If you google search "discontinuous constituent", you will see that various extensions to CFG have been proposed to describe discontinuous constituents. One of them is found in my favorite book on syntax, The Syntactic Phenomena of English by James McCawley, who begins his book with a proposed extension of CFG, since he is a proponent of discontinuous constituents, and gives many examples of discontinuous constituents in English in his book.

But my question is whether any extension to CFG is necessary. I have an answer to this question, "Not in some cases", which I will offer in due course.

  • I've studied many non-configurational languages (including Udi, Abkhaz, Eastern Armenian and a few Native American languages) and my answer would be "no", too, but that's partly because discontinuity is captured at the level of functional headship, which is orthogonal to phrase-structure. That said, I'm all ears to hear your reasoning.
    – Atamiri
    Commented Apr 14, 2020 at 14:11
  • I hope the answer includes diagrams to help us understand the reasoning. Commented Apr 14, 2020 at 15:12
  • @Atamiri, I don't know about functional headship, but my account applies most straightforwardly to heads only.
    – Greg Lee
    Commented Apr 14, 2020 at 18:51
  • @GregLee What I meant is the difference between functional heads and categorial heads in constraint-based approaches to syntax that use both context-free rules and dependencies. In such formalisms, discontinuous constituents pose no problems to the analysis at the level of phrase structure.
    – Atamiri
    Commented Apr 14, 2020 at 21:33
  • If deep structure is actually 3D, then crossing shadows on the 2D surface structure should not be surprising.
    – amI
    Commented Apr 15, 2020 at 2:50

2 Answers 2


Consider the verb-particle construction of English when the particle is separated from the verb: "Joe will look the number up." Instead of introducing some special notation to show that "look" and "up" are both parts of the verb, suppose we replace the phrase structure rules in (1) with the rule (2):

(1) VP -> V NP Particle
V -> look
Particle -> up

(2) VP -> look NP up

and note that "look up" is a verb by making a definition:

(3) The pronunciation part of a VP is a verb.

An implication of my proposal is that in syntactic structures, verbs cannot occur as constituents of anything other than a VP. Although one might think that infinitives and gerunds are verbs that occur as constituents of NPs, "To err is human", "Eating is fun", actually, these are both VPs, not verbs, as can be easily seen by using as examples verbs which have complements.

  • What about other discontinuous constructions such as topicalisation?
    – Atamiri
    Commented Apr 15, 2020 at 7:11
  • @Atamiri, Evidently my proposal does not account for those.
    – Greg Lee
    Commented Apr 15, 2020 at 8:09
  • I've never heard of VPs functioning as subject before. "To err" and "eating" may be VPs but surely they function as predicate of a subjectless clause.
    – BillJ
    Commented Apr 15, 2020 at 10:10
  • @GregLee Apparently, it only accounts for projective discontinuities. But it might also work for unbounded dependencies, it just rules out binary branching (which I'm not a fan of BTW).
    – Atamiri
    Commented Apr 15, 2020 at 10:15
  • @BillJ, My only point here was that they are VPs rather than verbs.
    – Greg Lee
    Commented Apr 15, 2020 at 13:15

https://www.aclweb.org/anthology/E87-1034.pdf seems to address this.

From a Prolog perspective, https://www3.cs.stonybrook.edu/~warren/xsbbook/node31.html

  • 2
    Hi, and welcome to Linguistics.SE! Links to support your answer with background information are most welcome, but a post containing only links is not a proper answer. Try to work at least a summary of the linked resources into your post so that people who cannot or don't want to click through can understand the gist of what's behind those links.
    – tripleee
    Commented Apr 24, 2020 at 8:45
  • @tripleee Thanks for the links. I think you may have missed the point of my answer. My proposal does not require extending context free grammar in any way, unlike the other theories I mentioned. The paper from The Netherlands appears to be yet another way to extend CFG. I do not extend CFG (making a definition is not a theoretical extension).
    – Greg Lee
    Commented Apr 24, 2020 at 20:34

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.