The critical-period myth: what 50 years of research actually says about adult language learning

It's the most quoted finding in popular language-learning writing, and it's also the most often misquoted. Children learn languages effortlessly; adults can't. The window closes at puberty. The proof is supposedly Lenneberg's 1967 monograph — except very few people who repeat the claim have read it.

What Eric Lenneberg actually proposed in Biological Foundations of Language was narrower: that native-like phonological mastery of a second language was difficult to attain after roughly age twelve. He did not claim adults could not learn languages. He did not claim grammar and vocabulary become inaccessible. He flagged one specific domain — accent — where biological maturation seemed to set a ceiling.

Half a century of replication has refined that picture in ways that matter for anyone learning a language at thirty-five.

1. What survives replication

Adult learners reliably retain detectable foreign accents, even after twenty years of immersion (Flege, Yeni-Komshian, & Liu, 1999). This part of Lenneberg's claim has held up.

Almost everything else has been narrowed.

A massive 2018 study from MIT collected English-grammar data from 670,000 native and non-native speakers — the largest sample of its kind. Hartshorne, Tenenbaum, and Pinker's central finding: there is no sudden cliff at puberty for grammar acquisition. There is a gradual decline, but it doesn't accelerate until the late teens — meaning a learner starting at fifteen has roughly the same ultimate ceiling as one starting at three. The drop-off only becomes steep after about seventeen, and the curve flattens but never goes to zero (Hartshorne et al., 2018).

2. What changes, and what doesn't

For adults beginning a second language after twenty, three things appear genuinely different:

Phonological detail is harder. The motor patterns required for native-like accents are difficult to retrain. Many adults do reach near-native phonology; most don't.

Implicit pattern induction is partially attenuated. Children pick up grammar rules from exposure without conscious instruction. Adults do this much less efficiently and benefit considerably from explicit teaching (DeKeyser, 2000).

Working-memory load on novel structures increases. Adult learners juggle more competing cognitive demands and have less spare capacity to absorb language passively.

These are real changes. They are also strikingly specific. They don't add up to "adults can't learn."

3. What adults have that children don't

Almost no popular language-learning content discusses this. Adults bring substantial assets to the process:

• Strategic awareness. An adult can recognize they are stuck and switch methods. A six-year-old can't.
• Domain knowledge. A doctor learning Spanish already knows ten thousand medical concepts. They need labels, not concepts.
• Metalinguistic awareness. Adults already have a first language whose structure they can compare against.
• Choice. Adults can engineer their input. They can pick podcasts, novels, conversation partners. They can decide when and how to be uncomfortable.

Bialystok and Hakuta's 1999 reanalysis of the immigrant literature found that when learning conditions are held equal — motivation, hours of practice, instructional quality — the age effect shrinks substantially. The advantage children show in observational studies is partly an artifact of how children are embedded in learning environments, not of their brains.

4. The practical implication

The "I'm too old" story isn't supported by the evidence in the form it's usually told. What is supported: adults benefit from instruction explicit enough to compensate for slower implicit pattern-induction, and from practice contexts dense enough to substitute for the immersion children get for free.

Not impossible. Not natural. Specifically engineered.