grammatical foundations

Interrogative CPs

We have seen in a number of previous places that the head and the specifier have a special relationship. In the IP where the subject sits in the specifier and the agreement morpheme sits in the head position, the two elements ‘agree’ with each other in terms of number and person features. Thus if the subject is a third person singular element, then the agreement will be third person singular and determine that the finite morpheme ‘s’ will show up on the verb. In the DP, the possessor sits in the specifier position and this can only be accompanied by a determiner that ‘agrees’ with it in terms of possession. We might assume therefore that the CP will be no different and that the head and its specifier will enter into an agreement relationship, presumably in terms of the [+wh] feature. So, if the specifier has this feature, so will the head. Of course, if the head has the feature, then it will project it to the CP and this will be interpreted as an interrogative clause, and hence the wh-specifier can indeed influence the interpretation of the clause in a round-about manner:

(30)CP

wh C©

C IP

In (30) the double arrow represents the agreement relationship between the head and its specifier and the single headed arrows represent the projection of the head’s features to the CP. By this path then the [+wh] on the wh-element ends up on the whole CP and a CP with a wh-specifier will be interpreted as interrogative. Thus, one motivation for wh-movement might be that there is need to interpret the clause as an interrogative.

This cannot be the whole story however. The following indicates that moving the wh-element to the front of the clause is not obligatory:

(31)A I voted for the Monster Raving Loony Party B you voted for who?

B’s response involves what is known as an echo question, in which a previously uttered sentence is more or less repeated and a part of it that was either not heard or not believed replaced by a wh-element. The meaning is quite clear: it is a request for someone to repeat or confirm the previous statement. This is very different from the meaning of a wh-question which is asking for information about a particular aspect of the sentence. Compare:

(32)a who did you talk to? b you talked to who?

One major difference in meaning between these two sentences is that the first presupposes the truth of the proposition: the speaker assumes that you spoke to someone is true. Thus this kind of question cannot be asked felicitously if the speaker doesn’t think that you spoke to anyone. This is not true of (32b), however. In repeating back the previously made statement, the speaker does not commit himself to either its truth or falsity: judgement about that is postponed until the answer is received. Clearly this difference in meaning has something to do with where the wh-element is situated.

251

Chapter 7 - Complementiser Phrases

If the wh-element is moved to the specifier of CP then the result is a wh- interrogative interpretation. If not then the sentence is interpreted as an echo question. We might think of this in terms of the interpretation of the wh-element itself. A wh- element interpreted in an echo question simply has the role of a pointer to the missing information. In a sense it is a ‘place filler’ that does little more than indicate what needs to be repeated. A wh-element that is moved into a the specifier of the CP is a little more complex in terms of its interpretation. The semantics of a moved wh- element is similar to that of a quantifier such as everyone or someone. These are called operators as they indicate a process that is needed to work out the meaning of the sentence that contains them. For example, consider the difference between the following:

(33)a Tim is tall

b someone is tall

The truth of (33) is fairly easy to establish. First we find the individual that the name Tim refers to and then we see if they have the property of being tall. The truth of the second statement is not quite so easy. For a start, there is no individual to whom the word someone refers and so it isn’t just a matter of checking to see if the person has the property of being tall. Instead we must go to the set of things that someone could potentially refer to (the set of people relevant to a conversation, perhaps) and go through each of them individually to see if they are tall. If at least one of them is tall, then the sentence in (33) is true. If none of them are tall then the sentence is false. Consider what who means in (32a). Like the quantifier someone the interrogative pronoun does not refer to a known individual. Instead, the hearer is asked to perform a process of going to the set of potential referents and finding those that if substituted for the wh-element would produce a true sentence:

(34)a who did you talk to?

b I talked to Tom – false = not the answer I talked to Dick – false = not the answer I talked to Harry – true = answer

Thus, a fronted wh-element is interpreted as an operator. Given that the difference between a wh-element that is interpreted as an operator and one that is simply used as an echoic device is that the former is moved to specifier of CP while the latter is not it seems that the movement plays a role in determining the interpretation of the wh- element as well as the interpretation of the clause that contains it. Let us assume the following interpretative principle:

(35)interpret a wh-element as an operator if it is in spec CP

There is one exception to the above principle however. Consider the following:

(36)who does Thelma think likes what

This is known as a multiple wh-question as it is a single question that asks for more than one piece of information. Note that both of the wh-elements may be interpreted as operators (the second one may be interpreted as an echo given the right intonation), in which case the answer to the question has to be a list of pairs ranging over likers and likees. The interesting point is that the second wh-element, although it may be

252

Interrogative CPs

interpreted as an operator, clearly has not undergone movement and so seems to violate the interpretative principle in (35). One thing is clear, however: the interpretation of this non-moved wh-element as an operator is dependent on there being a moved wh-element in the same sentence. If this were not a multiple wh- question, the wh-element would have to move. What we need then is to somehow tie the non-moved wh-element to the moved one. One possibility would be to claim that at some level of representation of the sentence which is relevant for semantic interpretation, multiple instances of wh-elements are interpreted as a single complex wh-element. Let us simply say that we indicate the interpretation of multiple wh- elements as a complex operator by coindexing them:

(37)who1 does Bill think likes what1

We can then alter our statement of the interpretative principle to fit this situation:

(38)interpret a wh-element as an operator if it is in spec CP or is coindexed with a wh-element in spec CP

The movement of the wh-element to spec CP therefore seems to have an interpretative motivation, which contrasts with the grammatical motivation of the movement to spec IP. There are other differences between the movements, which we will look at in the next chapter. We may, for now, simply identify the kind of Case filter/grammatically motivated movement as A-movement (A stands for ‘argument’ as it is only argument DPs which undergo it) and the kind of semantically motivated movement, such as wh- movement, as -movement ( means ‘not argument’).

3.3Inversion

Looking at inverted auxiliaries, we see that they all have one thing in common: they are all finite. An inverted auxiliary may be a modal, which is inherently finite, or an aspectual auxiliary which bears tense and we never get a non-finite form of the auxiliary in the inverted position:

(39)a could they be finished? b have they finished?

c are they starting again? d *having him been seen

This suggests that the auxiliary moves to the C position from the inflection position. This is straightforward with modals as they are generated in the inflection position at D- structure. Thus they undergo a movement from I to C in certain interrogative structures:

(40)CP C©

C IP

DP I©

I VP

253

Chapter 7 - Complementiser Phrases

The case of aspectual auxiliaries is similar to the modals. In the previous chapter we claimed that these auxiliaries are inserted elements that may move from the tense position to the I position when there is null agreement in I. Thus, if they are in I, they can also undergo movement from I to C:

(41)CP C©

C IP

DP I©

- v©

v vP

have seen my father

Main verbs are problematic however, as they do not appear to be able to move to the C position:

(42)a have you read the book? b *read you the book

c did you read the book?

As we can see, a yes–no question involving a main verb moving to the C position is ungrammatical and instead of the main verb moving to C what happens is that the dummy auxiliary do is inserted into the tense position, and from there it moves to C, via I. Of course, this is readily accounted for if main verbs do not move to I, as is the standard assumption. If they are never in I they cannot move to C without violating the head movement constraint. But we argued that main verbs can move to I and so it is not readily apparent why they cannot move to C. We will put this issue to one side until we have discussed the facts about I-to-C movement more fully. For the time being, then, we will concentrate on I-to-C movement as it involves auxiliary verbs.

3.4The interaction between wh-movement and inversion

The most important issue concerning I-to-C movement is why it happens. There are two main views on this. One is that I-to-C movement happens because there is a bound C morpheme in interrogative clauses and this triggers the movement of the auxiliary to support it in the same way that inflections trigger verbs and auxiliaries to move. The other view is that the movement happens in precisely the cases when there is nothing in C and there is a requirement that there must be. The first is perhaps the most intuitively obvious, but it faces a number of problems which make the other approach more attractive. Let us consider each proposal separately.

254

Interrogative CPs

We know that the complementiser is the head of the CP and that this provides the relevant features for the force interpretation of the sentence. Some complementisers are interrogative and some are declarative. Thus, when the clause is interrogative, presumably it has an interrogative complementiser and vice versa if the clause is declarative. In yes–no questions, therefore, we might expect the clause to have an interrogative complementiser. As we can see no complementiser in such sentences, we must assume it to be abstract. The assumption of such an element is supported by the fact that in some languages a morpheme with exactly these properties appears overtly. For example, in Japanese, a yes–no question is marked by the appearance of the morpheme ka at the end of the clause. Thus, the difference between (43a) and (43b) is that the first is declarative and the second interrogative:

(43) a Keko-wa sensei desu Keko teacher is ‘Keko is a teacher’

bKeko-wa sensei desu ka? ‘is Keko a teacher?’

In Japanese, the complementiser goes in the final position of the clause and so the question particle is suitably analysed as a complementiser. Thus, we might claim that English is similar to this, the main difference being that the question particle is covert in English:

C©

C IP

Q she is a teacher?

The simple assumption that this interrogative complementiser is a bound morpheme in English is enough to justify the movement of the inflectional element to support it:

A similar analysis can be provided for inversion that accompanies wh-movement to spec CP. As we discussed in the previous section, wh-movement to spec CP establishes an agreement relationship between the wh-element and the head of CP that

255

Chapter 7 - Complementiser Phrases

enables the clause to be interpreted as an interrogative. One way to realise this agreement relationship is in terms of the presence of an interrogative complementiser, i.e. the empty question particle, in the C position:

(46)CP

DP C©

-Q

Again, as this particle is a bound morpheme, it will trigger movement to support it:

(47)CP

DP C©

This all seems fairly straightforward. Unfortunately things are a little more complex and it appears that we do not always get inversion in question clauses. One place where we find that inversion doesn’t happen is in embedded questions:

(48)a I didn’t know [what he would say] b *I didn’t know [what would he say]

In some embedded contexts, it seems as though inversion is optional:

(49)a the boarder guard asked [why the tourist didn’t have a passport] b the boarder guard asked [why didn’t the tourist have a passport]

Note, however, these two sentences have very different meanings. In the first case the embedded question reports on the ‘content’ of the question that was asked. The question might not have been framed in these exact words, or even in English! In contrast, in the second case the embedded clause reports on the words that were actually used. Thus in the second case, the embedded clause actually echoes a previous sentence that has been uttered. When it was uttered, obviously this sentence was not an embedded clause but a main one, and hence it is not surprising that it follows the pattern of a main clause. Therefore, (49b) isn’t really an example of an embedded clause with inversion but it involves something that is a main clause, seemingly being used in an embedded context. The conclusion is that in real embedded clauses we don’t get inversion.

256

Interrogative CPs

But this is a puzzle. Why do we not get inversion in an embedded interrogative? If the account of why we get inversion in main clauses is as we have so far suggested, then we must conclude that the question complementiser either does not need binding in an embedded clause, or that it is bound by some other element. For example, we might assume that the question complementiser in an embedded clause is not the same complementiser that appears in the main clause. We know that we do not get the kinds of overt complementisers that introduce embedded clauses in main clauses:

(50)a *that Rachel is rich b *if you saw that

However, we have also claimed that main clauses are CPs and so have complementisers introducing them. Therefore there seems to be a distinction between main clause and embedded clause complementisers. If this is so, then we might claim that the main clause interrogative complementiser differs from the embedded interrogative clause complementiser in that the former is a bound morpheme while the latter is not:

But this is a rather ad hoc solution which doesn’t really tell us why things are this way and hence does not really have much by way of explanatory content.

A second proposal claims that the difference between main and embedded clauses is that while complementisers are allowed in embedded clauses they are not in main clauses. Thus, a main clause interrogative actually lacks an interrogative in the C position, which is therefore underlyingly empty. An embedded clause on the other hand differs in two ways from a main clause: complementisers are allowed in embedded clauses and the clause itself is selected as a complement by some predicate. We have seen how a predicate imposes selectional restrictions on its clausal complements and those predicates which take an interrogative complement will demand that the CP be marked as interrogative. The way to mark a clause as interrogative is to give it an interrogative head and this can either be overt, i.e. if or covert, i.e. Q:

(52)a Robin doesn’t remember [CP if [IP she bought bread]]

b Richard doesn’t recall [CP where Q [IP he left his horse]]

257

Chapter 7 - Complementiser Phrases

If we assume that Q, like if, is not a bound morpheme, then there will be no reason for inversion. However, in a main clause there can be no complementiser and hence Q like all other complementisers will be prevented from appearing. An interrogative CP still needs to be interpreted as interrogative however and hence it will need its head to have the [+wh] feature. We have seen how, when a wh-element moves to spec CP, it agrees with the head. If we make the reasonable assumption that nothing can agree with an empty head position, this necessitates something being in the head position of an interrogative clause. Given that complementisers are systematically excluded from this position, the only option is to move the nearest head into the C position and this will be the auxiliary in I:

(53)a *[CP what e [IP I can do]]

b *[CP what if/Q [IP I can do]] c [CP what can1 [IP I t1 do]]

(53a) is ruled out because the wh-element has nothing to agree with and hence the clause cannot be interpreted as an interrogative. (53b) is ruled out by the general exclusion of complementisers appearing in main clauses. This leaves (53c) as the only grammatical possibility.

While this analysis seems less ad hoc than the assumption of different interrogative complementisers in main and embedded clauses, there are still a number of questions left unanswered. For example, how are we to analyse yes–no questions under the assumption that complementisers are not allowed in main clause complementiser positions? What triggers the inversion in this case if there is no wh-element to agree with? The answer might be that there is a wh-like element in yes–no questions. Although we do not get inversion type yes–no questions in embedded clauses, for obvious reasons, it is still possible to have an embedded yes–no question. These are typically expressed by the element whether:

(54)I asked [CP whether I should bring some wine]

If we were to put the content of this embedded clause into a main clause it would be should I bring some wine, i.e. a yes–no question. Thus, (54) contains an embedded yes–no question. The wh-element whether is quite strange. In many ways it looks like a complementiser, being a word that appears at the front of an interrogative clause, rather like if. However, there are a number of reasons to believe that whether is not a complementiser. For one thing, it can introduce both finite and non-finite clauses:

(55)a I wonder [CP whether I should boil the eggs] b I wonder [CP whether to boil the eggs]

As we have seen, complementisers generally select for a particular type of CP complement, either finite or non-finite. Moreover, no other complementiser can introduce a non-finite clause with a missing subject:

(56)*they hoped [for – to win]

Whether can be coordinated with operators like the negative and wh-pronouns:

(57)a I’m not sure [whether or even when I should applaud] b they asked [whether or not they would be paid]

258

Interrogative CPs

No other complementiser can do this:

(58)a *I don’t know [if and when to stand up] b *she wondered [if or not to pack the bags]

Finally consider the fact that the word whether was used in Old English to introduce yes–no questions in main clauses:

Yet Old English did not have main sentences that started with complementisers, indicating that whether never was a complementiser even in former stages of the language.

But if whether is not a complementiser, what is it? It differs in one very large way from wh-elements and that is while wh-elements move out of a clause whether does not seem to. Wh-elements are all moved to the spec CP from some position inside the IP and therefore they are always associated with a ‘gap’ in the clause, filled, of course, by a trace:

(60)a [CP who1 did [IP you think I met t1]]

b [CP who1 did [IP he say t1 likes tennis]]

c [CP where1 did [IP you put the anti-tank missiles t1]]

But whether is not linked to any position inside the IP from which it has moved. This, I think, is understandable in terms of the functions of wh-elements. Most wh-elements are used in wh-questions and their function is to mark the clause as an interrogative and to indicate what the focus of the question is. (60a), for example is a question about the object, which is the position from which the wh-element derives. As we said, the function of whether is to mark a yes–no question and these questions focus on the truth of the sentence rather than on any particular piece of information it may carry. For this reason, then, whether does not need to hold a position in the clause, it merely indicates the clause’s question status. We may assume therefore that whether is a wh-element that is generated directly in the specifier of CP as a general interrogative operator. As it appears in embedded contexts, it will be accompanied by an interrogative complementiser Q with which it agrees:

(61)CP

whether C©

C IP

Q

What about main clause yes–no questions? Obviously in Modern English, whether is not allowed to appear in main clauses. But we might suppose that something appears in the same position in main clause interrogatives. This is clearly an operator like whether, only phonologically null. Standardly, null operators are denoted by Op:

259

Chapter 7 - Complementiser Phrases

(62)CP

Op C©

C IP

Like other main clauses the complementiser position is empty and yet the null operator must agree with something to ensure the correct interpretation of the clause. This then is the trigger for the auxiliary inversion we see in yes–no main clauses. We end up then with a very uniform analysis of interrogatives in English: they all have an interrogative operator in the specifier position of the CP which agrees with the head. In main clauses the head cannot contain a complementiser, so an auxiliary is moved to C to enter into the agreement relationship. In subordinate clauses however, complementisers can appear and so there is no need for auxiliary inversion:

A final issue we might mention in connection with this analysis concerns the embedded clauses. If embedded questions can have complementisers as well as wh- elements in their specifier positions, why do we never get them together?:

(64)a I never heard [Op if [they caught the burglar]] b I never heard [who Q [they caught]]

c *I never heard [who if [they caught]]

This is a puzzle for which I have no real account. Apparently it is a very general condition that a CP can contain either an overt operator or an overt complementiser. We will see that this extends to other clauses too. Moreover, it appears not to be violated by an auxiliary moving to C. The condition has been known as the Doubly Filled COMP Filter since (1977) when it was introduced by Chomsky and Lasnik. However, this stipulatory account has never been superseded by anything more explanatory. I will therefore adopt the Doubly Filled COMP Filter as a condition on the well-formedness of structures in lieu of a proper explanation:

(65)the Doubly Filled COMP Filter

no CP can have both an overt specifier and an overt complementiser generated in C

This will have to suffice until we gain a better understanding of this phenomenon.

260