An interesting issue has come up in one of my classes yesterday. I was making a point that grammatical rules make reference to structural notions and not to counted entities (this is also known as Structural Dependency Principle). To illustrate it, I used the famous example of yes/no question formation making reference to matrix auxiliary and not to first/second auxiliary. For instance, declarative sentences The beauty who was sleeping was kissed by the prince and The prince was kissing the beauty who was sleeping can be turned into yes/no questions by fronting the matrix auxiliary was: Was the beauty who was sleeping ___ kissed by the prince and Was the prince ___ kissing the beauty who was sleeping, respectively (here and below, the underscore marks the location of the element — here, the auxiliary — that has been fronted). Crucially, it doesn’t matter if the matrix auxiliary happens to be the first or the second one in the declarative sentence. A student told me that it is “illogical” to extract an auxiliary from the embedded clause if we want the whole sentence to be a matrix question. But it only seems illogical because we don’t do that. Here, I’d like to elaborate on this point.
But first, let’s consider the classification of questions. There are two types of questions: yes/no questions and content questions. The former can be answered with “yes” or “no” and the latter require to provide more contentful answers (“who”, “what”, “where”, “when”, “why”, etc.). In English, content questions are formed by fronting the wh-phrase (who, what, where, when, why, which boy, etc.). Another classifications distinguishes matrix questions from embedded ones. For example, the sentence Who said that Mary is married? is a matrix question with an embedded declarative that Mary is married. Conversely, John asked who Mary can marry is a declarative sentence with an embedded question who Mary can marry. Matrix and embedded questions in English differ as to whether they employ subject-auxiliary inversion: matrix questions do and embedded questions don’t.
Now we are ready to examine the issue of whether it is “logical” or “conceivable” to front anything out of the embedded clause in order to make the whole sentence a (matrix) question.
It is true that an auxiliary cannot be fronted out of a relative clause; hence, the ungrammaticality of *Was the beauty who ___ kissed by the prince was sleeping?. In fact, nothing can be extracted from a relative clause! In particular, in content questions the wh-phrase cannot be extracted from a relative clause either: hence the ungrammaticality of *Who was the beauty that was kissed by ___ ___ sleeping? . As it turns out, extraction from a relative clause is universally prohibited. This is known as the Complex Noun Phrase Constraint: a noun phrase where the head noun is modified by a relative clause is a complex unit out of which nothing can be extracted.
However, it is not true that an element cannot be extracted from an embedded clause and fronted to the beginning of the sentence to make the whole sentence a matrix question. It is possible to extract a wh-phrase (who, what, which…, etc.) from an embedded sentence to make the whole sentence a matrix question. For example, take the declarative sentence M said that James Bond is in love with Money-Penny. We can question the object of the embedded clause: Who did M say that James Bond is in love with ___?. What happens here is that an element is fronted from the embedded clause [that James Bond is in love with Money-Penny] — but it is the whole sentence that becomes a question! Notice also that the fact that the whole sentence is made into a matrix question is highlighted by the fact that subject-auxiliary inversion happens in the matrix portion of the sentence: did M say.
This fronting of a wh-phrase from an embedded clause to make the whole sentence into a matrix question is possible as long as two conditions are met. First, the fronted wh-phrase may not be a subject immediately following that. A subject can be questioned/fronted only if that is absent. Therefore, *Who did M say that ___ is in love with Money-Penny? is ungrammatical, but Who did M say ___ is in love with Money-Penny? is fine. Linguists call this the “that-trace effect”.
The second condition is that there cannot be another wh-phrase interfering between the fronted wh-phrase and its original position (i.e., the position where the answer would be). This can be illustrated with the ungrammatical question: *Who did M ask why James Bond fell in love with ___?. What has gone wrong with this question? The wh-phrase why is in between the fronted wh-phrase who and its original position indicated by the underscore. This is known as Wh-Island: the embedded question [why James Bond fell in love with X] is impermeable to extraction, and nothing can get out of it.
To recap, while (auxiliary) verbs can only be fronted from the matrix clause, wh-phrases can be fronted across clausal boundaries, as long as the necessary conditions are met.
Like this post? Please pass it on:
« Thus spoke Shakespeare! Or did he?
Da-da-da… — Duh! »