Why does Chomsky consider recursion in language to be a "narrow" ability unique to humans?

Question

There is a well-known classification of four varieties of grammars, differing in complexity, from unlimited to regular. These grammars correspond to four classes of automata in computer science:

Regular grammars correspond to finite automata;

Context-free grammars correspond to push-down automata;

Context-sensitive grammars correspond to linearly bounded automata;

Phrase-structured (unrestricted) grammars correspond to Turing machines.

IMO, this correspondence is astonishing and its consequences have not yet been fully revealed; except for this correspondence, I know of only one similar one - the Curry-Howard-Lambek correspondence (between computer science, logic and category theory).

However, Chomsky then makes a somewhat vague statement: he claims that the ability of "recursion" in language (I don't quite understand what is meant - maybe general recursion?) is inherent in the human genes and that other animals do not have this ability. With that in mind, I have 3 questions:

1. If by "recursion" Chomsky means that some class of grammars is capable of it, which one? Perhaps unlimited one or... what?

2. In order for recursion to manifest itself in a certain system, it only takes a simple feedback loop! However, this doesn't require any special genes or unique designs of the human brain - even in the most primitive nematodes, feedback loops do exist.

3. If there really are language constructs that humans can understand, but, for example, dolphins and monkeys cannot, then what are those constructs? What specifically "recursive" expressions are we talking about? (Or does Chomsky simply want to arbitrarily single out people as a separate group?).

Thanks in advance.

Answer 1

First you need to understand what recursion is, especially as applicable to linguistic structure. It is typically understood (defined) as the situation where a type α is defined in terms of type α. Recursion is a "theorem" from the situation where any type α is definable in terms of arbitrary type β, which is how it is relevant to language. We would have to ask Chomsky to clarify his specific thinking about recursion.

Also recall that the question that Chomsky is answering is, what cognitive mutation distinguishes humans from other (primate) species, and which plays the most significant role in saying what it means to "have language", which humans do and chimpanzees do not? If you disagree with the premises, maybe you can inquire about the notion of the human language faculty separately.

The question of how to model a hierarchy of types in terms of electronic circuitry is orthogonal to the logical observation that types, as they exist in human cognition, are definable in terms of other types, and a type is not just a collection of particular terminals. This is not a claim about some arbitrary theory or grammar and computation, it is a premise implicit or explicit in every theory of language.

Linguistic structures are composed, not emitted as un-analyzable primitives. For example, the sentence "The teacher explained that simplistic statistical models of language fail to account for the fact that we can understand and produce infinitely many utterances while also excluding as many utterances" is not spewed out via a look-up function that returns memorized utterance #434,854,672. Instead, we have a subject of which a proposition is predicated. That proposition itself contains a predicate which is defined in terms of another predicate, and so on. In old-style syntax, that means that S → NP VP, VP → V NP, NP → ... (various things: article+N, or comp+S).

Those familiar with Chomsky's rhetoric over the past 70 years are familiar with the problem that he offers vague hints rather than concrete claims, and the details of implementation are left to others to work out. IMO, the one significant failing of his most recent foray into evolutionary biology and linguistics fails to address the distinction between the faculty of reason, and the faculty of language. It's not just language that is unique to humans, it is something broader.