grammatical foundations

Chapter 8 - The Syntax of Non-Finite Clauses

(2)a [CP that [IP the sheriff will [vP [VP shoot the outlaw]]]] b [CP for [IP the sheriff [vP to [VP shoot the outlaw]]]]

But while there certainly are non-finite clauses which fit this pattern, there are others for which bits and pieces of the structure appear to be missing. In some cases the complementiser is not only absent, but obligatorily so, which is very unlike the finite clause which has an optional complementiser:

(3)a I said [(that) the sheriff’s forming a posse]

b I believe [(*for) the sheriff to be forming a posse]

In other non-finite clauses the subject appears to be obligatorily absent, again contrasting with the finite clause which always has a subject:

(4)a the sheriff tried [(*him) to ride the horse] b the sheriff said [*(he) rode a horse]

There are even some cases of non-finite clause where not only is there no evidence of a complementiser or an inflection, but there isn’t even a verb:

(5)I consider [(*for) the cowboy (*to) tough]

What is the best analysis for these clauses with obligatorily missing parts? We will argue in this chapter that the missing elements are mostly not just null, but absent. We will start by considering those clauses with missing complementisers.

Compare the following sentences:

(6)a the sheriff believes [that they are hiding in the hills] b the sheriff believes [them to be hiding in the hills]

One difference between the finite embedded clause and the non-finite one is that the former has a nominative subject and the latter an accusative one. We saw in chapter 5 how the finite inflection is responsible for assigning nominative Case to its specifier. A relevant question is where the accusative Case of the non-finite subject comes from: does the non-finite inflection assign accusative Case? The answer would appear to be no because not all infinitival clauses can have an accusative subject, and if the nonfinite inflection were able to assign accusative Case, we would expect an accusative subject to be a permanent possibility, just like the nominative subject is always possible in a finite clause:

(7)a *the outlaw attempted [him to escape]

b *the cowboy hoped [him to brand the cow] c *the town relied [him to keep law and order]

Moreover, in some non-finite clauses the subjects Case is dependent on the complementiser rather than the inflection. This can be seen by the fact that without the complementiser an accusative subject cannot appear:

(8)a [for him to shoot the sheriff] would not be wise b [to shoot the sheriff] would not be wise

c *[him to shoot the sheriff] would not be wise

282

Exceptional and Small Clauses

So it appears that the non-finite inflection does not have the capacity to assign Case as if it could, (8c) would be grammatical.

But if this is true, where does the accusative Case on the subject of the non-finite clause in (6b) come from? We know that accusative Case is assigned by light verbs in other situations, could the accusative Case come from a light verb in this structure? Consider the structure in more detail. The verb believe has an experiencer subject and hence there is a light verb which assigns this -role. The clausal complement of the verb sits in its specifier position and the verb will move to support the light verb:

While in principle this might be possible, it does place the Case assigning light verb and the DP to which it assigns Case in rather distant positions. This would not be advisable as on the whole it appears that Case assignment is a local affair and thus there are limitations on how distant the Case assigner and assignee can be. In particular, if the situation pictured in (9) were accurate, we would expect any light verb to be able to assign an accusative case to the subject of a clause appearing in the specifier of its VP complement. But this is not so:

(10) a *the outlaws think1-e [VP [CP him is on their trail] t1]

b *the outlaws hope1-e [VP [CP him to be on the wrong trail] t1]

Note that this has nothing to do with the finiteness of the complement clause, (10b) is just as ungrammatical as (10a) and the latter involves a non-finite clause. There is clearly a difference between verbs like believe which can take non-finite complement clauses with accusative subjects and verbs like hope which cannot. One observable

283

Chapter 8 - The Syntax of Non-Finite Clauses

difference between them is that hope can take a non-finite complement clause with a complementiser, but believe cannot:

(11)a the bartender hoped [for the sheriff to stop the fight]

b *the bartender believed [for the sheriff to be a coward]

One assumption that might solve a number of problems might be that with believe there is no CP, just an IP. This would account for the missing complementiser with such verbs, it would also place the subject of the non-finite clause a little nearer to the Case assigning light verb:

Recall from chapter 6 that the notion of government, which is relevant for Case assignment, imposes a restriction on what can be governed in that an element can govern ‘up to a point’. We left undefined what that point was in our previous discussion, but now it is important to be more precise. There are a number of ways that we might think of doing this, but one of the most intuitive is to suppose that certain nodes in a tree form barriers to government in that they ‘protect’ their constituents from government from the outside. Thus a governor may be able to govern up to a barrier, but not through a barrier. We know that a light verb can assign Case to the specifier of the VP, as this is the normal configuration in which accusative Case is assigned:

(13)vP v©

vVP

284

Exceptional and Small Clauses

This means that the VP doesn’t act as a barrier to government. From observations given in (8) above, we can conclude that the for complementiser assigns Case to the subject of its non-finite clause complement, as the DP subject is only grammatical when the complementiser is present. If this is so, then IP does not count as a barrier either as the complementiser can Case mark the subject through the IP:

(14)CP C©

C IP

But if neither VP nor IP is a barrier to government, then we expect that the Case marking relationship depicted in (12) should be perfectly possible. There is no reason however to believe that any element can assign Case from outside a CP to any element within the CP and indeed it is a standard assumption that CP does count as a barrier to government. This supports the assumption that the non-finite complement clause of a verb like believe has an IP status and is not a full CP.

Given that clauses are normally CPs, a clause which only has an IP status is an exception. Hence, such clauses are known as exceptional clauses and the verbs which take exceptional clauses as their complements, i.e. verbs like believe, are known as exceptional verbs. Finally the process of assigning Case to the subject of an exceptional clause is sometimes called Exceptional Case Marking, or ECM for short.

Below are a few examples of exceptional verbs:

(15)a the sheriff expects [the outlaws to be in hiding] b the horse supposed [the sheriff to be lost]

c the deputy assumed [his horse to be outside the saloon]

d the bartender understood [the horse to be brighter than the deputy] e the law requires [the sheriff to arrest the outlaws]

In all these cases the complementiser for would be ungrammatical if used to introduce the non-finite complement. All of these verbs also may have a finite complement clause, which is obviously a CP as the that complementiser can appear:

(16)a the sheriff expects [that the outlaws are in hiding] b the horse supposed [that the sheriff was lost]

c the deputy assumed [that his horse was outside the saloon]

d the bartender understood [that the horse was brighter than the deputy] e the law requires [that the sheriff should arrest the outlaws]

Thus it is only the non-finite complement of exceptional verbs that are IPs. Another observation that supports the claim that the subject of the exceptional

clause is Case marked by the light verb of the exceptional verb is that these verbs can undergo passivisation. Recall that in English it is only verbs which have a Case assigning light verb that can undergo the process, which we described as the

285

Chapter 8 - The Syntax of Non-Finite Clauses

replacement of the agentive or experiencer light verb with the passive morpheme which neither assigns a -role to the subject position nor assigns Case to the specifier of the VP. Consider what happens when an exceptional verb is passivised:

(17)a the outlaws1 were expect-ed [t1 to be in hiding] b the sheriff1 was suppose-ed [t1 to be lost]

c his horse1 was assume-ed [t1 to be outside the saloon]

d the horse1 was understood [t1 to be brighter than the deputy] e the sheriff1 was requir-ed [t1 to arrest the outlaws]

The subject of the non-finite clause moves to the subject position of the passive verb just like the object of a passivised verb does. Given that we argued that the motivation for the object’s movement was to get Case, having been robbed of its light verb Case assigner, it is reasonable to assume that this is exactly what is going on with the movement of the subject of the exceptional clause.

Let us take a look at this analysis in more detail. Suppose we take a structure similar to (12) and replace the light verb with the passive morpheme:

v©

The two immediate consequences of this are that the experiencer -role fails to be assigned, leaving the specifier position of the vP empty, and the subject of the exceptional clause is left without Case given our assumption that the passive morpheme is not a Case assigner and neither is the non-finite inflection. Thus if the subject were to stay in this position, it would violate the Case Filter, which demands all DPs to receive Case, and hence the sentence would be ungrammatical, which it is:

(19)*it was believed [him to be on their trail]

Note that if the complement clause were finite, there would be no problem:

(20)it was believed [that he was on their trail]

286

Exceptional and Small Clauses

In this case the subject of the complement clause gets nominative Case from its own finite inflection and so the replacement of the light verb with the passive morpheme makes no difference.

In conclusion, we can see that there is a class of verbs which subcategorise for exceptional non-finite IP complements. These clauses have subjects which are dependent on the light verb of the governing verb for their Case. Therefore these subjects bear accusative Case. Moreover these subjects are affected by the passivisation of the governing verb which robs them of their Case assigner. Hence they will undergo a movement to the subject position of the higher clause, just like the object of the passive verb does.

Before closing this section, we will point out that there are some verbs which might look like exceptional verbs, but which are probably not. The most notorious of these is the verb want. This verb appears in structures that are remarkably similar to those involving exceptional verbs:

(21)the mayor wants [the sheriff to support him]

Here we have a complement clause which is non-finite and has an overt subject. The fact that this subject can be replaced by an accusative pronoun him shows that the subject is in an accusative position, just like we get with an exceptional verb. However, unlike exceptional verbs, this verb cannot passivise:

(22)*the deputy1 was wanted [t1 to ride the horse]

If this were an exceptional verb, there would be no problem in passivising it. In this way, want behaves like a non-exceptional verb such as hope:

(23)*the outlaws1 were hoped [t1 to be caught]

Presumably the reason why the subject cannot move out of the non-finite clause is because this is a CP, not an IP. Besides, passivising the verb would not affect the nonfinite clause’s subject as this does not get its Case from the light verb, but from the complementiser. But if this is the reason why want does not passivise, then we must conclude that it has a CP complement, not an IP, i.e. it is not an exceptional verb. If the complement of want is a CP, then its subject must get its Case from a complementiser, but there is no complementiser visible in (21). However, the complementiser can be made visible by separating the clause from the verb by an intervening adjunct:

(24)the mayor wants very much [for the sheriff to support him]

As can be predicted, the same thing does not happen with exceptional verbs, whose complements are not CPs in the first place:

(25)a *the sheriff believes very much [for the bandits to have robbed the bank] b *the horse expects very much [for the deputy to feed it]

Thus we may conclude that want takes a CP complement with a complementiser that may be null. Why the complementiser is null when the clause is adjacent to the verb and why it becomes overt when it is not, is a complete mystery.

287

Chapter 8 - The Syntax of Non-Finite Clauses

1.2Clauses without IP

If some clauses can lack CPs, the question naturally arises as to how small clauses can get. Recall that clauses are structured in various layers, each of which adds something to the interpretation of the whole clause. But the basic proposition is expressed by a predicate and its arguments. As the predicate does not have to be a verb a basic proposition can be expressed with no verbal element at all:

Both expressions on either side of the hyphen state exactly the same relationships between the predicate and its arguments. The difference is that while the expressions on the right are grammatical English sentences, those on the left are not. Or at least, one might think so. But consider the following:

(27)a I consider [Tim tall]

b I require [Graham in the garden] c I believed [Steven a student]

Given that these ‘basic propositions’ can be replaced by a full clause with virtually the same meaning, it seems that we should view them as being clauses of one sort or another:

(28)a I consider [that Tim is tall]

b I require [Graham to be in the garden] c I believed [that Steven is a student]

But what kind of a clause lacks a VP, let alone a CP or an IP? Moreover, what is the categorial status of such clauses?

One of the earliest suggestions, which still has a certain appeal, is that these clauses are simply phrases with subjects (Stowell 1983, who termed these constructions Small Clauses):

(29)a I consider [AP Tim [A©tall]]

b I require [PP Graham [P©in the garden]] c I believed [DP Steven [D©a student]]

One argument which favours this analysis is that different verbs take different types of small clauses as their complements and what determines the type of the clause is the head of the predicate part of the clause following the subject. The verb consider, for example, takes an AP type and a DP type of small clause, but not a PP type:

(30)a I consider [DP him a liar]

b I consider [AP him untrustworthy] c *I consider [PP him in the garden]

The verb, order, on the other hand, takes PP types, but not AP or DP types:

288

Exceptional and Small Clauses

(31)a I ordered [PP him out of the room] b *I ordered [DP him a fool]

c *I ordered [AP him foolish]

The fact that heads subcategorise for their complements in terms of the complement’s category would seem to suggest that these clauses do indeed differ in terms of their categorial statuses.

Opponents of this view, however, tend to point to the fact that the predicate part of the small clause seems to have the status of a phrase and if the whole clause is the phrase with the subject in its specifier position, the predicate should have the status of an X©:

(32)AP

DP A©

certain of his position

One piece of evidence that the predicate has full phrasal status comes from the fact that they seem to be able to move to phrasal positions, without taking the subject along with them:

(33)[how certain of his position]1 do you consider him t1 ?

Within the confines of X-bar theory, one way to get both the subject and the predicate of the clause to be phrases, is to posit a head between them of which the subject is the specifier and the predicate the complement:

DP X©

e certain of his position

But this proposal claims that the head of the predicate is no longer the head of the whole clause and hence it no longer determines the categorial status of the clause. It would be difficult therefore to account for the observations of (30) and (31) where different verbs subcategorise for different small clauses in terms of the category of the predicate. A separate question concerning (34) is what the status of X is. Haegeman (1994) argues that this head is an agreement element, i.e. what we have been calling I. Thus small clauses, according to Haegeman, are IPs (AgrPs in her terminology):

289

Chapter 8 - The Syntax of Non-Finite Clauses

(35)IP

DP I©

I XP

Besides the problem that under this analysis the head of the predicate is not the head of the clause, a major problem facing it is that it forces us to assume that the inflection can subcategorise for a whole set of different complements, ranging from DPs to PPs. But functional heads do not normally display this amount of freedom in their complement taking abilities. In all the cases we have considered so far, the agreement head selects for a v/VP complement, i.e. complements with [–N, +V] features. But if I can select for DP, AP and PP complements as well, i.e. [+F, +N, –V], [–F, +N, +V] and [–F, –N, –V], this must mean that it imposes no categorial conditions on its complement whatsoever. This is not true as bare NPs, non-thematic vPs, IPs and CPs cannot act as predicates inside small clauses:

(36)a *I consider him [NP student] b *I thought him [vP have gone] c *I ordered him [IP will leave]

d *I consider him [CP that he will leave]

Stowell has countered the argument that the predicate part of the small clause is a full phrase by claiming that it only seems to behave like a phrase as the subject moves out of the subject position before the predicate itself moves. Thus the derivation of a sentence like (37) would follow the steps indicated below:

(37)a you consider [him how intelligent]

b you consider him1 [t1 how intelligent]

c [t1 how intelligent]2 do you consider him1 t2

The debate continues and we will not attempt to put an end to it here. Stowell’s analysis does seem to be able to address the problematic issue of the selection of small clause complements, but his analysis of how the predicate can appear to behave like a phrase rests on the validity of the suggested subject movement, and it is not at all clear that there is a well motivated position for the subject to move to, even assuming a more articulated structure of the vP.

2 Raising and Control

Another aspect of non-finite clauses that we have noted without much comment is their ability to have missing subjects. This is surprising in more sense than one. Finite clauses in English never have missing subjects:

(38)a *(this) is a mouse b *(he) has gone

This is so even in cases where there is no semantic subject:

290