As I understand, grammars boil down to the acceptability problem: Is an utterance acceptable to the users of some language X? How to differentiate acceptable from non-acceptable utterances? Those who offer a grammatical analysis of some language X aim to describe the rules that differentiate acceptable from non-acceptable utterances.

Different approaches then differ only in such aspects like the (1) format of the description (2) the definition of acceptability (3) "area" of language covered. For (2) for instance, some define acceptability as merely grammatical, and ignore the issue of semantic well-formedness, while some include the latter in their definition. For (1), some require a description using a language that is easily implementable in NLP, while some settle with natural language descriptions.

For (3), some only deal with acceptable sentences (traditionally "syntax"), while some widen the scope and deal with acceptable words or word forms (what we traditionally call "morphology"), and even acceptable sound combinations ("phonology") or graphic representations ("orthography").

I also am cognisant that some approaches might even be explanatory and predictive rather than descriptive. They aim to explain why certain acceptability rules exist based on non-linguistic factors. And that some might be prescriptive, aiming to promote a certain rule, rather than to describe what rules actually exist.

(1) This is what I understand, but I'd like to know if I am missing something: are there grammatical approaches that do/did not aim to deal with the acceptability problem at all in the first place?

(2) Also, am I missing something even within the acceptability problem approach?

My aim is to have the most inclusive view of all that has been done on grammar - all schools of thought, all languages - as much as possible.

2 Answers 2


A little history on this might be useful. In the early 20th century, syntax was thought to be defined basically in the way you describe your "acceptability problem." That is, the central question is "how can we predict which word-strings are (un)acceptable to speakers?"

However, what Chomsky did when he initiated Generative Grammar was change the question in an important way. We're no longer asking "how can we predict which word-strings are acceptable to speakers," but "what is knowledge of language." In other words, what's in a person's head when they know a particular language? From the new perspective, acceptability is not that informative on its own. What we're considering is the acceptability of certain pairings of form and meaning. For example, the sentence (i) can ask "when did the saying take place" OR "when did the fixing take place." But (ii) can ONLY have the first interpretation, "when did the wondering take place."

(i) When did you say that he fixed the car?
(ii) When did you wonder if he fixed the car?

So the question is a little more complicated than just "acceptable or not." We're concerned with possible form-meaning pairings. What is in a speaker's head that prevents them from assigning the "when did the fixing take place" reading to (ii)? The hypotheses that are made to explain (at a certain level) the (im)possibilities are often formulated in terms of syntactic structure.

Hope this is helpful.


In the theory of Generative Grammar, what you describe is roughly the theory of performance, but grammars are about competence, which is a more abstract ability. Performance is about observed behavior, such as a person's willingness to "accept" a string, the actual production of some string, the ability to assign a meaning to a string and to produce a string likely to cause another person to assign a particular meaning, and so on. In that theory, the grammar does not flinch at strings like "the horse the man I knew bought died", even though people usually don't accept or produce such a string. Instead, separate cognitive theories outside the realm of grammatical theory explain why such sentences are not accepted or produced.

Of course you can transform the generative view of grammar and its relationship to language to what you describe, under the rubric "it depends on how you define". The term "acceptability" is misleading as a description of the broadest phenomenon because it views language from the perspective of a particular kind of psychological experiment used to test ___, a perspective that focuses on subject perception and contextualization skills rather than on production.

It is not clear what Chomsky himself currently believes a "grammar" to be, but generally speaking, generative grammarians consider a grammar to have at least phonological and syntactic components (morphology tends to get the short shrift, being divided between syntax and phonology). Non-syntactic relations that are computed by the phonology are still part of grammar, they just aren't part of syntax (and vice versa). However, it is also true that in terms of the performance of linguists (not their competence), specialists in component X may deem a structure "grammatical" when then mean "generated by the X component, but not necessarily passing the requirements of all grammatical components. Norbert Hornstein has written a fair amount about the problem of acceptability vs. grammaticality in syntax, and the problem that people do sometimes (often) get confused over the two.

  • "it is not clear what Chomsky believes a grammar to be" i'm not too sure about this. from my perspective, it seems he thinks a grammar is a description of a speaker's knowledge of language, what he calls an "i-language." do you think that's accurate? Jun 2, 2020 at 1:52
  • That part is not in question. What is not clear is whether he considers things that are not syntax (maybe specifically "recursion") to be part of "grammar". My uncertainty is based on his somewhat-recent thinking as reflected in e.g. Hauser, Chomsky & Fitch.
    – user6726
    Jun 2, 2020 at 4:17

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.