Four Years, Maybe.
What Learning a Language Does to an Ageing Brain.
You have heard that speaking two languages delays dementia by about four years. That claim comes almost entirely from one type of study. When researchers looked only at the studies that followed people forward in time, the effect vanished. This is a post about what happened in that gap, and what is left standing after it.
Here is something you have almost certainly heard before. Speaking two languages delays the onset of dementia by four to five years. No drug comes close to that. It is free. Go and learn Italian. Or Turkish, an Arabic dialect and Hindi like me.
I wanted it to be true. It is the most attractive finding in cognitive ageing, because unlike almost everything else on the list it asks you to do something pleasant.
So I went to check it, expecting to write a post about how to learn a language for brain health. What I found instead was one of the cleanest examples I have seen of a research literature splitting in half depending on how the studies were designed.
The four-to-five year figure is real. It just comes from the other kind of study.
What follows is that gap, what causes it, which of the two answers to believe, and then the separate and much more hopeful question hiding underneath the whole argument: whether actually learning a language now does anything, which is not the same question at all.
The Gap
Thirteen studies in one systematic review. Sorted by design, they give opposite answers. Tap each side.
Mukadam, Sommerlad and Livingston, systematic review and meta-analysis of 13 studies: 5 prospective, 8 retrospective or cross-sectional. [1]
The explanation the review's authors give is the part worth slowing down for, because it turns a methodology point into something much more uncomfortable.
A retrospective study asks when the symptoms started. Researchers take people who already have a dementia diagnosis, sort them into bilingual and monolingual, and compare age at symptom onset. Onset is established by asking the patient and the family to remember.
A prospective study watches it happen. Researchers enrol people who do not have dementia, follow them forward, and record who develops it using standardised assessment. Nobody has to remember anything.
The authors attribute the discrepancy to confounding by education and to cultural differences in presentation to dementia services. [1] Read that second phrase again. Bilingual populations in many of these samples are immigrant and minority-language populations. If those communities reach a dementia clinic later, for reasons of access, language barriers at the point of care, or how memory changes are interpreted within a family, then the measured age at onset is later. Not because the disease started later. Because the appointment did.
Part of the famous protective effect may be a diagnostic delay wearing a lab coat.
I want to be fair to the original studies here, because they were not careless. Ellen Bialystok, Fergus Craik and Morris Freedman's 2007 paper, which reported a 4.1 year difference in symptom onset, is carefully done and honest about its design. [2] Suvarna Alladi and colleagues then ran a version specifically built to test the confounding objections, in Hyderabad, where bilingualism is normal across the population rather than concentrated among migrants, and found a 4.5 year difference that held independently of education, literacy, occupation, urban or rural residence, and immigration status. [3] That is a serious result and it is not easily dismissed.
Meanwhile Laura Zahodne and colleagues followed Spanish-speaking immigrants in New York prospectively and found no association between bilingualism and rate of cognitive decline or dementia risk. [4]
So the honest summary is not that the finding was debunked. It is that a well-designed retrospective study and a well-designed prospective study, in different populations, disagree, and the design difference tracks the disagreement with unsettling precision.
Six Studies, Sorted by What They Can Actually Rule Out
Not by whether they agree with each other. Filter them.
The design column matters more than the result column. A study that cannot rule out the obvious objection does not stop being evidence, it stops being decisive. [1][2][3][4][5][6]
Every version of this argument eventually reaches the same objection, and for decades nobody could answer it.
People who speak several languages might simply have started out with better cognitive function. You cannot tell whether the language caused the brain or the brain caused the language, because by the time anyone measures either, both are already there.
Answering that requires something almost nobody has: a measurement of the same people's cognitive ability before they learned the second language.
Scotland had one. In 1947, essentially every eleven-year-old in the country sat a standardised mental ability test. Sixty years later, Thomas Bak and colleagues retested 853 of them, then aged around 73, and asked whether bilinguals performed better than their own childhood scores predicted.
They did. Bilinguals performed significantly better than predicted from their baseline cognitive ability, with the strongest effects on general intelligence and reading. [5]
And the part that matters for you. The effect was present in people who acquired their second language in adulthood, not only in those raised bilingual. [5]
This is, to my reading, the single strongest piece of evidence in the whole literature, because it is the only design that removes the reverse causation objection rather than adjusting for it. It is still observational. Whatever led someone to pick up a language at forty might also be protective. But it is a much smaller loophole than the one it closes.
Which brings us to the question that got lost while everyone argued about lifelong bilinguals.
What If You Start Now
A different question from whether lifelong bilinguals are protected. Four studies that asked it.
Every study here is small. That is the state of the field, not a selection I made. [7][8][9][10]
The people whose brains supposedly benefit are the ones being told their languages are a problem.
Look at who the bilingual participants in this literature actually are. Immigrants. Minority-language speakers. People in multilingual regions of India and Canada and the United States. The protective effect, where it appears, appears in them.
Those same language repertoires, in daily life, get treated as deficits. Children arrive at school speaking two languages and are recorded as having limited proficiency in one of them. Adults get their competence assessed through their accent. Mothers are advised, still, to speak only the majority language at home so as not to confuse the child, advice that has no support in the developmental literature and has been circulating for generations.
And then the research establishment takes the same repertoire and calls it cognitive reserve.
The second thing worth being angry about is buried in the methodology. The leading explanation for why retrospective studies overstate the effect is cultural difference in presentation to dementia services. [1] That is a finding about unequal access to diagnosis, and it has been sitting inside a debate about bilingualism, treated as a nuisance variable, when it is a serious health inequity in its own right. A community that reaches a memory clinic years later than another community is not protected. It is underserved, and the delay has been quietly inflating an attractive headline.
And the third. Language learning as an intervention costs almost nothing, has no side effects, and has been studied in samples of a few dozen people. Compare that with what gets spent establishing whether a compound moves a biomarker. The cheap interventions stay under-evidenced not because they are unpromising but because nobody owns them.
Four things to hold lightly
The bilingual advantage in executive function, the mechanism most often invoked to explain all of this, does not survive scrutiny well. Minna Lehtonen and colleagues meta-analysed the adult literature and, after accounting for publication bias, found no reliable bilingual advantage across executive function domains. [6] If the proposed mechanism is that unreliable, claims resting on it should be too.
The intervention studies are tiny. The Swedish structural imaging study had a few dozen participants, and the trials of language courses in older adults are described by their own authors as pilots or feasibility studies. [7][8][9] None of them establish a cognitive benefit that persists.
Structural brain change is not the same as clinical benefit. Learning something intensively changes brain structure measurably. So does juggling. The presence of grey matter change tells you learning occurred, not that anything was protected. [7]
And cognitive reserve is a model, not a substance. It describes why some people tolerate more pathology before symptoms appear. Nobody can measure your reserve, and nothing in this post lets you know whether yours has changed. [10]
Starting a language in adulthood is one of the more demanding sustained cognitive activities available, it is social, and the one study that could rule out reverse causation found the effect in adult learners too. [5] That is a reasonable basis for doing it. Four guaranteed years is not.
The Swedish study measured structural change after three months of full-time intensive study at an interpreter academy, against a control group of demanding non-language students. Whatever produced that change, it was not fifteen minutes a day on a phone. [7]
Social isolation is itself on the Lancet Commission's list of modifiable dementia risk factors. A language learned alone on an app and never spoken to anyone strips out the component that has independent evidence behind it. This is speculation on my part and I am labelling it as such.
The advice to drop a home language for the sake of a child's development does not have developmental evidence behind it. Whatever the bilingualism and dementia literature eventually settles on, nothing in it argues for having fewer languages.
I set out to write the post that tells you to learn Italian because it buys you four years. I cannot write it. The four years comes from studies that cannot separate a later illness from a later appointment, and the studies that can find nothing.
What survives is smaller and, I think, more interesting.
Eight hundred and fifty-three Scots, tested as children in 1947 and again in their seventies, where the ones who had learned another language, including the ones who learned it as adults, did better than their own eleven-year-old selves predicted. That result does not depend on anybody remembering when the symptoms started. It is the only piece of this literature that closes the loophole rather than stepping around it.
It is still one cohort. It is still observational. It is the best thing here, and it says that something about doing this, at any age, is associated with holding on to more than you otherwise would.
Which, given that the alternative is sitting with the certainty of four years that does not exist, seems like a better reason to start. Not because a study promised you a number. Because it is difficult, and social, and you get to become slightly different, and the one design that could have ruled it out did not.
The pattern where a retrospective finding travels for fifteen years before the prospective data catches up is the same pattern running through most of what women are told about their bodies. The Art of Female Health is built on checking rather than repeating. nammu.academy/book
References
- Mukadam, N., Sommerlad, A., & Livingston, G. (2017). The relationship of bilingualism compared to monolingualism to the risk of cognitive decline or dementia: A systematic review and meta-analysis. Journal of Alzheimer’s Disease, 58(1), 45–54. https://doi.org/10.3233/JAD-170131
- Bialystok, E., Craik, F. I. M., & Freedman, M. (2007). Bilingualism as a protection against the onset of symptoms of dementia. Neuropsychologia, 45(2), 459–464. https://doi.org/10.1016/j.neuropsychologia.2006.10.009
- Alladi, S., Bak, T. H., Duggirala, V., et al. (2013). Bilingualism delays age at onset of dementia, independent of education and immigration status. Neurology, 81(22), 1938–1944. https://doi.org/10.1212/01.wnl.0000436620.33155.a4
- Zahodne, L. B., Schofield, P. W., Farrell, M. T., Stern, Y., & Manly, J. J. (2014). Bilingualism does not alter cognitive decline or dementia risk among Spanish-speaking immigrants. Neuropsychology, 28(2), 238–246. https://doi.org/10.1037/neu0000014
- Bak, T. H., Nissan, J. J., Allerhand, M. M., & Deary, I. J. (2014). Does bilingualism influence cognitive aging? Annals of Neurology, 75(6), 959–963. https://doi.org/10.1002/ana.24158
- Lehtonen, M., Soveri, A., Laine, A., et al. (2018). Is bilingualism associated with enhanced executive functioning in adults? A meta-analytic review. Psychological Bulletin, 144(4), 394–425. https://doi.org/10.1037/bul0000142
- Mårtensson, J., Eriksson, J., Bodammer, N. C., et al. (2012). Growth of language-related brain areas after foreign language learning. NeuroImage, 63(1), 240–244. https://doi.org/10.1016/j.neuroimage.2012.06.043
- Ware, C., Damnee, S., Djabelkhir, L., et al. (2017). Maintaining cognitive functioning in healthy seniors with a technology-based foreign language program: A pilot feasibility study. Frontiers in Aging Neuroscience, 9, 42. https://doi.org/10.3389/fnagi.2017.00042
- Bubbico, G., Chiacchiaretta, P., Parenti, M., et al. (2019). Effects of second language learning on the plastic aging brain: Functional connectivity, cognitive decline, and reorganization. Frontiers in Neuroscience, 13, 423. https://doi.org/10.3389/fnins.2019.00423
- Antoniou, M., Gunasekera, G. M., & Wong, P. C. M. (2013). Foreign language training as cognitive therapy for age-related cognitive decline: A hypothesis for future research. Neuroscience & Biobehavioral Reviews, 37(10), 2689–2698. https://doi.org/10.1016/j.neubiorev.2013.09.004