Sixty-one Chilean schoolchildren had a brain scan and filled in a food questionnaire. The frequent pastry eaters had a smaller memory centre, and fruit and vegetables showed no link either way

The paper has the kind of title that travels: 'The weight of junk food'. It was published on 11 October 2026 in Applied Psychology: Health and Well-Being, and it reports that Chilean schoolchildren who ate more pastries, fried food and soft drinks had smaller volumes in a handful of brain regions. That is what the data show. What they mean is a harder question, and the authors themselves call the work exploratory more than once.

The children came from the Cogni-Action Project, a study of 1,296 pupils aged 10 to 14 in fifth to eighth grade, run from the Pontificia Universidad Católica de Valparaíso between 2017 and 2019. Brain scans are expensive, so only a small group whose families volunteered had one. After removing children without a usable image, three whose scans were blurred by movement or had other quality problems, and one in whom the scan turned up a tumour nobody had known about, 61 were left, from ten schools. Their average age was 11.6 and 26 of them were girls. All were right-handed and none had a diagnosed psychiatric or neurological condition, because both can change how a brain looks on a scan.

Each child had a structural MRI lasting about four and a half minutes. From it the team measured total grey matter, the hippocampus (which the brain leans on for forming memories), the amygdala (involved in emotion), and a set of areas at the front of the brain that handle planning and self-control.

Diet was measured with a short Chilean questionnaire that asks how often a child eats nine kinds of food: fruit, vegetables, dairy, legumes and fish on the healthy side, and soft drinks, pastries, fried foods and added table sugar on the other. The researchers then split each food into two groups, children who had it up to three times a week and children who had it four times a week or more. By that cut, 74% were frequent eaters of fried food, 69% of soft drinks and 57% of pastries. A quarter ate fruit that often and one in six ate vegetables that often. On the questionnaire's overall score not a single child had a diet rated healthy. Eighty-seven per cent were rated unhealthy and the rest 'in need of changes'.

Then came the comparisons. Taking one food and one brain region at a time, and correcting for the number of tests run, exactly one link held up: children who ate pastries four or more times a week had a smaller hippocampus. Nothing else did, healthy or unhealthy.

The second analysis is the one behind the headline. Instead of testing foods one by one, it asks whether a blend of foods moves together with a blend of brain regions. For the four unhealthy foods it found one such pattern, and a fairly strong one (a correlation of 0.70): more pastries, fried food and soft drinks went with smaller hippocampus, amygdala and three of the frontal areas. Table sugar added almost nothing to the pattern. For the five healthy foods there was no pattern at all.

The authors read that lopsided result as a sign that, at this age, the harm from unhealthy food may outweigh any structural benefit from healthy food. That is one reading. Another is simpler. When no child in the sample eats well and only ten eat vegetables most days, there is very little difference between children on the healthy side for a statistical method to find.

Everything else about the design asks for the same caution. The children were scanned once and asked about food once, so nobody knows which came first. The authors point out that the frontal areas involved are the ones that help a person resist impulses, and it is at least as plausible that a child with less of that braking power reaches for the pastry as that the pastry shrank anything. The questionnaire records how often, not how much, and total calories were never measured. Body weight and fitness were collected but left out of the final adjustments, in a group whose average weight-for-age sat above the World Health Organization's line for overweight. Sixty-one children is a small number for a method that weighs four foods against ten brain regions at once. And these were families who agreed to a scan, in one part of Chile.

There is also a quieter point about what 'smaller' means. The paper does not say how much smaller in cubic millimetres, and the children were not given memory or attention tests for this analysis. An eleven-year-old's brain is in the middle of being trimmed and rewired, a process in which some regions are supposed to lose volume. A slightly smaller structure on one scan is not a diagnosis of anything.

So what is left? A first look, from a part of the world that brain-imaging research mostly ignores, suggesting that the kinds of food Chile's warning-label law is aimed at may be worth studying for what they do above the neck as well as around the waist. The team says the next step is to follow children over time with repeated scans and repeated diet records. Until that exists, this is a reason to ask the question, not an answer to it.