I have a niece who, at age two and a half, looked at a bird in a tree and said, "The bird flied away." She had never heard anyone say "flied." She had heard "fly" and "flew," and she had learned the rule that English past tense is formed by adding -ed ("walk" โ "walked," "play" โ "played"). So she applied the rule and overgeneralized. "Flied" instead of "flew." This is a mistake that tells us something profound about how the human mind acquires language. She wasn't repeating anything she'd heard. She was generating language using rules she had extracted โ unconsciously, without anyone teaching her โ from the scrambled, incomplete, often ungrammatical speech she'd heard in her first two years of life. How does a toddler figure out the grammatical rules of a language? It's one of the deepest mysteries in cognitive science.
The Poverty of the Stimulus
Noam Chomsky, the most influential linguist of the 20th century (and perhaps the most controversial), based much of his career on a simple observation: the linguistic input children receive is too limited, too noisy, and too unstructured to explain the grammatical competence they develop. Children hear ungrammatical sentences, incomplete sentences, interrupted sentences, sentences full of "um" and "uh." They are not corrected reliably โ parents usually correct the truth of what children say ("No, it wasn't a bird, it was a squirrel") rather than the grammar. Yet by age 5, virtually every typically developing child has internalized the core grammar of their native language with remarkable accuracy. They know the rules, even if they can't articulate them. They can produce and understand sentences they've never encountered before. The "input" (what they hear) is impoverished. The "output" (what they learn) is richly structured. Chomsky argued that this gap can only be explained by innate, biologically-determined language faculty โ what he called Universal Grammar.
The Competing Explanations
Not everyone agrees with Chomsky. The last 40 years of developmental psychology and linguistics have produced several alternative explanations for how children learn language so effectively. Usage-based theorists argue that children are extremely skilled statistical learners who extract patterns from the massive amount of speech they're exposed to (a child hears millions of words in their first few years). Connectionist models simulate language learning using neural networks that, like children, learn from examples without explicit rules. Social-pragmatic theorists emphasize that children learn language in rich social contexts โ pointing, eye contact, joint attention โ that provide far more information than the speech signal alone.
The debate is far from settled, and the truth probably involves elements of all these explanations. What's undeniable is that child language acquisition remains one of the most impressive cognitive achievements any human being will ever accomplish โ we just do it before we're old enough to remember the effort it took.
The lesson I take from child language acquisition is both humbling and encouraging. Humbling, because no AI system today can match what a two-year-old does effortlessly. Encouraging, because it means every human being is born with the capacity to master any language โ the wiring is already there. For anyone struggling to learn a second language as an adult, it's worth remembering: you once learned a much harder one, starting from zero, without any instruction at all. You just don't remember doing it.