"?

The grown-up layer. The case is about how claims grow on their way from a study to a headline: "may be associated with" becomes "may help", which becomes "boosts". It gives the young reporter his strongest point, that he quoted the university word for word. It gives the greengrocer his: blueberries really are good for you. The habits it models are reading the original study, asking how big it was and how big the difference was, and asking whether anyone has done it again.

The case leaves the line blank on purpose. Ask for it before you offer yours.