Stephen van Vlack

Sookmyung Women`s University

Graduate School of English

Introduction to Linguistics

Fall 2003

Answers for Cook and Newson, `Chapter 7b`.

1. What does it mean for two elements to be bound?

The definition of bounding is really quite simple. It revolves around the familiar items of c-command and co-indexation. An element will bind another if it c-commands that element and the two are also co-indexed. Co-indexation can only occur when the two items share the same set of agreement features. Look at example A, below.

A. John_i takes very good care of himself_i.

In the above example, John can be said to be bound with himself because John c-commands himself and the two or co-indexed. There are a lot more things that we can say about bounding in relation to local domains, especially looking at the example below, but the most important thing is just remember is what bounding means and that it is a structural relationship which ties one element to another in a structural way.

Now look at B, below.

B. John_i wants [to take very good care of himself_i].

Here we have bounding across a clause boundary. According to the rules of binding theory, John and himself in B. should not be able to be co--indexed. At this point, before you move onto the next question, I want you to think about how this might be made possible, for clearly our intuition tells us they deserve to be co-indexed. Obviously, following te example in B, we need to change our concept about what a local domain is and how binding actually works. Such considerations lead us to the next question.

2. How does Binding Theory relate to empty categories, including PRO?

It has been observed that empty categories, like traces, for example are subject to the same restrictions that certain elements in the framework of binding theory are subject to. As case in this point, look at the examples below.

C. It was believed the porcupine bit the armadillo.

D. The porcupine_i was believed t_i to bite the armadillo.

E. *The armadillo_i was believed the porcupine bit t_i.

According to examples above (C. is there to give you better idea of what the underlying representation of D. and E. might be like), it is possible to move a DP from the subject position of the dependent clause to the subject position of the preceding (in this case) matrix clause. Movement, as shown in E. is blocked, however, when the DP is in object position. We can see striking parallels to this in looking at the presumed behavior of anaphors as described in binding theory. This is shown in F and G below.

F. The porcupine_i believed himself_i to bite the armadillo.

G. *The porcupine_i believed the armadillo_j to bite himself_i.

So, is possible to see the traces involved in DP movement seem to follow the same restrictions applied to anaphor binding in binding theory.

Now, let's look at another type of movement: WH-movement. Look at H and I below.

H. Who_i did the elephant say the porcupine bit t_i?

I. Who_i did the elephant_i say the porcupine bit_i?

In the examples above, it is possible to bind traces with the Wh-word, but not with anything else. In example I., when we co-index who and its trace with the DP the elephant the sentence is rendered ungrammatical. It seems the trace of a wh-word can only be bound by that wh-word. Taking this one step further, it seems that wh-traces must be free from virtually everything. This is basically the same as the behavior of r-expressions of principal C of binding theory.

Now to move on, the empty category pro acts, not surprisingly, like a pronominal in binding theory. Its counterpart large PRO, however, behaves a little differently. Because PRO must be ungoverned, PRO can be both free and bound. The simplest way of showing this is to give examples of what PRO can be replaced by. Look at J and K below.

J. The armadillo_i wants PRO_i to eat the hedgehog.

K. The armadillo_i wants himself_i to eat the hedgehog.

In the examples above, we see that we can replace PRO with a reflexive or anaphor. Now look at the samples in L below.

L. PRO_i to bite the hedgehog_j would be a big mistake.

La. *PRO_i to bite the hedgehog_i would be a big mistake.

M. For one_i to bite the hedgehog_j would be a big mistake.

Ma. *For one_i to bite the hedgehog_i would be a big mistake.

Here, in L and M, we see that PRO acts as a pronominal, which must be free in its domain. Following this we, therefore, must come to the conclusion that PRO behaves like both a pronominal and anaphor in relation to binding theory. Since PRO is both bound and free in its governing category, it seems that PRO simply cannot have a governing category. The contradiction in being both bound and free erases the possibility of this. Now if PRO does not have a governing category than it cannot be governed.

Looking at all this we can easily see that empty categories, and particularly traces (and this is the most interesting thing for us) are able to be equated with some of the principles of binding theory in that different empty categories seem to behave like the different elements of binding theory. This is an important step for us to take in trying to reduce grammar to a small set of principles which function in different realms of language. It also serves to bring empty categories into the fold and explain how they really are part of language.

3. What is bounding?

In order to understand what bounding is it is a big help to actually think about the meaning of the word itself. The verb bound has two different meanings. The first meaning that most people know is bound as the past tense or participle form of the verb 'to bind'. The other, less familiar, meaning of bound is as a verb derived from the noun bound meaning 'a very far jump'. The verb bound, therefore, means to run by taking a series of long jumps as in the sentence, 'The elephant bounded across the room'.

The meaning of bounding in linguistic theory is almost the same. Bound, as we went over above with binding relations, indicates a special relationship that one element holds over another in a structure. Also, bounding has to do with movement. What it (bounding) basically says is that no element can move too far and that if an element intends to move far it must do so in stages. That is, movement often occurs in a series of short jumps, rather than one great leap from one location in a structure to another. To return now to the definition of the word bound that was presented above, we can say that elements often bound (or hop) through the structure of an utterance. If all this is hard to understand just think of bounding as a way of limiting the distance that elements can move in a structure.

4. How does the concept of bounding work to limit movement?

Central to the idea of bounding is the idea that there are restrictions on movement. These restrictions are called bounding notes. According to the principal of subjacency, an element can only move past one bounding note at a time. Look at the utterances in N and O below.

N. Who does the armadillo thinks the porcupine bit?

Na. Who_i does [AGRP the armadillo think [CP t_i [AGRP the porcupine bit ti]]]

O. *Who does the armadillo believe the elephant porcupine bit?

Oa. *Who_i does [AGRP the armadillo believe [DP the elephant [CP t_i [AGRPthe porcupine bit t_i]]]]

Looking at the examples above, we can see that in N. we can co-index who with two traces. This does not seem to be possible in example O. The difference in the structures between these two sentences is the inclusion of a DP object (the elephant) in the second clause. This DP object must be working as a block on the movement of the Wh-word. It is actually quite simple. The inclusion of two DP's in the same clause means that there are two bounding notes which the Wh-word would have to cross in order to get to its final position. Since there are two bounding notes, movement is blocked in the O. example.

5. What is proper government and how does it relate to the Empty Category Principle (ECP).

In order to understand how traces move we need to take a serious look at how they are governed. This will require us to expand and fine-tune our notion of government. We can do this through the Empty Category Principal (ECP).

We have already seen that there are rather large differences in how objects and subjects are governed. Objects are governed by the preceding VP, which is an overt lexical category. Subjects, on the other hand, are governed by AGRP, which is a functional category. Studies of traces have led to the ECP, which states that all traces must be governed by lexical heads. This condition is often stated in terms of proper government. One of the conditions for proper government is that the element in question is actually governed by a lexical ahead.

As long as we are dealing with the movement of objects we don't need to worry too much about government but for movement from subject position this is a major problem. We, therefore, have to rethink our rather limited idea of government as being licensed only by heads. Linguists have proposed that traces can be properly governed if they are governed by the elements with which they are co-indexed. This is called antecedent government because it is the antecedent which ends up governing the trace.

P. John loves his sister.

Pa. [CP [John_i AGPR loves_j [TP t_j [t_i VP t_j [DP his sister]]]]]

In P, above, we see that the NP (John) does govern its trace in Spec VP position. There's nothing which occurs in between John and its trace which could possibly block John from governing its trace. Likewise, if we look at VP movement we see how the verb bounces (bounds) from VP to TP and finally to AGRP position. It should be noted that the verb governs its trace in the TP position and that the trace in TP position governs the trace left in VP position.

In looking at governing in this way we're saying the government is a special relationship between two elements and that interference from another element (usually in between) will not be tolerated. Now look at Q below.

Q. Who will clean the mess?

Qa. [Who_i CP [t_i AGRP will [t_i VP clean [DP the mess]]]]

In Q we see that the wh-word who bounds from Spec VP to SPEC AGRP to finally settle in SPEC CP position. The wh-word has no problem governing is traces because there's nothing intervening or there's nothing which comes between them (all the parts) which could block the governing. Look, however, at R below.

R. * Who that will clean the mess?

R. is ungrammatical because the CP position is filled with a potential governor (that). The complementizer then blocks the Wh-word from governing is trace because the complementizer is a potential governor itself.

6. What is subjacency and how does it work?

We have already seen above how subjacency (together with the concept of bounding) works to restrict movement. In short, subjacency is the idea to movement can only occur across one bounding node at a time. We also said that the bounding notes for English (at least) are AGRP an DP. We then combined the idea subjacency with the ECP, and found that a very powerful tool was created in defining how movement is restricted. This is exactly what Rizzi set out to do it is concept of Relativized Minimality, as we shall see below.

7. How does the idea of Relativized Minimality work to limit movement?

The idea of Relativized Minimality works by trying to incorporate, in one system, the ideas of bounding and the ECP. Central to this idea is the concept the government is a special relationship that holds between two elements. This means the one element can be governed by only one element. Dual governing is not allowed. In this theory idea what can be a governor depends on what is being governed.

The first main thing we need to look at is the Head Movement Constraint (HMC). The HMC claims that heads cannot be moved over the top of other heads.

S. have_i the porcupine would t_i died

is ungrammatical while the utterance,

T. would the porcupine have died

is perfectly acceptable.

Similar phenomenon is reported in wh-movement and argument movement in that a wh-item cannot be moved over another wh-item and one argument cannot be moved over the top of another argument. This is shown in U and V below.

U. *Who_i do you find out what_j t_i bit t_j

V. *Bob_i seems it to be likely t_i to cry

But we can still move Wh-words over heads and arguments just as we can move arguments over heads and heads over arguments. It, therefore, seems that movement is blocked depending on the type of movement that is to take place. In order to understand relativized minimality we have to go back to the ECP. The proper governor for the trace of the head is, according to the ECP, the moved head. We can, therefore, conclude that only heads can properly govern the trace of a head. Any head coming in between the moved head and its trace will block government by that moved head. This us used to explain why heads cannot leap over other heads. This can be stated very simply as: different types of movement will be blocked by an element of the same type coming in between them.