How to Learn a Language as an Adult

What the critical period research actually shows

The strong version of the critical period hypothesis — that language acquisition shuts down at puberty — has not survived scrutiny. But something real sits underneath it.

The most statistically powerful study on the question is Hartshorne, Tenenbaum and Pinker (2018) in Cognition, which used an online grammar quiz that went viral to gather data from roughly 670,000 participants — an order of magnitude beyond anything previously possible in this field.

Their findings were more specific than either side of the popular debate:

• The ability to learn grammar remains close to constant until around age 17–18, then declines.
• Reaching fully native-like grammatical mastery appears to require beginning by roughly age 10, because full acquisition takes about a decade and the learning window closes in the late teens.
• Learning after that point is entirely possible; complete native-like mastery of the subtlest grammatical judgements is what becomes unlikely.

There is a second, well-replicated finding that cuts the other way and is rarely mentioned: in the early stages, adults and older children learn faster than young children. Adults have metalinguistic awareness, can be told a rule and apply it, have vastly larger conceptual vocabularies to attach words to, and can study deliberately. A motivated adult will typically outpace a five-year-old for the first year or two.

Where age genuinely bites hardest is pronunciation. The phonological system appears to be the most age-sensitive component, and adult learners rarely become indistinguishable from native speakers.

That matters less than people fear. The realistic target is intelligibility, not nativeness, and intelligibility is achievable at any age. Nobody has ever failed to be understood because of a slight accent; plenty of people have failed because they never practised speaking.

How many words you need

Vocabulary is the single strongest predictor of comprehension, and the useful thing about it is that word frequency in every language is extraordinarily skewed — a small number of words does a huge amount of work.

Research by Paul Nation and colleagues, examining the vocabulary size needed for adequate comprehension, gives approximate coverage figures:

Word families known	Approximate coverage
1,000	~75–80% of everyday spoken language
2,000	~90%+ of spoken discourse
3,000	~95% of spoken discourse
5,000	~95% of written text
8,000–9,000	~98% of written text — the level for comfortable unsimplified reading

Two conclusions follow directly.

The first 2,000 words are worth vastly more per word than any that follow. Learning them in frequency order is one of the highest-leverage decisions available. A frequency list for your target language is a better starting curriculum than any textbook's arbitrary thematic order.

The gap between 95% and 98% coverage is enormous in practice. At 95% you meet an unknown word every twentieth word — roughly one per line — which makes reading exhausting and guessing from context unreliable. At 98% it is one per fifty, and reading becomes pleasant enough to sustain. This is why intermediate learners so often stall: they can technically read, but it is not yet enjoyable, so they stop doing the thing that would carry them further.

How long it takes

The most credible public estimates come from the US Foreign Service Institute, which has trained diplomats intensively for decades and tracks how long it takes to reach professional working proficiency. Their categories, for native English speakers:

Category	Approximate class hours	Examples
I	600–750	Spanish, French, Italian, Portuguese, Dutch, Norwegian
II	~900	German, Indonesian, Swahili
III	~1,100	Russian, Hebrew, Hindi, Polish, Turkish, Vietnamese
IV	~2,200	Arabic, Mandarin, Cantonese, Japanese, Korean

Read these carefully, because they are routinely misquoted. They are class hours with expert instructors, small groups, full-time immersion in a structured programme, and substantial homework on top. They are not total hours, and they are certainly not what an evening learner should expect.

What they measure accurately is relative difficulty — and the variable is distance from what you already know, not intrinsic complexity. Mandarin is not a harder language than English; it is harder for an English speaker, because it shares almost no vocabulary, no writing system and no grammatical structure. Speakers of a related language have a large head start that has nothing to do with talent.

The realistic version for a part-time adult learner: a Category I language at an hour a day, done well, is a project of roughly two to three years to comfortable conversational ability. Category IV is considerably longer. Anyone promising fluency in three months is selling something.

A method that works

Four components, in rough order of how much they contribute.

1. Comprehensible input, in volume

The single largest predictor of progress is how much of the language you take in at a level you can mostly understand. Stephen Krashen's input hypothesis — that acquisition comes from understanding messages slightly above your current level — is contested in its strong form, but the practical evidence for extensive reading and listening is solid.

What matters is the "mostly understand" part. Input far above your level is noise, and input far below it teaches nothing. Aim for material where you follow the gist and meet unknown words at a rate you can tolerate.

This is where most adult learners under-invest by an enormous margin. Hundreds of hours of listening are required and there is no shortcut, which is why passive listening while commuting, cooking or exercising is so valuable — it is the only realistic way to accumulate that volume.

Graded readers, learner podcasts, dubbed shows you already know, and children's content all work well early on. Move to native material as soon as it is tolerable.

2. Deliberate vocabulary work, spaced

Input alone builds vocabulary slowly, because you need many encounters with a word before it sticks and low-frequency words appear rarely. Spaced repetition solves this precisely.

Practical guidance: learn words in context, not as isolated pairs — put a whole sentence on the card, so you acquire usage and grammar alongside meaning. Start from a frequency list. And add fewer cards than you want to; reviews compound, and an unmanageable daily queue is the most common reason people quit.

3. Speaking, earlier than feels comfortable

Producing language is a distinct skill from understanding it, and it does not develop for free. Retrieval under time pressure — which is what a conversation is — is exactly the effortful recall that builds durable access.

The obstacle is almost never ability. It is the discomfort of being bad at something in front of another person. The only fix is to start before you feel ready, since the feeling of readiness does not arrive on its own.

Tutoring marketplaces have made this cheap and low-stakes. A weekly conversation hour is worth more than several extra hours of app drilling, and it exposes the gaps that passive study conceals.

4. Enough grammar, and not more

Adults have a real advantage here and often overuse it. Being told that a case system exists and what it does saves months of confusion — that is worth doing.

But grammar study has steeply diminishing returns, and it is comfortable in the way re-reading is comfortable: it feels like studying and produces no fluency. Learn the structural outline early, then return to specific points when input makes you notice you do not understand something. Grammar is best used as a reference, not a syllabus.

Why intermediate is where people quit

Beginner progress is exhilarating because everything is new and the frequency curve is doing the work. Then it flattens, and the flat part is long.

Three things are happening at once. The words are rarer, so each new one appears less often and pays off less. Progress becomes invisible, since going from understanding 80% to 90% of a conversation does not feel like the leap from 0% to 50%, though it is a bigger achievement. And the material is still not fun — comprehension is not yet high enough for native content to be enjoyable, so motivation has to be supplied externally rather than by the content itself.

The way through is mostly structural rather than motivational. Increase input volume and accept partial comprehension, because the 95%-to-98% gap only closes with hours. Find content you would want in your own language, since a subject you care about makes the unknown words worth pushing through. Track something other than feeling — words known, hours listened, conversations had — because your sense of progress is the least reliable instrument you have at this stage. And keep speaking, because it is the only part that reliably still feels like it is improving.

Further reading: Hartshorne, Tenenbaum and Pinker, A critical period for second language acquisition, Cognition (2018). The vocabulary coverage research is Nation, How large a vocabulary is needed for reading and listening?, Canadian Modern Language Review (2006). The FSI difficulty rankings are published by the Foreign Service Institute.