Home > Positioning > Subjects > Nativism

Nativism

Nativism is the position that significant structure of the mind is present in advance of experience rather than drawn from it — that a learner brings to the world more than the world could have given it. As a doctrine about knowledge it is ancient; as a research programme it is modern, and its centre of gravity is language, where the claim takes its sharpest and most defended form: that a child could not acquire a language from experience alone, and that acquisition therefore rests on an innate endowment specific to language. Founded in that form by Noam Chomsky in the late 1950s and dominant in theoretical linguistics for decades, it turned the study of language toward the mind and made the child’s achievement — a complex grammar, acquired fast, from fragmentary evidence, without instruction — the central explanandum of the field. It remains live, defended, and contested, and its internal disagreements are as sharp as those between it and its critics. This page is chiefly about the linguistic programme; it marks the older lineage and the cognitive variants where they touch it.

The modern programme inherits an argument older than psychology — Plato’s slave boy in the Meno, Descartes’s innate ideas, Leibniz’s reply to Locke, Kant’s categories — the recurring shape being that experience underdetermines what the mind ends up with, so the mind must bring the difference. Chomsky claimed that rationalist lineage explicitly in Cartesian Linguistics (1966), a claim historians of the period have disputed as read back into its sources.

The vernacular version — “language is innate” — is not what anyone defends. No nativist claims a language is present at birth. The careful claim turns on a single word: a developed linguistic trait is innate if it cannot be explained by the learner’s sensitivity to the environment, but must be explained by a language-specific learning system. Domain-specificity is the whole dispute. Everyone grants that some innate structure is at work — a rock does not learn Dutch, and nativism’s most prominent opponents are explicit that infants bring a great deal to the task. What divides the field is whether what does the work is dedicated to language or general-purpose. Arguments that some endowment is needed do not by themselves establish nativism, and its proponents say so.

The poverty of the stimulus

The argument at the centre. Children come to know things about their language that domain-general learning could not have got from the limited and imperfect data they encounter, so something innate and specific to language must bridge the gap. The supporting observations are that children receive positive evidence only, that correction is rare and ignored, that different children on different data converge on the same grammar, and that no other organism in the same environment acquires it.

The cases the argument is built on each mark a place where a child’s knowledge appears not to be in the evidence: structure-dependence (the flagship — questions move the structurally correct auxiliary, never the linearly first, though the input is consistent with both rules), binding, the anaphoric one, island effects, phonological natural classes, and the fixing of word meaning. The burden the tradition acknowledges is threefold — establishing what adults know, what children encounter, and why general learning could not bridge the two — with a corresponding burden on critics to answer it for every phenomenon, not one.

A second, more concrete line of evidence turns on structure appearing where it was demonstrably not in the input at all: creoles arising within a generation from grammarless pidgins (Derek Bickerton’s language bioprogram), Nicaraguan Sign Language grammaticalising over successive cohorts of deaf children with no shared language, and the homesign systems deaf children invent without a model (Susan Goldin-Meadow). The counter-reading holds these draw on general cognition and interactional pressure rather than a bioprogram, and that cohort effects show cultural transmission at work.

What is claimed to be innate

Nativism’s answer to what is innate has moved in one direction across sixty years: downward. The rich innate rule schemas of transformational grammar (Aspects, 1965) gave way to the universal principles and binary switch-settings of Principles and Parameters (1981), and those to the Minimalist Program (from 1993), which strips the language-specific component toward a minimum — in the limit to Merge alone, a single operation combining two objects into a hierarchical set, with meaning and sound pushed outward onto the interfaces. The motive is not disputed within the programme: Universal Grammar must be rich enough to overcome the poverty of the stimulus and restricted enough to have plausibly evolved, and those goals pull opposite ways. Whether the reduction succeeds — whether what minimalism offloads has really been derived elsewhere — is contested, as is the syntax-first architecture itself, which Ray Jackendoff’s parallel architecture rejects in favour of several independent generative systems linked by interfaces.

Concepts

Merge combines objects; it does not supply them, and where the objects come from is the least settled part of the programme. That the words a child ends up with come from the environment is not in dispute (Leeds gives you cat, Ljubljana mačka); the mainstream nativist account of word-learning is accordingly about constraints — whole-object and taxonomic biases, mutual exclusivity, syntactic bootstrapping — not innate vocabulary. Chomsky’s own position is stronger and an outlier: lexical concepts are too complex to have been extracted from experience and so must in substance already be there, triggered rather than learned — extended, he says, even to carburetor. Jerry Fodor took this to its limit with the language of thought: thinking occurs in an innate, compositional Mentalese, concepts cannot be learned but only triggered, and so car is innate — a conclusion widely received (Putnam, Dennett) as the reductio of the position. Core knowledge (Elizabeth Spelke, Susan Carey) is the main developmental alternative and is nativist about cognition rather than language: a small set of evolutionarily old core systems (objects, agents, number, space, social partners), with language the mechanism that combines them into uniquely human concepts — the exact inverse of Fodor, for whom language externalises a conceptual system already complete. Against all of these stand accounts that deny thought is language-like at all — associationist and image-based representation (Prinz, Barsalou), connectionism, and Patricia Churchland’s eliminativism.

How the endowment arrived

If the faculty is biological it evolved, and nativists divide internally on how — the disagreement running among people who all accept the endowment. Chomsky and Robert Berwick argue for a sudden, minimal, recent origin: since the faculty reduces to Merge, one chance rewiring could deliver it, with no precursors, and language evolved as an instrument of thought rather than for communication. Pinker and Bloom argue the reverse from within nativism: grammar is complex functional design, and gradual natural selection for communication is the only known explanation for such design. The recursion-only hypothesis — Hauser, Chomsky and Fitch’s proposal that the uniquely human residue (FLN) might be recursion alone — is the live fault line, attacked by Pinker and Jackendoff and defended in reply, and it remains open. Bearing on the dispute from outside nativism are accounts that relocate the structure out of the genome: Terrence Deacon’s co-evolution of language and brain through niche construction, and Christiansen and Chater’s inversion, on which language evolved to be learnable rather than brains to learn language.

The challenges

That the stimulus is not that poor. Pullum and Scholz questioned whether the input is impoverished as claimed; corpus work has been offered to show the relevant patterns present more often than the argument allows. Berwick, Chomsky and Piattelli-Palmarini replied that the poverty of the stimulus stands.

That general learning suffices. Statistical learning, connectionist models, Bayesian inference, and construction grammar have each been proposed as domain-general mechanisms adequate to the task the argument reserves for a dedicated faculty.

That the endowment is not language-specific. Tomasello’s usage-based account is the sharpest form: it grants substantial innate structure — intention-reading, joint attention, shared intentionality — and denies any of it is dedicated to language, building grammar from use. Construction and cognitive grammar (Goldberg, Langacker, Lakoff) contest the autonomy of syntax alongside it.

That the universals are not there. The typological attack. Daniel Everett argued from fieldwork that Pirahã lacks recursion, against the recursion-as-universal claim, in a protracted and unresolved dispute; Evans and Levinson’s “The Myth of Language Universals” (2009) argued more broadly that the proposed universals dissolve under typological diversity. Generativists reply that Universal Grammar constrains the space of possible grammars, not the observable inventory.

Large language models. The current front. Piantadosi argues that models acquiring syntactic competence from exposure alone undermine the innateness claim; the standard reply is that they train on orders of magnitude more data than a child, and so do not touch a claim about the poverty of the child’s input. The exchange is ongoing.

Further objections are on the record without being developed here: that the adult competence attributed to speakers is mis-described; the circularity worry (Lasnik) that picking out linguistic units already presupposes linguistic categories; the explanatory-emptiness charge (Paul Churchland) that calling a trait innate names no mechanism; that development is gradual where parameter-setting predicts switch-flips; and that morphology and semantics track input frequency in the way empiricist learning predicts.

Key works and further reading

See also: Chomsky · Tomasello · Piaget · Behaviourism · Descartes · Putnam · Dennett · Quine