ok so I'm trying to understand the argument for an innate language faculty, and specifically Chomsky's opposition to the behaviorist model of language acquisition. Behaviorists thought that language acquisition could be accounted for by a combination of mimicry and positive feedback. My understanding is that Chomsky refuted this position, stating that behaviorism cannot account for linguistic creativity. In other words, if children were only able to repeat syntactic structures that they heard in their input, they would not be expected to to be capable of producing novel sentences.

But I don't understand why this is the case. Lets say I've heard and understood two sentences:

  1. I want to pet the dog
  2. The cat is purring

... and I later produce the sentence "I want to pet the cat". Is this an example of creativity? I guess this would assume an underlying understanding of syntax, I guess it's a bad example...? Idk bro I'm confusing myself, but it seems like I produced a novel sentence that I haven't heard before using the input I've been exposed to.

Do children really make sentences that can't be somehow derived from their input in this way? Children don't use syntactic structures that they literally have never heard before... right?

I'm definitely trying to play devils advocate with myself here, as I understand that the language faculty must have some element of creativity in order to account for languages existing in the first place (because how would a child learn from mimicking input when there was no input). I'm just trying to get a better mental understanding of where exactly children show such a massive gap between exposure and competence.

  • 1
    I;m not surprised. Just read what Geoff Pullum says about Chomsky in Wikipedia link
    – BillJ
    Aug 19, 2023 at 16:54

1 Answer 1


Neither the behaviorist model of 65+ years ago nor Chomsky's model were submitted to rigorous empirical testing in the realm of first language acquisition, because neither was made explicit enough that observations could prove or disprove a theory. The arguments were 99.9% conceptual, and 0.1% anecdotal. The anecdotes favored Chomsky's model, as did the conceptual arguments. The rules of the game completely changed over the last 2/3 century, indeed the neo-behaviorists have mostly won the day in my opinion.

The main advantage that the neo-behaviorists have exploited is abandoning the earlier rat-centric simplistic model of learning. Contemporary behaviorist learning is very complex and powerful. It retains from the earlier versions of the theory the commitment to a "general" theory of learning that does not assume any mental faculties that are specific to language, except perhaps those that pertain to physical input and output (for example, there may be a built-in capacity to recognize certain noises as "speech" rather than animal threat, but this is completely outside the domain of grammar).

Progress in computer hardware and software has now provided a more plausible response to the question "how could a mechanical model like behaviorism account for highly improbable utterances". This was the core of Chomsky's original argument, that myriad utterances have extremely low probability of having been produced before (actually, infinite improbability), therefore a simplistic "record and play back" model of language learning is clearly untenable. Harder to dismiss is a model that includes a component of analogy, where novel utterances can be created by analogy to mechanically-recorded utterances. This has always been a conceptual possibility (for thousands of years), what has changed is that improved computer tech has made it actually possible to create analogical systems, operating somewhat like what your example works.

Children are not particularly relevant to the theory of language acquisition, at least right now, instead the central question is, what does it take to teach a computer to create and interpret utterances? But to address your fundamental premise that "children don't use syntactic structures that they literally have never heard before", no, every child has literally heard "good dog" and "eat apple", or the equivalent before. The earlier view of the question sought to expand "the number of structures" in syntax to be "all possible utterances, stripped of specific words", and yes there are infinitely many of them, a list that cannot be memorized. Being mindful of the current much simpler model of syntactic structure, all you have is "a pair of two things, which is itself a thing that can be combined". Every child is exposed to the fundamental fact of syntax, there is only one "structure". Children don't learn compound structures, they learn how to handle compound structures.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.