He will support whichever candidate wins.

In the NP hypothesis, whichever candidate wins functions as the object of the verb support, and has the noun candidate as its head, so it makes sense to say that whichever candidate wins is a noun phrase (NP). Now, I wonder what N' is in the NP in X-bar theory.

Is N' candidate wins or the same as the NP whichever candidate wins?

Alternatively, how can whichever candidate wins be analyzed under the DP hypothesis?

  • It's a relative clause, no?
    – Draconis
    Dec 4, 2023 at 3:51
  • 1
    @Draconis The sentence is equivalent to He will support the candidate who wins, no matter which it is. Here, who wins is a clause, but the candidate who wins is not a clause but an NP.
    – JK2
    Dec 4, 2023 at 4:25
  • 1
    A paraphrase is not an analysis.
    – Alazon
    Dec 4, 2023 at 7:31
  • @Alazon The paraphrase does reveal a syntactic aspect of the OP, namely, that whichever candidate wins has the function not just of the relative clause who wins but of the antecedent candidate. The paraphrase is there to facilitate an analysis, not to replace one.
    – JK2
    Dec 4, 2023 at 9:36

2 Answers 2


The example has been misanalysed in the question.

Your question is about the object of the verb "support" in a structure "He will support X", with X = "whichever candidate wins". This phrase X is a relative clause of a special sort, a free relative clause or internally-headed relative clause. In generative syntax, it has the category CP. So the noun "candidate" is not the head. – In general, relative clauses, if introduced by a relative pronoun (here: whichever) have an invisible C as their head; otherwise the word "that" may be seen as an overt such C head (as in other that-clauses). Compare two simpler relative clause structures and then your above example, which is more involved:

  • (the candidate...) CP [ C:that [ I prefer -- ]]
  • (a candidate...) CP [who C [ - wins]]
  • CP: [whichever candidate C [- wins] ]

The last structure is introduced by a complex phrase, parallel to the structure in the question

  • CP [ [Which candidate] C:do [you prefer - ]] ?

So, "which(ever) candidate" is an NP or DP (depending on your theory), but the whole big thing is a clause, and "candidate" sits in its specifier position (SpecC).

There has been discussion, however, whether the whole relative clause construction, if it has the role of a direct object, should be a kind of NP/DP. Generative syntacticians have proposed structures in which the clause is embedded under an invisible nominal head, for example, it has been analysed as a determiner followed by a relative clause (e.g. in Grosu (2016) Nat Lang Linguist Theory 34:1245–1280, DOI 10.1007/s11049-016-9333-0):

  • DP: [ D CP:[whichever candidate wins]]

But this way of constructing a DP is different from what you assumed in your question...

  • 1
    Thanks. Dennis Ott (2011) argues this: "the dual nature of FRs is a consequence of cyclic Transfer. FRs are clausal in that they are headed by a C head; however, since this C head does not bear any interpretable features, it does not ‘‘survive’’ Transfer, leaving its specifier as the only remaining material at the next cycle (matrix vP)." research.rug.nl/files/62651382/… Doesn't this mean that whichever candidate is the head of the free relative?
    – JK2
    Dec 5, 2023 at 2:30
  • Is Ott's analysis the background assumed in your question? Is this what you call "the NP hypothesis"? Then your question should be rewritten in order to focus on that analysis. Ott's paper is interesting and it sheds light on the double nature of the wh-phrase in FRs – but it is far from clear how it relates to "X-bar theory". It essentially assumes the presence of unanalysable material at later cycles. Maybe this is exactly your question? It would be a question about "phase theory" as a whole then. (And is it plausible that a rel-clause has no clause-type features...?)
    – Alazon
    Dec 5, 2023 at 6:21
  • 1
    "Free relative clause" is a misnomer. Although it contains a clause it's clearly an NP. The "clearly" here arises from the fact that such fused relative NPs or whatever we're going to call them, can be singular or plural and require corresponding verbal agreement. Consider: "Whichever pens we used were discarded" and "Whichever pen we used was discarded". That's why the only vaguely plausible clausal analysis is in your second answer. A better approach would be to simply accept that it's an NP containing a clause and that the head is candidate! Dec 9, 2023 at 13:34
  • "The head is the N", yeah, but the question was about whether the rest that we see is a dependent of that head, and here we're in trouble again.
    – Alazon
    Dec 9, 2023 at 16:28
  • JK is right that "whichever candidate wins" is an NP, not a clause. There's plenty of evidence to support.
    – BillJ
    Dec 10, 2023 at 10:37

Moving to a second phase :) of dealing with the question – in a comment to the other answer, JK2 cites an analysis of free relatives that proceeds in cycles, following some version of Chomsky's later theories. I'm writing a second answer so comments can be attached to it.

The phase idea assumes that the first round of derivation proceeds in the way I sketched in my first answer, and then only the C-head and specifier would be visible when the relative clause starts interacting with the outside world of the clause. The idea is that only the "edge" of the clause interacts with the outside, so the part visible from outside would only be the second version below:

  • CP:[whichever candidate C [- wins] ]
  • [ [NP/DP whichever canddate] ((C)) ... wins ]

Furthermore, the idea says that C in relative clauses does not have an interpretable clause type (as questions would have it) and is subsequently lost and all that remains visible is the specifier, and since this is an NP-or-DP, the whole thing will have the properties of an NP/DP instead of CP. Interestingly, as Ott (2011: 186) mentions in a footnote, the whole thing could also become PP-like for the same reason:

(I’ll live) [FR[PP in whatever town]i you live ti]

The "in" must be inside the relative clause because the verb "live" needs it (in the trace position t-i). The matrix occurrence of "live" also needs its complement to be a PP... So, descriptively speaking, the free relative clause always behaves just like its specifier.

In my view, the question formulated by JK2 still doesn't really arise, and cannot be answered in terms of "X-bar theory". As I understand the phase stuff (not my speciality), the rest is in the shade at later cycles and its structure is inactive, so to speak. I don't see a reanalysis of the whole phrase structure so as to make the material of the core clause a dependent of the relativiser. I still see standard NP/DP "whichever candidate" or PP "in whatever town" and no other phrase structure accessible to analysis. But this is more like a question: does phase theory assume that? Maybe there are experts of that framework around, who can enlighten us?

  • I'm not an expert, but as far as I understand the paper, the whole free relative construction boils down to an NP/DP, so I think the question is still valid.
    – JK2
    Dec 8, 2023 at 5:26
  • But there's no reanalysis of the internal structure. It's only about visibility of the "edge", isn't it?
    – Alazon
    Dec 8, 2023 at 9:47
  • The answer is difficult to understand. It may be as coherent as it can be given the theoretical assumptions underlying the question. For me, though, the fact that the analysis of such a frequently occurring and seemingly simple structure can be so difficult to penetrate casts doubt on the plausibility of theoretical apparatus assumed. Jan 6 at 6:01

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.