Actual human natural language is meaningfully recursive, because any grade school student understands the recursive properties of certain toy sentence examples in that they could "go on forever" as a set of arbitrarily larger complex sentences.
The commonality of mathematical grammars (as mathematical objects in their own right) and natural language is that both have computational properties such as recursive structure - or ambiguity, which was discussed in the paper.
You're confusing the object language (human natural language) and the meta-language (formal grammars - ala Chomsky).
Formal grammar theory relies on recursive formulae to define what is a a language (formally set of sentences), natural language itself does not use/require recursion as understood in the same way. Some natural languages support grammatical features like sub-clauses but that's only "recursive" in the sense that attempts to describe such features (using formal language grammars) use recursion when in fact and in practice nesting sub-clauses do not "go on forever".
An analogy with computer programming languages - programming languages exist that do NOT support recursion (pre-Algol) but nonetheless require recursion to define their formal grammar. For example, a programming language can support limitless nested lexical scoping (loosely analogous to sub-clauses in natural language) WITHOUT supporting recursion (which requires a stack and managing frames).
For some programming languages (functional for example), recursion is such a key feature that it's inconceivable that the language could be useful without it.
This is not the case for natural languages. Most of the people I interact with daily have English as a second language and rarely or never use sub-clauses (which are not recursive but are often used to claim that language is recursive) so from practical experience, I know that English as a communication tool looses very little expressive power when users do not use sub-clauses.
I'd definitely say all language use must rely on recursion, you're just highlighting our tendency to halt the pronunciation of it early at arbitrary points due to linguistic norms. Like, AFAIU this was Chomsky's whole impetus, and I really don't think anyone has answered better than he: how can a finite system compute linguistic meaning in finite time, otherwise? I'd guess the "natural language" answer would be along the lines of "it just comes naturally who knows", in which case I'm staying I'm on my side of the battle lines ;)
Either way, I think you'd win a Nobel for proving that complex recursive processes aren't going on inside the head of someone speaking in natural language. We can discuss the likely situation all day, but directly measuring this in the moment seems well beyond our technical capabilities
I think Chomsky is a linguist really more famous and respected among non-linguists than actual experts in the field. I don't think modern linguistics - particularly applied linguistics for example in the area of language acquisition - references Chomsky.
His theories are unfalsifiable and un-empirical and thus, for me, unscientific. This discussion is typical. Chomsky's definition of "recursion" - like that of a "language organ" - has become completely plastic. In his initial work, it was deeply tied to the same notion in formal grammar theory but is constantly redefined and evolving as the initial notion becomes more and more dubious. The "human language is unique in that it is _recursive_" theory has no practical application and has provided no path for advancement in the field of language pedagogy for example.
Watch me chew up a bunch of karma dismissing Chomsky's theories of language. It's what often happens with comments that question his theories of language - I don't know why it's seemingly impossible to just admire his politics without feeling the need to vigorously defend his 60 or 70-year-old theories of language which have long been abandoned in the field.
It would genuinely shock me if Chomsky wasn't the most cited author for works published today even if you exclude works that are dismissive of him.
The reason the specifics of Chomsky's proposals are changing is that it isn't a solved problem and he and others are continuing to try to advance the understanding by revising proposals in light of better arguments, simpler explanation and (contrary to what you say, see [1] which lists some examples) experimental evidence. In Chomsky's most recent book he himself quotes Wheeler: "Surely someday, we can believe, we will grasp the central idea of it all as so simple, so beautiful, so compelling that we will all say to each other, ‘Oh, how could it have been otherwise! How could we all have been so blind so long!’". I think people have a hard time understanding this because it is one of the few nascent sciences and therefore no one has figured out the fundamental theories.
And if some specific proposal like X-bar theory is falsified or supplanted by something else that doesn't falsify the core ideas of universal grammar or the minimalist program. You say that it isn't possible to provide evidence that these are false but Moro for example conducted experiments comparing hierarchical and linear language processing that potentially would have provided reasonable evidence against language being hierarchical had they shown linear languages were processed similarly (instead they support it).
The reason you should be downvoted is that you made a bunch of evidence free assertions; You could have said "Neil Bohr's work has been completely falsified and is of dubious value" with the same amount of supporting argument.
Well he calls that whole field “not real linguistics” basically, so I think it makes sense that they wouldn’t accept his theories! Whatever you call the two things, I feel like
1. “what are the neural structures and programs that produce language” and
2. “what languages have humans used to communicate in specific historical contexts, and how do these practices relate to each other?”
…are both worthwhile questions, and they obviously inform each other. But don’t you see where he’s coming from by calling the former “the biological study of language production” and the latter “the anthropological study of language usage”?
I’m interested to see you usually loose points! That brings me joy as a (non-academic, self-employed) chomskian, even tho it’s obviously infuriating people are still downvoting based on differences of opinion in 2024. My view is the opposite of yours: almost every corner of the internet has dismissed Chomsky as unempirical and also disproven by recent cross-cultural studies (??) and LLMs, from Hinton to the Stanford Phil page on linguistics to the Wiki and Britannica articles. I frequently see anti-Chomsky papers on HN and in my lit reviews from the connectionist/positivist camp in general, but that’s obviously a terribly biased survey. Perhaps we all have confirmation bias to see our “team” as the underdogs…
In terms of debating specifics (no pressure to respond, your outlook is very valid):
He’s still being cited a lot (160K in 2023 compared to recent peak of 210K in 2017), but of course that could all be refutations and/or non-linguists. Ultimately I absolutely agree that applied linguistics has far more practitioners in the US (the “west”?) today, but I don’t think that’s necessarily any kind of refutation of the merits of theoretical work.
I’m currently engaged in implementing general ideas of Chomsky’s into a cognitive model for AI, and I definitely think it’s a more productive theory than you’re giving it credit for. For one, the two basic observations “basically all linguistic inference is done by procedurally merging symbols and signs” and “what separates humans from animals is the ability to introspectively index and merge symbols and signs effectively” are essential pillars of my whole paradigm — without them I wouldn’t event know where to start! I think it’s much better than some competing “grand narratives” for human sciences like Behaviorism or Postmodern Relativism, but of course it’s far from perfect and surely has discouraged some valuable work over the years. I agree that it isn’t empirical, but only because it’s a guiding idea/standpoint/perspective and not what Chomsky would call a “scientifically defined term”. UG/nativism/minimalism gradually loose or gain parsimony in relation to our body of evidence over time, but they’re not consistent or specific enough to be fundamentally proven or disproven anytime soon IMO
Well, I don't share this insistence that natural language is somehow special or privileged. Natural language has a tree structure and nobody need to learn about formal grammars to comprehend that. A schoolchild understands this much, without having to understand recursion in a rigorous way.
Furthermore, as I mentioned before, formal grammars are artificial languages in their own right. The key thing is that the only way a human being can understand these languages (e.g., still at the level of syntax, independent of recursive computations like in functional programming which you did mention) is through some kind of recursive reasoning about their grammar.
And to keep the discussion at the forest instead of the trees, it is weird that human cognition comprehends these tree structures, recursive structures, or fractal structures, relatively naturally and intuitively. If I had said fractal would you have argued that natural languages don't exhibit fractal properties? We are talking bird's-eye level view of things here.
I also believe in the Church-Turing thesis plus the Chomsky hierarchy of computable languages (the idea that languages are models of computation, and that some language classes contain other classes in a hierarchy of what is computable -- see for instance the formal class of Recursive Languages), meaning that natural languages are not based on magical computations exceeding the power of universal computation; furthermore, as another subcommenter (see both users bbor and foobarqux) put it differently to you but if you are committed to a position that natural human language can only be approximated by artificial languages defined by formal grammars then you can never have a humanly intelligible account of the computational features of actual human language. It would be like arguing a laptop is not "meaningfully" like a Turing machine because a laptop has finite number of atoms as is thus really technically a finite state machine -- it's throwing the emergent/abstract baby out with the empirical bathwater.
Clauses within clauses are not "recursive", no more than "let x = 1 in (let y = 2 in (let z = 3 in x + y + z)" represents recursion in some hypothetical programming language.
Programming languages and human languages existed before people felt the need to define them using formal grammars. Yes most formal grammars use recursion but that's a property of the system you are trying to use to describe (meta) the structure of the object language - not part of the object language itself.
Try to use Haskell without recursion, it's effectively useless. But using an Algol derived language (Java, C, C++, etc.) without recursion is fine and hardly restricts the power of the language at all. Try to use English (or any other indo-European language - I've no knowledge of others) without sub-clauses and it remains a very powerful and useful communication tool.
Agreed: if you're willing to be as limited with composing in english as you would be with programming in C, lack of recursion is not an issue.
(there's a bit of a parallel, in that XVII english writers, not having readers as pressed for time as our current variant, make far more use of nested sub-clauses than we tend to; similarly, CPL, not having been as pressed for space as either BCPL or C, made far more use of nested sub-expressions than its successor languages)
You are saying that text on a (finite) page is not recursive (because recursion is a computation, not the output of that computation). But we are interested in the computations that the language faculty performs not the output.
> Clauses within clauses are not "recursive", no more than "let x = 1 in (let y = 2 in (let z = 3 in x + y + z)" represents recursion in some hypothetical programming language.
But that is an example of a recursive context free grammar. The parser needs to use a stack or be defined recursively to parse things like that.
>> Some natural languages support grammatical features like sub-clauses but that's only "recursive" in the sense that attempts to describe such features (using formal language grammars) use recursion when in fact and in practice nesting sub-clauses do not "go on forever".
The commonality of mathematical grammars (as mathematical objects in their own right) and natural language is that both have computational properties such as recursive structure - or ambiguity, which was discussed in the paper.