3.2.3 Traces

In the previous sections of this chapter we have concerned ourselves with the positions in which elements originate at D-structure and the positions they move to at S-structure. We might term these positions the Extraction site and the Landing site of a moved element. What is the status of the landing site at D-structure and the extraction site at S-structure? In (58) we introduced a restriction on movement called the Structure Preservation principle, which states that movements are not allowed to alter the basic X-bar nature of a structure. The result of this restriction is that the structure cannot be much different between D- and S-structure. In particular, we would not expect landing and extraction sites to appear at one level of representation but be absent at another. Thus, if we consider a passive structure in which the object moves to subject position, we can expect the object and subject positions to be present at both D- and S-structure:

These representations indicate that the positions marked [_DP e] are present in the structure, but empty and thus the movement does not change the structure but merely moves things about within the framework it provides.

There is reason to believe however, that the two empty positions in (92) are not the same as they display different properties. Consider the empty subject position in (92a). As this is a position to which something moves it would not be reasonable to think of it as being filled by some other element before the movement takes place. If there were something in this position at D-structure, presumably it would have to be deleted to allow the object to move into the position as general principles of structure do not allow two elements to occupy the same structural position. But if this element is always deleted, how could we ever be aware of its existence, let alone its nature. Moreover, if it were possible to delete elements in a structure to allow others to move into the vacated positions, we would expect far more movement possibilities that we actually observe. We would be able to move an object into a subject position of any verb, not just the passive ones:

Obviously, this is a situation we want to avoid and so we need to strengthen the Structure Preservation principle to prevent things in a structure from being deleted willy-nilly. Suppose we assume that lexical material that enters a structure cannot be altered by a movement. This maintains the Structure Preservation principle given that the lexical items that are inserted into a structure determine that structure to a great extent through notions of projection and selection, but it also prevents the deletion of lexical information once it has been inserted into a structure. This principle is called the Projection Principle:

What this means is that anything that is inserted into a structure from the lexicon cannot change from one level of representation to another. If a verb is put into a structure, nothing can delete or alter this verb, turning it into a noun, for example. Also no movement can alter a verb’s selectional properties: a transitive verb will remain transitive at D-structure and S-structure even if the object is moved.

Under these assumptions, it must be that a passive verb loses its subject before it enters into a structure. There are numerous ways in which we might suppose that this can happen, but we will put the matter aside until we are in a better position to understand it. The general point is that as a result of being passivised, a verb fails to assign a Θ-role to its subject and hence this position is absolutely vacant at D-structure. So with regard to the empty subject position in (92a) we can say that this position, whilst being present in the structure, is simply devoid of any contentful element and hence is vacant to be moved into.

Now consider the nature of the empty object in (92b): the extraction site of the moved object. By the Projection Principle, this object position must remain in the structure and cannot be deleted otherwise the transitive verb would find itself without an object and hence would be sitting in the structural position of an intransitive verb. As this would alter the lexical nature of the verb, we conclude that it would be impossible. Yet if this position were simply vacant, like the subject position is at D-structure, we might expect that it could be the landing site for some other moved element. This, it turns out, is not true at all. Consider the following analysis which indicates several movements step by step:

Both of these derived structures seem to be grammatical, but importantly they do not mean the same thing. In (95b) Fred is interpreted as the one who is helped and the interrogative pronoun who is the one doing the helping, as is indicated by the D-structure in (95a). But in (95c), Fred is the one doing the helping and who is the one helped. Under the assumption that Θ-roles are assigned at D-structure, it cannot be the case that (95c) was formed from the D-structure (95a), but must be related to another D-structure, i.e. (96a):

The fact that (95c) cannot be interpreted in the same way as (95b) leads us to conclude that the movement indicated in the former is impossible and that the object cannot move into the vacated subject position.

The overall conclusion of this discussion then is that the empty positions that are present at D-structure are of a different nature to the empty positions present at S-structure which are created by movements: D-structure empty positions are vacant to be moved into, S-structure empty positions are not. Obviously this demands an explanation.

One possible account of the nature of empty extraction sites is that they cannot act as landing sites for subsequent movements because they are occupied. As two elements cannot occupy the same position and as we are not allowed to delete material, this would block movement into this position. There are two problems that this assumption faces: what element occupies the extraction site and why cannot we see it? Given the above discussion, there is no choice as to the identity of the element that occupies this position: it must be the moved element itself. No other element could either be moved into this position or be inserted into it from elsewhere without drastically changing the lexical information represented by the D-structure and this would violate the Projection Principle. But then we seem to be forced to accept that one element can occupy two positions.

We can get some understanding of this situation if we make the following assumption: when an element moves, it leaves behind a copy of itself in the extraction site. This copy is called a trace and is envisaged to be identical to the moved element in terms of its grammatical and semantic properties. Thus the category of the trace, its role in the thematic structure of the sentence and its referential properties are the same as the moved element. The main way in which the trace differs from the moved element is that the trace has no phonological content and hence is unpronounced. We have already considered the possibility of phonologically empty but grammatically active elements when we discussed imperatives at the beginning of this chapter. A trace is another such element.

Traces are typically represented by a t, which bears an index which it shares with the moved element, both to link the trace and the moved element and to demonstrate that they have the same reference:

The S-structure representations here demonstrate the movement of an interrogative pronoun from two different D-structure positions, marked by the trace. In (97a) who moved from object position and hence the sentence is interpreted as a question about the one who was helped. In (97b) on the other hand who moves from the subject position and hence the question is about the one who does the helping.

There are two views concerning the nature of traces. One is that a trace is related to but independent of the moved element. From this point of view a trace is a little like a pronoun referring to the moved element:

In these examples, the trace and the pronoun sit in object positions and refer to the subject in much the same way. From this perspective, the trace can be seen as having properties of its own independent of the moved element. We will see that there is some truth to this idea. However, traces are not like pronouns in one important way. As we see in (98b), the pronoun and its antecedent represent two different arguments, though they both refer to the same individual. Thus Harry is interpreted as the one who does the helping and himself is interpreted as the one who gets helped. With the trace in (98b) however, there is only one argument interpreted here: the one who was cheated. Because the verb is passivised, the subject’s Θ-role is not present and so no element is interpreted as agent. Thus, it is as though the trace and the moved element share the same Θ-role, which strictly speaking should not be possible due to the Theta Criterion. From this perspective, the trace and the moved element seem to be interpreted as a single element, capable of bearing a single Θ-role. This single element is, however, spread out across a number of positions in a structure.

The notion of a chain might be useful here. We can see a moved element and its associated traces as a single object made up of several parts: like a single chain is made up from different links. Extending this analogy further, we can then refer to the different parts of a movement chain as the links of the chain. Thus, the movement in (99a) can be said to contain the chain represented in (99b):

This chain has two links: this sentence and the trace. We say that the moved element is at the head of the chain, while the trace is at its foot.

There is one more point we need to make in connection with positions involved in movement. Above we said that the landing site for a moved object in a passive structure is an empty position: the subject. This seems to make sense given that the subject position is an integral part of the sentence, without which we would not have a complete sentence. However, it is not always easy to identify a landing site as something that exists before the movement takes place. For example, consider the case of PP preposing, in which a PP is moved to the front of the clause:

This PP does not move to the subject position, which is already occupied by Petra. The position it moves to is to the left of the subject and it is not really a position that we easily claim to be integral to all sentences given that in many it is not filled at all.

Furthermore, this position has a number of things in common with an adjunct, in that an unlimited number of elements can undergo preposing:

If we were to propose that these movements put the moved elements into empty positions, we would have to suppose the existence of an indefinite number of empty positions at the front of the clause which sit there waiting for something to move into them. This does not seem a reasonable assumption.

Moreover, there are instances that we might want to analyse as a case of movement where an element moves to a position that is blatantly not empty. For example:

In (102b) it might be claimed that the negative element moves to the position occupied by the modal auxiliary and the two somehow join together to become a single word. Again there is something like an adjunction formed in this case with a word being created from two independent words, like a compound noun (in this case the ‘compound’ is a verb).

Because of examples such as these it has been proposed that there are two types of movement. One, known as substitution, moves an element into a vacant position. The other, called adjunction, creates an adjunction structure by the movement. Thus, PP preposing might be argued to move the PP to a position adjoined to the left of the clause:

Similarly we might propose that some movements can move words to adjoin to other words, as is the case of the contracted negation:

The irrelevant aspects of the analysis, such as the extraction site of the negative, need not detain us here. The important point is that the negative is moved from one position to a site adjoined to the modal.

One might wonder if adjunction conforms to structure preservation as it does seem to alter the structure from its D-structure condition. However, it should be noted that adjunction does not alter lexically determined aspects of structure and so is perfectly compliant with the Projection Principle which supersedes structural preservation. Moreover, adjunction is something which X-bar theory allows for and hence to create an adjunction structure is not to create something that violates the possible X-bar nature of the structure. In this way, adjunction movement does not radically alter structure and can be seen as structure preserving.