Chapter 8: Transformations (III)


8.0. Introduction

The postulation of transformational rules in the previous two chapters not only enabled us to generate a host of grammatical sentences in English that are otherwise excluded by our PSG, but also provided important insights into the nature of those sentence types which otherwise could not have been captured had we attempted to accommodate those sentence types by expanding on our PSG. In this chapter we examine three more constructions in English that call for the postulation of transformational rules. These constructions are so common in our day to day use of English that our discussion of English syntax would be seriously incomplete without including them.


8.2. Wh-questions

Among the most common type of sentences in English are "wh-questions" like the following:

(1) a. What is John buying?
b. What have you done this week?
c. Who did you talk to?

Can these sentences be generated by our PSG plus the transformations we have posited up to now. The answer appears to be yes at first sight, if we consider these sentences to be examples of S', and recall the possible contents of Comp:

(2) Comp ----> that, whether, if, for, which, why, Q, . . .

A reasonable expectation is then that a wh-question might be directly generated by the PSG with the following deep structure:


The surface structure can be obtained once SAI, DO Support and Affix Hopping have applied.

However, this way of generating a wh-question misses an important fact concerning the construction. In the examples in (1), each sentence contains a transitive verb whose object is nevertheless missing from the object position. This state of affairs must obtain just in case the sentence starts with a wh-word. Without the wh-word, a sentence with a missing object would be ungrammatical:

(4) a. *John bought.
b. *You have done this week.
c. *Mary talked to.

And conversely, when a wh-word appears, no object can appear after the verb:

(5) a. *What did John buy the book?
b. *What have you done the work this week?
c. *Who did Mary talk to Jane?

The state of affairs we observe here suggests that a wh-word is first generated in the base (at D-Structure) as a constituent within S, and then moved to the Comp position by a transformational process, which we shall call Wh-Movement. The deep structure underlying (1b) should be:


We can assume that in the deep structure the only relevant element in Comp is the Q morpheme, which has the effect of triggering Wh-Movement and, in the above examples, also SAI. In fact, we can even say that Wh-Movement brings a wh-word into Comp, and substitutes it for the Q-morpheme, thereby giving "content" to the latter. (If no wh-word is moved there, the Q would be interpreted with a "yes-or-no" content, and if embedded, spelled out as whether or if.)

The reason for the hypothesis that wh-questions are derived by movement is essentially the same as that for postulating the passive transformation. By assuming that surface wh-questions in English have deep structures like (6), we can maintain important generalizations concerning subcategorization, s-selection, etc., by applying those generalizations at the D-Structure level.

A wh-question can be a direct question, as in each of (1). A direct question is a direct request for information, and the listener is expected to provide an answer to it. Such a question is formed by preposing the wh-word to the beginning of the entire (main) sentence. A wh-question can also be formed by preposing a wh-word to the beginning of an embedded clause, as in (7) below:

(7) a. I want to know what John is buying.
b. I wonder what you have done this week.
c. I wonder who you talked to.

Clauses like What John is buying, what you have done this week, and who you talked to are called indirect questions. An indirect question does not request information from the addressee. Rather, it serves as an argument of a verb that s-selects it. In (7) the indirect question is the complement of the matrix verb wonder or know. The matrix sentence that embeds an indirect question, as a whole, may be a declarative sentence. Each sentence in (7), taken as a whole, is a declarative sentence.

Note that, unlike direct questions, indirect questions do not require SAI. We shall assume that the structure of an indirect question is otherwise the same as that of a direct question. In the following structure, the shaded portion is the domain of the indirect question embedded as an object of wonder. When the wh-word is moved to the position of Q, an indirect question is formed, as in each of (7).


An interesting property of wh-questions is that the wh-word at the beginning of a given clause need not originate as an argument of that clause. It can originate in a position much more deeply embedded, as in (9):

(9) a. What did John think that I like?
b. What do you think that Mary believes that you have done?
c. Who do you suppose I should talk to?

In these cases, we assume that the wh-word is moved across sentence boundaries, from within a subordinate clause to the beginning of a superordinate clause.


In all the examples we have seen, there is overt evidence that the wh-word has been moved away from their original, deep-structure position. However, there are wh-questions where the clause-initial wh-words appear not to have moved. This is the case when a subject NP is wh-questioned, directly or indirectly:

(11) a. Who will be here?
b. Who borrowed my book?

(12) a. I want to know who will be here.
b. John wonders who borrowed my book.

One possible hypothesis we can draw here is that subject wh-questions do not involve any movement. (Note also that the SAI also appears not to have applied.) There are two reasons, however, to assume that even subject wh-words do move, albeit vacuously, in the formation of a wh-question. The first reason is theoretical. We have assumed that wh-words are moved because they are attracted by the Q morpheme the sentences contain. There appears to be no reason why wh-subjects are exempt from this requirement, so theoretically it is desirable to keep the theory simple and maximally general, by assuming that subjects do move, though in these cases the movement is "string-vacuous" (i.e. invisible on the surface).

The second reason for assuming subject wh-movement is empirical. In fact, there is independent evidence that subject wh-phrases do move. This manifests itself when an embedded subject is directly questioned, as in (13):

(13) a. Who do you guess __ will be here?
b. Who do you think __ borrowed my book?

In each case above, the initial wh-word has been moved from the subject position of the embedded clause. We shall henceforth assume that all wh-words, subject or object, are affected by Wh-movement in the formation of wh-questions.

Indeed, not only arguments but also adjuncts may be moved in forming questions. The wh-adjuncts may originate from the main clause, as in (14), or they may originate from the embedded clause to form indirect questions, as in (15):

(14) a. When did you arrive?
b. Where did you see him?
c. How did you fix the car?
d. Why did you buy those knives?

(15) a. I know when he arrived.
b. She told me where you saw him.
c. Nobody knows how you fixed the car.
d. They wonder why you bought those knives.

Note that questions like the following are ambiguous:

(16) a. When did he say (that) he was hurt?
b. Where did you tell me that you saw him?
c. How did you guess that he fixed the car?
d. Why did you say that he bought the knives?

(16a) may be understood to be a question about when he made the statement that he was hurt, or one about when, according to what he said, he was hurt. Within our model of grammar, this ambiguity follows quite easily. The ambiguity of (16a) arises, because the sentence can be derived from two different deep structures, depending on whether when originated in the matrix clause or the subordinate clause, as in the structures below (details omitted):

(17) a. b.

In transformational grammar, then, both synonymy and ambiguity that arise from structural reasons can be captured quite straightforwardly. Synonymy arises when one deep structure diverges into two different surface structures due to the application of different transformations. Ambiguity arises, on the other hand, when two different deep structures converge on the same surface form.

Note that the ambiguity of sentences like (16) provides an important argument for the postulation of wh-movement, and against an approach that base-generates wh-words directly in their surface, sentence-initial position. Can you substantiate this point by showing that, under this latter approach, there isn't an easy way to structurally capture the ambiguity displayed in (16)?


8.3. Topicalized Sentences

The following sentences differ from normal declarative sentences in that they take the form of a topic followed by a comment about the topic.

(18) a. John's articles, I will never read.
b. That he will rise among the best, I have never doubted.
c. John I am sure will be angry when he finds out about it.
d. This book, I thought you told me that you don't like.
e. John said he would pay the bill soon, and pay the bill, I am sure he never will.

In (a) and (b), an object NP and a complement S, respectively, have been topicalized. In (c) and (d) an embedded subject and object, respectively, have been topicalized. And in (e) a VP is fronted as the topic of the sentence. These sentences bear resemblance to the wh-questions, with the exception that they do not actually involve a wh-word, nor the process of SAI. We shall assume that they are derived via a process of Topicalization. Again, as before, we can assume that the topic is moved from within a sentence into the domain of S', under Comp. The arguments pertaining to subcategorization and s-selection for Wh-movement and Passivization apply equally for the postulation of Topicalization as a movement rule. You should be able to run through one or two arguments on your own.


8.4 Relative Clauses

The embedding structures we have considered up to now generally involve subordinate clauses that play the role of an argument for a verb. Such a clause can be a complement (of verbs like say, think, etc.), or it can be a subject, as illustrated below:

(19) a. That you were late surprised me.
b. What he will say is unclear to me.

A subordinate clause can be used as an (adverbial) adjunct as well:

(20) a. John arrived before Bill left.
b. I will give you the money if you promise me to get lost forever.

In all these cases, a subordinate clause is a constituent of S or VP. In addition, a subordinate clause can be a constituent of an NP as well. There are two common types of subordinate clauses embedded under NPs: NP complements and relative clauses. A clause may be used to complement a noun much in the same way it may be used to complement a verb. Thus, just as the S' "that he won the prize" is a complement of the verb think in (21a), it is a complement of the noun thought in (21b):

(21) a. John thinks that he won the prize.
b. The thought that he won the prize pleases John.

More examples of NP complementation are given in the (b) sentences, alongside the examples of VP complementation:

(22) a. John believes that he was right.
b. John has always held the belief that he was right.

(23) a. The author contended that men are not created equal.
b. The author's contention that men are not created equal has aroused controversies.

In addition to being used to complement NPs, subordinate clauses can also be used as adjuncts to NPs. The most common such clauses are the relative clauses:

(24) a. The man who you saw yesterday is my brother.
b. The book which you bought is very expensive.

In (24a), "who you saw yesterday" is a relative clause modifying the head noun man. In (24b), we have the relative clause "which you bought" modifying book. Relative clauses differ from NP complements in that serve to provide additional information about the noun they modify, but are not required for the completion of the noun's meaning. Relative clauses are on a par with adjectives in their relationship with the head noun, and are sometimes called adjectival clauses, in parallel to adverbial clauses which are adjuncts to VPs or sentences.

One important structural difference between NP complements and relative clauses is that the latter, but not the former, involve the movement of a relative pronoun to the beginning of S'. Although the situation is more complicated, we shall assume for the moment that both are S' dominated by NP following their head nouns. The main difference is that NP complements contain the complementizer that, whereas relative clauses require a relative pronoun in the position of the complementizer:

(25) NP complementation structure

(26) Relative clause structure

We also assume that, in the derivation of relative clauses, a process of Relativization moves the relative pronoun from its deep structure position into its surface position in C(omp). Thus, the deep structure of (26) is actually one in which the relative pronoun who occurs as the object of saw. A more detailed deep structure of the subject NP in (26) is (27):


Note that the process of movement, which we call Relativization, bears resemblance to Wh-Movement discussed above. Relative pronouns also generally are wh-words. Both objects and subjects may be relativized, just as they can be wh-questioned:

(28) a. The woman who loved John was betrayed by him.
b. The man who saw me yesterday is my brother.

And relativization can go "long distance" just as Wh-Movement can:

(29) a. This is the story which I told you that you should hear.
b. The man who I believe Mary thought she had seen has disappeared.

In addition, not only can an argument be relativized; an adjunct can be relativized as well:

(30) a. This is the day when he arrived.
b. This was the reason he was late.
c. We will meet at the restaurant where we last met.


8.5. Summary and Conclusion

In this chapter, we have seen that wh-questions, relative clauses and topic-comment sentences in English provide evidence for the postulation of movement transformations. Sentences containing such structures are not base-generated by the PSG as they are on the surface, but are better understood if we look beyond the surface into their deep, underlying structures.

More generally, in this and the previous two chapters we have demonstrated the power of a grammar that incorporates a transformational component in capturing certain important insights, or linguistically significant generalizations, that represent part of our knowledge of grammar, a window into the nature of the human mind.

In the beginning of this book, we started our inquiry into the nature of language by reflecting upon what we know when we know a language. We saw that much of what we normally take for granted includes a large amount of our tacit knowledge that, upon closer examination, appears quite extraordinary and yet clearly systematic. We then proceeded to develop a grammar of English on the basis of our own observations, and have since found various ways to explicitly describe, or formally characterize part of that tacit knowledge, thereby providing formal insights into the nature of that knowledge. The ability to formally characterize what is normally taken for granted makes the difference between experts and laymen, and between science and no science.

What we have learned in these eight chapters is, however, a very small step towards the goal of understanding the nature of language as a mirror of the human mind. By discussing our tacit knowledge in explicit, formal terms, we have often encountered, and put aside, many problems we have not been able to answer. One positive outcome of this introduction, we hope, is that it has enabled us to realize how very little we actually know about the language that we thought we knew well, and how much more we still do not know. This is a good start, because it is only by being aware of the many problems that we can begin the task of building a theory of grammar that explains them.


Homework 8