Patrick D. Elliott

A note on if-questions

October 15, 2016 Tags: linguistics

If-clauses are weird. They have a couple of prominent uses: as the antecedent clause in a conditional if…then… statement (see (1)), and as an embedded question (1).

  1. If Britta is taking this class, then I’m not.
  2. Jeff asked if Britta is taking this class.

Here, I’m interested in the latter cases, which I’ll call if-questions. Their distribution is constrained in some interesting ways. Under rogative1 predicates, if-questions seem, more-or-less, to be in free variation with whether-questions.

  1. Jeff {wondered|asked} {if|whether} Britta is taking this class.

Elsewhere however, the similarity between whether-questions and if-questions breaks down. Unlike whether-questions (and interestingly, much like that-clauses), if-questions aren’t permitted as the complement of a preposition.

  1. Britta thought about {whether|if|that} she should attend class.

Furthermore, if-questions generally seem pretty bad as subjects, unlike both whether-questions and that-clauses, which are attested, under certain conditions, in this environment.

  1. {Whether|*if} Jeff attends depends on whether Britta attends.
  2. That Britta is attending class worries Jeff.

These facts are clearly telling us something interesting about the syntactic properties of if-questions vis a vis other varieties of embedded clause, maybe specifically about their case properties. I don’t want to speculate more about this here.

Moving on, if-questions show some extremely interesting behaviour under responsive predicates, such as know and admit. These facts were first uncovered and analysed in a 2001 paper by Adger & Quer (2001)2, who claim that if-questions exhibit polarity sensitivity under responsive predicates. The judgements are subtle, but the following contrast is fairly illustrative.3

  1. *?Jeff told me if Britta left.
  2. Jeff didn’t tell me if Britta left.

Adger & Quer develop a theory tailored to derive this kind of contrast. I have some issues with their account, but I’ll leave that for another post (maybe!). For the time being I just want to note that the predicate imagine poses an interesting problem for the distribution of if-clauses, and its certainly outside of the scope of Adger & Quer’s account. First, note that imagine doesn’t embed whether-questions at all (9). The polarity of the matrix clause doesn’t seem to make a difference (10). However, when the matrix clause is an imperative (11), or otherwise a request directed at the addressee (12), an embedded if-question is licensed, despite the fact that an embedded whether-question still isn’t.

  1. Jeff imagined {whether|if|that} Britta left.
  2. Jeff never imagined {whether|if|that} Britta left.
  3. Imagine {*whether|if|that} Britta left.
  4. Can you imagine {*whether|if|that} Britta left.

Adger & Quer of course don’t consider this kind of data, but it’s difficult to see how one could extend their account to this, since as far as I can see, they predict if-questions to be licensed in a subset of the environments that whether-questions are.

Zooming out a bit, how the hell do kids acquire these kinds of subtle distributional facts, given that whether-questions and if-questions seem, roughly, to mean the same thing?

More selectional puzzles soon! And, as always, my deepest apologies to the non-linguists.


Adger, David & Josep Quer. 2001. The syntax and semantics of unselected embedded questions. Language 77(1). 107–133.

Heim, Irene Roswitha. 1982. The semantics of definite and indefinite noun phrases. PhD thesis.

  1. Heim (1982) classifies predicates that embed questions, broadly, as either rogative or responsive. Rogative predicates are those which may embed interrogative clauses, but not declaratives, like wonder and ask. Responsive predicates on the other hand may embed both interrogatives and declaratives, such as know and tell.↩︎

  2. This is behind a paywall (, but you can apparenly download the .pdf here for free.↩︎

  3. This isn’t much discussed much by Adger & Quer (2001), but (8a) is almost certainly acceptable under the reading that can be paraphrased as “If Britta left, then Jeff told me (that Britta left)”. I assume that this is a possible analysis of (8a), since tell independently allows Null Complement Anaphora, as in “Britta left. Jeff told me.”↩︎