In English, there are many different verbs which can combine with clausal complements. These verbs can be further sub-categorised as to whether they embed a propositional that-clause, or an embedded question. Some verbs, such as believe, may only embed a proposition (i). Other verbs, like wonder, may only embed a question (ii), and still other verbs, like know, may embed either a proposition or a question (iii).

  • i. a. John believed [that Mary came]
    b. *John believed [who came]
  • ii. a. *John wondered [that Mary came]
    b. John wondered [who came]
  • iii. a. John knows [that Mary came]
    b. John knows [who came]

My question concerns the properties of nouns that can combine with clausal complements, and how they relate to the paradigm above. Note that there are many nouns, such as belief that can combine directly with a propositional clausal complement. They do not require a prepositional element (iv), unlike Noun Phrase complements to Nouns (v).

  • iv. a. The belief [that evolution is false] is thankfully extremely rare.
    b. *The belief of [that evolution is false] is thankfully extremely rare.
  • v. a. *The photograph John is being developed.
    b. The photograph of John is being developed.

Surprisingly, based on my tentative investigations, nouns which combine with questions are rather uncommon, and those that can combine with questions require a prepositional element (vi). In this respect question complements to nouns seem to behave more like DPs than that-clauses.

  • vi. a. *The knowledge [who came] is extremely valuable.
    b. The knowledge of [who came] is extremely valuable.

My question has an empirical aspect and a theoretical aspect: (1) The empirical question is to what extent this set of facts is unique to English. Are there other languages where interrogative clauses can freely combine with nouns? What happens in languages with case-marking for example? (2) The theoretical question is why question complements to nouns would behave more like Noun Phrase complements than that-clause complements.

  • Proposition is a logical term and question a pragmatic one. Are we talking syntax here or what? And what about non-finite complements like infinitive and gerund clauses? Is this only about finite that and embedded-Q complements?
    – jlawler
    Commented Jul 26, 2014 at 17:27
  • @jlawler If it helps, take every instance of 'proposition' and replace it with 'that-clause', and every instance of 'question' and replace it with 'interrogative clause'. I'm mainly interested in finite clauses, yes.
    – P Elliott
    Commented Jul 26, 2014 at 20:26
  • Well, generally the rule is that nouns formed from complement-taking verbs inherit the complement licenses of the verb. This applies to both tensed and non-tensed complements. Tensed noun complement clauses are one of the two constructions that Ross discussed in the Complex Noun Phrase Constraint part of his dissertation (the other was relative clauses).
    – jlawler
    Commented Jul 26, 2014 at 22:23

1 Answer 1


Concerning the first question, German behaves similar to English regarding the content clauses that nouns can and cannot take, e.g.

 1a. die Tatsache, dass er das gemacht hat
     'the fact that he did that'

 1b. *die Tatsache, wann er das gemacht hat
     'the fact when he did that'

 2a. *die Frage, dass er das gemacht hat
     'the question that he did that'

 2b.  die Frage, warum er das gemacht hat
      'the question why he did that'

German has morphological case, but I am not aware that this fact has any influence at all on the distribution of content clauses (that's what I've encountered these clauses called, i.e. content clauses). Whether or not a given noun can take a declarative or interrogative content clause is an idiosyncratic trait of the noun; it is a lexical matter, i.e. part of the valency of the noun. Certainly semantic considerations play a role in determining valency traits, but my guess is that semantics will get you only so far.

The second question receives a clear answer that is purely syntactic in nature (as opposed to lexico-semantic). that-clauses have a distribution that is more limited than embedded wh-clauses. In particular, wh-clauses can easily appear as the complement of a preposition, whereas that-clauses really cannot, e.g.

 3a  *We looked at that he did that.
 3b.  We looked at why he did that.

 4a. *They talked about that he did that.
 4b.  They talked about why he did that.

 5a. *the discussion of that he did that
 5b.  the discussion of why he did that

 6a. *We were struggling with that he did that.
 6b.  We were struggling with why he did that.

These data motivate the stance that prepositions can freely take wh-clauses as complements, but not that-clauses. Of course the pertinent question is why this should be the case. I believe there is a solid syntactic explanation.

This explanation resides with the analysis of that-clauses and wh-clauses (matrix vs. embedded). If one assumes that the head/root of a that-clause is the subordinator that and the head/root of an embedded wh-clause is the wh-element, then the explanation comes into view. The subordinator that does not have the status of of a nominal, whereas wh-elements do (because of their wh-feature). In other words, prepositions are subcategorizing for a nominal dependent; that-subordinators do not have nominal status, whereas wh-subordinators and wh-elements do have this status.

This explanation relies crucially on the stance that the wh-element is the head/root of the embedded wh-clause. So for instance, in the following embedded wh-clause, the wh-element is the head/root of that clause:

7. the explanation about why he did that

By assuming that why is the head/root of the clause why he did that, we can produce a coherent syntactic explanation of the acceptability judgments in 1-7. The wh-feature on why renders why the equivalent of a nominal category, which means that it can be the immediate dependent of a preposition. Since the subordinator that lacks the wh-feature, it cannot be the immediate dependent of a preposition.

I address this issue (i.e. the analysis of matrix and embedded wh-clauses) directly in two of my/our publications. I would be happy to provide copies to anyone who is interested (my email is [email protected]).

And one further point: the overarching issue is touched on briefly in an (in my view convincing) argument against syntactic movement. Observe the following acceptability contrast -- the examples are taken from Wikipedia:

 8a. *We talked about that he was sick for days.  
 8b.  That he was sick, we talked about for days. 

This acceptability contrast consitutes what is known as a movement paradox; it is originally due to Joan Bresnan, I think. If topicalization really involved syntactic movement, then both of these sentences should be acceptable. The fact that 8a is bad demonstrates that the topicalized clause in 8b cannot have originated in its "base" position shown in 8a.

  • Thanks for the great answer Tim. 2b seems to show that a noun can combine with an interrogative content clause without a prepositional linker, unlike in English: "the question why he did" is ungrammatical for me. Interesting. It's also an interesting idea to capture the 'nominal' character of embedded questions by saying that the wh-element is the head - but then how do you distinguish embedded questions from free relatives? The following is ambiguous between the two: "John knows who just arrived" (John knows the answer to who arrived/John is acquainted with the person who arrived).
    – P Elliott
    Commented Jul 27, 2014 at 14:39
  • The NP "the explanation why he did that" is fine for me. English and German are the same in this area. Embedded questions and free relatives have the same basic syntactic structure. In both cases, the wh-element is the head/root. Yes, the ambiguity you point to is definitely real, but the source of the ambiguity lies with the matrix predicate. The predicate "know" can take an NP complement (e.g. "I know the answer") or a wh-complement (e.g. I know what happened). In the case of n free relative complement, the free relative is the eqiuivalent of an NP Commented Jul 28, 2014 at 5:08
  • In the case of a wh-complement, the wh-clause also has the status of an NP, but the NP is now marked with a wh-feature. Note that in German, both the embedded wh-clause and the free relative clause have VF (verb final) word order, a fact that demonstrates that the two have the same basic syntactic structure. Commented Jul 28, 2014 at 5:12
  • Thanks for the explanation. What i still don't understand is how embedded questions headed by a non-argumental wh are still treated as NPs within your system, e.g. whether he left. This still requires a prepositional case-assigner: the question of whether he left (or not). If you have written something elaborating on this system, please do send me it ([email protected]).
    – P Elliott
    Commented Jul 28, 2014 at 11:01
  • The formulation "the question whether he left or not" is acceptable for me. The preposition is optional. In fact I prefer to leave it off. In my publications, I present arguments to the effect that the wh-expression is the root of the embedded clause that it introduces. Thus embedded questions, free relative clauses, normal relative clauses, they all have the wh-expression as their head/root. I have not, however, argued explicitly that the wh-expression has the status of a nominal, although that assumption is a logical extension of my analysis. I'll send you a paper. Commented Jul 28, 2014 at 12:40

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.