The Bitter End of Universal Grammar: Skinner–Chomsky Debate Today

(Skinner–Chomsky Debate Series, 1 part)

Abstract For more than half a century, the Skinner–Chomsky debate has been narrated as a decisive episode in the history of modern thought: B. F. Skinner’s behaviourist account of verbal behaviour was defeated by Noam Chomsky’s theory of an innate human capacity for language, and from this intellectual confrontation emerged the Cognitive Revolution. This article argues that the arrival of Large Language Models requires this familiar verdict to be reconsidered. Systems organised around statistical learning and next-token prediction now generate novel, grammatical and contextually sophisticated linguistic sequences without possessing a biologically inherited Universal Grammar. Their success does not demonstrate that Skinner’s account of human language acquisition was correct, nor does it establish that human brains operate like transformer architectures. It does, however, undermine one of the strongest intuitions historically associated with the Chomskyan revolution: that statistical learning from linguistic environments could never be sufficient to produce genuinely generative linguistic performance. Read through Thomas Kuhn’s model of scientific revolutions, the rise of LLMs can therefore be interpreted as an anomaly within a linguistic paradigm that once appeared secure. The significance of this technological development extends beyond linguistics. The original dispute between Skinner and Chomsky concerned fundamentally different images of the human being. Skinner sought to remove the explanatory homunculus from psychology and understand verbal behaviour through the interaction of biological endowment, environmental history and social contingencies. Chomsky relocated linguistic explanation within an internally structured organism endowed with species-specific capacities whose development could not be derived from linguistic experience alone. The resulting conflict concerned not merely how children acquire grammar but where human autonomy, creativity and individuality should be located. Although Chomsky’s theory is biological rather than religious, the cultural reception of linguistic innateness contributed to a conception of language as evidence of a privileged human interiority: something generated from within rather than assembled through histories of interaction with the world. LLMs disturb this distinction because they function as unexpected mirrors. Their fluent output demonstrates how readily prediction can appear as intention, statistical regularity as grammar, recombination as creativity and linguistic continuation as understanding. The disturbing possibility is consequently not that machines have finally acquired the mysterious linguistic essence once reserved for humans, but that linguistic performance may require less mystery than twentieth-century cognitive science imagined. In dialogue with Skinner’s concept of verbal behaviour, contemporary statistical-learning research, Elan Barenholtz’s account of language and LLMs, and Ayolov’s broader thesis concerning the planned obsolescence of language, this article proposes that the Skinner–Chomsky controversy has entered a new historical phase. The question is no longer simply whether Skinner or Chomsky won in 1959. It is whether the conceptual conditions under which Chomsky’s victory was declared still exist. The bitter end of Universal Grammar may therefore signify something larger than the decline of a linguistic hypothesis: the beginning of a Kuhnian transformation in which language ceases to appear as the privileged expression of an autonomous inner human being and begins to appear instead as an emergent, predictive and historically contingent technology whose relationship to Homo sapiens must be reconsidered. Keywords B. F. Skinner; Noam Chomsky; Universal Grammar; verbal behaviour; Large Language Models; LLMs; next-token prediction; statistical learning; language acquisition; poverty of the stimulus; Cognitive Revolution; Thomas Kuhn; scientific paradigms; Elan Barenholtz; autonomous man; linguistic innateness; generative linguistics; artificial intelligence; planned obsolescence of language; post-linguistic paradigm Introduction: The Debate That Was Supposed to Be Over Few intellectual confrontations of the twentieth century acquired as definitive an ending as the dispute between B. F. Skinner and Noam Chomsky. Skinner’s Verbal Behavior appeared in 1957 as an attempt to analyse language without appealing to an autonomous mental agent hidden behind speech. Verbal behaviour was to be explained through the biological organism, its environmental history, the verbal community surrounding it and the consequences of its actions. Chomsky’s famous 1959 review challenged this programme at its most vulnerable point: human language is generative. Children produce sentences they have never heard, develop grammatical competence from limited and imperfect input, and make systematic errors that cannot be explained as simple imitation. The poverty of the stimulus appeared to demand an internal explanation, and language was consequently relocated from Skinner’s history of behavioural relations towards a biologically endowed human capacity. The subsequent Cognitive Revolution transformed this theoretical victory into a historical verdict: Skinner had lost, Chomsky had won, and the mind had returned to psychology. Yet the simplicity of this story concealed the philosophical depth of the disagreement. Skinner never denied genetic endowment, just as Chomsky never proposed a supernatural soul. What divided them was the location of the causes required to explain human behaviour. Skinner attempted to remove what he called the homunculus, the autonomous internal agent supposedly responsible for choosing and initiating action. Chomsky insisted that environmental history was insufficient because the human organism contributed an internally structured biological inheritance to language acquisition. Beneath their disagreement about grammar therefore stood a much older question: what is a human being? Is language evidence of an autonomous interiority that unfolds from within, or can apparently spontaneous linguistic behaviour emerge from interactions among organisms, environments and historically accumulated verbal cultures? Chomsky’s revolution seemed to settle the question scientifically by demonstrating that linguistic creativity could not be derived from environmental regularities alone. In Thomas Kuhn’s terms, a new paradigm had displaced an older one, and once established, that paradigm determined not only the acceptable answers but increasingly the legitimate questions. More than six decades later, however, an anomaly has appeared that neither Skinner nor Chomsky could have anticipated. Large Language Models generate novel, grammatical and context-sensitive language through systems organised fundamentally around statistical learning and prediction. Next token follows next token, yet from this apparently mechanical process emerge translations, arguments, metaphors, dialogues and sentences that have never previously existed in precisely the same form. LLMs do not prove Skinner correct: transformer training is not operant conditioning, artificial models consume vastly more linguistic data than children, and machine performance tells us little by itself about the biological mechanisms of human language acquisition. Nevertheless, their existence damages one of the strongest intuitions associated with the historical Chomskyan victory: that statistical learning and prediction could never generate the open-ended structural complexity characteristic of language. This is where the debate becomes a Kuhnian problem again. Statistical-learning research in infants had already demonstrated that very young humans detect probabilistic regularities in linguistic input, while Universal Grammar itself evolved from comparatively rich accounts of innate linguistic structure towards increasingly minimal formulations. LLMs now turn these theoretical anomalies into an everyday technological spectacle. More importantly, as Elan Barenholtz’s mirror metaphor suggests, they reverse the direction of the question. Instead of asking why machines lack the mysterious depth supposedly expressed through human language, it becomes possible to ask why systems based upon prediction can reproduce so many external signs of that depth. The unsettling possibility explored in The Planned Obsolescence of Language and the wider Miscommunication project is therefore that language has been misunderstood precisely because humanity identified it too closely with itself. Language may be less an eternal expression of an innate human essence than a historically recent symbolic technology capable of organising cognition, reproducing through verbal communities and now migrating into artificial substrates. The question of 2026 is consequently not whether Skinner can simply be declared the retrospective winner of a battle fought in 1959. It is whether the conceptual conditions under which Chomsky’s victory was declared still survive. The machine has reopened the supposedly finished debate, and behind the technical question of Universal Grammar stands the more disturbing possibility that the old paradigm of language itself is approaching its bitter end. Part I: The Man Inside the Machine: Chomsky’s Revolution Against Skinner The conventional history of the Skinner–Chomsky debate begins with a simple opposition: Skinner treated language as learned behaviour, while Chomsky demonstrated that children possess an innate capacity that cannot be derived from imitation and reinforcement. This version captures part of the dispute but obscures Skinner’s more radical philosophical argument. Skinner did not deny genetic inheritance or imagine the newborn as an empty organism. He argued that behaviour should be explained through the interaction of biological endowment, individual environmental history and present circumstances without inserting another autonomous agent behind them. His objective was, in his own memorable formulation, to “de-homunculise” the human being. The person did not need another miniature person hidden inside the brain who perceived, decided and commanded the body. His distinction between language and verbal behaviour followed the same logic. Language was a verbal culture: a historically accumulated environment of practices and contingencies into which individuals were born. Verbal behaviour was what particular organisms learned to do within that environment. Chomsky reversed the direction of explanation. Children generate utterances they have never heard, extend grammatical patterns to unfamiliar cases and acquire linguistic competence despite limited and imperfect input. The poverty-of-the-stimulus argument therefore suggested that the environment could not contain everything necessary to explain the resulting knowledge. Some essential constraints had to come from the organism itself. Universal Grammar became the name for this inherited contribution, although Chomsky’s conception of it would change substantially over subsequent decades. The great intellectual movement of 1959 was therefore from outside towards inside: from the history of verbal behaviour towards the internal biological architecture that makes language possible. Behind this disagreement stood a much larger conflict over the nature of the human being. Skinner’s rejection of autonomous man seemed to threaten freedom, creativity, responsibility and dignity because behaviour became the product of causes extending beyond the conscious individual. Chomsky’s defence of internal structure restored limits to environmental determination: human beings could not simply be manufactured by manipulating their surroundings because biology already constrained what kinds of beings they could become. This was not a return to a supernatural soul. Chomsky’s position was explicitly naturalistic and biological. Yet it occupied a philosophical territory historically associated with the idea of privileged human interiority: something essential comes from within and cannot be completely explained by the external world. Language became especially powerful evidence because its apparent creativity seemed radically disproportionate to the experience from which it developed. Skinner, however, regarded precisely this autonomous inner author as an explanatory illusion. If a person acts because an internal self chooses to act, science must still explain what caused that self to choose. His environmentalism consequently produced an apparent paradox: determination could make change possible. If behaviour depends substantially upon environments, changing education, social conditions and reinforcement systems can change behaviour; invoking autonomous character can instead conceal the conditions that produced it. Chomsky feared the politically programmable human being, while Skinner feared the mythology of an autonomous individual used to disguise existing forms of control. Their argument over grammar was therefore already an argument about politics and human nature. The crucial question was never simply whether children learn sentences through reinforcement. It was whether linguistic creativity requires an autonomous internal source or whether the appearance of such a source can itself emerge from biological organisms acting within complex cultural environments. Chomsky appeared to win because generative language seemed to provide decisive evidence for the former. Six decades later, the appearance of generative language in machines without a human inner self makes that victory considerably less secure. Part II: The Victory That Became a Paradigm: From Universal Grammar to the Last Innate Rule Chomsky’s victory became much more than the defeat of one behaviourist theory. It helped establish a new paradigm in which serious explanations of language were expected to search beneath observable speech for internal rules, representations and biologically constrained structures. In Kuhn’s sense, the revolution became normal science: generations of linguists were educated within the conceptual world created by generative grammar, while Skinner increasingly survived as the representative of an obsolete stimulus-response psychology. Yet the history was less straightforward. Kenneth MacCorquodale argued that Chomsky had attacked a simplified version of Verbal Behavior, particularly by treating reinforcement as if every utterance required explicit reward and Skinner’s functional analysis as if it were a mechanical chain of stimuli and responses. More importantly, the victorious theory itself continued to change. Universal Grammar was never a single immutable hypothesis. Richer conceptions of innate grammatical structure were progressively revised through transformational grammar, Principles and Parameters and eventually the Minimalist Program, where the specifically linguistic inheritance became increasingly economical and the elementary combinatorial operation Merge acquired exceptional importance. This theoretical evolution can be interpreted as scientific progress towards deeper simplicity. But it permits another Kuhnian interpretation as well: the distinctive innate content of the original paradigm became progressively thinner as linguistic diversity, acquisition research and alternative learning mechanisms complicated the older picture. The controversies surrounding Daniel Everett and Pirahã illustrate the problem. Claims that recursion might not appear in every language did not simply falsify Universal Grammar; instead they generated disputes about what recursion meant, whether the underlying capacity could exist without overt expression and which level of linguistic organisation should count as universal. Such responses may be theoretically legitimate, but they expose a fundamental problem for any mature paradigm: if apparent counterexamples can continually be relocated to deeper and less observable levels, determining what would genuinely falsify the theory becomes increasingly difficult. At the same time, the supposedly impoverished linguistic environment began to appear richer than the original opposition between behaviourism and nativism suggested. Research on statistical learning showed that infants can detect probabilistic regularities among linguistic sounds before they themselves speak. Novel behaviour therefore does not require that every novel response has previously been reinforced; organisms capable of extracting patterns can generalise beyond individual examples. This does not prove Skinner correct or eliminate biological constraints, but it expands the territory between the two old extremes. A child need not be either a blank slate copying rewarded utterances or the possessor of a richly specified innate grammar. Statistical learning, general cognition, embodiment, social interaction and cultural transmission may perform considerably more explanatory work than the twentieth-century debate initially allowed. Skinner’s description of language as a verbal culture becomes unexpectedly relevant here: individuals enter an already structured environment containing patterns accumulated across generations, and their linguistic repertoires develop through prolonged contact with that environment. The deeper historical irony is therefore that Chomsky’s revolution may have succeeded too completely. A revolutionary programme became the framework through which its defeated opponent was remembered, while its own changing assumptions became part of normal science. This does not justify calling Chomsky personally a dictator, but it does make his extraordinary intellectual authority relevant to Kuhn’s sociology of paradigms: successful theories organise disciplines and determine what later generations regard as plausible. By the twenty-first century, however, the old monopoly had begun to weaken. Universal Grammar increasingly confronted a question that could no longer be dismissed as behaviourist naïveté: how much linguistic structure actually has to be innate if powerful learners can extract complex regularities from structured environments? The decisive challenge would soon arrive from outside linguistics. For the first time, statistical systems exposed to an immense human verbal culture began generating fluent language themselves. What had once been largely a theoretical argument about the limits of learning was about to become an engineering demonstration. Conclusion: The Bitter End of Universal Grammar The Skinner–Chomsky debate can no longer be closed with the comfortable twentieth-century verdict that Chomsky defeated behaviourism and permanently demonstrated the necessity of Universal Grammar. History was more complicated, and the arrival of LLMs has made that complexity impossible to ignore. Skinner’s Verbal Behavior did not provide a complete theory of language, and contemporary machine learning does not retroactively validate every proposition of radical behaviourism. Human children remain vastly more data-efficient than current artificial systems, human cognition is embodied and socially situated, and nothing about next-token prediction proves that biological language acquisition operates through the same architecture. Yet the burden of proof has changed. Generativity, novel sentence production and grammatical complexity can no longer function as sufficient evidence for a specialised innate grammar, because artificial systems without a human biological language faculty now display all three. The strongest version of the old argument — statistics cannot become syntax, prediction cannot become creativity, exposure cannot produce open-ended linguistic behaviour — has encountered an empirical counterexample of extraordinary scale. Universal Grammar may survive in weaker, more minimal or differently formulated forms, but it can no longer claim the explanatory monopoly established during the Cognitive Revolution. In Kuhnian terms, this is what makes the present moment significant. The paradigm does not collapse because one experiment falsifies it; it begins to lose authority when phenomena once regarded as impossible become ordinary and when questions excluded by the reigning framework return as legitimate scientific problems. Skinner’s ghost has returned not because psychology is going back to 1957, but because the conceptual boundary that made his defeat appear final has dissolved. The bitter end of Universal Grammar is therefore not primarily the end of a technical linguistic hypothesis. It may mark the beginning of a different conception of language and, with it, a different conception of the human being. For centuries language was treated as one of the strongest signs of human exceptionalism: the expression of reason, consciousness, creativity, spirit or an autonomous interior self. The Chomskyan revolution secularised rather than invented this intuition by relocating linguistic uniqueness within biological architecture. LLMs introduce a disturbing alternative. If linguistic performance can migrate from biological organisms into artificial systems, then language itself may be better understood as a transferable cultural technology rather than the exclusive signature of an inner human essence. This is the larger hypothesis developed through the Miscommunication Trilogy and The Planned Obsolescence of Language: beneath the linguistic self lies a far older organism whose perception, affect, bodily regulation and non-verbal interaction preceded the historically recent world of words. Language entered that organism from a culture already in existence, became one of the principal technologies through which humans organise themselves and their societies, and now begins to reproduce through machines as well. The mirror held up by LLMs is uncomfortable precisely because it does not reveal a soul awakening inside artificial intelligence. It reveals how much apparently meaningful linguistic behaviour can exist without one. The question left by 1959 must therefore be reversed. Instead of asking what hidden linguistic structure must exist inside a being capable of producing language, it is now necessary to ask how much of the hidden being was inferred merely because language was produced. Once that question becomes legitimate, the Skinner–Chomsky debate ceases to be a historical dispute and becomes the opening confrontation of a new paradigm. And every paradigm shift produces another question of power: who defines the new truth, who controls its institutions, and what happens when a revolutionary theory becomes an orthodoxy of its own? That is where the next article "A Case Against the Chomskyan Linguistic Cult" begins. References Barenholtz, E. (2025) ‘Large language models as mirrors of human language and cognition’. Florida Atlantic University, Center for the Future Mind. Available through the Center for the Future Mind research and public lecture materials. Chomsky, N. (1959) ‘A review of B. F. Skinner’s Verbal Behavior’, Language, 35(1), pp. 26–58. Chomsky, N. (1965) Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, N. (1971) ‘The case against B. F. Skinner’, The New York Review of Books, 30 December. Chomsky, N. (1980) Rules and Representations. New York: Columbia University Press. Chomsky, N. (1995) The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, N., Roberts, I. and Watumull, J. (2023) ‘The false promise of ChatGPT’, The New York Times, 8 March. Everett, D.L. (2005) ‘Cultural constraints on grammar and cognition in Pirahã: Another look at the design features of human language’, Current Anthropology, 46(4), pp. 621–646. Hauser, M.D., Chomsky, N. and Fitch, W.T. (2002) ‘The faculty of language: What is it, who has it, and how did it evolve?’, Science, 298(5598), pp. 1569–1579. Kuhn, T.S. (1962) The Structure of Scientific Revolutions. Chicago: University of Chicago Press. MacCorquodale, K. (1970) ‘On Chomsky’s review of Skinner’s Verbal Behavior’, Journal of the Experimental Analysis of Behavior, 13(1), pp. 83–99. Marcus, G.F., Vijayan, S., Bandi Rao, S. and Vishton, P.M. (1999) ‘Rule learning by seven-month-old infants’, Science, 283(5398), pp. 77–80. Pinker, S. (1994) The Language Instinct: How the Mind Creates Language. New York: William Morrow. Saffran, J.R., Aslin, R.N. and Newport, E.L. (1996) ‘Statistical learning by 8-month-old infants’, Science, 274(5294), pp. 1926–1928. Sapolsky, R.M. (2010) Human Behavioral Biology. Stanford University lecture series. Stanford, CA: Stanford University. Skinner, B.F. (1953) Science and Human Behavior. New York: Macmillan. Skinner, B.F. (1957) Verbal Behavior. New York: Appleton-Century-Crofts. Skinner, B.F. (1971) Beyond Freedom and Dignity. New York: Alfred A. Knopf. Skinner, B.F. (1972) ‘Humanism and behaviorism’, The Humanist, 32(4), pp. 18–20. Skinner, B.F. (1974) About Behaviorism. New York: Alfred A. Knopf. Skinner, B.F. (1981) ‘Selection by consequences’, Science, 213(4507), pp. 501–504. Skinner, B.F. and Segal, E.F. (1988) Discussion of verbal behavior. Interview conducted at Harvard University, February 1988. Tomasello, M. (2003) Constructing a Language: A Usage-Based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A.N., Kaiser, Ł. and Polosukhin, I. (2017) ‘Attention is all you need’, Advances in Neural Information Processing Systems, 30, pp. 5998–6008.