I've noticed that phrase structure rules have been very inconsistent over my studies. I've seen NP = (det)(adj)N ; NP = (det)N(PP); these definitions seem to change with context. Is it just because there's no strict phrase formalisms in language? Or am I missing something?

  • 2
    No, it's because there are many possible phrase structures. Typically there are dozens of "NP =" and "VP =" rules in any PS grammar that attempts to cover more than a toy language.
    – jlawler
    Commented Dec 6, 2013 at 5:30
  • Then why make phrase structure rules in the first place? I understand trying to represent the relationship between constituencies in trees, but phrase structure rules seem rather frivolous. Commented Dec 6, 2013 at 6:00
  • 2
    @Farting, yes, PS rules in the traditional form are rather silly. The number of those rules that you would need to start to cover the combinatorial properties of just a few lexical items is enormous. A better approach is to focus on the actual lexical items and how they combine with other lexical items. Commented Dec 6, 2013 at 6:17
  • 2
    The reason why they're presenting them to you is that they're baby steps toward syntactic rules: Start with an input and change it somehow in a deterministic fashion to produce an output. It's just an instructional strategy, and probly not a very good one.
    – jlawler
    Commented Dec 6, 2013 at 21:06
  • I am having second thoughts about the meaning of your question. Are you concerned with the fact that there are several PS rules NP= ... in the same PS grammar for English, or with the fact that these rules are not always the same, depending on the book, the chapter, the exercise or the teacher ? My answer assumed the latter.
    – babou
    Commented Dec 6, 2013 at 22:31

1 Answer 1


There are many ways to describe a given language, even in a given formalism. I would think that at any time, a language is always inconsistent. Diachronic evolution seems a good reason to believe that. Latin has slowly evolved into Italian, Spanish, Catalan, Roumanian, Portuguese and French (what did I forget?). Actually there have been many more dialects and variants that did not all survive. And the evolution has been progressive. How can you expect to have consistency in such an unstable context?

I do not know of such experiments, but I would guess that no two people will agree everywhere on the grammaticality of some sets of sentences.

But that shows most for the visible (surface) syntax of languages. One task of linguistics is to identify deeper and more stable concepts that structure the language in a more consistent way.

What matters for surface syntax, that you are considering, is more to understand adequate and tractable formalisms that will let you describe the language (almost) adequately, and develop techniques to attain from that the deeper structure. Distinct surface descriptions in the chosen formalism can hopefully give you similar understanding.

Actually, even the choice of the formalism can be open, and there are many competing. Even if you consider only phrase structure, the variant you have been using seem to be the context-free variant of Chomsky. But others, with interesting descriptive properties have been developped, such as tree-adjoining-grammars, to give an example. Then these formalisms can form a skeleton to be completed by more detailed information (features).

Also relations with the lexicon have to be established, and structural information and features can be attached to the lexicon. Interestingly, when choosing or designing a formalism, independently of the language to be described, there will be tension between the "phrase structure skeleton" and the lexicon, regarding where some of the structural information should be placed. Emphasizing the lexicon can also be done by lexicalizing the phrase structure skeleton (which is what Greibach normal form does for context-free languages).

And there are other very different kinds of formalisms, such as categorial grammars.

What matters is to acquire the knowledge, understanding and effective usage of natural language description tools. Then you use these tools to give the best description you can of the natural language structure. We do not all have to agree on what might be the best description of the structure of a natural language (hence the variations you complain about), but we need common descriptive tools to understand each other when discussing these issues. The formalisms are (part of) the language needed by linguists to exchange and discuss meaningfully about the natural language.

  • The answer provides way more information than the question required. My guess is that Farting has no idea what most of this answer is talking about. Commented Dec 6, 2013 at 19:15
  • 1
    Though it's pretty much all useful information, and -- who knows? -- somebody else might read it, too. Or chance upon it while doing a search for several of the terms. We're not really dealing with individual requests here in principle; we're making informative texts. The variety of questioners and their wildly diverse presuppositions as shown on ELU means that there's almost always a mismatch between what the questioner is asking and the information they actually need, whether they requested it or not.
    – jlawler
    Commented Dec 6, 2013 at 20:10
  • @Jlawler, the question asks about "phrase structure rules". Nowhere in the answer is the term "phrase structure rules" or "PS rules" even used. In short, the answer misses the target almost entirely. Commented Dec 6, 2013 at 20:37
  • Though the reply was a bit advanced, I think it helped answer my question. I was seeing PS rules as more of a strict generative rule, rather than a descriptive tool. Commented Dec 7, 2013 at 5:56

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.