Four Languages, a Brain 13 Years Younger: Read the Caveats

A finding presented in Europe this summer has travelled further than most cognitive-ageing research, and the reason is a single striking number: a brain estimated to be thirteen years younger than the calendar says. It is a real result from a serious group. It is also a conference presentation on a modest sample, describing an association, and almost every version you will have read dropped at least two of those three qualifications.

What was measured, and how

The work was presented at the Federation of European Neuroscience Societies Forum on 6 July 2026 by Lucia Amoruso of the Basque Center on Cognition, Brain and Language in San Sebastian, with collaborators at institutions in Chile, Argentina and Ireland. The instrument was magnetoencephalography, or MEG, which records the very small magnetic fields produced by electrical activity in the brain, at a much finer time resolution than most imaging methods offer.

The team trained a machine-learning model on MEG recordings from 728 people to predict chronological age from patterns of resting brain activity, then applied that model to a separate group of 144 participants. The output is a brain age: what the model would guess your age to be from your brain activity alone. Subtract the real age and you get a gap, negative when the brain looks younger than the birth certificate.

• Bilingual speakers: brain age roughly six years below monolinguals.
• Trilingual speakers: roughly seven years below.
• Speakers of four languages: roughly thirteen years below.
• Independently of the count, greater proficiency and earlier acquisition of a second language were associated with a larger negative gap.

That last point is the most interesting part of the study and the part most reporting omitted. The effect behaved like a dose rather than a category. If multilingualism were simply a marker for some other advantage, you would not necessarily expect depth of experience to track the outcome on its own.

Four reasons to hold the number loosely

The result is genuinely interesting and it is not yet a finding you should act on. The reasons are structural rather than a criticism of the researchers, who have been careful in how they described it.

• It was presented at a conference. Conference abstracts are not peer-reviewed to the standard of a journal article, and details routinely change between presentation and publication.
• The estimation group is 144 people. Once that is split by number of languages spoken, the subgroups behind a thirteen-year figure are small, and small subgroups produce large, unstable estimates.
• The design is observational. People who speak four languages differ from monolinguals in education, occupation, migration history, income and social contact – all of which independently track brain health.
• Brain age is a model output, not a measurement. Change the training data or the algorithm and the number moves. It has no clinical threshold and no established meaning for an individual.

Cognitive reserve, and why the idea is plausible anyway

The framework behind this kind of work is cognitive reserve: the observation that people with similar amounts of physical brain pathology can differ substantially in how much cognitive impairment they show. The proposed explanation is that a lifetime of demanding mental activity builds more flexible processing, so that function is maintained for longer as damage accumulates. Education, occupational complexity and social engagement have all been studied as contributors.

Managing several languages is a reasonable candidate on that logic. A multilingual speaker is continuously selecting one language and suppressing the others, which recruits executive control – the same broad system involved in attention, inhibition and switching. Whether this constitutes exercise in any useful sense is exactly what remains unsettled, and it is a close cousin of the transfer problem that has dogged commercial cognitive training, which we set out in our review of what twenty years of trial data show about brain training.

Cognitive reserve also has an uncomfortable implication that rarely survives the summary. If reserve works by maintaining function while pathology accumulates silently, then people with more of it are not spared the disease – they reach the point of visible impairment later, with more damage already present. Some of the bilingualism literature reports precisely that pattern: later symptom onset without a corresponding difference in the underlying condition. That is a genuine benefit measured in years of ordinary life, and it is not the same claim as a healthier brain.

It is also why the direction of causation is so hard to pin down here. Learning and maintaining several languages to a high standard requires the kind of memory, attention and verbal facility that a slower-ageing brain would supply anyway. An observational study finds the association whichever way the arrow points, and no amount of statistical control fully separates the two. Only a trial that assigned people to language learning could, and no such trial has run at anything like this timescale.

What this is not: an IQ finding

Brain age and IQ are different things and the distinction gets lost quickly in coverage. An IQ score estimates current reasoning performance against a reference sample of people the same age, on a scale where the average is set to 100 and the standard deviation – a measure of spread – is usually 15. A brain-age gap is a model's residual, expressed in years, derived from resting brain activity and not from any task performance at all. Nothing in this study reported an ability score, and a younger-looking brain is not a claim about reasoning.

It is also not a prescription. The study cannot tell you that starting a language at forty will subtract years from your brain age, because it did not test that and its design could not have. What it adds is one more observational thread to a long-standing hypothesis about mentally demanding lifelong activity, measured with an unusually precise instrument. Our piece on when different abilities actually peak covers what is much better established about cognitive change across the lifespan.

How to read the next brain-age headline

• Ask whether the work is published or presented. A conference talk is a preview, not a record.
• Ask how many people the headline subgroup contains, not how many were in the study overall.
• Ask whether the comparison is randomised or observational. Almost every lifestyle-and-brain finding is the latter, and the confounders are usually obvious once named.
• Ask what brain age is being derived from, and remember that no two brain-age models agree on an individual.
• Ask whether any cognitive performance was measured at all. Frequently none was.

The honest position is that learning languages is worth doing for reasons that need no neuroscience, that a large and careful literature makes lifelong mental engagement a plausible contributor to cognitive resilience, and that a thirteen-year figure from a conference presentation on 144 people is an intriguing signal rather than a fact about you. For a case where the long-run evidence is far stronger, our report on childhood IQ scores and later-life health shows what a genuinely settled longitudinal finding looks like – and you can take the test or open the percentile calculator if you want a number of your own.

Common questions