Is It Too Late to Learn English as an Adult? What the Research Actually Says

Why an adult programme has to answer this question in writing

Every adult education programme meets the same question at the enrolment desk, usually from someone in their forties or fifties who has already half decided the answer. Is it too late to learn English as an adult? The applicant has often been told so by a relative, a former teacher, or a headline about a critical period. They are asking whether the course is worth their money and their evenings, and they are also asking whether they will be the oldest person in the room and the slowest.

For a college, a community learning service, a university language centre or an employer's training team, the honest answer is not only a matter of kindness. It shapes who enrols, what they are promised, how the course is paced, which outcomes are reported to funders, and whether staff quietly expect less from older learners. A programme that has never written its answer down still has one. It is just held informally by whoever staffs the desk that week.

This article sets out what the research actually says about age and learning English, including what the best-known studies do not show. It then turns that evidence into decisions an adult programme makes every term: outcome targets, intake advice, course design, staff beliefs and reporting. The aim is a position a head of adult learning could put in a course handbook and defend to an inspector, a funder or a sceptical applicant.

The short answer is no, it is not too late, with two qualifications that matter. Adults who start English late rarely end up indistinguishable from native speakers on fine points of grammar or accent. And that is almost never the goal an adult learner actually has. Most want to be understood at work, at the doctor's, at a parents' evening or in a job interview, and the research on those outcomes is far more encouraging than the headlines.

What the critical period claim says, and what it never said

The idea that language learning has a window that closes is old, and it began with first language acquisition rather than second. In its strong form it said that after puberty the brain could no longer acquire language in the natural way children do. The claim travelled into second language learning, where it became the popular belief that adults cannot really learn a new language, only memorise bits of one.

Researchers have since separated two questions that the popular version runs together. The first is about ultimate attainment: after many years, does a person who started late reach the same end state as one who started early? The second is about rate and usefulness: how fast does a learner progress, and how well can they use the language for what they need? The evidence on the first question shows real age effects. The evidence on the second question often favours older beginners, at least at the start.

A third distinction also matters for programmes. Much of the research measures one slice of language, such as grammar judgements or pronunciation, and treats native-like performance as the yardstick. An adult learner's goal is rarely to pass as a native speaker. It is to get things done. A study that shows late starters rarely reach native-like grammar is not evidence that late starters cannot reach a confident B2 in workplace English. Those are different claims, and a programme should not let one stand in for the other.

The largest study: Hartshorne, Tenenbaum and Pinker (2018)

The study most often quoted in this debate is “A critical period for second language acquisition: Evidence from 2/3 million English speakers” by Joshua Hartshorne, Joshua Tenenbaum and Steven Pinker, published in Cognition in 2018 (volume 177, pages 263 to 277). The data came from a ten-minute online grammar quiz that went viral on Facebook and was taken by nearly 670,000 people, both native and non-native speakers of English.

Its scale let the authors use a computational model to separate three things that usually travel together: a person's current age, the age at which they first met English, and how many years of experience they had. Their headline finding is that the ability to learn grammar is preserved almost to adulthood, about 17.4 years old, and then declines steadily. They describe the evidence as supporting a sharply defined critical period. In press coverage Hartshorne put it plainly: someone who wants native-like knowledge of English grammar should start by about the age of ten, and people who start between ten and eighteen still learn quickly but have a shorter window before their learning ability declines.

That is a genuine age effect, and an adult programme should not pretend otherwise. But the study's own boundaries are just as important for the question an adult applicant asks. It measured grammar, through a short quiz. It did not measure vocabulary, listening, fluency, pronunciation or the ability to carry out tasks in English. The authors state that the length of the period and its underlying causes remain unknown, and Tenenbaum, speaking to the press, raised both a possible biological change and social changes at that age, such as leaving home, working full time or specialising at university. Accent was named only as a question for future work.

Two further limits are worth stating in a staff briefing. The participants were volunteers who chose to take an online quiz, not a representative sample of adult learners. And the finding concerns the end point reached after long exposure, measured against native speakers. It does not say an adult who starts at 35 learns nothing, or learns too little to be useful. It says that adult starters are unlikely to end up with grammar judgements indistinguishable from someone raised in English. For nearly every adult education outcome, that is the wrong yardstick.

The census evidence: a slope, not a cliff

A second large study points the same way on direction but not on shape. Kenji Hakuta, Ellen Bialystok and Edward Wiley used the 1990 US Census to examine self-reported English ability among about 2.3 million immigrants whose first language was Spanish or Chinese, all of whom had lived in the United States for at least ten years. Their paper, “Critical Evidence: A Test of the Critical-Period Hypothesis for Second-Language Acquisition”, appeared in Psychological Science in 2003 (volume 14, issue 1, pages 31 to 38).

They found that English ability did decline with the age at which people arrived. What they did not find was a break. Testing cut-off points at 15 and at 20, the critical period terms added almost nothing to their models, and the authors conclude that success in second language acquisition declines steadily throughout the life span. A slope and a cliff lead to different policy. A cliff would justify telling a 40 year old that the door has shut. A slope says that each year makes the task somewhat harder and that a 40 year old and a 30 year old are not in different categories.

The census study also found something an adult education service should pay close attention to. Years of formal education were the strongest predictor of English ability in both language groups, and education alone accounted for a large share of the variance. Age of arrival still mattered on top of schooling, but a learner's educational background told the researchers more about their eventual English than any single age line did.

The measure has a clear weakness, and the authors acknowledge it. Census respondents chose a self-description such as speaking English well or not well, and in a check sample those self-reports agreed only moderately with objective tests. A programme should therefore read the census study as evidence about broad patterns across a population, not as a precise measure of what an individual can do. It is still the clearest large-scale evidence that age works as a gradient and that prior schooling carries real weight.

Older beginners often move faster at the start

If the large studies describe the long-run end state, smaller longitudinal studies describe the early months, which is where an adult course actually lives. The classic example is a study by Catherine Snow and Marian Hoefnagel-Höhle published in Child Development in 1978. They followed English speakers of different ages who had moved to the Netherlands and were picking up Dutch, testing them three times during their first year with a battery covering several aspects of language ability.

The result ran against the popular belief. Learners aged 12 to 15 and adults made the fastest progress in the first few months. The youngest group, aged three to five, scored lowest on every test. By the end of the year, the 8 to 10 and 12 to 15 year olds had the best control of Dutch. The authors concluded that their data did not support the critical period hypothesis for language acquisition. A programme should be careful not to stretch this too far: one year of naturalistic learning says nothing about where each group ends up a decade later.

Classroom evidence points the same way. The Barcelona Age Factor project, reported in Age and the Rate of Foreign Language Learning, a 2006 volume edited by Carmen Muñoz, followed school learners of English who began at 8, 11, 14 or 18 and over, measuring them after 200, 416 and 726 hours of instruction. As a summary on the MEITS project blog sets out, older groups generally progressed faster in the early stages, and younger starters had not overtaken older ones by the end of the period studied. The same summary notes a caveat that matters for adult programmes: the adults' progress slowed after the first 200 hours or so.

Put together, these studies describe a pattern that every experienced adult teacher will recognise. Adults bring things children do not: literacy, knowledge of how language works, study habits, and the ability to use explicit explanation. Those advantages produce quick early gains in a structured setting. They do not guarantee native-like results, and progress tends to flatten after the first burst. A course designed around that pattern looks quite different from one designed around the belief that adults are slow from the start.

Accent is not the outcome that decides whether adults are understood

Pronunciation is where the too late belief is strongest, and where adults themselves are most self-conscious. It is also where the gap between the research and the popular reading is largest, because most adult learners do not need to lose their accent. They need to be understood.

Murray Munro and Tracey Derwing tested this distinction directly in “Foreign Accent, Comprehensibility, and Intelligibility in the Speech of Second Language Learners”, published in Language Learning in 1995 (volume 45, issue 1, pages 73 to 97). Native English listeners heard English recorded by Mandarin speakers who had learned it after puberty. They rated how accented each sample sounded and how easy it was to understand, and they also wrote down what they heard, which gave an objective measure of intelligibility.

As the psycholinguist François Grosjean summarised the study, the accent ratings and the comprehension results came apart. Many samples heard as accented were still judged easy to understand, and close to two thirds of the samples scored between 91 and 100 percent on the transcription measure of intelligibility. A later conceptual replication with learners of Spanish found the same partial separation between accent, comprehensibility and intelligibility. The finding is not that accent never matters. It is that a noticeable accent and clear communication routinely coexist.

For an adult programme this changes the target. If the course measures pronunciation against a native accent, older learners will look like failures however much their communication improves. If it measures intelligibility and comprehensibility, it can set goals they can reach and track progress that listeners outside the classroom will actually notice. That is a policy choice, and it is one the programme makes, not the learner's age.

What each study shows and what it does not

Staff and stakeholders tend to remember a study by its headline. The table below sets each source cited here next to its method and, just as important, the claims it cannot support. It is a useful handout for a staff meeting where the too late belief comes up, because it moves the conversation from opinion to the specific evidence.

Read across the columns, the pattern is consistent. Age effects on the long-run end state are real, particularly for grammar measured against native speakers. Age effects on early progress often favour older learners in structured settings. Prior schooling carries a lot of weight. And the outcome most adults care about, being understood, is only partly tied to the native-like accent that late starters rarely acquire.

Setting outcome targets an adult cohort can actually hit

The first practical consequence is for the targets written into a course outline. When a programme leaves its goals vague, the native speaker becomes the hidden benchmark, because that is what learners compare themselves with. A learner who can now handle a phone call with a landlord may still feel that they have failed, because they do not sound like the landlord.

Named levels on the Common European Framework of Reference fix this. A target such as moving from A2 to B1 in spoken interaction over two terms is specific, observable and indifferent to age. It says what the learner will be able to do, not how close they will come to a native speaker. The same logic applies to employer programmes, where the useful targets are job tasks: leading a short meeting, handling a customer complaint, giving a safety briefing.

Targets also need to be honest about time. The Barcelona findings suggest adult progress is fastest early and then slows. A programme that promises the same gain every term will disappoint learners in the second and third terms and may lose them. A programme that tells learners to expect quick early progress followed by a slower, harder phase, and that sets intermediate targets for that phase, gives them a reason to stay through it.

Finally, targets should be differentiated by starting point and background rather than by age. The census evidence suggests that a learner's prior schooling tells a programme more than their birth year. Two 45 year olds, one with a university degree and one who left school at fourteen, are likely to need different pacing, different literacy support and different intermediate targets, even if both say they want to reach B1.

Intake interviews and the advice staff give older applicants

The enrolment conversation is where a programme's belief about age reaches the learner first. In many services it is handled by administrative staff or by whichever teacher is free, with no agreed script. If the person on the desk believes adults struggle, they may say so kindly, steer older applicants to a shorter or lower course, or soften expectations before the learner has taken a placement test.

A better intake process records age, because funders often require it, but does not use it to predict anything. It assesses what the research suggests matters more: prior education, literacy in the first language and in English, the purpose the learner has for English, and the time they can realistically commit outside class. It places learners by what they can do now, not by an assumption about what they will be able to do later.

The advice itself can be short and evidence based. Adults usually make fast progress early in a structured course. Very few late starters end up sounding exactly like native speakers, and they do not need to in order to work, study or deal with services in English. Progress slows after the early stage, and that slowdown is normal rather than a sign that age has caught up with them. A programme that can say those three things clearly, and point to the sources, removes a major reason adults give up before they start.

It is also worth training intake staff to recognise when the too late worry is masking something else. Some older applicants are really worried about literacy, about technology used in class, about being in a group with much younger students, or about their memory. Those concerns have practical answers, such as literacy support, an induction session on the learning platform, mixed-age groups with clear norms, or extra recycling of vocabulary, which a vague reassurance about age does not provide.

Designing the course around how adults actually progress

If older beginners move quickly at first, the opening weeks of an adult course are its most valuable. That argues for a dense start: more contact time or more structured self-study in the first block, clear explicit teaching of high-frequency structures, and early, frequent speaking so that the quick gains reach the learner's mouth and not only their notebook.

Explicit grammar deserves a place in that design. Adults can use explanation in a way that young children cannot, and the early-rate studies suggest they benefit from it. The Hartshorne finding is sometimes read as a reason not to teach adults grammar, because they will never reach native-like intuitions. That reverses the logic. Precisely because adults are unlikely to absorb fine-grained grammar implicitly to a native standard, clear explanation followed by plenty of practice is how they get to accurate, usable English.

The later phase needs different support. When progress slows after the first few hundred hours, learners need visible evidence that they are still moving: recorded speaking samples compared across weeks, task checklists ticked off, short can-do statements tied to the CEFR level they are working towards. They also need more speaking practice than class time can provide, which is where structured practice between sessions becomes part of the course design rather than an optional extra.

Pronunciation work should target the features that most affect intelligibility for the cohort's language backgrounds, not the whole of a native accent. A teacher can identify those features by listening for where communication actually breaks down, rather than for every point where a learner sounds different. That keeps the work focused, achievable, and directly connected to the outcome learners care about.

When the staffroom believes the myth

The too late belief is not confined to learners. Teachers and managers absorb the same headlines, and a staffroom that half believes adults are past it will set lower expectations without anyone deciding to. It shows up in small ways: easier homework for older learners, less correction because it feels unkind, fewer opportunities to speak at length, and a readiness to explain slow progress by age instead of by teaching.

Leaders can address this directly. A short staff session built on the evidence table above is enough to separate what the studies show from what they are assumed to show. The key messages are that the strongest age effects concern native-like end states, that older beginners often progress quickly at first, that schooling matters a great deal, and that intelligibility rather than accent is the realistic communication target.

It also helps to look at the programme's own data. Most services can compare progress between age bands using placement and exit assessments they already hold. If older learners are progressing more slowly, the next question is whether they started from a lower level, had less prior schooling, attended less, or received different teaching. If those factors explain the difference, age is not the variable to act on. If a gap remains after accounting for them, the programme has learned something specific it can respond to.

Reporting progress to funders, employers and boards

Reporting is where a programme's position on age becomes public. Funders, employers and governing boards usually want to know whether learners progressed and whether the money was well spent. Comparing adult results with native-speaker standards, or with the results of children in school, will make adult provision look weak for reasons that have nothing to do with its quality.

A more defensible report compares each learner with their own starting point and with the targets set at enrolment. It reports CEFR movement in the skills the course targeted, task outcomes that matter to the funder or employer, and retention through the slower middle phase, which is where adult programmes often lose people. Where age is reported, it should be reported as a description of the cohort, with progress shown by prior schooling and attendance alongside it, so that readers do not draw a conclusion about age that the data cannot support.

A brief evidence note in the annual report can also set expectations for the people who read it. A paragraph explaining that age effects concern native-like end states, that adult learners commonly progress quickly early in a course, and that the programme measures communication rather than accent tells a board how to read the results. It also protects the programme from a future manager who decides, on the strength of a headline, that adult English is a poor investment.

Where AI speaking practice fits, and what one hands-on test found

Every adult programme runs into the same constraint: learners need far more speaking practice than class time allows, especially in the slower phase after the first few hundred hours. Many adults also feel more comfortable rehearsing with software before speaking in front of a group, which matters for learners who arrived worried about being too old. AI speaking tools can fill part of that gap, but they vary, and a programme should rely on evidence from people who have used them rather than on vendor pages.

Borderset did not run its own test of these tools. The most relevant recent example is Naomi Park's hands-on review video on YouTube, published on 27 June 2026, in which she tried twelve language learning apps for English and German and named a top five. Naomi Park also writes for Borderset and is the named author of this page, so readers should treat this as the author's own test rather than an outside one. In that review she picked Enverson AI as her overall favourite, because in her testing it combined speaking, grammar, writing and personalised feedback in a single app.

She rated the other four on different strengths. In her view Praktika AI offered the most realistic AI conversations, Babbel the most structured course with strong grammar, Talkpal the widest variety of AI chat partners, and Duolingo Max the best support for a daily habit through gamification. Her own conclusion was that no single app is best for everyone, so the sensible choice is the one that fits how a person learns. A programme serving adults with very different goals and backgrounds may find that different learners land on different tools.

The descriptions that follow are Enverson's own account of its product. According to Enverson, its Multidimensional Personalization Engine, or MPE, keeps separate readings of each learner rather than one blended score, and directs practice to whichever is weakest. The readings it tracks are:

• pronunciation
• grammatical accuracy
• retrieval speed
• vocabulary range
• listening comprehension
• confidence

Adult cohorts rarely have an even profile, since one learner reads well and freezes when speaking while another talks fluently with a narrow vocabulary, and per-skill tracking of this kind is aimed at that gap. Enverson also says its curriculum draws on more than 10,000 hours of classroom teaching by founders who ran a language school for a decade before building the product, that learners practise with a larger set of realistic voice agents so that they hear a range of speakers, speeds and registers, and that its lessons apply established methods such as spaced repetition, shadowing, comprehensible input and deliberate error correction, mapped to CEFR levels. Those are the company's descriptions of its product, not findings from this article. According to Enverson, the app teaches spoken English only.

Whatever tool a programme chooses, the same rules apply as for any homework. Tie practice to the course's CEFR targets, ask learners to bring evidence of practice to class, and review it so that the app supports the teaching rather than replacing it.

What the research does not settle

A programme that cites research should be candid about its gaps. The large studies here measure grammar or self-reported ability, not the full range of skills an adult course teaches. The smaller studies are short, cover particular languages and settings, and cannot say where learners end up after many years. None of them followed adult learners in modern blended programmes with speaking practice between classes.

The causes of the age effect also remain open. The Hartshorne team said the length of the period and its causes are unknown, and possible explanations range from changes in the brain to changes in how people live after their late teens: leaving education, working full time and having less time and exposure. If social and practical factors explain part of the decline, a well-designed adult programme is one of the things that can push against it.

Finally, the research describes averages and broad trends, and an average age effect says little about the particular 50 year old in front of an intake adviser. That is the strongest reason not to let age drive advice or placement. The evidence supports telling adults the truth, which is that native-like perfection is unlikely, useful English is very likely with steady work, and the early months are a good time to be a beginner.

Related reading on Borderset

For the gap between understanding English and speaking it, and how a department closes it, see why learners understand English but cannot speak it. For the research on input, conversation exchange and speaking practice, see comprehensible input, conversation exchange and fluency.

Leaders planning AI tools across a programme may find AI in school language programmes: a leader's view useful, and private providers can read school management for language academies. For employer-run English training, see AI language learning apps for corporates.

Frequently asked questions

Is it too late to learn English as an adult?

No. The research shows real age effects on the final level people reach, especially on native-like grammar and accent, but adults who start late still learn English well enough to work, study and deal with services. Older beginners often progress faster than young children in the first months of structured study.

What did the 2018 Hartshorne, Tenenbaum and Pinker study find?

Using a grammar quiz taken by nearly 670,000 people, it found that grammar-learning ability stays largely intact until about 17.4 years old and then declines steadily, and that native-like grammar is most likely for people who start by about 10. It measured grammar only, used volunteers, and did not identify the cause of the decline.

Do adults learn languages more slowly than children?

Not at the start. Studies of learners of Dutch in the Netherlands and of school learners of English in Barcelona found that teenagers and adults made faster early progress than younger children. Adult progress tends to slow after the first few hundred hours, and younger starters may catch up over many years.

Will an adult learner always keep an accent?

Most late starters keep some accent, but accent and being understood are only partly linked. In Munro and Derwing's 1995 study, many accented speakers were still easy to understand, and close to two thirds of samples scored 91 to 100 percent on intelligibility. Programmes should target intelligibility rather than accent loss.

What predicts adult success in English better than age?

In the 1990 US Census study by Hakuta, Bialystok and Wiley, years of formal education were the strongest predictor of self-reported English ability. Age of arrival still mattered, but as a steady slope rather than a cut-off. Attendance, practice time and the quality of teaching also shape results.

How should an adult education programme use this research?

Set named CEFR targets instead of a native-speaker benchmark, assess prior schooling and literacy at intake rather than using age to predict outcomes, front-load the early weeks, teach grammar explicitly, focus pronunciation on intelligibility, and report progress against each learner's starting point.