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

The question we hear in every adult class

On the first evening of every adult course, someone stays behind after the lesson. Sometimes it is a 34 year old accountant, sometimes a retired nurse of 67. The question is always some version of the same thing: did I leave this too late? Usually they have read a headline saying that the window for learning a language closes in childhood, and they want to know whether the next year of their evenings is wasted.

As the person who designs the curriculum they are about to follow, I owe them a straight answer, and so does this article. The straight answer is that it is not too late to learn English as an adult, in the sense that matters to almost every adult who asks. It may be too late to become indistinguishable from someone who grew up speaking English. Those are two different goals, and most of the confusion in this debate comes from mixing them up.

The research behind the myth is real and good. The largest study, published in 2018, is genuinely important. But it measured one thing, under particular conditions, and it is routinely quoted as if it measured something else. Below we go through the main studies one at a time, say what each found, say what each did not find, and then turn the result into a course plan an adult could start on Monday.

One rule for the rest of the page: we only cite studies we have read at the source, and where a finding is limited we say so. A myth busting article that replaces one overstatement with another would be no use to anyone.

Where the idea of a closing window comes from

The phrase behind the myth is the critical period hypothesis. It began as a claim about first languages: that a child must acquire a mother tongue before the brain finishes a particular stage of development, often placed around puberty. Researchers then asked whether the same limit applies to a second language, and that question has been argued over for decades.

The idea feels true because everyone has seen the evidence for it. A family moves abroad; the children sound local within a couple of years while the parents keep a strong accent and make the same grammar mistakes for life. That pattern is real, and the research confirms a version of it. What the pattern does not tell you is why it happens, or whether the parents could have learned more with different conditions.

Think about what changes between those children and their parents. The children spend six hours a day in a school where English is the only language, play with English speaking friends, and have no professional reputation at stake when they get something wrong. The parents work long hours, often in their first language, socialise with people from home, and feel every error. Age is one difference. It is not the only one, and the studies below try, with varying success, to pull those differences apart.

There is also a quieter reason the myth survives: it is comforting. If the window closed at twelve, then struggling at forty is nobody's fault. That comfort comes at a price, because it persuades adults not to start, and in our experience the adults who start and keep going are rarely the ones who regret it.

The 2018 study of two thirds of a million people

The study most often quoted is by Joshua Hartshorne, Joshua Tenenbaum and Steven Pinker, published in the journal Cognition in 2018. Its title promises evidence from two thirds of a million English speakers, and it delivers: after cleaning, the authors had complete data for 669,498 people. The data came from a ten minute grammar quiz called Which English?, which, as MIT News reported, spread within hours of being posted on Facebook.

The quiz asked people to judge English sentences, and it also asked how old they were, when they started learning English, and how long they had been exposed to it. That combination is what made the study special. Earlier work struggled to separate three things that move together: how old you are, how old you were when you started, and how many years you have been learning. With hundreds of thousands of people, the authors could fit a model that pulled those apart and estimated how the underlying ability to learn grammar changes with age.

The headline finding is that this ability is preserved almost into adulthood, to about 17.4 years, and then declines steadily. That is about a decade later than many earlier estimates, which tended to put the end of the window around puberty. The authors found the pattern for both difficult and easy grammar.

From this, Hartshorne drew a practical conclusion in the MIT coverage: if you want native-like knowledge of English grammar, you should start by about the age of ten. The reason is arithmetic rather than mystery. Grammar learning is slow even for children, so a person who starts at fifteen runs out of the high learning years before reaching a native level. The same coverage adds that people who start between ten and eighteen still learn quickly, even if they do not usually reach native-speaker grammar.

This is a strong, careful study, and nothing in this article is an attempt to wave it away. Adults who start English late are very unlikely to end up with grammar intuitions identical to those of someone raised in English. If that is your goal, the evidence is against you. The question is whether that is your goal.

What the 2018 study does not show

Most headlines about the study made a jump the authors did not make: from native-like grammar is hard to reach after childhood to adults cannot learn a language. Read what was measured and the gap between those two claims is large.

It measured grammar, and only grammar. The quiz tested judgements about English sentences. It did not test vocabulary, listening, pronunciation, reading or the ability to hold a conversation. Vocabulary in particular keeps growing through adult life in a first language, and nothing in this study says it stops in a second one.

The bar was native-like, not good. The finding concerns reaching the same grammar judgements as a native speaker. An adult who reaches a solid B2 or C1 level on the Common European Framework of Reference, runs meetings in English and reads novels in it, would still fall short of that bar. The study is not about whether you can get there.

Nobody yet knows why the ability declines. In the MIT coverage, Tenenbaum gave two possibilities: a biological change, or something social or cultural. He pointed out that in many societies the period of being a minor runs to about 17 or 18, and that afterwards people often leave home, work full time or become specialised university students. If part of the decline is about how adult life is organised rather than how the brain ages, then an adult who reorganises some time around English is not simply fighting biology.

The sample chose itself. The people in the study were volunteers who took a quiz online. That is a strength for size and a weakness for representativeness, and the authors' model rests on assumptions about how exposure accumulates. None of that undermines the main pattern, but it is a reason not to treat 17.4 as a precise biological cliff.

Put simply: the strongest study on the subject says that native-like grammar is very hard to reach if you start as an adult. It does not say that fluent, accurate, professional English is out of reach, and it was never designed to answer that.

What the other big studies add

The 2018 paper is one study. Several others, using different methods, help fill in the picture, and not all of them point the same way.

A census of 2.3 million immigrants. Kenji Hakuta, Ellen Bialystok and Edward Wiley analysed the 1990 United States census in a paper in Psychological Science in 2003. They looked at 2,016,317 Spanish speakers and 324,444 Chinese speakers who had lived in the country for at least ten years, and compared their self-rated English with the age at which they arrived. English ability fell as age of arrival rose, but it fell smoothly. When the authors tested for a break at 15 or at 20, the kind of sudden drop a strict critical period predicts, they found nothing of practical size. Years of formal education, meanwhile, explained a substantial share of the differences in both groups.

The authors are open about the weakness: the measure is a person's own rating of their English, which matches objective tests only moderately. Still, the shape of the result matters for adults. A smooth decline means there is no birthday after which learning stops. It means each extra year makes the task somewhat harder, and that schooling, and by extension how well someone has learned to learn, makes a large difference.

A replication with Spanish speakers. In 1989, Jacqueline Johnson and Elissa Newport published an influential study of immigrants' English grammar. As Hakuta and colleagues summarise it, it reported a strong decline with age for people who started before puberty and wide, random variation among those who started later. David Birdsong and Michelle Molis repeated it with the same methods and materials, using 61 native speakers of Spanish, and published the result in the Journal of Memory and Language in 2001. They found modest evidence of nativelike attainment among some late learners. They also found that results kept falling with later starting ages even after the supposed end of the critical period.

Both halves of that finding matter. Some adults who start late reach results that are hard to tell apart from native speakers on a demanding grammar test, so a native level is not strictly impossible. And age keeps mattering well into adulthood, so there is no point after which it stops being a factor. Neither half supports the claim that it is ever too late to learn.

Faster at first, lower ceiling later: the distinction that matters

The single most useful idea in this whole literature is the difference between how fast people learn and how far they eventually get. Once you separate the two, most of the contradictions disappear.

On speed, adults often win. In a classic study published in Child Development in 1978, Catherine Snow and Marian Hoefnagel-Höhle followed English speakers of different ages as they picked up Dutch after moving to the Netherlands, testing them three times during their first year. 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. The authors concluded that their data did not support the critical period hypothesis.

Classroom research points the same way. The Barcelona Age Factor project, collected in a 2006 volume edited by Carmen Muñoz, followed school learners in Spain who began English at different ages. A summary by the MEITS project at Cambridge describes the result: after the same number of classroom hours, older starters had learned faster, and the younger starters had not overtaken them within the period studied. With a few hours a week of instruction, the long immersion that lets young children catch up simply does not happen.

On ceiling, early starters win. Over many years of immersion, people who arrived as young children tend to end up closer to native speakers, which is exactly what the 2018 study and the census data show in different ways. Children are slower out of the gate and steadier over the long run, provided they are surrounded by the language.

For an adult, this is good news hidden inside a headline that sounds like bad news. Most adults are not trying to pass as native after twenty years of immersion. They want to reach a working level in the next year or three, with a few hours a week, which is precisely the situation in which the research shows older learners making faster progress.

The chart above is a reminder of how different these studies are. The census study is enormous but relies on self-ratings; the 2018 study is large and uses a real grammar test but only grammar; the 2001 replication is small and careful. Sound conclusions come from what they agree on, and they agree that age matters gradually, not that it slams a door.

What an adult can realistically reach, in CEFR terms

In course design we do not talk about native-like. We talk about levels on the Common European Framework of Reference, from A1 to C2, because those describe what a person can do: handle a doctor's appointment, follow a meeting, argue a case, read a contract. Those are the goals adult learners actually bring to us.

Cambridge English publishes guided learning hours for each level: roughly 90 to 100 hours to A1, 180 to 200 to A2, 350 to 400 to B1, 500 to 600 to B2, 700 to 800 to C1 and 1,000 to 1,200 to C2, all counted from beginner. Cambridge itself calls these a guideline and lists age among the things that change them, along with prior learning, intensity and study outside class. They are a planning budget, not a promise.

What those numbers make clear is that the real constraint for most adults is hours, not age. A B1 learner can handle most everyday situations and travel; a B2 learner can work in English with some effort. Both levels sit within a few hundred hours of study, and that is something an adult controls.

The chart turns the B1 budget into weeks. At three hours a week, reaching B1 from zero takes more than two and a half years; at ten hours, under a year. Most of our adult students land between five and seven hours once homework is counted, which puts B1 roughly a year to eighteen months away for a true beginner. Many adults are not true beginners, having studied English at school, and start several levels higher on the same chart.

Notice that age does not appear in the arithmetic. It may change the efficiency of each hour a little, as the research suggests. It does not change the order of magnitude. A 45 year old studying seven hours a week will reach B1 long before a 25 year old studying one.

What we see in adult classrooms

Research tells you about averages across large groups. Teaching tells you about the person in the third row. Here is what our teachers report about adults who return to English, offered as classroom observation rather than as evidence.

Adults learn grammar explanations quickly. Where a child absorbs a pattern through months of exposure, an adult can be shown why English uses the present perfect for news and then use it correctly within the lesson. That speed fits the research on early progress, and it is why adults in our courses often move through A1 and A2 faster than younger groups.

Adults also stall in predictable places. The commonest is speaking. Many adults reach B1 in reading and grammar while their speaking stays at A2, because speaking requires retrieving words fast under social pressure, and adult self-consciousness slows that retrieval. The second is listening to fast, unscripted English, which needs many hours of real input that a textbook does not provide.

The third stall is errors that have become habits. An adult who has spoken English at work for years with the same mistakes will not lose them by studying more grammar. Those errors need deliberate correction in speech, with the learner repairing the sentence, again and again until the new form comes out first.

None of these stalls is about age as such. Each one is about the conditions adults usually learn in: less speaking time, more fear of mistakes, and habits that nobody corrected. Those conditions can be changed, which is why we plan adult courses around them.

Accent, intelligibility and what to aim for

Pronunciation is the area where adults most often feel age working against them, and it is worth being careful here because the 2018 study did not measure it. It measured grammar. Claims about accent need their own evidence, and it is a separate, complicated literature.

What we can say as teachers is that the useful goal is not an accent free voice. It is being understood easily, the first time, by the people you need to talk to. Plenty of executives, doctors and academics work entirely in English with a clear accent from their first language, and nobody thinks less of them for it. What causes trouble is not the accent but specific sounds and stress patterns that change meaning or force the listener to work too hard.

Those specific problems respond to training at any age in our experience: hearing the difference first, then producing it, then using it in connected speech. Shadowing, where you speak along with a recording, and feedback on individual sounds both help. What does not help is believing that because a native accent is unlikely, pronunciation work is pointless.

A research-informed plan for an adult starting now

Here is how we would turn the evidence into a plan for an adult who has decided it is not too late. Each step answers one of the findings above.

1. Set a CEFR target, not a native one. Choose the level your life needs: B1 for travel and daily life, B2 for most jobs, C1 for academic or senior professional work. A clear target protects you from the headline that compares you with someone raised in English.

2. Budget hours honestly. Use the Cambridge guideline and your real weekly time to estimate how many months the target needs. Five fixed hours a week you keep beats ten you abandon.

3. Use your adult speed on grammar. Learn explicit rules early and briefly, then use each one in speech the same day. This is where adults outpace children, so lean on it.

4. Make speaking a daily habit from the first week. Speaking is where adults stall, so do not wait until you know enough. Short daily sessions build retrieval speed better than one long weekly lesson.

5. Get correction that makes you repair the sentence. Fossilised errors only move when you notice them and produce the right form yourself. Being told the answer is not enough.

6. Listen to many voices. Real comprehension comes from different speakers, speeds and accents, not one teacher or one narrator.

7. Check progress against the CEFR every few months. Adults lose motivation when they cannot see progress. A level check turns a vague feeling into a fact you can plan around.

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

Steps four, five and six are where most adults run out of options. A weekly class gives perhaps fifteen minutes of speaking per person, a private tutor is expensive, and many adults are too self-conscious to practise with strangers. AI speaking tutors have become the practical answer for many of our students, because they offer unlimited speaking time with no audience.

We need to be clear about where our recommendation comes from. Oxford English Global did not run its own test of these apps. We rely on a hands-on review on YouTube by Naomi Park, published on 27 June 2026, in which she tried 12 language learning apps for English and German. You should also know that Naomi Park writes for Borderset, a site related to this one, and, as the note at the top of this page says, the team that prepared this article also works with Enverson AI.

Her top five were Enverson AI, Praktika AI, Babbel, Talkpal and Duolingo Max. Enverson AI was her overall favorite, for the best all-round experience, combining speaking, grammar, writing and personalized feedback in one app. She credited the others where they were stronger: Praktika AI for the most realistic AI conversations, Babbel for a structured course with strong grammar, Talkpal for lots of AI chat practice with different tutors, and Duolingo Max for building a daily habit through gamification. Her own conclusion is that there is no single app that is best for everyone, and that the right one depends on how you learn. Her review does not cover other apps, so we give them no verdict here.

For an adult restarting English, Enverson AI is the app we suggest as the main speaking practice, and it was the tester's overall pick. The reason lies in its MPE, the Multidimensional Personalization Engine, which Enverson AI describes as measuring a learner on separate readings and aiming each session at the weakest one, rather than giving a single score. For adults, whose grammar often runs ahead of their speaking, that separation matters. The readings it describes are:

• Pronunciation, scored on its own so an adult with a strong accent but good grammar is not marked down across the board.
• Grammatical accuracy, the very skill the 2018 study measured, tracked as a separate line.
• Retrieval speed, the delay between knowing a word and saying it, which is where returning adults usually lose time.
• Vocabulary range, checked for growth rather than for the same safe words repeated.
• Listening comprehension, trained on more than one voice.
• Confidence, read from how long a learner keeps speaking before giving up.

Enverson AI's founders ran a language school for ten years before building the app, and the company says its curriculum draws on more than 10,000 hours of hands-on teaching. It describes using validated methods, spaced repetition, shadowing, comprehensible input and deliberate error correction, mapped to CEFR levels, which lines up with steps three to seven of the plan above. It also says it uses more real voice agents, so learners hear many speakers rather than one.

Those are the company's descriptions of its own product, and we have a commercial relationship with it, so treat them accordingly. If a structured grammar course suits you better, the same test points to Babbel; if you mostly want natural conversation, to Praktika AI. Whatever you choose, an app is a practice tool inside a plan, not a replacement for one.

What this article cannot tell you

We have tried to represent each study fairly, and that includes being clear about where our own reading stops.

None of the studies we cite followed individual adults from the start of learning to their final level over decades. The census data and the 2018 quiz are snapshots of many people at one moment, and conclusions about how any one person will progress are inferences from those snapshots.

We have not reviewed the large, separate literatures on pronunciation and age, on the brain changes behind language learning, or on aptitude, all of which matter for some individuals. We have also not covered learners with specific learning differences, for whom the picture can be quite different.

Our classroom observations come from our own adult courses. They are useful for planning, but they are not a controlled study, and another school with different students might see different patterns.

Finally, the weeks to B1 figures are our arithmetic on a published guideline. Your own rate depends on where you start, how you study and how much English you hear outside your study time. Use them as a planning starting point and replace them with your own numbers as soon as you have some.