grammatical foundations

Raising and Control

(39)evidently *(it) seems that the electrician found a mouse

The verb seem is one which takes a clausal complement but it has no thematic subject. In this case the subject position is filled by a meaningless it, known as a pleonastic subject. This subject, like all subjects in English finite clauses is obligatory. This suggests that the obligatory nature of the subject is more than a semantic condition that arguments need to be realised. In fact there seems to be a grammatical requirement that clauses have subjects. This condition has been called the Extended Projection Principle (EPP). Recall from chapter 3 that the Projection Principle ensures that the lexical properties of heads are projected into the structure at all levels of syntactic representation. Thus if a verb requires an object as a lexical property, it must have an object at D-structure and at S-structure. The Extended Projection Principle claims not only this, but that the subject position must be present at all levels of structural representation and moreover that it must be filled by something at S-structure. Of course, under usual circumstances there will be something in the subject position at S- structure as an argument of the verb will move there for Case reasons. But even if there is no argument inside the VP in need of Case, the subject position must be filled by the insertion of a pleonastic subject:

But non-finite clauses are different as they do not always have subjects:

(41)a he appears [- to have left] b they want [- to leave]

c [- to leave now] would be rude

How are these clauses able to escape the EPP? Note that it would in fact be ungrammatical to fill these positions with a pleonastic subject:

Recall also that the -Criterion requires that -roles be assigned to arguments. While some verbs may take implicit arguments which are not actually present in the structure but are ‘understood’ at a semantic level, these arguments are always complements and never subjects:

(43)a he is eating a sandwich b he is eating

c *is eating a sandwich

How are the non-finite clauses in (41) able to satisfy the -Criterion if there is no subject to assign the -role to?

The answer to all these problems is that the non-finite clauses in question do not lack subjects at all, they simply do not have pronounced subjects. One argument in favour of this assumption is that in different non-finite clauses there may be different types of unpronounced subjects. The argument is that one absent subject ought to be exactly the same as another absent subject and only if they are present could they

291

Chapter 8 - The Syntax of Non-Finite Clauses

possibly differ from each other. To see how these subjects differ from each other, consider the following facts:

(44)a Tim seems [ - to be tall] b Robin wants [ - to be rich]

(45)a it seems [Tim is tall]

b *it wants [Robin is rich]

(46)a *Tim seems [Tina is tall]

b Robin wants [Rupert to be rich]

(47)a *[ - to be tall] is what Tim seems b [ - to be rich] is what Robin wants

(48)a *it seemed to Larry [ - to look after himself] b we persuaded Larry [ - to look after himself]

(49)a *Peter seemed [that [ - to be a pilot] would be exciting] b Peter thinks [that [ - to be a pilot] would be exciting]

Consider the two sentences in (44). They both appear to have missing subjects and in other ways they seem to be similar. However, even at this point we can see that the two missing subjects are not entirely equivalent. In (44a), the missing subject is referentially identical to the subject of the higher clause, Tim. This subject is not semantically related to the verb of its own clause: Tim is not the one doing the seeming. We have already seen that verbs like seem do not have a subject of their own and often have pleonastic subjects, as we see in (45a). Thus the missing subject of the non-finite clause and the overt subject of the higher clause share a single -role assigned from the lower predicate tall. In other words, they represent a single argument. This contrasts starkly with the situation in (44b), where the missing subject and the overt subject of the higher clause bare completely different -roles: Robin is the thematic subject of want and the missing subject is the thematic subject of rich. The two subjects are coreferential, but they are independent elements in exactly the same way that a pronoun and its referent are independent elements:

(50)Henry thinks [he is happy]

In this case, Henry is the one doing the thinking and he is the one who is happy. If the pronoun refers to Henry then the interpretation is ultimately that Henry is happy (or at least this is what he thinks). But if the pronoun refers to someone else, then the interpretation is not that Henry is happy. However, the two elements are independent, regardless of what their referential properties are. The same is true of the overt subject and the missing subject in (44b). (45b) demonstrates that there really are two independent arguments in this construction as the subject of the higher predicate cannot be spelled out as a pleonastic element. In contrast, this is exactly what is possible in (45a), demonstrating that there really is only one argument here. The same point is made the other way round in (46). In this case we see that with a verb like seem, a different

292

Raising and Control

argument cannot be realised in the two different subject positions as there is only one - role involved. A verb like want, on the other hand, can realise arguments overtly in both subject positions as there are two independent -roles. So, one kind of missing subject shares a -role with another element in the sentence while the other kind of missing subject has a -role all of its own. The next three examples demonstrate that the two different kinds of missing subject have different referential properties. The nonindependent type of missing subject which shares a -role with its antecedent, must be lower in the structure than its antecedent. Hence it cannot be part of a structure which is raised to a higher position and the ungrammaticality of (47a) follows. The independent type of missing subject on the other hand can, under certain circumstances be higher in the structure than its antecedent, hence the grammaticality of (47b). The contrast in (48) again shows a difference in the referential properties of the two missing subjects. In (48a) we see that it is impossible for the dependent missing subject to refer to an object: they are always associated with subjects. In (48b) we have an independent missing subject, it being the one who is doing the looking after. As we can see, it is capable of referring to the object as the ultimate meaning is that Larry will be the one ‘looking after himself’. Finally, (49) shows that the dependent type of missing subject cannot refer out of the subject clause of another clause, where as the independent missing subject can.

Summarising, the dependent type of missing subject shares a -role with its antecedent and is fairly restricted in its referential properties, always being below its antecedent which is a subject in the immediately higher clause. The independent missing subject has its own theta role and demonstrates far more flexible referential properties. Its antecedent can be a subject or object, higher or lower, close by or more distant, given the right circumstances. If the missing subjects in both cases were the result of an absence rather than a presence of something unpronounced, it would be rather difficult to account for these differences. If there really is something in these positions, then the evidence suggests that there they come in different types which have different properties.

Of course, the fact that the dependent missing subject shares a -role with its antecedent can lead to only one conclusion, given the theory of -role assignment discussed in chapter 2: the missing subject and its antecedent are a single argument. In other words, this kind of phenomenon is the result of movement and the missing subject is a trace:

In reality then, the subject is not ‘missing’, it has just moved. By the same conditions, we cannot treat the independent missing subject as the result of movement: it bears a-role different from its antecedent and so they do not represent a single argument, but two different ones. Moreover, at D-structure these two arguments must also be in different positions as different -roles are assigned to different positions at D- structure. Hence we conclude that this kind of missing subject is present at D-structure. As we see in (51), this is not true of the other ‘missing’ subject. Because this element has many properties in common with a pronoun, i.e. it bears an independent -role but can be referentially dependent on something else, it is often referred to as PRO.

In the next sections we will look at these two different elements individually.

293

Chapter 8 - The Syntax of Non-Finite Clauses

2.1Raising

As we have seen, with certain verbs a subject which is generated in one clause can move into the subject position of a higher clause. This movement is known as raising as the subject always moves from a lower clause to a higher one and never the other way round. The conditions on raising are that the moved element must originate in a non-finite clause and it must be the subject of that clause. Thus, we do not find raising out of finite clauses or raising directly out of object positions:

(52)a The electrician1 seems [t1 to have found a mouse] b *The electrician1 seems [t1 has found a mouse]

c *A mouse1 seems [the electrician to have found t1]

In a number of ways this is similar to the kinds of movements we have seen previously which take a DP from one position and move it to a subject position, such as the movement of the subject from the specifier of a VP or the movement of an object of a passive or unaccusative verb. Those kinds of movements, we saw, were motivated by the fact that the DP started off in a Caseless position and hence in order to satisfy the Case filter it had to move into a Case position. The subject of a finite clause is a Case position as this is where nominative Case is assigned to by the finite inflection. In (52a) we see a DP that is moved into the subject of a finite clause, and so it may be that this movement is also Case motivated. If this is so, we expect to find that the position it moves from is a Caseless position. Is this prediction accurate? Consider the relevant structure in a little more detail.

A first issue to decide on is whether the embedded clause has the status of a CP or an IP. We saw in the previous section that some verbs select for IP non-finite complement clauses while others do not. The question we need to answer, then, is whether verbs like seem are exceptional verbs or not. There is reason to believe that these verbs are exceptional, as they never take a non-finite complement with a for complementiser:

(53)a *it seems [for the electrician to have found a mouse] b *it appears [for the mouse to be dead]

A possible explanation for this fact could be that the clause is an IP and hence there is no position for the complementiser. Let us assume this to be correct.

At D-structure the complement clause will sit in the specifier of the verb:

seem

The verb will move from its original position to support some inflection, depending on what is present. If there are aspectual morphemes, the lowest will be supported by the verb, if not the verb will move to either the (null) tense, if there is a modal, or all the way to the I position if there is a bound agreement morpheme as well:

294

Raising and Control

(55)a … has seem1-ed [VP [IP the electrician to have found a mouse] t1] b … will seem1- [VP [IP the electrician to have found a mouse] t1] c … seem1-ed [VP [IP the electrician to have found a mouse] t1]

It is important to realise that, as the verb has no subject of its own, there will be no light verb to assign a -role to the subject. As we know, it is the light verb which is responsible for assigning Case and hence as there is no light verb, there will be no Case assigned. The subject of the embedded clause also cannot receive Case from inside this clause as the inflection is non-finite and non-finite I does not assign Case. Thus, we can conclude that, despite the exceptional status of the embedded clause, its subject will not be assigned Case and if it remains in this position it will violate the Case Filter. Raising this subject to the next clause satisfies the Case Filter as it can get Case from the finite I of this clause:

Here, the verb first moves to the tense position, and then into the I to support the bound tense and agreement morphemes. The subject in its D-structure position is Caseless, so it moves into the vacant specifier of the IP where it is assigned nominative Case.

Next, consider the restrictions on the movement shown in (52b) and (c). The subject of the embedded clause cannot undergo raising if it is in a finite clause, or in an object position. In both of these cases, the DP is sitting in a Case position, therefore regardless of any other restriction, there would be little point in it moving to the specifier of the higher clause as once it has satisfied the Case Filter, it does not need to do so again. We might assume a kind of laziness to the system (some call it ‘economy’) such that if something doesn’t need to happen, it will not happen. If the clause is finite and hence the subject gets nominative from the finite inflection, then the higher subject position will be unfilled. It is under these circumstances that the EPP will force the insertion of a pleonastic subject:

295

Chapter 8 - The Syntax of Non-Finite Clauses

In the case that the clause is non-finite, although the object will not move out of its Case position, the subject will of course need to get its Case by moving to the higher subject position, as we saw in the examples above.

The properties of a verb that allow it to be involved in raising structures are quite specific. First it must lack a light verb which is responsible for assigning a -role to the subject and a Case within the VP. Without this light verb the subject position will be vacant and hence available to be moved into. If a verb has such a light verb, it will not be able to take part in raising structures for the simple reason that the subject position will be filled already and moreover, if the lower subject cannot get Case from within its own clause, it will be able to get it from the light verb. Second, it must take a clausal complement. Without the clausal complement, the subject of this clause will not be able to ‘raise’. Moreover, the complement clause must be capable of being nonfinite, given that raising only happens from non-finite clause subject position, for reasons we have just discussed. A verb which has no subject of its own, but cannot select for a non-finite clause will always have a pleonastic subject and will never be involved in raising. A possible verb that fits this pattern is emerge:

(58)a it emerged [that the mouse was shocked] b *the mouse1 emerged [t1 to be in shock]

Another structure which bears a remarkable similarity to raising structures concerns the passive exceptional verb. From what we know about the properties of exceptional verbs and the process of passivisation, it can be predicted that they will behave very much like raising verbs. As we know, an exceptional verb can take a nonfinite IP complement. Normally there will be an accompanying light verb and this will assign Case to the DP subject of the complement clause. When we passivise a verb, we replace the light verb with the passive morpheme, which neither assigns a -role to the subject, nor a Case to the complement. This, then, is the same set of properties that raising verbs have. We can see that such verbs do indeed behave like raising verbs:

296

Raising and Control

(59)a it was believed [CP that the electrician was scared of mice] b the electrician1 was believed [IP t1 to be scared of mice]

When the exceptional verb has a finite complement, the subject of this clause will not move as it gets Case from its own finite inflection, making movement unnecessary. When the clause is non-finite however, its subject will not receive Case from the nonfinite I and moreover will not get it from the light verb of the exceptional verb as this will have been exchanged for the passive morpheme. Thus movement will be necessary:

Certain adjectives can also appear in raising structures. As adjectives do not assign Case, if an adjective takes a non-finite complement, the subject of that complement will not get Case and will therefore have to move. Furthermore, if the adjective does not assign a -role to its subject, the subject position will be underlyingly vacant and will therefore either need to be filled by a pleonastic element or by a DP moving into it:

(61)a it is unlikely [that the mouse survived]

b the mouse1 is unlikely [t1 to have survived]

One more point can be made concerning raising and raising-like structures. As this movement allows a DP to escape the confines of the clause that it originates in, we might wonder how far that DP can move. The following datum seems to suggest that a subject can be raised over quite a distance:

(62)the builder1 seemed [to be unlikely [to be considered [t1 to be very skilled]]]

297

Chapter 8 - The Syntax of Non-Finite Clauses

In this example, the subject starts off in the lowest clause as the subject of the adjective skilled. It then moves out of three clauses to the subject position of the raising verb. In principle, then, it might appear that there is no limit to how far a subject may raise. However, it is interesting that in order for this to happen, each predicate between the original clause and the final landing site of the raised subject must be either a raising predicate or a passive verb and moreover each intervening clause must be non-finite and have a vacant subject position. If any of these conditions is not upheld, the sentence is ungrammatical:

(63)*the builder1 seemed [that the electrician believed [t1 to be incompetent]]

The grammatical (62) raises the problem of how it can be grammatical with so many clauses but only one visible subject. The EPP demands that all clauses have subjects and so we might expect that this sentence ought to be ungrammatical. All these problems can be solved if we assume that the subject does not move in one go, but moves from clause to clause, stopping off in each subject position:

(64)the builder1 seemed [t1 to be unlikely [t1 to be considered [t1 to be very skilled]]]

In this way, each clause is provided with a subject, the trace, and hence the EPP can be satisfied. The ungrammaticality of (63) demonstrates that when a subject raises, it cannot actually be moved too far. Looking at what is possible and what is not possible with such movements, there is something similar about the restriction to the restriction we have noted concerning head movement. Recall that he Head Movement Constraint demands that heads do not move over the top of other heads. It appears that the restriction on subject movement is that it cannot cross over the top of another subject. A general way to express both these restrictions is to claim that a moving element cannot move over the top of a like element. This principle, known as Relativized Minimality, was introduced by Rizzi (1990) as a way of accounting for locality conditions on movement. The following diagram might help to make clear how the principle works:

What this depicts is a situation in which an element Z is moving to a position X over the top of another element Y. Given the structure preserving nature of movement, X and Z will be of the same type, i.e. both phrases or both heads, but if Y is of the same type too, the then movement is not allowed. Thus, a head cannot move over a head and a subject cannot move over a subject.

2.2Control

Let us now turn to the other non-finite clause with an apparently missing subject, in which there is in fact a phonologically empty pronoun. There are a number of interesting points to be made about this element. The first is that although it is obviously a DP, it has a much more limited distribution than normal DPs. The other matters of interest concerning the empty pronoun PRO are the limitations on its

298

Raising and Control

referential properties, as it does not appear to behave like other pronouns in this respect. We will consider these points separately.

PRO can be found in the subject position of non-finite clauses:

(66)a we attempted [PRO to work the machine] b they tried [PRO turning the wheel]

c I painted the ceiling [PRO balanced on a chair]

It does not ever appear in the subject position of a finite clause, the object position or the object of a preposition:

(67)a *the message said [that PRO would self-destruct in five seconds] b *I congratulated PRO

c *the guard spoke to PRO

Note that there is nothing semantically wrong with these sentences. As PRO is a pronoun that refers to some other element in the sentence, the meaning of the sentences in (67), were they grammatical, would be perfectly understandable:

(68)a the message said [that it would self-destruct in five seconds] b I congratulated myself

c the guard spoke to himself

Why is it that PRO cannot appear in these positions? One relevant observation is that these positions are those to which Case is assigned. The non-finite subject positions, in which PRO is allowed, seem to be Caseless as overt DPs cannot appear there and so presumably they violate the Case Filter:

(69)a *we attempted [Sid to work the machine] b *they tried [Tony turning the wheel]

c *I painted the ceiling [Bob balanced on a chair]

If PRO must avoid Case positions, we predict that we should not be able to go in the subject position of the non-finite complement of a exceptional verb, as this is a position assigned Case by the light verb of the exceptional verb. This expectation is indeed fulfilled:

(70)a *he believes [PRO to be rich]

b *I suppose [PRO to drive a Trabant]

Similarly, PRO will not be able to appear with a for complementiser, which we have argued assigns Case to the subject position of the non-finite clause that it introduces:

(71)a *she hoped [for PRO to be on TV.]

b *we were anxious [for PRO not to be late]

In such cases, if the complementiser is absent, the sentence is grammatical:

(72)a she hoped [PRO to be on TV.]

b we were anxious [PRO not to be late]

So it seems that PRO cannot appear in Case positions and is therefore in complementary distribution with overt DPs, which of course must sit in Case positions.

299

Chapter 8 - The Syntax of Non-Finite Clauses

However, this conclusion is problematic both conceptually and empirically. On conceptual grounds, it is odd to say the least that there should be a principle stating that all DPs must have Case and then to find out that there is one DP that not only does this not apply to, but exactly the opposite holds of it and it cannot have Case. The empirical issue is that the assumption does not account completely for the distribution of PRO as there are places which are not Case marked, and so could not support an overt DP, but in which PRO cannot appear either. One such place is the subject position of the non-finite complement clause of a raising verb or a passive verb:

(73)a *it seems [PRO to be rich]

b *it was believed [PRO to have gone]

One possible solution to both these problems would be to claim that PRO doesn’t avoid Case positions per se, but has to sit in special Case positions which up to now have been assumed not to be Case-marked, but in fact might be assigned a special Case, applicable only for PRO. Chomsky and Lasnik (1993) proposed that PRO must sit in special Case marked positions. They argue that the subjects of certain non-finite clauses are not Caseless but that what they term ‘Null Case’ is assigned to them. Only PRO can bear Null Case and Null Case is the only Case that PRO can bear. Thus PRO will not be able to sit where overt DPs go as these will be Case marked with something other than Null Case. Moreover no overt DP can sit in a position in which it would be assigned Null Case as this is not ‘strong’ enough to satisfy the Case Filter. The good thing about this assumption is that it predicts complementary distribution between overt DPs and PRO but does not force us to assume that PRO can occupy any position in which we cannot find an overt DP. From this perspective, then, PRO cannot sit in a position to which no Case is assigned, as in (73).

So far, I have remained uncommitted about the status of the clause that contains PRO: is it a CP or is it an IP? Under both assumptions that PRO cannot sit in Case positions or that it can only sit in Null Case positions we have to ensure that the place where it can be found is not assigned a full Case from an element outside the clause. We have seen that as PRO cannot be the subject of an exceptional clause it must be assumed that this is not possible. One way to ensure that nothing else can assign Case to the place occupied by PRO is to assume that it is protected by a CP. Recall that a governor can govern up to a CP, but not through it as CP acts as a barrier to government. For this reason then, we will assume that all clauses containing a PRO subject are CPs and not IPs.

Turning to the referential properties of PRO we find that this is quite complex. To see how PRO behaves, we should first consider how other pronouns behave in terms of reference. There are two types of referential pronouns which behave differently with respect to each other. Compare the following:

(74)a Sue said Lucy likes her

b Sue said Lucy likes herself

In (74a) the pronoun her can either be taken as referring to Sue or someone not even mentioned in the sentence. Note that it couldn’t possibly refer to Lucy. In contrast the pronoun in (74b) can only refer to Lucy and cannot refer to someone not mentioned or to Sue. We call the first kind of pronoun a pronominal and the second kind anaphors.

300