The brain's insulation can turn against the wire

Scottish children sat an intelligence test in 1947. Decades later, some of them donated their brains. In those who declined fastest, the myelin was thicker, not thinner, and the cells that made it look less like bystanders than culprits.

On 4 June 1947, 70,805 Scottish schoolchildren sat the same mental test on the same morning. Nearly sixty years later, researchers in Edinburgh traced 1,091 of the ones born in 1936 and started testing them again, at 70, and again, and again, on memory, processing speed and spatial skill, out to age 82. Of them, 866 came back for follow-up testing after 70. Almost everyone slipped a little. What made this cohort extraordinary was that the slipping could be measured against a person's own eleven-year-old self, and that a subset of them agreed, at the end, to give their brains to science. That is the unglamorous machinery behind a finding published in Nature Medicine on 25 August 2026 by Georgina Craig, Veronique Miron and colleagues at the University of Edinburgh, the UK Dementia Research Institute and St Michael's Hospital in Toronto.

The expectation, when the team went looking at white matter under an electron microscope, was straightforward: faster decline should mean less myelin, the fatty sheath that insulates nerve fibres and lets signals travel fast. They found the opposite. In donors whose cognitive trajectory had been steepest, the myelin was thicker, the myelinated axons were smaller, large nerve fibres were depleted, and oligodendrocytes, the cells that build myelin, were more numerous rather than fewer. Crucially, according to the Edinburgh team's account of the work, these features tracked the rate of decline over the years, not a person's score at any single sitting. Thicker insulation of poor quality, wrapped around shrinking wires: not a brain that had run out of repair crew, but one where the repair crew had started building badly.

The gene expression work pointed at a single switch. Oligodendrocytes in the steepest decliners showed downregulation of NRF2, a transcription factor that governs hundreds of genes protecting cells from oxidative damage, and immunostaining confirmed fewer NRF2-positive oligodendrocytes. Association, of course, is not cause, and the team knew it. So they went to mice, deleting NRF2 specifically from oligodendrocytes in aged animals. The mice reproduced the human picture, thicker myelin and a shift towards myelination of smaller-diameter axons, and their cognitive performance failed to improve over time the way it otherwise does. That is a mouse result, in mice, and it establishes sufficiency in a rodent, not in a person. But it is the step that turns a correlation in donated tissue into a candidate mechanism.

"We have always considered oligodendrocytes as purely beneficial," Craig said in the university's announcement of the work, and the field's assumption really has run that way: myelin as armour, myelin loss as injury. This study inverts the framing without discarding it. Myelin is still essential. What the data suggest is that the cells making it can, with age and with NRF2 running low, produce a product that harms the fibre it is meant to protect. For families watching a parent's processing speed fade year by year, that is a different mental model of what is happening inside the skull, and a different place to aim a drug.

Which brings the inevitable question. The NRF2 pathway is not untouched pharmacological territory; the authors note that drug-induced NRF2 activation with dimethyl fumarate has improved cognitive function in rodents and in multiple sclerosis, and the Edinburgh team raise repurposing as a possibility for later life. Possibility is the operative word. No one in this study took a drug. There is no trial here, no dose, no outcome in an ageing population, and the LBC1936 is a general cohort of community-dwelling Scots, not a dementia cohort, so nothing in it tells you whether any of this alters the course of Alzheimer's disease, vascular dementia, or any other diagnosis. Anyone wondering about a medicine they already take, or one they have read about, has exactly one correct next step, and it is a conversation with their own clinician.

Seventy-nine years after a nationwide school test, the answers those children gave are still doing work, now in the form of a hypothesis about why some brains hold their speed and others do not. The next question is whether a pathway that can be switched in a mouse can be switched safely, usefully, and measurably in a person. That will take trials, and trials take years.