A central distinction between dependency grammars (DGs) and phrase structure grammars (PSGs, also known as constituency grammars) is the understanding of the initial division of the clause. Traditionally, PSGs divide the clause into a subject NP and a predicate VP. The predicate VP corresponds to a finite VP constituent - there are also nonfinite VP constituents, of course. DGs reject this initial binary division, which means they reject the notion that finite VP exists as a constituent (but they acknowledge nonfinite VP constituents). Thus in order to evaluate the two competing approaches to syntactic structure, the empirical evidence that can be brought to bear on the (non)existence of a finite VP constituent is central.

My question is as follows: What theory-neutral evidence can be produced to either verify or refute the existence of a finite VP constituent? I already have a solid opinion about this matter. I am interested, however, in learning how others respond when exposed to the question.

  • 1
    I don't understand why you need to choose one; clearly they both have uses, and clearly they can both be right simultaneously. Just as one can simultaneously believe that look in look at the lamp is an intransitive verb, while lamp is the object of the preposition at, and that look at is a transitive verb, with lamp as its direct object. A stereo view is often useful.
    – jlawler
    Nov 28, 2013 at 21:31
  • 3
    Thanks for your comment. The sentiment you express can be construed as implying that there is no truth in science. The one theory is every bit as good as the next. It's all a matter of preference. Nov 28, 2013 at 22:01
  • 2
    No, it's only that all syntactic theories are incomplete, contradictory, and full of untested or untestable presuppositions; in Sapir's phrase, all grammars leak. In that case, the more the better, if they work in some domain. If they don't work in any domain, then of course the correct thing to do is chuck them and find better ones. And, btw, how you construe what others say is up to you, as always.
    – jlawler
    Nov 28, 2013 at 23:29
  • 1
    @jlawler, your words are a breath of fresh air!
    – Alex B.
    Nov 29, 2013 at 16:50
  • 1
    @TimOsborne If there are people claiming that there's a maximal projection headed by the verb which is also tensed, please forward the reference, as i'd be interested in taking a look.
    – P Elliott
    Nov 29, 2013 at 20:24

3 Answers 3


Comparatives may provide evidence in support of finite VP being a constituent in English, but the evidence is quite theory dependent. In her 1983 paper ‘Comparative Ellipsis: A Phrase Structure Analysis’ (which can be downloaded from her website http://www.swarthmore.edu/SocSci/Linguistics/xling12napoli.html), Donna Jo Napoli distinguishes between two types of comparatives. Belonging to the first type are comparative clauses like those in (1) (taken from Napoli (676: 4)), which can be accounted for by processes that are known to be involved in other types of sentences with missing material in them.

(1) a. Mary wrote more books than John did
     b. Mary loves Fellini more than John, Bertolucci
     c. John would lie to Sue sooner than Bill would to Jane
     d. I organize more than I actually run her life

(1a) is an example of VP-ellipsis; (1b) of gapping; (1c) of pseudogapping; and (1d) of Right Node Raising. To the second type of comparatives belong those that cannot be accounted for by general processes; the examples in (2) are taken from Napoli (679: 8, 9) (my representation of the missing/unpronounced material doesn’t follow her account).

(2) a. John sent books to more people than [ComparativeClause [S Sue sent books to ∅x many people]]
     b. John sent books to more people than [ComparativeClause [S [NP Sue] sent books tox many people]]

In (2a-b), x many people designates the material that is obligatorily not pronounced in the comparative clause, and the striked-through portion designates material that is not pronounced but could have been pronounced. Napoli argues that the pronounced material after than must be a constituent of the embedded comparative or else form a constituent together with the material that is obligatorily unpronounced. She supports her claim by contrasting the grammatical sentences in (2) with the ungrammatical ones in (3) (again taken from Napoli (679: 8, 9)).

(3) a. * John sent books to more people than Sue sent books tox many people
     b. * John sent books to more people than Sue sent books tox many people

Napoli assumes that then is a coordination word in (2), and that it may coordinate phrases belonging to categories other than the sentence, among them verb phrases; (4) is again taken from her paper:

(4) I eat more than drink

Now it seems to me that we can also find comparatives where the pronounced material following than is a finite verb followed by its direct object:

(5) He more often eats cakes to gain weight than [ComparativeVP [VPx often drinks water to gain weight]]

If we accept Napoli’s generalization, this means that the sequence finite verb + direct object must be a constituent of the comparative VP in (5), as is shown in (6).

(6) He more often eats cakes to gain weight than [ComparativeVP [VPx often [VP drinks water] to gain weight]]

  • 1
    Great answer - thanks. It's not even necessary to look at comparatives to make this argument, it goes through with standard coordination too, i.e. John eats cakes and drinks water. Assuming that only constituents can be coordinated, this shows that the finite VP is a constituent. @TimOsborne, in one of the comments above, contends that this isn't true, but the counter-example given: Susan writes and Bill tells the jokes is a clear case of Right Node Raising, which has its own peculiarities, and is generally analysed as involving ellipsis and/or movement.
    – P Elliott
    Dec 5, 2013 at 17:14
  • 1
    Just to illustrate the point further that coordination cannot "identify most any string of words as a a constituent": (i) [John bought a yellow and Frank stole a yellow] car (ii) [John does know that Bill and Frank doesn't know that Bill] is sleeping. (i) and (ii) sound pretty clearly ungrammatical to me, which shows that not just any strings can be coordinated.
    – P Elliott
    Dec 5, 2013 at 17:20
  • @Shai, Thanks for the answer. You are reaching to a sophisticated argument and complex data to attempt to prove the existence of finite VP as a constituent. But basically what your insight boils down to is that since finite VP can be coordinated, it must be a constituent. The data you provide are from comparatives and comparatives often involve coordination. This is basically what Napoli points out. I am familiar with her article, and I have my own article that explores such data (in NLLT, 2009). I will happily share this article with anyone who wants to take a look. Dec 5, 2013 at 19:19
  • @Shai, the fact that coordination (in comparatives or otherwise) identifies finite VP as a constituent is the only solid type of data from constituency tests that can be produced suggesting that finite VP should be a constituent. But as I stated and briefly illustrated above, coordination identifies most any string of words as a constituent. An appeal to RNR is indeed how PSGs attempt to overcome the data, but by doing this, the PSG reasoning has become circular. It is circular to argue that coordination identifies constituents and then when it doesn't, to augment it with a further mechanism. Dec 5, 2013 at 19:30
  • 1
    @Shai, one last comment: Your examples (5) and (6) show the the adjunct "to gain weight" being deleted/elided from the comparative clause. This is a general trait of adjuncts in coordination, e.g. [the books on the shelf] and [the folders] -- the folders are understood as being on the shelf, too; [Tom arrived yesterday] and [Susan left] -- Susan is understood as having left yesterday. Dec 5, 2013 at 20:00

Your question is methodologically misconstrued - any hypothesis can be verified or refuted only within a certain paradigm/framework, cf.

"scientific statements can only be made within a theory" (Bierwisch 1971).

Igor Mel'chuk (Mel'cuk 1988: 12) puts it very nicely - the following applies to any syntactic theory:

By its logical nature, dependency formalism cannot be "proved" or falsified. [..] Dependency formalism is a tool proposed for representing linguistic reality1, and, like any other tool, it may or may not prove sufficiently useful, flexible or appropriate for the task for which it has been devised; but it cannot be true or false [emphasis mine - Alex B.]

There can be no theory-neutral evidence to refute or verify the existence of a finite VP. As a matter of fact, just in your question there are the following assumptions: the existence of a VP, finite VP, constituents etc., cf.

"constituents are neither essential nor fundamental to linguistic structure" (Langacker 1997: 9)

see also Carnie 2008: 18 "we will find many instances where these [constituency - Alex B.] tests can give false results and results that are contradictory with the output of other tests."

  • Thanks for the answer, but I think more is needed. In order to be understood, I use a language that most syntacticians and grammarians can understand, hence the term "finite VP". This situation is forced because that is the state of the field. Let me restate the question in more neutral terms: What evidence exists that there is a node in the structure that dominates what PSGs take to be a finite VP constituent (and nothing more)? Is there any? What is it? Nov 29, 2013 at 5:30
  • "most syntacticians and grammarians can understand" doesn't mean agree with/subscribe to. The way you rephrase your question doesn't help at all - again, you seem to assume that all syntactic theories have nodes, constituents, and phrases.
    – Alex B.
    Nov 29, 2013 at 16:00
  • 1
    In other words, while we may understand what you say, we don't necessarily share the presuppositions required to agree.
    – jlawler
    Nov 29, 2013 at 16:55
  • Thanks guys for responding and for the citations from the literature. I'm not asking for anyone to agree; that gets boring quickly. I'm trying to see what evidence can be produced to support a fundamental assumption of much work in syntactic analysis. So far, no cogent evidence has been produced here. The strategy seems to be, rather, to invalidate the question. It's a simple question and it should have a simple answer. Nov 29, 2013 at 17:58
  • 1
    @Alex B. Thanks for quoting Mel'cuk. Here's another quote by Mel'cuk that seems relevant at this juncture: "Even though it sounds a bit too Whorfian, I am fairly sure that PS-syntax could not have been invented and developed by a native speaker of Latin or Russian…To promote PS-syntax, one has to be under the overall influence of English, with its rigid word order and almost total lack of syntactically driven morphology." (Mel’čuk 1988:4) Nov 30, 2013 at 1:31

The existence of a VP constituent is language specific. There are detailed constituency tests. One of the definitions of a language being configurational is the existence of a VP.

The main evidence comes from sentence topology. In German, for example, no finite VP can be identified because the sentence splits into Vorfeld, Mittelfeld, etc., that is, it's organized along a completely different scheme.

  • 1
    Thanks for the answer. You mention constituency tests. Which constituency tests deliver evidence for the existence of a finite VP constituent in English? Examples? Most constituency tests in fact do not support the existence of finite VP in English (topicalization, clefting, pseudoclefting, proform substitution, answer fragments, passivization, omission). The only clear exception is coordination, but coordination would identify most any string of words as a a constituent, e.g. [Susan writes] and [Bill tells] the jokes. Nov 28, 2013 at 22:11
  • What about "do (so)"-substitution? I ate some cake, and Joe did too. Are you saying that this doesn't demonstrate a finite VP constituent because what's really going on is elision of a non-finite VP after did?
    – TKR
    Nov 29, 2013 at 0:06
  • Yes, the "do-so" test provides evidence for the existence of nonfinite VP as a constituent, not for finite VP, e.g. "Larry didn't say it, but Sam did (say it)", or "A: Sam didn't say it. B: He did so (say it)". Auxiliary verbs introduce VP-ellipsis, and "do" can be an auxiliary verb. The confusion with the test comes from the fact that auxiliary "do" needs contrastive emphasis to appear. If this contrastive emphasis is not intended, then the VP-ellipsis becomes almost obligatory, e.g. "??Larry solved the problem, and Sam did solve the problem, too." Nov 29, 2013 at 4:51

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.