Recursion and the Pirahã Debate: Can One Language Break a Universal?
What Daniel Everett's fieldwork revealed, and what it did not
My articles are free for everyone to read, engage with and share with others
Take the sentence the woman saw the man. Now extend it: the woman said that the man believed that the child had left. The same basic operation, repeated several times, produces a sentence of greater complexity. Grammatically, there is no fixed ceiling. A finite vocabulary and a finite set of rules can generate an indefinitely large number of sentences.
But grammar and processing are not the same thing. A grammar may permit indefinitely deep embedding in principle, while human comprehension fails much sooner in practice. This is clearest with centre-embedded relative clauses, where one sentence is placed inside another before the first one has finished: The dog that the cat chased ran away is manageable; The dog that the cat that the boy saw chased ran away is already hard. Push the structure further and comprehension usually collapses.
This property is called recursion. More precisely, it is the capacity to apply a rule to its own output, producing structures that can in principle extend without a fixed upper bound. In 2002, Marc Hauser, Noam Chomsky and W. Tecumseh Fitch proposed in Science that recursion may be the only component of the narrow faculty of language that is uniquely human.
That was a strong claim. In 2005, Daniel Everett’s work on Pirahã turned it into a public controversy.
A Language That Does Not Embed
Daniel Everett is a linguist and anthropologist who spent decades with the Pirahã, a community of approximately 300 to 400 people living along the Maici river in the Brazilian Amazon. He arrived originally as a missionary. He left, eventually, as a convinced materialist. What he documented in between became one of the most debated datasets in modern linguistics.
His 2005 paper in Current Anthropology argued that Pirahã lacks recursive syntactic embedding. There are no centre-embedded clauses. Sentences are not inserted inside other sentences. There are no relative clauses of the kind that allow one proposition to be subordinated within another.
Everett’s explanation was cultural rather than grammatical. The Pirahã, he argued, operate according to an immediacy of experience principle: the culture restricts assertions to events that are directly witnessed by the speaker or by a living member of the community. Embedded clauses about reported speech, hearsay, and events beyond direct experience are culturally suppressed. The grammar, on this account, reflects the epistemology.
His paper also claimed that Pirahã lacks number words, has no perfect tense, and uses only relative brightness distinctions rather than fixed colour terms. These claims generated their own debates. But the recursion claim was the one that struck directly at Hauser, Chomsky, and Fitch.
Three Separate Questions
The dispute that followed is routinely treated as a single argument. It contains at least three distinct questions, and conflating them is where most of the confusion enters.
The first question is descriptive. Does Pirahã lack the familiar surface forms of recursive clausal embedding found in languages such as English? On this narrower point, Everett’s claim is strong. His Pirahã data do not show centre-embedded clauses, relative clauses, or complement clauses of the kind usually used to illustrate syntactic recursion..
The second question is analytical. Does that surface absence mean recursion is absent from the grammar as a whole? This is a much harder question. Nevins, Pesetsky and Rodrigues argued in 2009 that Everett’s analysis was not the only possible account of the data. They challenged several details of his description and argued that recursive structure might still be present in less obvious parts of the grammar. The question is therefore not simply whether Pirahã has English-style embedding. It is whether recursion is absent from all relevant grammatical levels.
The third question is theoretical. If Pirahã really lacks recursive clausal embedding, does that refute Universal Grammar? Not automatically. Hauser, Chomsky and Fitch distinguished between the broad faculty of language and the narrow faculty of language. Their claim concerned an underlying computational capacity, not the prediction that every language must display the same surface structures.
The implication is that this computational operation does not entail identical surface realisations across all languages. A biological capacity for recursive computation does not require every language to deploy it in every possible structural position. Everett’s own explanation for the absence in Pirahã is a cultural constraint. That appeal to culture, if accepted, may itself be consistent with an underlying computational capacity that the community’s grammar does not activate. The capacity is present; the deployment is suppressed.
This is an important counterargument which itself raises a legitimate counter-concern: if UG can accommodate the absence of recursive surface syntax by appealing to cultural constraints on deployment, it becomes harder to specify what evidence would count against it. That is a question about the theoretical architecture of UG, not just about Pirahã.
The Broader Argument
Everett developed his position considerably in Language: The Cultural Tool (2012). There he moved well beyond the recursion observation to a larger thesis: language is not a biological endowment but a human invention, shaped entirely by culture and use. Grammar emerges from communicative practice. There is no prior mental architecture waiting to be triggered.
This challenges UG at a more fundamental level than the Pirahã data alone. The generativist response has been consistent: a cultural tool account cannot explain why all known human languages share deep structural properties that no community designed or was taught. Children acquire language without input sufficient to explain the result. The poverty of the stimulus argument applies here as directly as anywhere. A culturally transmitted tool should, in principle, show more cross-linguistic variation than is actually observed.
What the Debate Has Clarified
The recursion debate has not been resolved. But it has produced something useful.
Before 2005, the claim that recursion is the defining property of human language was stated with considerable confidence. After Everett, that claim required more precise articulation. What exactly does UG assert about recursion? Does it predict surface recursive embedding in all languages, or only the underlying computational capacity? If the latter, what would count as evidence that the capacity is absent? The debate forced these questions into the open.
What the Pirahã case has not established is that Universal Grammar is false. What it has established is that Universal Grammar must be stated carefully enough to be testable. If recursion is a universal property of human language, linguists need to say where it must appear, what counts as evidence for it, and what would count as evidence against it.
Although not settled, the Everett controversy has a real and genuine value. It did not refute Universal Grammar, but it exposed a problem in how the recursion claim had sometimes been stated: a universal cannot remain useful if its absence in a grammar can always be reclassified as non-use rather than counter-evidence.
My articles are free for everyone to read, engage with and share with others
©Antoine Decressac, 2026
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, pp.1569–1579.
Everett, D.L. (2005). ‘Cultural Constraints on Grammar and Cognition in Pirahã.’ Current Anthropology, 46(4), pp.621–646.
Nevins, A., Pesetsky, D. and Rodrigues, C. (2009). ‘Pirahã Exceptionalism: A Reassessment.’ Language, 85(2), pp.355–404.
Everett, D.L. (2012). Language: The Cultural Tool. Profile Books.
Chomsky, N. (1986). Knowledge of Language: Its Nature, Origin, and Use. Praeger.