HOW WE LEARN TO SPEAK: THEORIES OF ACQUISITION
Language acquisition is often described as a developmental marvel because it seems to happen quickly, reliably, and with little formal teaching. Within a few years, children move from single-word utterances to sentences with tense, agreement, and complex word order. Yet the apparent ease hides a serious puzzle sometimes called the logical problem of language acquisition: how do learners infer intricate grammatical systems from input that is incomplete, noisy, and full of false starts? Adults rarely provide explicit rules, and the speech children hear is shaped by context, emotion, and casual conversation rather than careful instruction. The central problem, therefore, is not whether children learn from experience—they clearly do—but what learning mechanisms could plausibly build such structured knowledge from messy evidence.
A prominent early answer came from behaviourism, associated most strongly with B.F. Skinner. In this framework, language is a kind of verbal behaviour learned the same way as other habits: through imitation, practice, and reinforcement. Children repeat forms they hear, and caregivers respond more positively to utterances that “work,” thereby strengthening those patterns. Positive reinforcement need not be formal praise; it can be attention, compliance with a request, or simply the continuation of an interaction. Behaviourism captured an important truth about learning in social environments: feedback and repetition matter. However, critics argued that imitation and reinforcement alone struggle to explain children’s creativity. Young learners routinely say things they have never heard—such as “I goed” or “two mans”—suggesting that they are not merely copying but constructing rules and then applying them broadly, even when the result is non-standard.
Noam Chomsky’s nativist critique turned this difficulty into a theoretical revolution. Chomsky argued that the input available to children is too limited to specify the grammar they ultimately master, an argument often summarised as the poverty of the stimulus. Learners receive relatively few explicit corrections, and they are rarely told which sentences are impossible; yet they converge on highly constrained grammatical systems. To explain this, nativist accounts proposed an innate Language Acquisition Device (LAD) and, more broadly, Universal Grammar: a set of built-in expectations about what human languages can look like. On this view, experience does not create grammar from scratch; it triggers and fine-tunes pre-existing options. Nativism does not claim that exposure is irrelevant—children must still hear a language—but it claims that biology supplies a structured starting point that makes rapid rule extraction feasible.
A different tradition, often labelled social interactionism, emphasises that language is learned in relationships rather than in isolation. Caregivers do not usually lecture children on syntax, but they do provide richly supportive contexts for meaning. Child-directed speech—sometimes nicknamed “motherese”—tends to have exaggerated rhythm, clearer intonation contours, and shorter clauses, all of which can make segmentation and attention easier. Jerome Bruner described how adults build scaffolding around a child’s attempts to communicate: they manage turn-taking, repeat key phrases, reformulate unclear utterances, and create predictable routines (such as book-reading or mealtime talk) where words map onto shared actions and intentions. Interactionist approaches argue that these structured exchanges are not mere background conditions; they actively shape what the child is able to notice, practise, and refine.
In the late twentieth century, researchers began to show that learners are also powerful pattern detectors. Work associated with Jenny Saffran demonstrated that even very young infants can use statistical learning to locate structure in continuous speech. In experimental settings, babies exposed to streams of syllables can track which syllables tend to co-occur and where transitions are least predictable—signals that often mark word boundaries. Crucially, this mechanism does not require explicit rule instruction or conscious hypothesis testing. Instead, the brain acts as a probabilistic learner, extracting regularities across many exposures. Statistical learning offers a bridge between “rule-based” and “experience-based” views: complex linguistic organisation can emerge from sensitivity to distributional patterns, provided the learner has sufficient exposure and memory to accumulate evidence.
Another long-running debate concerns time: are there biological windows during which language learning is especially efficient? Eric Lenneberg’s Critical Period Hypothesis proposed that the brain is particularly prepared for language in childhood and that acquisition becomes harder after maturation. Evidence often cited includes the general decline in ultimate attainment for late second-language learners and the severe outcomes when early linguistic input is absent. Importantly, the critical period claim is not that adults cannot learn languages at all, but that effortless, native-like mastery is less common because neural systems become less plastic and because adults rely more on explicit strategies that can interfere with automatic patterning. This perspective links language learning to broader developmental constraints rather than to a single learning mechanism.
Most contemporary accounts adopt a synthesis rather than a single-cause explanation. Biological preparedness may constrain what can be learned and when; social interaction may structure input and motivate communication; statistical learning may extract patterns that support vocabulary and grammar; and reinforcement may shape habits and strengthen successful forms. The modern view is sometimes summarised as “nature via nurture”: experience is essential, but it works through brain systems that are themselves shaped by evolution and development. The continuing challenge is to explain how these components interact—how attention, prediction, memory, social goals, and biological constraints combine to produce fluent language in real-world settings.