grammatical foundations

Interrogative CPs

3.5Subject questions

The last issue concerning interrogatives we will discuss concerns the difference between wh-questions which focus on the main clause subject and all other kinds of wh-questions. With most wh-questions it is fairly easy to see that movements take place as elements such as objects and adjuncts do not appear in their expected positions, but at the beginning of the clause:

(66)a who1 did they execute t1

b when1 was the meeting scheduled t1

Inversion is also possible to detect as the auxiliary and the subject end up on the opposite sides of each other. But when it is the main clause subject that is the focus of the question, things are no longer so clear cut. The word order is consistent with at least three analyses:

(67)a [CP e [IP who can save the world]]

b [CP who1 e [IP t1 can save the world]] c [CP who1 can2 [IP t1 t2 save the world]]

In (67a) both the wh-subject and the auxiliary are in the IP and do not move to the CP. In (67b) the wh-subject moves to the specifier of CP but the auxiliary does not move and in (67c) both the subject and the auxiliary move. But which one is correct and how can we know? From all that we have said above, one might hope that (67c) is accurate as only in this is the CP specifier filled with a wh-element and the head filled by something it can agree with. However, doubt is cast on this conclusion from the following phenomena:

(68)a who did you meet b who met you

It seems that for some reason, to which we return shortly, main verbs cannot move to C. So when a wh-element moves to the specifier of CP and it requires some element in the C position to agree with, the dummy auxiliary is used and hence we get do- insertion. However, when the subject is the focus of the question, there is no do- insertion indicating that nothing has to move to C. If this is a general condition then it suggests that no element moves from I to C in subject questions and hence that (67c) is not correct. Opinions differ as to the correctness of (67a) or (b), but obviously both are problematic for the straightforward analysis of interrogatives.

Before we consign (67c) to the waste bin however, let us see if it might be salvaged. We have argued that main verbs do not differ in their positions from auxiliary verbs except in the case of negation where the main verb cannot move over the negative head, but as auxiliaries are inserted into tense, the presence of negation does not affect them. Other than this, though, both main verbs and auxiliaries alike can occupy the I position (contra standard wisdom). If this is a general fact, then it is possible that main verbs can raise to C just as auxiliaries do and hence (68b) might be analysed as involving I-to-C movement of the main verb:

261

Chapter 7 - Complementiser Phrases

(69)CP

DP C©

As a structure this is straightforward. It contains a wh-element in the specifier of CP so that the whole clause can be interpreted as an interrogative. Moreover it has something in the C position for the wh-element to agree with. If we accept this as basically correct, we then have (68a) to account for: why can the verb not move to C when anything but the subject moves to spec CP? The answer would appear to have to do with the subject: when this does not move to spec CP the main verb cannot move past I.

We have seen restrictions like this before. The adverb, for example must be above the verb and the negation must be above the verb but below the tense. It seems that the verb must be below the subject and hence when the subject is in spec IP the verb can be no higher than I. However, when the subject moves to spec CP the verb can move to C and still remain lower than the subject. When the verb cannot move from I to C there is no choice other than to insert the auxiliary do. However, it appears that the dummy auxiliary is not enough by itself to provide the wh-element with something to agree with, thus the auxiliary must be inserted in to tense and move from there to I and from there to C to provide enough semantic content to support the agreement. The verb stays behind in the VP:

(70)CP

wh C©

262

Relative Clauses

To conclude this section. It seems that the distribution of elements in an English sentence is partly due to structural conditions imposed by X-bar theory and the selectional requirements of certain heads, partly due to the morphological properties of certain heads and partly due to general ordering requirements affecting certain elements. Specifically we have seen the effects of the following conditions:

(71)a verbs follow their subjects

b negation follows the finite tense and precedes the verb c adverbs (follow the finite tense and) precede the verb

The brackets around the condition in (71c) indicate that this seems to be a preference rather than a rigid condition and that if it is impossible for the adverb to both follow the tense and precede the verb, both are in the same place for example, then the condition may be relaxed.

4 Relative Clauses

There is another construction in English which looks like an embedded interrogative clause, but which is very different in interpretation. This clause acts as the modifier of the noun inside the DP:

(72)a I asked [who you met] b the man [who you met]

Such noun modifier clauses are called relative clauses.

4.1The position of the relative clause inside the NP

As relative clauses seem to modify nouns we can assume that they occupy a position within the phrase headed by the noun. There is evidence that this is so as the noun and the relative clauses can be pronominalised and coordinated separately from the determiner:

(73)a the [mistake that I made] was much bigger than that one

b these [components that we make] and [boxes that we pack them in]

There are two possible adjunction sites inside the NP: the N©and the NP itself. There is some evidence that relative clauses are adjoined to the N©.For example, if it is the case that the adjectival phrase is adjoined to the N©, then the fact that the relative clause can be adjoined lower still than APs indicates that they must also be adjoined to N©and not to the NP:

(74)my favourite [places I like to go] and [people I like to visit]

In this example, the AP favourite modifies both places I like to go and people I like to visit, indicating that these are below the adjective:

263

Chapter 7 - Complementiser Phrases

However, there is a slight complication in that there are more than one type of relative clause. Those that we have been looking at so far are known as restrictive relative clauses. Semantically these tend to focus on one element out of a set of possible referents. For example, the components that we make focuses on a particular set of components out of a larger set of components which are distinguished by the fact that we make them. Thus, the purpose of the relative is to ‘restrict’ our attention to a certain element or elements out of a possible range of elements. By contrast, nonrestrictive relative clauses simply add extra information about the referent of the noun being modified:

(76)the earth, which is 93 million miles from the sun

Note here, there is not a range of possible referents for the noun earth and the relative clause restricts our attention to one of them. There is just one earth being spoken of, and the fact that it is 93 million miles from the sun is given as information about this object.

Restrictive and non-restrictive relative clauses also differ from each other in terms of their internal properties, something which we will discuss more fully in the next section. For now, let us just note that restrictive relative clauses may begin with a that whereas non-restrictive relatives never do:

(77)a the man [that you met] (as opposed to all the other men)

b *the earth [that is next to Mars] (as opposed to all the other earths!) c the earth, [which is next to Mars]

The two kinds of relative also differ in their prosodic properties. In written form the non-restrictive relative is followed by a comma, which indicates a slight pause between the noun and the relative. There is no pause between the noun and its restrictive modifying clause however.

It may be that the two clauses also differ syntactically and indeed it is often assumed that non-restrictive relative clauses are more distant from the noun than are restrictive relatives. This is supported by the following observations:

(78)a the [man [who you met]] and [woman [who you haven’t]] b this [man [who you met]] is taller than that one

(79)a my [mother [who you met]] and [father [who you didn’t]] b *my [mother [who you met]] is taller that his one

264

Relative Clauses

Both kinds of relative clauses can be coordinated with other like constituents showing that they both form a constituent with the noun that they modify. But only the restrictive relative clause and its noun can be pronominalised by one. A possible explanation for this is that one only pronominalises N©s and the string consisting of a noun and a non-restrictive clause is not an N©. This suggests that while restrictive relatives adjoin to the N©, non-restrictive relatives adjoin higher up, perhaps to the NP:

Having established the external distribution of the relative clause, let us move on to look at some of its internal properties.

4.2A comparison between relative and interrogative clauses

As pointed out, in many ways the relative clause has many properties in common with a wh-interrogative. But relative clauses are not interrogative, but declarative. This is clear both from their interpretation and the fact that they may start with the complementiser that which as we have seen, introduces declarative clauses. Thus, the wh-element which starts relative and interrogative clauses seems to have a different set of features: an interrogative pronoun is [+wh] and a relative pronoun is [–wh]. One might think therefore that they are entirely different lexical elements. This is supported by the fact that in some languages there are differences between relative and interrogative pronouns. In Hungarian for instance, there is a systematic difference with relative pronouns beginning with a-:

Even so, there is still an obvious relationship between the two and so it probably would not be wise to claim them to be completely separate. It may be that wh- pronouns are lexically unmarked for the feature [±wh] and get this through the agreement with the appropriate complementiser, though this suggestion does not entirely square with the claim made earlier that in main clauses there is no complementiser for the interrogative pronoun to agree with.

Whatever the relationship between interrogative and relative pronouns, it still needs to be acknowledged that there are different possibilities in both types. For example, in English although what is a perfectly good interrogative pronoun, its use as

265

Chapter 7 - Complementiser Phrases

a relative pronoun is somewhat stigmatised, being seen as a characteristic of ‘uneducated’ speech. Dialects and sociolects that make use of ‘what relatives’ also do not use the relative pronoun in a way consistent with the interrogative pronoun. While what as an interrogative pronoun has a ‘non-human’ aspect to it in any dialect, in that you couldn’t point to a person and ask “what is that” without being offensive (“who is that” would obviously be more appropriate), what-relatives are often used to modify nouns with human referents:

(82)a a man [what I know]

b this bloke [what I was telling you about]

In standard English, however, what-relatives are not accepted even for modifying nouns with non-human referents:

(83)a *the book [what I read] b *an idea [what I had]

The only acceptable use of a what-relatives in standard English is in relatives which appear to lack a modified noun, what are sometimes called headless relatives:

(84)a [what you should do now] is …

b [what I say and what I do] are two totally different things c [what I don’t understand] is …

To conclude on this issue, what as an interrogative pronoun and what as a relative pronoun are used in very different ways in all dialects.

There is another interesting difference between the use of interrogative and relative pronouns which shows that there is a possibility available for a relative pronoun that is not generally available with interrogatives. The restrictive relative is often noted to come in three different forms. One starts with a wh-relative pronoun and is called a wh-relative. Another starts with the complementiser that and is called a that-relative, and the third has nothing in front of the subject and is called a zero relative. These are exemplified below:

(85)a the man [who I paid] b the man [that I paid] c the man [I paid]

The wh-relative resembles an interrogative in more than just the fact that it is introduced by a wh-element, but also in the process which apparently forms the two structures. In both cases, the wh-element starts off inside the clause and moves to the specifier of CP. Thus, both types of clauses start with a wh-element and have a corresponding ‘gap’ in the position it was moved from. The gap contains the trace of the moved wh-element:

(86)a I wonder [who1 Sherlock suspects t1] b the butler [who1 Sherlock suspects t1]

Interestingly, although they do not start with a wh-element, both that-relatives and zero relatives contain a gap in the same place that they would if they did have a wh- element:

266

Relative Clauses

(87)a the butler [that Sherlock suspects -] b the butler [Sherlock suspects -]

We should first counter a myth about that-relatives that prevails from traditional grammars. In these it is common to find that at the beginning of the relative clause referred to as a relative pronoun, thus suggesting that it be given the same treatment as wh-elements. If this is true, then this element originates inside the clause and moves to the specifier of CP, as do wh-elements. But there are many reasons to believe that this word is not a wh-element but is, as appearances predict, a simple complementiser. Firstly, note that as would be predicated on the assumption that it is a complementiser, that is only ever used in finite clauses and although wh-elements can marginally be used in non-finite clauses, that never is:

Instead, we can get a for complementiser in non-finite relatives, as would be expected:

(89)a a man [for you to contact] b a place [for me to stay]

Another argument that that is not a relative pronoun in that-relatives is that it does not behave like a wh-element with respect to prepositions. Note the following two possibilities with a wh-relative:

(90)a the house [which1 I live in t1] b the house [in which1 I live t1]

When the wh-element is part of a PP it has the option of moving alone, a strategy known as preposition stranding, or of taking the whole PP with it, a strategy known as pied-piping (after the story of the Pied Piper of Hamlin, who played his pipes and the rats followed him – the connection between prepositions and rats is, however, mysterious). If that were a relative pronoun, we might expect the same options to be available in that-relatives. But this is not true:

(91)a the house [that I live in -] b *the house [in that I live -]

One explanation for why we do not get pied piping with a that-relatives is that that is not a relative pronoun and did not originate in the gapped position and hence the preposition could not be pied-piped by it.

If that is a complementiser in the that-relatives, then that-relatives and zero relatives are alike in that they do not contain a wh-element and the difference appears to be the standard ability of the complementiser to be overt or covert in a finite clause:

(92)a I said [(that) I was reading a book] b the book [(that) I was reading -]

The fact remains however, that there is still a gap in these relative clauses. What is the nature of this gap? In many ways it has exactly the same nature as the gap in a wh- relative. Consider this a little more closely: in a wh-relative, the relative clause acts as

267

Chapter 7 - Complementiser Phrases

a modifier of the noun by relating the noun to a position in the relative clause itself. Thus in (93a) the noun is modified by being interpreted as the object of the verb in the relative clause, whereas in (93b) the noun is modified by being interpreted as the subject of the relative clause:

(93)a the team [which we beat t] b the team [which t beat us]

The relationship between the noun and the relevant position is not a direct one, however: it is not the noun which moved out of this position otherwise we would end up with a somewhat circular structure with the noun being part of the relative clause that is part of the NP headed by the noun! The relationship between the noun and the position in the relative clause is mediated by the relative pronoun: it is this element that originates in the relevant position and this pronoun is referentially dependent on the noun:

(94)a the team [which1 we beat t1] b the team [which1 t1 beat us]

We might claim that this is exactly what the function of the relative pronoun is in a relative clause. Indeed, if there is no wh-element related to a gap, or if the wh-element does not move to create the gap, then the relative clause is ungrammatical:

(95)a *the team [we beat Liverpool] b *the team [we beat which]

If this is true, then the same must be true of all relatives, including that and zero ones. The fact that we cannot detect a wh-element in these relatives suggests that they should be analysed as containing an empty wh-element, similar to the empty operator in yes– no questions, but this time behaving like the type of wh-element that originates inside the IP and moves to the spec of CP:

(96)the team [Op1 (that) we beat t1]

Thus the difference between relative clauses and interrogative clauses in this respect is that relative clauses can use the null operator in ways not possible in an interrogative, i.e. as a referential operator rather than a non-referential one which is associated with the truth of the expression.

Having enumerated several differences between interrogative and relative clauses we can now ponder the question of whether these differences show a fundamental distinction between the constructions, or whether they fall out from other considerations. Let us start with why the wh-element undergoes movement in both constructions. We have said that the interrogative wh-element undergoes movement

268

Relative Clauses

because the clause needs to be interpreted as interrogative. This clearly cannot be the reason for the movement of the relative pronoun as relative clauses are not interrogative. The reason why a relative pronoun moves is presumably something to do with its function as a mediator between the modified noun and a position inside the relative clause and again this seems to differ from the wh-interrogative as wh-elements in interrogatives do not act as mediators.

However, if we take one step back from the details, we can see some striking similarities between the reasons for wh-movement in both types of clauses. For one thing, both movements have semantic rather than grammatical motivations. Moreover, the reason why the wh-element moves in an interrogative is to enable the CP to be interpreted as a question. The reason why the wh-element moves in a relative clause is to enable the CP to be interpreted as a modifier.

Finally by moving to the specifier of the CP, the wh-element is interpreted as an operator in both interrogative and relative clauses. The fact that one is interpreted more like a quantificational operator, like quantificational pronouns such as everyone or someone, while the other is interpreted like an anaphoric operator, which is referentially dependent on some other element in the sentence, like a reflexive pronoun such as himself, falls out due to the different functions of questions and relative clauses: one asks a question and the other modifies a noun.

What about the use of empty operators in relatives as compared to their limited use in interrogatives?:

(97)a the idea [Op1 (that) I had t1] b *I asked [Op1 (if) you had t1]

Again this may be entirely due to differences in the use of these constructions. As a relative clause modifies a noun making use of an anaphoric operator, there must be an antecedent for the operator to take its reference from. This antecedent, i.e. the modified noun, can provide us with the content of the operator and hence this is recoverable even if we cannot see the operator itself. With a question, however, as there is no antecedent, the content of the operator has to be visible on the operator itself and hence the null operator cannot be used in this way. The null operator used in yes–no questions is clearly non-referential and hence has no specific content to be recovered. In this situation then the null operator can suffice.

In conclusion then, it seems that the differences between interrogative and relative clauses are mainly to do with their different functions. Syntactic and other differences may be derivable from these. Certainly, given the above discussion, it is not really surprising that they have very similar internal organisations and employ very similar processes in their formation.

269

Chapter 7 - Complementiser Phrases

5 Other fronting movements

So far we have concentrated on movements which seem to specifically involve complementisers and their projections. However, there are movements in English which appear to move an element out of the IP but which do not involve the CP. This indicates that there are positions between the CP and the IP. In this final section of this chapter we will briefly investigate this part of the English clause structure.

5.1Topicalisation

So far we have claimed that AP modifiers of VP and of the clause are adjoined to various places inside the IP. There is another position in which we can find both VP and sentential adverbs which appears to be outside the IP altogether:

(98)a certainly, no one saw the thieves get away b quietly, the robbers made their get-away

In some ways, the initial position for adverbs is similar to the wh-position, which is unspecified for the kind of wh-element that can occupy it:

The initial position of the adverb seems also unrestricted in that any kind of adverb can occupy it. The reason why the specifier of CP is so unrestricted is because elements move to this position from positions compatible with their status (e.g. from subject, complement and adjunct positions). Taking this into account, we might therefore claim that the initial adverb position is a position to which various kinds of adverbs move. This is backed up by the observation that this position is not only reserved to adverbs, but a whole range of elements seem to be able to occupy it:

(100)a [PP on the train], I saw Biggs –

b [NP Biggs], I remember seeing – on the train c [VP see Biggs on the train], I certainly did –

These fronted elements are often referred to as topics, as they represent information that is already part of the discourse, or can be assumed to be readily retrieved by the participants in the conversation from the context or from general knowledge (often called ‘old’ information). Note that out of context, these expressions often sound strange, but given a context in which the topicalised element has already been introduced, they greatly improve:

(101)a considering all the places that I saw the robbers (i.e. on the platform, in the engine and on the train), on the train, I saw Biggs, …

bof all the people that I recognised (i.e. Biggs, Smith and Jones), Biggs, I saw on the train, …

cI expected to see Biggs on the train and [see Biggs on the train], I certainly did

270