questions - Can words like "what" be the subject of a sentence?

In a question like "Who hears a noise?", is the subject of the sentence who?

I can think of a few tests for subjects like: "the subject is the phrase that inverts with the auxiliary to form a question". But this is a question and there is no auxiliary. We could say that's because it's a subject, but that would be just to presuppose that it's a subject in the first place.

I can think of tests like: "the form of the verb will agree with the number of the subject", but on the other hand who does not seem to have fixed grammatical number, and it sometimes seems like who might actually be agreeing with noun phrases later in the sentence:

• Who are the most prolific writers of our age?


• Who is the most prolific writer of our age?

There are also tests like "the subject is usually the first noun phrase in the sentence". But of course if we have a question like:

• Who have you bitten?

... you, not who seems to be the subject. The first noun phrase test is not very good.

I also know that some theories of grammar say that there is a gap in wh- questions that the wh- word is extracted from. So can this gap for instance be the real object of a question? If so, is there a gap in questions like mine functioning as subject?[See community wiki post in the linked to question]

How can we show whether who is really the subject of my original example question? And how about other wh- words like what? Can they (also) function as subjects?

Answer

Both sides of this question have been argued in the linguistic literature. "who" could be a topic, and then the sentence structure would be

[who [ __ hears a noise]]

on the analogy of other wh-questions with a wh-word moved to the top of the structure and leaving a gap, __, where it once was. Or, perhaps questions whose subject is a wh-word simply don't need to be changed by moving the wh-word to the front, because, well, it's already there. In that case, we have

[who hears a noise]

and "who" is simply a subject.

There are a number of grammatical theories which do not permit stating grammatical relations in terms of word order, but only structural relationships, so such theories would presumably not recognize the logic of the argument that a wh-word subject must remain a subject in a question, because it is already at the front of the sentence, where it needs to be. (Such "order free" theories are Chomsky's latest theories (I think), dependency grammar, relational grammar, and GPSG.)

You might think that the subject-agreement in the verb "hears" with the singular "who" shows that "who" is a subject. And perhaps that is evidence, but a follower of the east coast school of linguistics would assume that the gap created by extracting the subject is a "trace" which is coindexed with the former subject, so the verb agreement can still be correctly described.

There is an argument for the second no-movement treatment quoted and discussed in the book Generalized Phrase Structure Grammar, and the argument itself (if I recall correctly, due to Pauline Jacobson) is based on so-called parasitic gaps. In the Wikipedia entry for Parasitic gap, this illustration is given:

Which explanation did you reject __1 without first really considering __2?

where the first gap must be higher up in the structure tree than the second (parasitic) gap. The relevance of parasitic gaps here is that they give us a diagnostic for detecting a gap in subject position. A subject gap, since it is highest in the structure of its clause, should license a parasitic gap elsewhere in the clause. But if there is no subject gap, there won't be any parasitic gap, because parasitic gaps depend for their existence on a gap higher up.

So, we can construct a test case from the above parasitic gap sentence by making it a passive whose subject is the wh-expression:

*Which explanation __1 was rejected by you without first really considering __2?

This is ungrammatical. If there were really a subject gap, it should have been okay. So, we can conclude that there is no subject gap and that a wh-subject in a question remains in place. I know of no evidence on the other side of this question, so that is my conclusion.