Bilingualism and the brain: a genuinely contested area
Whether speaking two languages protects the aging brain: prospective studies did not find protection from dementia while retrospective studies suggested later symptom onset (Mukadam 2017), and bilinguals were on average 3.45 years older at symptom onset in one meta-analysis (Li 2025).
Do not take up a language expecting dementia protection.
- new word-finding or language difficulty that is unexplained and progressive
What the evidence says
The sources are meta-analyses and systematic reviews of bilingualism, dementia and second-language learning, and they differ by design. Mukadam 2017 (meta-analysis) reports a combined odds ratio of dementia of 0.96 for bilinguals against monolinguals in prospective studies, not significant, and concludes that prospective evidence did not find that bilingualism protects from cognitive decline or dementia, while retrospective studies, prone to confounding, suggested later symptom onset. Li 2025 (meta-analysis) reports bilinguals on average 3.45 years older at symptom onset, with six incidence-rate studies not providing robust evidence for a preventive effect, and concludes that bilingualism has potential effects in delaying onset but not preventing it. van den Noort 2019 (systematic review) reports delays of 4 to 5.5 years in some studies and no evidence in others, and concludes that no firm conclusions can be drawn; Deví Bastida 2020 reports 7 of 10 studies suggesting a positive relationship and calls bilingualism a probable protection factor. On language learning in adulthood, Ware 2021 (systematic review) reports improvements in attentional switching, inhibition and working memory and concludes that there are some indications of cognitive benefits in later life, especially for executive functions; Gkintoni 2025 reports neuroplastic adaptations correlated with improvements in working memory, attentional control and cognitive flexibility; Carthery-Goulart 2023 (meta-analysis) reports that speaking more distant language pairs positively influenced monitoring performance in seniors. Whether learning a language protects against dementia is not settled by these sources; learning one for connection, culture and enjoyment does not depend on it.
- 'Bilingualism protects against dementia' (prospective studies did not find protection, while retrospective studies suggested later symptom onset; Mukadam 2017)
- 'Only childhood bilingualism counts' (adult language learning is worthwhile for its own goals regardless of the cognitive debate)
- Does not prevent or delay dementia or Alzheimer's disease
- Does not raise IQ or guarantee cognitive benefits
- The protective claim is scientifically contested, not established
- Agreement
- Direction recorded for 0 of 7 sources; not yet recorded for the rest.
- The claim's outcome
- Outcome not yet recorded
- Certainty
- No source states a certainty rating.
- How the sources were read
- Some sources are based on abstracts, not full texts.
What each source measured and concluded (7)
- What it measured
- Relationship between bilingualism and age at dementia onset, and bilingualism and dementia incidence rate
- How it was read
- based on the abstract, not the full text
- What the review says it includes
- Bayesian meta-analysis
- Studies and participants, as stated
- 18 age-at-onset studies and 6 incidence-rate studies
- What it measured
- Relationship between bilingualism and Alzheimer's disease as a protection factor / cognitive reserve
- How it was read
- based on the abstract, not the full text
- What the review says it includes
- Systematic review
- Studies and participants, as stated
- 10 studies
- What it measured
- Neuroplasticity and cognitive outcomes (executive function, working memory, attention) and socio-emotional processes
- How it was read
- based on the abstract, not the full text
- What the review says it includes
- Systematic review (including RCTs and longitudinal studies)
- Studies and participants, as stated
- N=80 studies
- What it measured
- Improvement in attentional switching, inhibition, working memory, and increased functional connectivity.
- How it measured it
- Cognitive tests and neuroimaging techniques (validated).
- How it was read
- read in full text
- What the review says it includes
- Systematic review of 9 interventional studies
- Studies and participants, as stated
- N=622 participants across 9 studies
- What it measured
- Cognitive reserve-enhancing effect of lifelong bilingualism.
- How it measured it
- Behavioral and neuroscience research methods.
- Its conclusion, in its own words
- “Although some evidence was found for a cognitive reserve-enhancing effect of lifelong bilingualism and protection against dementia, to date, no firm conclusions can be drawn.”
- How it was read
- read in full text
- What the review says it includes
- Systematic review
- Studies and participants, as stated
- N=4946 (bilingual) + 4524 (monolingual) = 9470 participants for cognitive decline; N=2794 (bilingual) + 4207 (monolingual) = 7001 participants for dementia across 34 studies
- What it measured
- Executive function, specifically monitoring performance.
- How it measured it
- Executive function tasks (validated).
- How it was read
- read in full text
- What the review says it includes
- Systematic review and meta-regression
- Studies and participants, as stated
- N=1417 participants across 18 studies
- Heterogeneity, as stated
- Exclusion of studies with heterogeneous language pairs.
- What it measured
- Cognitive decline or incident dementia
- How it was read
- screened only, not extracted
- What the review says it includes
- Systematic review and meta-analysis (random-effects)
- Studies and participants, as stated
- 13 articles (from 1,156); prospective n=5,527
Each source below is shown with its study design and a link to the original study, or, where it is not a completed research paper, with what kind of record it is; those are not counted. This is the audit trail.
Why do some records show a record type instead of a study design?
A study design is shown for a completed, peer-reviewed paper. A registered protocol, a preprint, a conference abstract, a referee report or a book chapter has not been through peer review as a finished paper, so it is labelled for what it is, and it is counted in no source count. Where a record could not be confirmed as a completed paper at all, it says so rather than guessing. They all stay in the list, and their DOI still resolves, because the count of what was screened is part of the audit trail.
- Protocol / registration A registered plan for a review or trial, not its results.
- Preprint Posted by its authors before peer review.
- Conference abstract A conference summary rather than a peer-reviewed paper.
- Peer-review report A referee's report on a paper, not the paper itself.
- Guideline / position stand A recommendation derived from evidence, which is a step further from the data than the evidence itself.
- Book chapter A chapter in a scholarly book rather than a journal paper.
- Type not recorded Our record does not say what kind of publication this is, so it is counted in no source count. That is a gap in our data, not a judgement about the research.
“Abstract only” means the review read that paper’s abstract, not its full text; “screened only” means it was included at re-screening and not extracted. Both stay in the trail with their DOI, because how a source was read is part of the audit.