Critical Periods in Language

A baby who cannot say “milk” can already lose the ability to hear a sound contrast from another language. By roughly 10 to 12 months, infants start tuning their ears toward the phoneme boundaries of the language around them. The trade is brutal: the brain gets better at the sounds it needs, and worse at sounds it does not hear.

How it works

Eric Lenneberg gave the modern idea its spine in Biological Foundations of Language in 1967. His claim was not that adults cannot learn languages. It was sharper: some parts of language learning, especially native-like acquisition, depend on a maturational window.

The cleanest evidence comes from phonemes. In 1984, Janet Werker and Richard Tees tested infants on sound contrasts that matter in Hindi and Salish but not in English. English-learning infants around 6 to 8 months could still discriminate those contrasts. By 10 to 12 months, many no longer did.

That is not decay in the simple sense. It is compression. The infant brain groups speech sounds into useful buckets, then routes future hearing through those buckets.

The case in numbers

Jacqueline Johnson and Elissa Newport’s 1989 study tested Korean and Chinese speakers who arrived in the United States between ages 3 and 39. Performance on English grammar declined with later arrival during childhood, then became much more variable after puberty.

Hartshorne, Tenenbaum, and Pinker scaled the question in 2018 with data from 669,498 people who took an online grammar test. Their model estimated that grammar-learning ability stays high until about age 17.4, then declines. That result softened the simple “puberty wall” story, but it did not erase the age effect.

Layer of language	Early advantage	Adult path
Phoneme hearing	Starts shifting by 10-12 months	Training can help, but L1 filters persist
Accent	Strong childhood advantage	Often fossilizes after years
Grammar	High through childhood, maybe into late teens	Learnable, but slower and variable
Vocabulary	No hard early cutoff	Adults can build large vocabularies

The sharp line is not “children can, adults cannot.” The line is cost. Children spend years inside the signal. Adults must notice the signal while already hearing it through a prior map.

What's contested

Researchers still argue over whether language has one critical period or several sensitive periods. Phonology, syntax, morphology, and vocabulary do not age at the same speed. A 35-year-old can learn 5,000 new words; the same learner may still miss a vowel contrast that a 9-month-old could hear.

The other fight is mechanism. Is the decline caused by brain maturation, less input, classroom methods, identity, motivation, or interference from the first language? The honest answer is plural. Biology sets the slope, but exposure quality and social need change the curve.

Why this has to do with other realms

Critical periods are an information problem before they are a language problem. The child’s brain has to reduce a noisy acoustic stream into categories. That links this page to concept information theory: compression creates power, but compression also throws away recoverable detail.

It also rhymes with concept transformers. A language model trained on one distribution carries that distribution into every future prediction. Humans are not transformers, but both systems show the same warning: early data shapes later perception.

Abhishek's take

The part that grabs me is not that children learn faster. It is that they learn by deleting distinctions. A good mind is not an empty container; it is a filter, and every useful filter has a cost.

Where I've used this

I see this in the tools I write for the buying floor: early labels decide what later models can notice. If the first taxonomy is lazy, the system keeps hearing new products through old buckets.

Tags: #language #cognition #childhood #learning #phonetics #neuroscience

Key sources

Lenneberg, Eric H. Biological Foundations of Language (1967) - the canonical critical-period argument in modern linguistics.
Werker, Janet F., and Richard C. Tees (1984), “Cross-language speech perception: Evidence for perceptual reorganization during the first year of life” - the infant phoneme-tuning result.
Johnson, Jacqueline S., and Elissa L. Newport (1989), “Critical period effects in second language learning” - age-of-arrival evidence for grammar outcomes.
Hartshorne, Joshua K., Joshua B. Tenenbaum, and Steven Pinker (2018), “A critical period for second language acquisition” - large-scale grammar data with 669,498 participants.

Abhishek