The CP and Split CP cross

advertisement
The CP and split CP cross-linguistically1
Elly van Gelderen, February 2003
In this article, I first argue that languages differ as to whether their clausal complements constitute
CPs or split CPs in the sense of Rizzi (1997). Then, I show that complements to factive verbs (e.g.
regret) are typically split CPs whereas those to non-factive verbs often are not. Thus, in Dutch and
German, only complements to factive verbs are split CPs; in Italian and English, complements to all
verbs are. English, however, has the added problem that the split CPs with factive verbs are nontransparant.
As many have shown, even as early as Kiparsky & Kiparsky's (1970) original article on
factives, factives can divided into emotives (e.g. regret) and non-emotives (e.g. discover) and the
latter pattern with the non-factives (e.g. believe). In work by Hooper & Thompson (1973), that
group is referred to as assertives. I then reformulate my proposal on split CPs in these terms. Nonassertives (regret) may have a split CP complement, whereas assertives (discover and believe) have
a single one in most languages. As mentioned, in English and Italian, the CP is always split.
The outline is as follows. Section 1 provides some background on the split CP by looking at
Rizzi (1997; 2001) and Cinque (1999). Section 2 examines cross-linguistic differences as to the
presence of the split CP, whereas section 3 shows that matrix verbs also influence the selection of a
split CP. Section 4 shows that the split CP is introduced in Middle English but that not until the
19th century did clauses become intransparent.
1
Background: the split CP
A relatively `traditional' notion of the clause is that it consists of three layers: the C(omplementizer)
P(hrase), the I(nflection) P(hrase), and the V(erb) P(hrase). The CP provides a connection to a
matrix sentence or contains an indication of speaker attitude; the IP includes grammatical
information such as tense and agreement; and the VP accomodates the arguments. The tree for this,
as in (1), is very familiar by now accounting for the word order in languages across the world:
1
1.
CP
.
C'
C
IP
.
I'
I
VP
Each of the layers in (1) has been argued to be split. The VP is split in Larson (1988), the IP is in
Pollock (1989), and the CP in Rizzi (1997). An extreme position is Cinque (1999) who has 40 or so
levels in a clause.
In 1.1, I will briefly review the reasons for assuming expanded CPs. I also show there is
evidence of movement of a complementizer and wh-element from the lower part of the CP to the
higher part. In 1.2, I examine Cinque's expanded structure, and justify the use of Rizzi (1997).
1.1
Rizzi (1997) and the expanded CP
In Lasnik & Saito (1992) and McCloskey (1991), topics and sentence adverbs are seen as adjoined
to the main clause. Adjunction is not very restrained and therefore it is good to eliminate it from the
grammar. One way to do this is to think of the `adjoined' elements as occupying separate functional
categories, as part of an expanded CP. During the last decade, Hoekstra (1993), Rizzi (1997; 2001),
and Cinque (1999), to name but a few, have suggested such an expanded functional projection to
accommodate the complementizer and other material appearing on the left edge of the sentence, as
in (2):
2.
... Force ... (Topic) ... (Focus) ... Fin IP
(from Rizzi 1997: 288)
Rizzi (2001: 289) assumes (3), where Int is interrogative. However, some of the Topic positions can
be eliminated by allowing a head to move to the different head positions:
3.
Force (Top*) Int (Top*) Foc (Top*) Fin IP
As is well-known, topics do not trigger V-movement in English, as (4) shows, whereas they
do in the other Germanic languages, e.g. (5) from Swedish:
2
4.
5.
a.
Stewed pears we always avoid.
b.
*Stewed pears avoid we always.
Kompott äter vi aldrig
Stewed-fruit eat we never
This difference receives a nice explanation under a more expanded CP, as in (2): in English, the
head of topic need not be filled, but in Swedish, it does. Recent work using a split CP to account for
Verb movement is e.g. den Dikken (2002). It also accounts for the fact that even though topics
generally precede wh-elements, in embedded sentences the reverse holds (e.g. Kiss 1995: 12), as (6)
shows:
6.
?I don't know whati to Mary ti we should give ti
(from Den Dikken, p. 4, but my traces are added)
This may mean that the wh-element moves from a lower to a higher position in embedded
sentences such as (6). This move to the ForceP will be shown to result in wh-clauses being able to
follow prepositions. As will be shown below, in language change, the complementizer that also
changes from being in the lower part of CP to the higher part.
Rizzi (1997: 283) argues that "the complementizer system [is] the interface between a
propositional content (expressed by the IP) and the superordinate structure (a higher clause or,
possibly, the articulation of discourse, if we consider a root clause)". Therefore, the CP contains
elements that look outside which he calls Force (following Chomsky 1995), and those that look
inside, which he refers to as Finite. In addition, the CP in (2) (optionally) accommodates Topic and
Focus. I am assuming a topic is old information, and a focus is new, but not much hinges on this
since there are very few sentences with both in the CP. See Rizzi (1997) for more. Languages in
which both topic and focus can appear sentence-initially display the sequence of (2), e.g. (7), from
Zulgo, a Chadic language, and (8) from Modern Greek:
7.
8.
mekele ka, gat na azla sigwe ya
mekele TOP he FOC he-took money FOC
`As for Mekele, it is he who took the money' (from Haller & Watters 1984: 30, but tones
left out)
Ti ðoulje mu se kanenan ðen tin embistevome
3
the work mine to noone not it entrust
`I don't entrust my work to anyone' (from Condoravdi & Kiparsky 2002, p. 4).
Rizzi maintains that, in Italian, the finite complementizer che is in Force. The evidence for
this comes from topicalization and left dislocation. The finite complementizer precedes the Topic,
as in (9), from Rizzi (1997: 288):
9.
Credo che il tuo libro, loro lo apprezzerebbero molto
believe that the your book they it appreciate much
`I believe that they would appreciate your book a lot'.
A tree for a sentence such as (9) would look like (10):
10.
VP
V
ForceP
.
Force'
Force
TopicP
.
Topic'
Topic
credo
IP
que
il tuo libro
loro lo ....
Poletto (2000: 129) says that speakers accept che on the other side of the topic as well, which may
mean optional movement of che from Fin to Force. She (2000: 174) argues either verb or
complementizer move this way depending on the setting features of the matrix verb. This is
reminiscent of movement of what in (6) and shows some clauses are split.
The English finite complementizers that and for behave like che, as (11) (taken from
McCloskey 1991) and (12) show. This is less true for factive verbs, as in (13), a fact which will
become important later on in section 4:
11.
12.
13.
She maintains that Irish stew she sort of likes t.
... for Irish stew I sort of like.
?I regret that Irish stew she sort of likes t.
4
This means that and finite for are in Force not in Fin2. Historically, this has probably always been
the case for finite for, which is first attested as a finite complementizer around 1200, as in (14) and
(15). Since topics and adverbs follow for, the latter is in Force. That is now in Force but in Fin
probably till about 1500, as (16) and (17) show:
14.
15.
16.
17.
For hardely I hym heete
for indeed I threaten him (York, 11, 286)
For frenshippe we haue foune. (York, 10,12)
She loved Arcite so
That [when that he was absent any throwe],
Anon her thoghte her herte brast a-two (Chaucer, Benson p. 377: 93-5)
and forþi we clepeð him fader for þat he us feide here
`and therefor we ... him father because he ...'
(Trinity Homilies HC-ME1)
The historical change of that from Fin to Force is compatible with contemporary ones such as that
of che, as described by Poletto, and Modern Greek, as in (44) below.
1.2
The `cartographic' approach
In this section, I briefly review Cinque and show why a less expanded CP is preferable.
Cinque (1999), in what has been coined the `cartographic' approach (see e.g. Cinque 2002),
argues that CP and IP do not suffice because adverbs need to be accomodated. The CP needs to
accomodate speech act (frankly, honestly), evaluative ((un)fortunately), evidential adverbials
(allegedly, evidently), and modal affixes in certain languages. The full range of the three CPadverbials is given in (18). I have added an IP-adverbial as well, namely the epistemic one:
18.
Mood speech act Mood evaluative Mood evidential Mod epistemic
frankly fortunately
allegedly
(from Cinque 1999: 107)
probably
Cinque (1999) does not use Rizzi's categories, unfortunately, but testing the compatibility of these
adverbs with topics and focus, one finds (19) and (20). Even though (19) seems slightly odd, it is
5
acceptable to native speakers with honestly in ForceP and those books in the topic. (20)
nevertheless presents a problem in Cinque's approach, since frankly is higher in the tree than
surprisingly, but the two cannot occur together as in (20):
19.
20.
?Honestly, those books, he should have read before class.
*Frankly, surprisingly, he read those books.
The examples Cinque gives with multiple adverbs are as in (21), but one of the adverbs is inside the
IP:
21.
Honestly, I am unfortunately unable to help you. (p. 33)
This suggests that the English CP is restricted.
Corpus data show the same as judgement data. In the 100 million word British National
Corpus (hence BNC), frankly and fortunately never cooccur; neither do fortunately and allegedly;
or allegedly and frankly (even though the adverbs occur frequently by themselves). Thus, matrix
CPs such as (24) and (25) show that the CP contains a ForceP with the sentence adverb and a
TopicP with a topic.
Cinque says very little about subordinating conjunctions, which is surprising since these
originally motivated the use of C and CP (Chomsky 1986). The BNC gives three instances of that
preceding frankly, as in (22):3
22.
She has told Paul that frankly she's lapping up the attention (BNC KBF 8830)
This should not be possible under Cinque's approach since the Specifier of the Speech Act adverb,
frankly, should be highest, and in (22), the complementizer that is.
Sentences such as (22) are perfectly compatible with Rizzi's account where that is in Force
and the adverb in the topic position. The ungrammaticality of (24) and (25) above is again in
accordance with Rizzi but not with Cinque.
2
Crosslinguistic Differences regarding the split CP
6
In this section, I will show that there is quite some variation in whether or not languages expand the
CP or not. Non-finite clauses typically do not, even though they do in Rizzi's account, since they are
reduced clauses (section 2.1). There is quite a lot of other variety, e.g. in whether topics are
incorporated (section 2.2), and whether there are multiple complementizers (section 2.3). In
concluding (table 1), I make a distinction between evidence for the split CP from actual
complementizers and from topic/focus material present in the CP.
2.1
Infinitivals
As to infinitival complements, Rizzi (1997; 2001: 288) argues that the infinitival di occupies Fin,
since it follows the topic, as in (23), from Rizzi:
23.
Credo, il tuo libro, di apprezzarlo molto
believe the your book for appreciate-it much
My focus is not Italian but other non-split possibilities exist, e.g. having di in I. Rizzi (2001: 287)
allows for the possibility of a single head in CP: "[w]e may think of Force and Finiteness as two
distinct heads closing off the complementizer system upward and downward respectively (and
perhaps coalescing into a single head in the simple cases)".
In English, the infinitival complementizer for in (24) is similar to Italian di in (23) in that a
topic cannot follow it. The reason is that for needs to be adjacent to the subject for Case reasons.
However, it is unlike di in that a topic cannot precede it either, as the ungrammaticality of (25)
shows. This is unexpected if for is in Fin:
24.
25.
*I expect for [his work] the president to do.
*I expected [his work] for the president to do.
If we test the position with other non-finite complementizers, the result is the same, i.e. topics
cannot precede them. Sentence (26) is taken from the BNC as a typical instance of in order to.
When transformed into (27), it becomes ungrammatical. Other topicalizations are impossible as
well. The same is true with non-finite whether, as in (28) and (29), formed from (30):
26.
There are several hurdles that must be overcome in order to successfully enter the UK job
market (BNC HCO 74).
7
27.
28.
29.
30.
*... that must be overcome in order [the job market] to successfully enter.
*The Office of Fair Trading considered [these takeovers] whether to refer to the
Monopolies and Mergers Commission.
*The Office of Fair Trading considered whether [these takeovers] to refer to the
Monopolies and Mergers Commission.
The Office of Fair Trading considered whether to refer these takeovers to the Monopolies
and Mergers Commission (BNC ALV 695).
Thus, non-finite complementizers in English do not fit readily in the expanded CP. Only the nonfinite complementizer can be present, not a topic or focus4. This is not unexpected given their
reduced clausal status.
2.2
Differences in finite clauses: incorporation of topics
Not all languages have such an expanded CP as in (10) above and I'll briefly look at that now where
topics are concerned.
Abraham (1997: 39) argues that there is no TopP or FocP between CP and IP in German,
whether the C is finite or non-finite. In Dutch, the same is true, and I will just give the Dutch
examples: (31) is comparable to (9), but neither order of complementizer and topic is acceptable,
and so is (32) with an adverbial clause, and (33) a factive verb complement which will be shown to
more readily allow split CPs in a number of languages:
31.
32.
33.
a.
*Ik geloof dat jouw boek ze waarderen
I believe that your book they appreciate
b.
*Ik geloof jouw boek dat ze waarderen
I believe your book that they appreciate
`I believe they appreciate your book very much'.
*Ik ging naar de winkel dat boek om te kopen
I went to the store that book for to buy
a.
b.
*Ik vergat dat jouw boek ze waarderen
I forgot that your book they appreciate
*Ik vergat jouw boek dat ze waarderen
I forgot your book that they appreciate
`I forgot they appreciate your book very much'.
8
The ungrammaticality of these sentences changes when the complementizer dat is left out. This is
only possible with non-factives, as in (34), the complementizer-less counterpart of (31), where the
subject (but any topic could go) moves to Spec CP and the verb moves to C. This confirms single
CP status. I'll come back to this in section 3:
34.
Ik geloof ze waarderen jouw boek.
I believe they appreciate your book
2.3
Double Complementizers
Even though embedded topics, as in (31), are limited in Dutch, there is still evidence for a split CP
in that multiple complementizers occur, as in (35). Extraction is possible, as in (36):
35.
36.
Ik weet niet wie of dat 't gedaan heeft.
I know not who if that it done has
Wie weet ik niet of dat 't gedaan heeft?
Who know I not if that it done has5
Double complementizers are only possible with verbs such as vergeten `forget', betwijfelen `doubt',
and afvragen `wonder'. Like for in English, it seems quite restricted and indicates a doubt or unreal
situation, as in (37):
37.
Ze gedraagt zich of dat ze hier de baas is
She behaves herself as-if that she here the boss is
`She behaves as if she is the boss' (from ANS 664 but with dat added).
Even though Dutch does not allow embedded topics, as (31) shows, it has a small class of verbs
that allows a split CP.
Barbiers (2002) argues that in Dutch, there is a ForceP in factives which allows whembedding, as in (38), but not in non-factives, as in (39):
38.
Jan zal meedelen welke boeken dat Marie leest.
9
39.
Jan will announce which books that Mary reads
*Jan zal denken welke boeken dat Marie leest.
Jan will think which books that Mary reads (Barbiers 2002: 50-51)
He correlates the ForceP with the presence of the complementizer of `whether' in factives, as in (40)
and their absence of non-factives, as in (41):
40.
41.
Jan zal meedelen of (dat) Marie deze boeken leest.
Jan will announce if that Mary these books reads
*Jan zal denken of (dat) Marie deze boeken leest.
Jan will think if that Mary these books reads (Barbiers 2002: 50-51)
I will come back on the use of factives but I think (38) and the others are a subclass of factives. ANS
(p. 645; 1151ff) talks about the difference between dat and of as certain versus uncertain, so factive
versus non-factive. This can also be seen in (37) above, and (42) with verb-second:
42.
Het scheelde geen haar of hij was verdronken
It missed no hair or he was drowned
`He had almost drowned' (ANS 648).
Adding a topic by changing (42) into (43) results in a grammatical construction. This shows that the
verb has moved to Fin and of is in Force:
43.
Het scheelde geen haar of gisteren was hij verdronken
It missed no hair or yesterday was he drowned
So, Dutch allows an expanded CP with those matrix verbs that have a double C. These are not
regular factives, however, in that rather than expressing an emotion, they express doubt.
2.4
To Split or to not
Because of the incompatibility of topics with non-finite for, as in (24) and (25) above, I assume
Modern English non-finite CPs have only one CP, with only the head filled. For finite
constructions, Rizzi's structure provides insight since finite for is in Force in (12), preceding the
10
topic, as is that in (11). Older English has two complementizers, as in (17), in the same way as
Dutch, one in Force and one in Fin. Dutch does not allow embedded topics in the way Modern and
older English do.
There is some evidence that in Modern Greek, there are three Cs, as argued in Roussou
(2000). Since oti can either precede or follow the topic/focus in sentences with tha in the lowest C,
there may be three Cs in (44), or tha may be in T:
44.
nomizo (ta mila) oti (ta mila) ðen tha ta fai o petros
think-1S (the apples) that (the apples) not future eat-3S the Peter
`I think that Peter won't eat the apples'. (Roussou 2000: 76).
It may also be the case that oti moves from Fin to Force. Thus when it follows the topic, it is in Fin
and when it precedes, it is in Force. This is similar to what happens historically with English that,
with wh-clauses as in (6) above, and in Italian varieties with che.
In table 1, I summarize the possibilities. There is in principle a `no-no' setting, but not in the
languages reviewed here:
____________________________________________________________
Italian/English Dutch
Greek/Middle English
__________________________________________
multiple Cs
no
yes
yes
topics in CP
yes
no
yes
____________________________________________________________
Table 3.1: Split CPs
3
Complement selection
Matrix verbs (partly) determine the structure of the verb in the complement. Thus, crosslinguistically, causatives do not have CP complements and verbs of saying do not have VPcomplements. This is of course not completely true since languages choose to be more CP, IP, or
VP oriented. In the present section, I examine a possible relationship between the meaning of a verb
and the presence of a split CP.
In 3.1, I'll first look at the traditional effects of factive and non-factive matrix verbs on the
11
structure of their complements, as argued early on in the work of Kiparsky & Kiparsky (1970),
hence K&K. K&K divide factives into emotives and non-emotives, mainly because of the ability of
the former to have a for-to and subjunctive complement. I discuss several characteristics of factives,
such as harder to move out of it through passivization, raising, and wh-movement (hence loosely
called non-bridge verbs), but, using arguments put forward by Hooper & Thompson (1973) and
others and using the BNC, I show these characteristics typically appear with emotive factives, but
not with the non-emotives (also called semifactives). H&T suggest `assertion' is the relevant
distinction between verb groups. None of these use the split CP to account for the differences: a
split CP, in particular one that can have the fact that, creates more barriers in the complements of
emotive factive verbs.
3.1
Factives, non-factives, emotives, and non-emotives
The distinction between factives and non-factives is well-known since the work by K&K (1970). I
will provide some background on these, as well as on K&K's distinction for factives between
emotives and non-emotives that has been accepted in a number of later works. I then relate the
difference to the split CP.
K&K (1970: 159), talking about complements to factive verbs, call their independence a
"syntactic insulation". They (1970: 162) show that the complements to factive verbs are more
insolating than those of non-factives. The difference between them is both semantic and syntactic.
With factive verbs, as in (45), the truth of the complement clause is presupposed by the speaker,
unlike with non-factives, as in (46):
45.
46.
I forgot that he left a key.
I believe that he left a key.
A partial list of factive and non-factives is given in (47):
47.
Factive: regret, be aware of, grasp, comprehend, ignore, forget, make clear, resent, know,
realize, deplore, see.
Non-factive: suppose, assert, allege, assume, claim, believe, conclude, say, think, assume,
deem, be possible, be true.
Both: anticipate, acknowledge, suspect, report, remember, admit.
(cf. Kiparsky & Kiparsky 1970: 145; 163).
12
Some syntactic differences involve it-pronominalization (with factives), ECM-constructions (with
non-factives), paraphrasing through the fact that (with factives), passive raising (with factives),
embedded topics (marginal with factives), and wh-extraction of the subject and adjunct (with nonfactives). I'll argue that certain factives have a split CP with one of the layers blocking extraction.
K&K divide the factives in emotives and non-emotives, as in (48):
48.
Emotive:
factives: regret, resent, deplore, be important
non-factives: intend, prefer, be unlikely
Non-emotive:
factives (later called semi-factives): forget, be clear, be aware of
non-factives: say, suppose, conclude, seem, be likely (K&K 169-170)
The main reason for this is that emotive complements can be for - to clauses or have subjunctives,
even though, as I'll show later, not all of them do.
Starting with it-pronominalization and paraphrasing through the fact that, these are possible
with factives, as in (49) and (50):
49
50
a.
b.
a.
b.
They regret it that Hittite is extinct (no examples in the BNC)
*They believed it that Hedwig had brought that message.
Many Americans regret the fact that they now discuss sport with the passion that
200 years ago they brought to every day debates... (BNC-ABD 752)
*Many Americans believe the fact that they now discuss ...
Hooper & Thompson (1973: 481) show that the subclass identified by K&K as non-emotives
cannot have these, as (51) and (52) show:
51.
52.
*I see the fact that the Bruins lost.
*I know the fact that you're not speaking to me.
(from H&T 481)
A search in the BNC confirms this for see but not for other non-emotives such as realize and
13
discover where two instances of each occur, as in (53) and (54):
53.
54.
... was quick to realize the fact that Cubism had not been the product of ... (BNC GUJ 863)
I thought that Sarah had discovered the fact that we were both, ..., being unfaithful to her
(BNC HD6 746)
ECM-constructions and passive raising occur with non-factives, as in (55) and (56), and
embedded topics are mostly ungrammatical with factives, but grammatical with non-factives, as in
(57):
55
56
57
a.
b.
a.
b.
a.
b.
I believe him to be nice.
*I regret him to be nice.
He was believed to be pro-French. (BNC-CRK 998)
*He was regretted to be pro-French.
John believes that this book Mary read often.
*John regrets that this book Mary read often.
Again, not all factives work alike. The non-emotive know allows ECM, as in (58), which occurs
very frequently, but not the non-emotive forget:
58.
59.
nor did I ever know him to be disloyal (BNC CD9 1620)
*I forgot him to have done that.
Emotives can have subjunctive and for-to clauses, as (60) and (61) show:
60.
Asserting, what was palpably untrue, that `There are probably few people in India who do
not sincerely regret that you should have made it impossible for any government to leave
you at liberty', he ... (BNC C90 1004)
61.
I regret for you to be in this fix (KK 169)
Having shown that factives versus non-factives is too broad, I'll discuss the assertive versus nonassertive difference.
14
2.2
Assertive
It has long been noted (Vikner 1995: 70 for an overview; Haider 1986) that non-factives allow
verb-second (V2) order in their complements in German, but that factives don't. The same is true in
Dutch, as (62) shows:
62.
ik geloof/*betreur gisteren is ie gevallen
I believe/regret yesterday is he fallen
This is similar to English topics, in (57) above, repeated here as (63). They are allowed with nonfactives, but disallowed with factives:
63
a.
b.
John believes that this book Mary has read many times.
?/*John regrets that this book Mary has read many times.
In English (63a), the complementizer that appears in Force and this book in the topic. This is true in
Danish as well, as in (64) from Vikner (1995: 67):
64.
Vi ved at denne bog har Bo ikke læst.
We know that that book has Bo not read
This means Danish has a split CP with at in Force and har in Fin. Since English doesn't allow V2,
and the topic would have to be in a separate position in the expanded CP, it means that the CP is
split in these cases. In German and Dutch, many argue that V2 is movement of the topic to the Spec
of CP and of the finite verb to Fin. Therefore, (62) could have that structure too, i.e. of a single CP.
This would mean complements to non-factive verbs in German and Dutch would be a single CP,
but complements to factive ones would be split and allow for double Cs6.
Meinunger (2002) shows that the verbs that allow V2 complements in (62) above are those
of saying, thinking, and the non-emotive factives (verbs of `discovery'). The ones that don't are
emotive factives, negative verbs, such as nicht glauben, imperatives, questions, and causative verbs,
such as verursachen, schaffen. He follows H&T in referring to these as assertive versus nonassertive. The volitional verbs, he argues, show the characteristics of the non-assertive verbs. He
relates this to H&T's observations on root transformations in English. These are allowed with
almost the same matrix verbs except that verbs of wanting are left out. I agree with this distinction,
15
even though it may be hard to fit assertion in a non-split CP (it would be typically something of
ForceP). Meinunger's verb classes end up as follows with semi-factive an alternative name for nonemotive factives7:
Assertive:
saying
thinking
semi-factive
Non-assertive:
emotive factive (resent, regret)
negative verbs
causative
volition
Assertive verbs allow V2, lack subjunctives, and display root transformations; non-assertive verbs
show the opposite characteristics.
How does this new division fare with K&K's prototypical features, such as the fact that and
it after the verb. A (semi-)factive such as see should have the fact that under K&K's account. In the
BNC, there are 115,200 instances of see but none where a form of see is followed by the fact that.
Realize(d), another (semi-)factive, has two instances, and the other semi-factives have some too.
This is problematic for the division, any division in fact. I would like to suggest that, as in the case
of which verbs select for to, there is an element of chance, as to whether and how the ForceP is
filled.
Meinunger doesn't rely on a split CP, but I'll argue that the (finite) CP is always split with
non-assertive predicates, such as regret, and that there is a the fact that blocking the ForceP. This
ForceP would also block movement to the topic position in English, perhaps because the entire
proposition is presupposed, i.e. the topic. English complements to assertives would also be split
since topicalization is allowed after that, but Dutch and German complements are single CPs since
topics cannot appear after dat/daB, as in (65) and (66):
65.
*Ik geloof dat deze boeken zij niet waarderen
I believe that these books they no appreciate
66.
*Ik geloof dat deze boeken waarderen zij niet
If German and Dutch assertive complements also had split CPs, (65) and (66) wouldn't be
ungrammatical.
16
3.3
A structure
A structure of emotive factives is given in (67), with the ForceP `blocked' by a covert or overt
element. This structure is a hybrid between a nominal and a clausal structure, and is proposed for it
with extraposed clauses in e.g. Stroik (1991):
67.
ForceP
Spec
(the fact)
(it)
Force'
Force
that
...
(64) would account for (49) and (50) readily. It would also connect the fact that both wh-clauses, as
in (68), with the wh in Spec ForceP (see (6) above), and the fact that/it that, as in (69), occur as
prepositional objects:
68.
69.
I was aware of what he said.
I was aware of the fact/it that he resigned.
Infinitival complements are reduced clauses, but in the case of emotive factives that have a structure
as in (67), they will never reduce to an IP completely. This explains why ECM-verbs and passives
are impossible (cf. e.g. Massam 1986), but for-to infinitives and subjunctives are. Thus, him in
(55a) but not in (55b) can be Case marked; he in (56a) can move to the subject position of the main
clause since there is no double CP boundary in between, but not in (56b). Volition verbs do not
have a structure as in (67) but one where the CP is nevertheless never reducible since an
irrealis/future marker for is present.
To recap the English situation, factives (all real factives and some semi-factives) occur with
the fact that and it that. This shows their structure is as in (67). The other verbs do not. ECM is
typical for non-factives except verbs of saying which never reduce (say, claim, assert). Within the
ECM verbs, there are two kinds, verbs of belief lack a C in the infinitive and those of volition have
one:
Assertive
semifactive
(discover)
Non-Assertive
factive volition
(regret) (want, intend)
think say
(believe)
17
the fact some
it that
Emb Top
ECM
Raisingyes
Subjunctive
for-to
no
some
yes
yes
yes
no
no
no
no
yes
yes
yes
no
no
yes
no
yes
no
no
yes
no
no
no
no/yes yes
no
no
no
yes
no
yes
yes
yes
Table 2: English complements
4
The split CP in the History of English
In Old English, there is not much evidence for a split CP: there are no double complementizers8 or
embedded topics. In this section, I provide some evidence for when the split CP is introduced.
Double complementizers start to appear in the 13th century (see (17) above); embedded topics and
prepositions followed by that-clauses appear a little later in the 15th century. This typically occurs
in adjunct positions. The fact that, as in (50) above, and it that do not appear regularly till the 19th
century.
4.1
Evidence for a split CP?
The first instance of for as a C that the OED provides is from 1123 (from the Peterborough
Chronicle). Not long thereafter, combinations such as for that, as in (17) above and (70), till that, if
that, while that start to occur:
70.
I trowe I loved hym best, for that he Was of his love dangerous to me
(Chaucer, Benson p. 112)
This means for is in Force and that remains in Fin, but the CP is split.
It is harder to date the introduction of embedded topics exactly. There are many in Chaucer
(lC14) and in the York plays (lC15). Again, these first appear in adjunct position, as in (71) to (73):
71.
But of his face I can not seyn the hewe/ For sikerly his face shone so bryghte (Chaucer,
18
72.
73.
Benson, p. 594, ll. 162-3)
Ther may swich cause ben .../ That hardily thou wolt thiselven saye (Idem, p. 577, ll. 13056)
And thus I lyved ful many a day/ That trewely I hadde no ned (Idem, p. 345 ll. 1252-3)
So by 1400, that may move to Force.
The introduction of complements starting with the fact that is quite late. According to the
OED, the phenomenon starts in the 19th century, and the first instances are factive, as in (74) to
(76), but slow to get started (e.g. none in Darwin's Origin of Species or in Jane Austen's novels):
74.
75.
76.
I would not agree to the fact that ennui prevailed (1803, from OED entry for fact)
... ought to be made aware of the fact that among the reigning Sovereigns, [they] have not ...
(1851, from OED entry for fact)
We cannot ignore the fact that aeroplaning is beginning to progress as a pastime (Observer
1927, from OED entry for aeroplane)
There are, however, some early instances, as in (77). The texts in which these occur interestingly
have a that-clause after a preposition, as in (78). This provides some evidence for the beginning of a
structure as in (67) above in the 15th century:
77.
78.
79.
We have done evyll that we have not taken surete (1489 Caxton, from OED entry for that)
I shalle not leue the goo, withoute that thow hold to me (1484 Caxton, from OED entry for
that)
After that I understande by your wordes (1475 Caxton, entry for after)
When do sentences with an it, as in (49) above, start appearing? Heralding objects occur in
OE, as Visser calls them, but are problematic in terms of their analysis. Considering Visser's
examples (p. 460) they occur with all kinds of verbs, as with the semi-factive understand in (80),
and others such as hear and believe, as in (81). This is unexpected if they are really like the
structure in (49). In addition, the heralding objects typically precede the verb, as in (81), and hence
do not contribute to an expanded CP:
80.
and þæt georne understandan, þæt ðær symble heofonlicra engla neawest bið
19
81.
and that eagerly understand, that there feast of-heavenly angels nearest is (Wulfstan Polity
252.15, from Visser 460)
þæt ne gelyfdon þte liffruma ... ahafen wurde.
that not believed that giving ... elevated was (Crist 656 from Visser 460)
If OE is a more paratactic language than its modern counterpart, one expects an independent object
with a separate clause. Sentences such as (80) are therefore not like (49) above, but contain two
separate clauses. The introduction of the modern variant of (80), one that provides evidence for (67)
is quite late. It doesn't occur in the Early Modern English section (up to 1710) of the HC; in
Shakespeare's First Folio (1623), there are none, but there is an instance of a prepositional object as
in (82):
82.
I will swear by it that you loue mee (Much Ado)
So, by 1400, a split CP is present: double complementizers occur since the 13th century,
embedded topics since the late 14th century, and prepositions precede that-clauses since the 15th
century. The Spec of ForceP is not `filled' up with it/the fact until later.
4.2
Do matrix verbs matter?
Having shown that the split CP started with adjuncts in Middle English, I'll show that assertive and
non-assertive complements are very similar in OE in terms of whether they take an indicative or
subjunctive complement. It isn't until Late ME when non-assertives start to allow for-to
complements and, as shown above, not until the 19th century that the fact that appears in the
complement clause.
The core set of non-assertives, the emotives, in English consists of loans from French, e.g.
resent, regret, deplore. OE counterparts are sorgian `be sad' and sweorcan `be troubled'. When they
occur with CP complements, these are either indicative or subjunctive:
83.
hi þa sorgodon þæt hi sceoldon heora gewunan forlætan
they then grieved that they should their ... leave (DOE, seg 30)
Volitional and causative verbs are also non-assertive and show single CPs, as in (84) and
(85). In OE, they have a CP complement, as in (84) and (85), which for both can be either
20
indicative, as in (84), or subjunctive, as in (85):
84.
85.
Ic gedo ðæt ðu forgitst
`I make you forget'. (Alfred, CP 207/11)
Ic wille ... þæt þu forgyte þæt ic þe nu secge I want you to forget ... (Byrhtferth's Manual
154/14, Visser 841)
Assertive complements such as to the semi-factive forgietan occur with a clausal
complement in the indicative, as in (86), and subjunctive, as in (87):
86.
87.
hig forgæton ðæt hig hlafas namon
The forgot that they breads took (from Visser 832)
ðæt hie forgieten hwider hie scylen
that they forget whither they should-SUBJ [go] (Pastoral Care 387.8, HC)
Another semi-factive is ongietan `to understand'. Looking through the Dictionary of OE texts, it
occurs most often in the Pastoral Care, and is often complemented by a that clause. There is one
instance where it might have a structure as in (67) with willan comparable to fact in Modern
English:
88.
he ongiete ðone ufancundan willan ðæt he ...
(Pastoral Care 51.7)
Perception verbs are assertive. In OE, see occurs mainly as a mental perception verb, as in
(89), or with a `see to it' meaning, as in (90):
89.
Gesihst þu nu þæt þa rihtwisan sint laþe 7 forþrycte
Seest thou now that the virtuous are hated and oppressed (Boethius, III, 4)
90.
Gesih ðu þæt nænigum menn ðu coeðe
See you that (to) man you talk (Lindisfarne MK I, 44, from Visser 838)
These complements are subjunctive. There is a possible 13th century direct perception example, i.e.
(91), but it is interesting that the first form of `see', sehe, is in the subjunctive and hence the `seeing'
21
is less direct. The second form, seoð, is a present plural which is more direct:
91.
Hwa þat sehe þenne hu þe engles beoð isweamed, þat seoð hare suster swa sorhfulliche
afallet
Who should see that the angles are grieved and who see their sister so sorrowfully
overthrown (Hali M Titus 23, 233-5)
A last class of assertives I'll look at is the class of saying. Verbs of saying occur with a
subjunctive (as well as indicative) complement in OE, as in (76), unexpected under the division
proposed above:
92.
þæt hi cwædon þæt he God wære
that they said that they God were (Hom I 190.32, from Mitchell I: 8)
Farkas (1992): 70) notes the same for Romance languages, and says there is a change of meaning,
e.g. between (93) and (94) from Romanian:
93.
Ion a spus ca Maria qa plecat
Ion has said that Maria has left
94.
Ion a spus ca Maria sa plece imediat
Ion has said that Maria SUBJ leave immediately (from Farkas 1992: 70)
So far I've shown that in OE assertive and non-assertive complements do not have a
different form: neither take double complementizers or embedded objects and both allow
subjunctives. In the remainder of this section, I show that when for-to complements start to occur,
there is a difference and since for in finite clauses is in ForceP, this may be the reason that the CP is
used for non-finites as well.
As is well-known from K&K (see section 3.1 above), emotive factives such as be important
and regret allow for-to complements. This use "is hardly ever met with before" the late 19th
century, according to Visser (2244-45). There are, however, instances he mentions, as in (95), and
(96) from the OED. The ealiest I have found is (97):
95.
and wishing for those hands to take off his melancholy bargain (1681 Dryden, from Visser
22
96.
97.
2248)
as I expect for my reward to be honoured with Miss Sophia's hand as a partner (1766
Goldsmith, from the OED entry for interest)
moche he lofde echn(e) cniht. þat lofde for to segg(e) riht
`Much he loved every knight who loved for to say the truth'. (1275, Layamon, Otho, 5523)
The complement to each verb and adjective has to be looked at separately since they all differ. The
history of important is interesting in that its use as a predicative adjective with a complement is
very late. It too is a loanword from French but quite a late one, namely late 16th century (the verb
import is from the early 16th century), and is initially only used attributively. Even 18th and 19th
century texts, such as (Bishop) Berkeley, (David) Hume, (Emily) Bronte, and (Jane) Austen only
contain attributive use, so the subjunctive use, as in (98), must have arisen late. Some early
American texts (e.g. Samuel Adams' 18th century Writing) just have subjunctives, as in (98), not
yet infinitival complements:
98.
Some of our military gentlemen have, I fear, disgraced us; it is then important that every
anecdote that concerns a man of real merit among them, and such I know there are, be
improved, as far as decency will admit of it, to their advantage and to the honor of a colony,
which, for its zeal in the great cause, well as its sufferings, deserves so much of America
(Samuel Adams, letter to E Gerry 1775, Cushing ed, University of Virginia e-texts)
George Washington (in the 37 volumes of his work written between 1745 and 1799; University of
Virginia E-text) uses important for with an infinitives seven times, as in (99):
99.
'Tis almost as important for us to know what does not happen as what does happen (1780,
Vol 19)
Concluding section 4, I have shown that OE has no split but that it starts gradually in the
13th century by having double complementizers, then embedded topics in the late 14th, and for to
very gradually from the 13th century. The relevance of prepositions preceding that-clauses is that
this is possible at a time when the split CP comes in but, once it/the fact become available, they are
needed.
23
5
Conclusion
Clauses consist of three layers: the CP, IP, and VP. In this paper, I have examined the structure of
the CP more. It turns out that languages choose (`set their parameters') as to how expanded a
particular layer is, in this case CP. I show that typically infinitival clauses are not split and that
languages differ as to how to expand the CP, by double complementizers or through embedded
topics. I suggest a structure to account both for verbs that have a more insulating complement
clause, namely (67) above for English. This structure accounts for the presence of the fact/it that,
impossibility of embedded topics, as well as for the properties of reduced clauses. Verbs that have a
structure as in (67) reduce in a different manner from those that don't and typically those verbs are
non-assertives.
References
Abraham, Werner 1997. "The base structure of the German Clause", in German: Syntactic
Problems. Werner Abraham and Elly van Gelderen (eds): 11-42. Tübingen: Niemeyer.
Algemene Nederlandse Spraakkunst (ANS). 1984. G. Geerts et al., eds. Groningen: Wolters
Noordhoff.
Barbiers, Sjef 2002. "Remnant Stranding and the Theory of Movement", in Artemis Alexiadou et
al. (eds): 47-67. Dimensions of Movement. Benjamins.
Beadle, Richard (ed.) 1982. The York Plays. London: Edward Arnold.
Benson, Larry 1987. The Riverside Chaucer. Boston: Houghton Mifflin Company.
Brook, G. L. & R. F. Leslie (eds). Layamon: Brut. Oxford: Oxford University Press, 1963 edition
of EETS 250.
Chomsky, Noam 1986. Barriers. MIT Press.
Cinque, Guglielmo 1999. Adverbs and Functional Heads. Oxford: OUP.
Cinque, Guglielmo (ed.) 2002. Functional Structure in DP and IP. Oxford: OUP.
Condoravdi, Cleo & Paul Kiparsky 2002. "Clitics and Clause Structure", Stanford Ms.
Dikken, Marcel den 2002. "On the Syntax of WH-movement". CUNY Ms.
Farkas, Donka 1992. "On the semantics of subjunctive complements". Paul Hirschbühler & Konrad
Koerner Romance Languages and Modern Linguistic Theory. Amsterdam: Benjamins.
Haider, Hubert 1986. "V-Second in German". Hubert Haider & Martin Prinzholm (eds)
Verb
24
Second Phenomena in Germanic Languages. Foris, Dordrecht, 49-75.
Haller, Beat & John Watters 1984. "Topic in Zulgo". Studies in African Linguistics 15.1: 27-46.
Hoekstra, Eric 1993. "Dialectal variation inside CP as parametric variation", Linguistische Berichte
5.
Hoekstra, Eric & Jan Wouter Zwart 1994. "De Struktuur van de CP". Spektator 23.3: 191-212.
Hooper, Joan & Sandra Thompson 1973. "On the Applicability of Root Transformations".
Linguistic Inquiry 4.4: 465-497.
Kiparsky, Paul & Carol Kiparsky 1970. "Fact". Manfred Bierwisch & Karl Heidolph Progress in
Linguistics, Mouton, The Hague, 143-173.
Kiss, Katalin 1995. Discourse Configurational Languages. New York: Oxford University Press.
Larson, Richard 1988. "On the Double Object Construction". Linguistic Inquiry 19.3: 335-391.
Lasnik, H. & M. Saito 1992. Move [Alpha], Cambridge: MIT Press.
Massam, Diane 1985. Case Theory and the Projection Principle. MIT PhD.
McCloskey, Jim 1991. "Verb Fronting, Verb Second and the Left Edge of IP in Irish". talk,
Stuttgart workshop on Comparative Germanic Syntax.
Meinunger, André 2002. "Verb position, verbal mood, and the anchoring (potential) of sentences".
MS.
Poletto, Cecelia 2000. The Higher Functional Field. OUP.
Pollock, J-Y. 1989. "Verb Movement, UG and the Structure of IP". Linguistic Inquiry 20: 365-424.
Rizzi, Luigi 1997. "The Fine Structure of the Left Periphery", in Elements of Grammar. Liliane
Haegeman (ed.): 281-337. Dordrecht: Kluwer.
Rizzi, Luigi 2001. "On the position `Int(errogative)' in the Left Periphery of the Clause". In Cinque
& Salvi Current Studies in Italian Syntax. Amsterdam: Elsevier: 287-296.
Rochette, Anne 1988. Semantic and Syntactic Aspects of Romance Sentential Complementation,
MIT diss.
Roussou, Anna 2000. "On the left periphery" Journal of Greek Linguistics 1: 65-94.
Sedgefield, W. 1899. King Alfred's Old English Version of Boethius. Oxford: Clarendon.
Stroik, Tom 1991. "Expletives Revisited", Linguistic Analysis 21: 23-33.
Sweet, Henry 1871 [1934]. King Alfred's West-Saxon Version of Gregory's Pastoral Care. London:
Oxford University Press.
Vikner, Sten 1995. "Verb Movement and Expletive Subjects in the Germanic Languages". Oxford.
Visser, F. Th. 1963-1973. An Historical Syntax of the English Language. Leiden: Brill.
25
Notes
1.
In this paper, I use several electronic copora and texts. The
British National Corpus (BNC) is a 100-million word corpus of
written and spoken contemporary British English. It is available
on the web, as well as on CD. I also use the Helsinki Corpus (HC),
the Dictionary of Old English texts (DOE), and the OED online. In
cases where I don't take the examples from the OED or from Visser,
I have provided references.
2.
Notice that, unlike topicalization in (1), left dislocation
is only possible in a main clause as in (i) and not in a
subordinate as in (ii):
i.
ii.
Those books, I read them.
*I know that those books I read them.
3.
Native speaker judgements show that sentences such as (i) and
(ii) are unacceptable:
i.
ii.
*I know that frankly she left
*I know frankly that she left.
4.
In the BNC, there are 3 instances of for
unfortunately, but these all involve finite for.
followed
5.
See also
sentences.
for
Hoekstra
&
Zwart
(!!1994:
198)
by
similar
6.
The Dutch translation of `the fact that' het feit dat is very
rare in sentences comparable to (50). ANS does not mention this at
all, and searching the Dutch newspapers in Lexis Nexis for the
past 10 years, I find over 40 after betreuren `regret' but they
all seem quite formal:
i.
Het algemeen bestuur van de Gemeenschappelijke Regeling ...
betreurt het feit dat de centrumgemeente Hulst ...
`the central administration of the ... regrets the facts that
the community of Hulst ...' (De Stem, 3 February 1994).
7.
I ignore adjectival predicates such as be possible.
8.
There are some instances of þæt followed by þe but these have
been shown to be a Specifier followed by a Head.
26
Download