Complementizer optionality and topicalisation/adjunction in complement clauses

As discussed in chapter 1, structures headed by the copy o f a lower head not only feature in the representation o f verb (s-)movement. On the assumption that no abstract functional heads are available whose specifier could contain adverbs or topicalised phrases and that

'* S o m e sp eak ers seem to m arginally allo w em bedded inversion into the h igh est position in certain contexts, as in (i) from M cC lo sk ey IÇ l92:(8l). (i) I D o e s the ch air often know w hich o f th e p ro n io tio n s d id the D e a n su p p o rt?

T his story im p lies the diachronic prediction that English should have acquired com p lem en tizer optionality at the sam e tim e as it lost “ V -to -I” m ovem ent. S ee Arnold 1995, 1997 for relevant d iscu ssio n .

there is no categorial projection, the need to place an adjunct can create structures very similar to s-m ovem ent structures. Given the assumption that phrase structure is at most binary branching, the presence o f an adjunct can trigger the creation a new head position if all otherwise suitable adjunction sites already have two daughters. As in the case o f verb movement, this additional head position, if it doesn’t correspond to an additional lexical item, must inherit its properties from its daughter. Given the structure in (20a), adjunction can give rise to the structure in (2 0b), where the new root node inherits the categorial information o f the lower node V :

(20) a. V b. V

/ \ / \

A B Adjunct V

/ \

A B

This structure o f course parallels that o f verb movement discussed in the previous section. Since the two highest nodes in (20b) are copies, the must form a chain. The formation o f a chain means that this structure, which is dominated by head that is not in the root position o f its chain, will be subject to the chain condition in (1) above. This case must be distinguished from a situation where a head that doesn’t already have two daughters is a v a i l a b l e .I n such a case, it may be possible that the adjunct can be added as a daughter to this head, if no other constraint would be violated. The relevant structures are given in (21). In (21a), the verb V is dom inated by a non-branching head F. As shown in (21b), placement o f an adjunct in such a structure is possible without the creation o f a new head position. In such a structure, the two highest nodes belong to different categories, and consequently no chain formation is necessary. The chain condition will not apply to a structure like (21b). Structures like (21b) can be found for example in Greek where a focused constituent in an embedded clause may precede complementizers like oti (Roussou 2000). This is illustrated in (2 1c).

(21)a. F b. F

I / \

V Adjunct V

/ \ / \

A B A B

c. F ovam e [c o Y a n n is oti tha m e apatisi] fe a r -1SG the-NOM Y annis that w ill me-ACC cheat-3SG ‘I fear that Y an n is w ill cheat on m e .’ (M arika Lekakou, p .c.)

Exam ples that correspond to the structure where a new head is created in (20b) are illustrated here. In the example in (22a), the yesterday and, in (22b), the topicalised object p asta are adjoined to the top o f the clause. Since neither the adverb nor the fronted object enter into a selectional relation with the root node, this node inherits the category from its (clausal) daughter unchanged;

(22) a. ‘Yesterday John cooked pasta.’ b. ‘Pasta, John cooked.’

T T

/ . \ / \

Yesterday T Pastaj T

/ \ / \

Johuj cooked ^ 02# John; cooked 0# 02#

/ \ / \

ti cooked 0] 02# ti cooked 0| 02#

/ \ / \

pasta cook ed 0 | 02 tj cook ed 0 | 02

As in the case o f verb m ovem ent discussed in the last section, the root node and its daughter m ust form a head chain because they are copies. And again, the membership in a complex chain o f the root node o f these structures affects the positions in which they can occur. In particular, the chain condition in (1) implies that these structures are not selectable, and therefore that they cannot sit in a position to which a 0-role is assigned. These two situations are illustrated here in a schematic way, adjunction in (23a), topicalisation in (23b):

T his m ay be the ca se o f lan gu ages that have special particles in adjunction structures, and it is the default in theories that a llo w abstract fun ctional heads.

(23) a. * v e ,02# b. * v e ,02# / \ / \ T V 0102 T V 0 | 02 / \ / \ A d ju n ct T Topicj T / \ / \

Subji V 0i#02# Subji V 0i#02#

/ \ / \

ti V 01 02# tsubj V 0 | 02#

/ \ / \

O bj V 0 1 0 2 tj V 0 I 0 2

It is then expected that even in a language without verb m ovem ent like English, complementizers should become obligatory if adjunction to the highest head in a complement clause occurs.

There are o f course very popular alternatives to such analyses o f topicalisation and adjunction. In particular, Rizzi 1997 and Cinque 1999, respectively, argue that topicalised constituents and adverbials and other adjuncts occur in the specifier positions o f dedicated functional heads, not in adjoined positions. Simplified structures in a projection-free representation are given here:

(24) a. Facij b. Ftop / \ / \ A djunct T TopiCj T / \

7

\ Subji V 0|# 0 2 # Subji V 0 |# 02# / \ / \ ti V 01 02# ti V 01 02# / \ / \ Obj V 0102 tj V 0| 02

While these proposals o f course differ from the present one in their assumptions about clause structure and functional heads, the most important thing for the present argument is the fact that in these analyses, the root node does not form a chain with its daughter. That is, the heads F^dj in (24a) and Fjop in (24b) have different categories from their daughters and consequently no head chain needs to be formed. Since there is no head chain, the chain condition in (1) does not block selection o f such structures. This m eans that on such a view there is no obvious reason why selection o f a clause with an initial adjunct should be 82

impossible, and consequently this view doesn’t entail that there should be a restriction on the om ission o f an otherwise optional complementizer.

The argum ent is then the same as in the previous section. The present proposal predicts that adjunction and topicalisation should block omission o f an otherwise optional complem entizer in an embedded clause. The standard alternative makes no such prediction.^' The empirical accuracy o f this prediction is then again a good indication o f the relative m erit o f the different theories.

Again, the present proposal gets it right. Where topicalisation or initial adverbs occur in a complement clause that, cannot be omitted:^^

(25) a. I hope that this book you will read. (Doherty 1997:3a)

b. This proves that Joyce h e ’d read but that Yeats he hadn’t. (Doherty 1997:3a) c. She prayed that next Wednesday the check would arrive. (Doherty 1997:12a) d. We m aintain that in London a nice flat is hard to find (Doherty 1997:12c) e. She swore/insisted/thought that (,) most o f the time (,) they accepted this

solution. (Grim shaw 1997:(43a))

(26) a. *I hope this book you will read. (Doherty 1997:7a)

b. *This proves Joyce h e ’d read but that Yeats he hadn’t. (Doherty 1997:7a) c. *She prayed next W ednesday the check would arrive. (Doherty 1997:15a) d. * We m aintain in London a nice flat is hard to find (Doherty 1997:12c)

(Doherty 1997:15c)

e. *She swore/insisted/thought (,) most o f the time (,) they accepted this solution. (Grim shaw 1997:(43b))

And again, an alternative that p osits a covert com plem entizer in place o f a m issin g th a t/q iie w ou ld pattern with the standard proposal.

T here are sim ilar ex a m p les w here a com p lem en tizer is not obligatory: ( i ) / ’m su re the le c tu re b y K a y n e he w ou ldn 't m iss. (W atanabe 1993:1 4 5 n 2 3 (attributed to N oam C hom sky)) (ii) H e s a id no m a tte r w h at s ta n d he ta k es it w o u ld b e m is c o n s tr u e d th a t h e w as sy m p a th etic to on e o r the o th e r o f the R e p u b lica n s. (E lsn ess 1984:524). 1 g u e ss here the ‘matrix c la u se ’ must be a parenthetical (as I g u e ss in the sen ten ce im m ediately preceding this parenthesis m ust be, actually).

W hile the present proposal predicts the effect o f adjunction and topicalisation in a straightforw ard way, it is not clear how this effect could be captured in the alternative approaches.

In fact, this point is even more serious in view o f the result o f the previous section, where the effects o f verb movement on complementizer optionality were discussed. In both cases, the present proposal outperforms the alternatives. But beyond that, this proposal also can account for the fact that verb movement and adjunction and topicalisation affect com plem entizer optionality in the same way. Since verb movement and adjunction (through m ovem ent or base generation) give rise to the same structural configuration, their identical behaviour with regard to selection and especially the distribution o f that is expected. It is once m ore not at all clear how this generalisation could be captured in different proposals.

The sim ilar behaviour o f movement o f the verb to the highest head in a complement clause and adjunction to the top projection in a complement clause has already been noted by M cCloskey (I9 92:I2n9,19) (see also Kayne 1994:28). However, M cC loskey’s concern is not with the distribution o f the complementizer that, and his CP recursion analysis is not designed to provide an analysis o f it. Furthermore, while the parallel restrictions on the availability o f em bedded inversion and embedded adjunction are both related to whether or not the projection targeted by these operations is s-selected, they are ruled out by independent conditions in M cCloskey’s proposal. Following Rizzi and Roberts 1989:107, embedded inversion is ruled out by the assumption that I-to-C m ovem ent is substitution and substitution into a selected position violates the Projection Principle. Embedded adjunction is ruled out by a condition called Adjunction Prohibition formulated specifically to rule out adjunction to selected constituents.^^

In Grimshaw 1997, the similarity o f inversion in embedded clauses and adjunction in embedded clauses follows from a single constraint called Pure-EP:

(27)N o adjunction takes place to the highest node in a subordinate extended projection; and no m ovem ent takes place into the highest head o f a subordinate extended projection. (Grim shaw 1997:394)

S ee also C h om sky 198 6 :6 , 16, R ochem ont 1989. M cC loskey 1992:11 su g g e s ts that the Adjunction Prohibition m ay be related to the Projection Principle, but he d o e sn ’t investigate th is p ossib ility.

A unified account o f the two phenomena is o f course desirable, and this constraint is clearly designed to capture this generalisation. However unless this constraint can be derived from independently m otivated principles o f the grammar it amounts to no more than a restatem ent o f the facts. As Grimshaw 1997:395 points out, this formulation o f Pure-EP suggests that there could in fact be two separate constraints, one on adjunction and one on verb movement. W hether this would be a more accurate analysis is an empirical question, and, as Grimshaw notes, at least for English, there seems to be no motivation for a separation o f the two constraints. This can be taken as a further argument against Pure-EP, whose form ulation invites such a separation.

In the next section, it will be discussed how this approach to complementizer optionality can be extended from complement clauses to a type o f clause that is not selected, namely relative clauses.

4 Com plem entizer optionality in relative clauses

In document Phrase structure and derived heads (Page 80-86)