In this wiki on "Subject–verb inversion in English", there are two approaches suggested to dealing with subject-verb inversion.

(1) Maintaining the traditional subject-predicate division of the clause (S → NP VP)

(2) Rejecting the existence of the finite VP constituent.

But neither is perfect, says the wiki:

In approach (1), "one has to assume movement (or copying) on a massive scale."

In approach (2), "this analysis does not capture the obvious dependency between the main verb and the inverted subject."

I'm not quite sure what the latter exactly means, but this sounds like a drawback of the second approach. (Please enlighten me on this as well in your answer.)

Now, what if a verb phrase (VP) can be defined to include a subject? Then, I think we should be able to analyze subject-verb inversion without (1) assuming movement (or copying) on a massive scale or (2) failing to capture the obvious dependency between the main verb and the inverted subject -- whatever that means.

[i] Under the bush crouched Bill.

[ii] Bill crouched under the bush.

In sentence [i], for example, if "crouched Bill" can be defined as a VP that includes a subject "Bill", then the sentence can be easily analyzed as follows:

S → PP + VP

VP → V + NP

In other words, this approach doesn't treat sentence [i] merely as an inversion of sentence [ii], but it treats sentence [i] as having a unique VP that includes a subject.

Has this kind of approach ever been suggested? If not, what do you think about it, compared to the other two approaches mentioned above?

  • How do you distinguish subject from object, then, if both follow the verb? Node labels are not the same thing as constituents, and nobody expects either to stay the same in derivations. After all, in English the simplest definition of a verb phrase is that it's just a clause without its subject.
    – jlawler
    Apr 16, 2018 at 19:18
  • @jlawler So you're with approach (1)? If so, you don't mind having movement "on a massive scale"? I think movement is a better theory for subject-auxiliary inversion where you only have one auxiliary movement. And the simple definition of a VP being a clause less its subject necessarily assumes that a subject is not a complement of a verb. So, you must think that a subject is not a complement of a verb. Right?
    – JK2
    Apr 17, 2018 at 5:15
  • In English there’s a rule: VP -> V NP (NP). The first NP is typically the primary object but in the case of inversion it can be the subject. Phrase structure alone isn’t enough to capture grammatical roles, it’s all about mapping the roles to the V’s dependents.
    – Atamiri
    Apr 17, 2018 at 11:23
  • No, I don't think that a subject is a complement of a verb. I use the term complement only to refer to clausal arguments of a predicate; if they're not clauses, I never use the term complement, because (like head) it's unclear and ill-defined. And I didn't say I was "with approach (1)"; that's not my theory, it's yours.
    – jlawler
    Apr 17, 2018 at 20:44
  • 1
    H&P use terminology in their own way, which is their privilege since they wrote an entire grammar. However, their terminology is not standard and most people don't follow it. Consequently, the notion of complement they're referring to is not standard, either.
    – jlawler
    Apr 18, 2018 at 14:28

2 Answers 2


I know of several theories which, in some sense, put subjects inside English VPs:

  • English as a VSO language. James McCawley proposed that in English deep structures, subjects immediately follow verbs. His article appeared in Language.
  • Case Grammar. In Charles Fillmore's Case Grammar, subjects arise through a transformation that moves one of the several "deep case" constituents that go with a verb into surface subject position.
  • 2PSG is my theory, based on some ideas of Relational Grammar. The main parts of sentences are verbs with dependencies of various degrees of obliqueness. Inversion constructions of the sort that interest you arise when an oblique verb dependency becomes less oblique than other verb arguments which, together with the verb, make up the VP. The least oblique dependency is that of root sentences, which is why inversions occur only in unembedded sentences. See 2PSG here (a deleted question).

In reply to a comment, here is a 2PSG example tree for the example [i] "Under the bush crouched Bill":

           [S0 under the bush crouched Bill ]  
                     /                   \  
        [S0 PP0 crouched Bill ]         [PP0 under the bush ]  
            /             \                    /           \  
      [S0 S1 ]   [S1 PP0 crouched Bill ]  [PP0 PP2 ]  [PP2 under the bush ]  
                   /            \                       /     \    
  [S1 PP0 crouch Aux1 Bill ]  [Aux1 ed ]   [PP2 under NP2 ]  [NP2 the bush ]  
              /              \                               /    \  
 [S1 PP0 crouch Aux1 NP1 ]   [NP1 Bill]           [NP2 NP1 ] [NP1 the bush ]    
           | by Raising
 [S1 NP1 crouch Aux1 PP2 ]  
  • Forms in the derivation tree are either lexical or derived from forms below
  • The derivational rule of Raising makes a variable less oblique
  • Other steps in the derivation are by the derivational rule of Substitution
  • Substitution substitutes a string of constants (phonemes) and variables for a variable. A tree node assigns the string as a value of the variable it replaces.
  • A variable is a grammatical type (e.g., S, NP, PP, Aux) and a grammatical relation (e.g., 0 for speech acts, 1 for subjective, 2 for objective, 3 for indirect objects and other obliques). Grammatical relations are taken here to be levels of obliqueness (as in HPSG).
  • The counterparts of forms in other grammatical theories are context free phrase structure rules, or tree nodes in Categorial Grammar.
  • Thanks. So do you think treating crouched Bill as a VP as suggested can be better than the other two approaches, at least in subject-verb inversion?
    – JK2
    Apr 17, 2018 at 2:13
  • No. I don't think there is such a thing as a VP, except in a derivative sense.
    – Greg Lee
    Apr 17, 2018 at 3:05
  • So you're agreeing with approach (2), right? Then, what do you think about the wiki's criticism of approach (2), namely, "this analysis does not capture the obvious dependency between the main verb and the inverted subject."?
    – JK2
    Apr 17, 2018 at 4:59
  • No. None of the theories I mentioned is a dependency analysis, and, anyhow, I don't understand the reasoning of the wiki.
    – Greg Lee
    Apr 17, 2018 at 6:06
  • So you don't agree with either (1) or (2) or mine. Then, could you show me how you would analyze [i] Under the bush crouched Bill.?
    – JK2
    Apr 17, 2018 at 6:08

In generative syntax, the fronted PP, under the bush, is usually analyzed as Spec,TP whereas the light subject, Bill, is in situ in VP, because Spec,TP is already occupied by the fronted PP and it cannot move to SpecvP because crouch is an unaccusative verb. The phenomenon is known as locative inversion.

See e.g. Culicover and Levine 2001 for further details.

  • Does that mean the "generative syntax" agrees with my analysis?
    – JK2
    Apr 19, 2018 at 4:16
  • 1
    @JK2 Are you aware of the Subject-VP Internal Hypothesis? It was proposed at least 30 years ago. You may want to check out this very short video youtu.be/y49jynm23pQ
    – Alex B.
    Apr 19, 2018 at 14:04
  • Thanks for the video. I've seen some of his videos before! Now I think I see where you're coming from. In any case, this generative syntax based on subject-VP Internal hypothesis does agree with my own analysis or at least very similar to it, right?
    – JK2
    Apr 19, 2018 at 15:47
  • Is ‘crouch’ an unaccusative vP like ‘be’ in other instances of locative inversion, e.g. “under the bush is Bill”? If it isn’t, then isn’t ‘Bill’ supposed to be generated in SpecvP? If the latter is the case, the correct word order could not be achieved because ‘crouch’ would move from V to v and ‘Bill’ would remain in SpecvP, yielding “under the bush Bill crouched.” Apr 21, 2018 at 5:59
  • @Morphosyntax No, you're very mistaken. See 6.4. Unaccusative subjects (pp. 223-224) in Adger 2003.
    – Alex B.
    Apr 21, 2018 at 18:12

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.