Dialectal Variation inside CP as Parametric Variation

Eric Hoekstra, P.J. Meertens Institute,
Royal Netherlands Academy of Arts & Sciences

W. Abraham & J. Bayer (eds) Linguistische Berichte. Sonderheft 5. Dialektsyntax, 161-179.

1. Theory of the Sentence

The sentence is nowadays generally considered to be the maximal projection of the complementiser C (Pesetsky 1982, Chomsky 1986). Correspondingly, sentences are analysed as CP's. Spec-Head agreement and X'-theory interact to allow not only for a head position but also for a Spec position. This neatly accounts for examples like the following:

(1) a. Hij weet welke jongen dat je gezien hebt
he knows which boy C you seen have
b. Hij weet hoe of je dat moet doen
he knows how C you that must do

Here a Wh-phrase is followed by a complementiser.1 The complementiser must be right-adjacent to the Wh-phrase. This follows directly if the Wh-phrase functions as the Spec of CP. The complementiser must be left-adjacent to the following IP. This follows if IP functions as the complement of C.2
Pursuing this line of investigation, the following sentence suggests that there is more structure to CP than has been supposed up till now:

(2) a. Dat is niet zo gek als of dat hij gedacht had
that is not so strange C1 C2 C3 he thought had
b. Dat is lijk of dat hij had gedacht
that is C1 C2 C3 he had thought

Both sentences exhibit three complementisers. The (a)-sentence is marginally acceptable in Standard Dutch, and it is found in some dialects, such as Frisian Dutch. The combination of complementisers exhibited in the (b)-sentence is found in West-Flemish and French Flemish (De Rooy 1965:24 and map VIII).3
The phenomenon of triple complementisers directly supports the central claim of this paper:

(3) There are three different heads

Each complementiser fills a separate head position, accounting for the observed facts.
Each complementiser is associated with its own semantic effect. Consider the following sentences (Standard Dutch, Frisian Dutch):

(4) a. Het feit dat/*of/*als hij op tijd is
the fact C3 *C2 *C1 he on time is
===> C3 in declarative contexts
b. De vraag of/*dat/*als hij op tijd is
the question C2/*C3/*C1 he on time is
===> C2 in WH-contexts
c. Niet zo gek als/*of/*dat ik gedacht had
not so strange C1/*C2/*C3 I thought had
===> C1 in comparative contexts

The (a)-sentence shows that the complementiser "dat" is found in declarative contexts. The complementiser "of" is associated with Wh-contexts. Complementisers like "als" and "lijk" are associated with what I refer to as comparative contexts, since a comparison is involved.
Before continuing, I will first show that sequences of complementisers cannot be analysed as words inserted into C.

2. An alternative possibility

The question is whether the sequence "als of dat" really consists of three words filling three functional head positions. It could also be argued (exploring a suggestion made by Eric Reuland) that there is only one word, "alsofdat", which fills the head position C, realising a specific semantic value. Under this approach, "alsofdat" and "lijkofdat" would be words inserted in the head position C. If this approach is correct, we need not hypothesize the existence of three functional head positions.
The problem with this approach is that that a cluster of complementisers does not behave as one word. To illustrate, conjunction may split up the sequence of complementisers (cf. De Rooy 1978):

(5) a. Ik vraag me af [of [dat Ajax de volgende ronde haalt en dat Celtic verslagen kan worden]]
I wonder C2 C3 Ajax the next round reaches and C3 Celtic beaten can be
"I wonder whether Ajax will make it to the next round and whether Celtic can be beaten"
b. Ongeacht [of EN [dat het regent EN dat het waait, zullen we je met de fiets komen opzoeken]]
regardless C2 both C3 it rains and C3 it blows, shall we you with the bike come visit
"Regardless of whether there is wind and whether there is rain, we will go visit you by bike"
(6) a. Ze doen [als [of [dat hij koning is en dat zij koningin is]]]
They act C1 C2 C3 he king is and C3 she queen is
b. Ze doen [als [of dat hij koning is en of dat zij koningin is]]
They act C1 C2 C3 he king is and C2 C3 she queen is
(7) a. * ... [hoe [-wel hij koning is en/of -wel zij koningin is]]
al though he king is and/or though she queen is
b.* ... [in-[dien hij koning is en/of -dien zij koningin is]]
if he king is and/or if she queen is

Disjunction splits up a sequence of two complementisers in (5). Conjunction splits up a sequence of three complementisers in (6). The example in (7) shows that neither conjunction nor disjunction can split up words. The paradigm provides an argument against the idea that "alsofdat" is a word realising a value of C.
If each functional head projects to a maximal projection, then these facts follow directly. (5) and (6a) involve a conjunction of C3P's. (6b) involves a conjunction of C2P's.
A second argument is that a Wh-phrase can split up a sequence of complementisers in Frisian Dutch. This is shown below:

(8) a. ?? Ik weet niet hoeveel rijker Jan is dan wie of dat jou gezegd heeft dat hij is
I don't know how much richer Jan is C1 who C2 C3 you said has that he is
b.?? Niet zo rijk als wie of dat jou verteld had zei je dat Jan was
not as rich C1 who C2 C3 you told had said you that Jan was
(9) a. ?? Ik weet niet hoeveel rijker Jan is dan wie jou gezegd heeft dat hij is
I don't know how much richer C1 who you said has that he is is
b.?? Niet zo rijk als wie jou verteld had zei je dat Jan was
Not as rich C1 who you told had said you that Jan was

The second and the third complementiser are optional in Frisian Dutch. Interestingly, the grammaticality judgments for these examples are independent of whether these extra complementisers are present or not. Suppose the complementisers would form a word. Then (8) should be crashingly ungrammatical, since a Wh-phrase is moved into a word. But (8) is not crashingly ungrammatical. Furthermore, (8) should be much worse than (9), since (9) would not involve Wh-movement into a word. But this is not the case, as the sentences of (8-9) have much the same, marginal, status. The absence of any significant contrast indicates that the three complementisers do not form a word.
The third argument is that if the three complementisers form a cluster inserted into C, then Wh-words should occur always to the left of the cluster. In (9a), however, the Wh-phrase occurs to the right of the complementiser. Placement to the left leads to heavy ungrammaticality:

(10) ** Ik weet niet hoeveel rijker wie dan jou voorgelogen heeft Jan nu eigenlijk is
I don't know how much richer C1 who you lied-to has Jan now in fact is

It is a cross-dialectal fact that (10) is much worse than (9a). Thus the idea that a cluster of complementisers is a word inserted into C leads to serious problems.
These problems are all avoided if each complementiser corresponds to a head. The null hypothesis is that each head projects to a maximal projection.4 Hence it is a possibility that the middle complementiser, associated with Wh-contexts, allows its Spec to be filled by a Wh-phrase, independent of whether C1 and C3 are lexically filled.
It has been shown that a complementiser sequence cannot be analysed as one word which is inserted into C. It follows that a complementiser sequence must involve three different functional heads. Below, we will show that in main clauses each head may have an overt specifier, and that the semantic nature of the specifier must be compatible with the semantic nature of the functional head.

3. V-movement into COMP

By the standard analysis of Verb-second, Verb-movement is movement into COMP (Den Besten 1989 and the references there). However, if there are three head positions available then there are three potential landing sites for V-movement.
Head-movement is generally assumed to be triggered by a feature (Koster 1986, Pollock 1989, Chomsky 1992). The presence or absence of this feature is subject to parametric variation. Let us call this feature "bind", since a positive value entails that the verb must land in that position. "Bind" entails lexicalisation. Thus it is a morphologically visible feature. Being visible, the value of "bind" is learnable, which is a welcome conclusion for a feature of which the value is parametrically fixed.
The following facts show that the C3 position, associated with declarative contexts, triggers V-movement in Standard Dutch and in most dialects:

(11) a. Piet zag ik gisteren op straat
Piet saw I yesterday on street
b. * Piet ik zag gisteren op straat
Piet I saw yesterday on street
===> C3 triggers Verb-Second and topics

The C3-position triggers V-movement. The Specifier licensed by that position typically hosts topics. Thus I will refer to the C3 projection (C3P) as a topic projection (ToP), and to C3 as To.
Consider next the C2 position, which is characteristically associated with Wh-contexts:

(12) a. Wie zag hij gisteren op straat
who saw he yesterday on street
b. * Wie op straat zag hij gisteren
yesterday on street saw he yesterday
===> C2 triggers Verb-Second and Wh-words

C2 also triggers V-movement. Wh-phrases occur in the Spec position which is licensed by C2. I will refer to C2 as Wh, and to C2P as WhP.
Let us turn next to the C1 position:

(13) a. * [Hoe sneller Jan loopt] is hij des te eerder thuis
how faster Jan walks is he the sooner at-home
b. [Hoe sneller Jan loopt], des te eerder is hij thuis
how faster Jan walks (C1) the sooner is he at-home
(14) a. * [Hoe snel Jan ook loopt] wint Piet toch
however fast Jan PTC walks wins Piet nevertheless
b. [Hoe snel Jan ook loopt], Piet wint toch
however fast Jan PTC walks C1 Piet wins nevertheless
===> There is a C which does not trigger V/2: C1

The semantic context is one of comparison, as in the (a)-sentence, or of explicit denial of comparison, as in the (b)-sentence. The C-position does not trigger V-movement, contrary to To and Wh, although its Spec position may be filled by a clause. This is allowed by our system of principles and parameters, yet it is problematic for any attempt that conflates To, Wh and C1 into one C-position.

4. On the parametrisation of the binding properties of functional projections

In this section, I will outline a view of the parametrisation of sentence structure, and discuss some questions which inevitably come up, even if they are not the subject of this paper. Following Zwart (1992), I assume that Ag must be bound by a higher head in Dutch. The binding relation is defined below:

(15) A binds B iff A governs B and A is coindexed with B

This binding relation can be realised either by Head-movement to A or by base-generating an expletive in A which binds B. Interestingly, as we will see in section 8, V-movement leaves variables whereas coindexing with expletives does not. Coindexing with expletives is an S-structure licensing process. Correspondingly, V-movement will be shown to obey the Coordinate Structure Constraint (Ross 1967) whereas coindexing with an expletive does not.
The dependent nature of Agr can now be expressed as follows, rephrasing the analysis of Hoekstra & Marácz (1989), Zwart (1992):

(16) Agr must be bound

Where complementisers bind AGR, they prevent V-movement to AGR, as below:

(17) Ik denk dati Jan [AGR]i morgen op bezoek komt
I think that Jan [e] tomorrow on visit comes

(16) accounts for the fact that many dialects of Dutch exhibit complementiser agreement (Van Haeringen 1939, Goeman 1979, Zwart 1992). This is just a reflex of the binding relation between the complementiser and Agr.
As we saw, the heads Wh and To trigger V-movement in the presence of a Spec. The involvement of Spec suggests that we should not talk about heads but about pairs (Spec, Head). This is natural pair to refer to since they are united by Spec-Head agreement. Parameters are now filled in as follows:

(18) a. Wh-agreement must bind
b. To-agreement must bind

From (18), it follows that V-movement must apply if the WhP or the ToP is "switched on":

(19) a. * Waarom hij jou niet bellen wou
why he you not phone wanted
b. * Wie hij niet kent
whom he not knows
(20) a. * Daarom hij jou niet bellen wou
for that reason he you not phone wanted
b. * Die hij niet kent
him he not knows

The heads Wh and To trigger V-movement. Both must be binders, so that Wh and To cannot cooccur. However, the absence of topics in embedded clauses must still be accounted for. We propose to require, in addition, that the pair (Spec, To) must be ungoverned. Again, this is a natural pair since it is united by Spec-Head agreement. Hence we require:

(21) Topic-agreement must be ungoverned

This requirement can only be met in root contexts. In embedded contexts, the head To is present but the topic agreement will be lacking. Hence it is not possible to license topics in embedded sentences, as illustrated below (cf. note 1):

(22) a. * Ik weet waarom [die jongen dat [ ik niet t mag]]
I know why that boy To I not like
b. Ik weet waarom dat ik die jongen niet t mag
I know why To I that boy not like
(23) a. * Het feit [Jan dat [t lachte]] verbaasde me
the fact Jan that laughed surprised me
b. Het feit dat Jan lachte verbaasde me
the fact that Jan laughed surprised me

The reason for (21) is not clear. It may, however, be related to the following conjunction of facts. (i) topics are not formally marked in Dutch (ii) topics are sensitive to the preceding sentential context. In the absence of a formal marking of topics in Dutch, it is not unreasonable to assume that the functional projection in which topics are licensed must occur in a position from where the preceding sentential context can be accessed.
Binding between two heads is effected through V-movement, or through the insertion of an expletive, a subordinator like "of".5 Below we will show that dialectal variation can be described as parametric variation involving the binding properties of functional heads.

5. Variation involving ToP

Topics fail to trigger inversion in several West Flemish and French Flemish dialects (Vanacker 1968, Debrabandere 1976):6

(24) Brugge, West-Flemish (WF), (W2:357)
Lanks te weg, ol de slunsen en sluffers dat i tegen kwaamp worren de zine
along the road all the villains and tramps that he PTC met became the his
(25) Oostende, WF (W2:364)
Zoender entwat te zeggen an zen boer, Wansje loat zen zwiins en geel de boel
without something to say to his farmer, W. leave his pigs and all the mess
(26) St. Sylvestre-Kappel, French-Flemish (RND 6: N105)
Bleke viezen me zieni fele gimeer uliern
pale calves we see not many here
(27) Kaasteren, French-Flemish (W2:393)
Over vele jaren der wos a boer
ago many years there was a farmer

This variation involves the parametric setting of ToP with respect to V-movement. In these dialects, the head To is not required to be a binder anymore. Hence V-to-To is not found.
The principle of economy (Chomsky 1992) predicts that V-movement to To is ruled out in these dialects because it is not necessary. In general, economy rules out optional movements. However, the following sentences show that inversion is not disallowed in these dialects; it is optional:

(28) Brugge, West-Flemish (WF), (W2:357)
Nor lange genoeg geschossebrost t'en, kwaamp i in e lankt wor ...
after long enough partied to have came he in a country where
(29) Oostende, WF (W2:364)
Oord en keer ier, zeit en i zen eigen ...
once there, says he to himself

The general problem is that a lot of processes are optional in language. Every case of optionality is a counterexample to economy, unless there is an independent explanation.
A plausible independent explanation for optionality could be local bilingualism. The following thought experiment makes this clear. We know that an adult can be a native speaker of two languages. Thus he will have two grammars. Suppose the grammars are identical except for the binding properties of ToP. It will appear as if inversion is optional. But, in fact, it is merely the choice of grammar which is optional. Depending on the grammar chosen, inversion will be obligatory or ruled out. The advantage of this view is that we can maintain that there are no optional syntactic rules. The independent explanation for apparent counterexamples to economy relies on the well-established fact that human beings can be native speakers of several languages. At the same time, bilingualism limited to one rule of grammar makes it possible to account for the process of language change. Such an explanation would be in the spirit of generative grammar which assumes that the language faculty is rich enough to allow for the creation of more than one subsystem of UG.7
If a natural language phenomenon (e.g. the optionality of inversion) corresponds to two simultaneously existing parametric values, we can also account for a lot of fuzziness. It is well-known that syntactic change is gradual and not abrupt. This follows directly if we see the change as involving two co-occurring parameter values, where a linguistic community gradually makes more use of one of the two values. Another example is dialect boundaries. As noted by Marinel Gerritsen, dialect boundaries are fuzzy, and not sharply delineated, as one would expect if there is a one-to-one relation between language and grammar. However, once we see language as the product of possibly coexisting parametric values, gradual differences with respect to, say, inversion signifies a gradual difference in the frequency of use which is made of a parameter value. Choice of parameter value, where two coexist, will typically be determined by extralinguistic factors.
To conclude, the lack of inversion after topics in West Flemish and French Flemish dialects is directly accommodated as parametric variation in the binding properties of the functional projection ToP. The optionality of inversion is plausibly analysed as an example of local bilingualism, that is bilingualism with respect to the binding properties of ToP.

6. Variation involving WhP

The complementiser "dat" fails to prevent V/2 in embedded contexts (Overdiep 1932,1940). Two examples are given below:

(30) a. Akersloot (North Holland), De Rooy (1965:127):
Daar raakte ik zo van onder de indruk dat ik had alles vergeten thuis
that impressed me so much that I had everything forgotten at home
b. Frisian Dutch
Het was zo koud dat ik ben maar naar binnen gegaan
it was so cold that I am but inside gone
(31) Groningen, De Haan (1988):
a. Hai zee dat hai har zien buurman zitten zain
he said that he had his neighbour sit see
b. Frisian, De Haan (1987):
Douwe sei dat hy hie my sjoen
Douwe said that he had me seen

This phenomenon typically occurs in two semantic contexts. It is found in consecutive clauses, as in (30), and it is found in clauses introduced by bridge verbs, as in (31). Geographically, the phenomenon is especially found in the North, in Frisian and in the dialect of Groningen.
Consider first the category of the complementiser "dat" in examples like the following from Standard Dutch and most dialects:

(32) a. Ik denk dat Jan hem zag
I think that Jan him saw
b. Wiei denk je [ti dat ik zag]
who think you that I saw

"That" could be assigned to either C, Wh or To. The element of comparison seems to be absent in this type of clause. Hence we may assume that the embedded clause is a WhP, with WH-movement proceeding through the Spec of Wh. If embedded clauses are WhP's then "dat" must be generated in Wh. In the normal case, "dat" will bind To and Ag, blocking V-movement in embedded clauses.
I will suppose that in the relevant dialects bridge verbs and consecutive clauses exceptionally allow a Wh-head to lose its government properties, hence its binding properties. Thus Wh will no longer govern ToP. As a result, topic agreement is allowed. We now expect that the complementiser "dat" may be followed by a non-subject topic, followed by the verb. An example from Frisian is given below:

(33) Hy sei dat moarn soe hy nei Ljouwert ta
he said that tomorrow would he to L.
(34) It wie sa kâld dat syn handen koe hy net mear waskje
it was so cold that his hands could he not anymore wash

This facts support the idea that "dat" occurs in Wh, and that it has lost its binding properties.
This account is supported by the behaviour of weak pronouns. Weak pronouns may not cliticise to "dat", if embedded V-movement has taken place (De Haan & Weerman 1976, Van der Meer 1988):

(35) Groningen (De Haan 1988)
a. Hai zee dati e [e]i zien buurman zitten zain har
he said that he his neighbour sit see had
b. * Hai zee dat e [har] zien buurman zitten zain [e]i
he said that he had his neighbour sit see
(36) Frisian (De Haan 1988)
a. Hy sei dati er [e]i syn buorman sitten sjoen hie
he said that he his neighbour sit see had
b. * Hy sei dat er hie syn buorman sitten sjoen
he said that he would it indeed do

This makes sense because "dat" does not bind Agr in these examples. The precise statement about cliticisation can now be formulated as follows:

(37) Cliticise weak pronouns onto the first lexical host
to the left of Ag

In the (a)-sentences, the first lexical host to the left of Agr is the complementiser. In the (b)-sentences, the first lexical host is the verb in To, and not the complementiser. If (37) is correct, we have an argument supporting the claim that the verb in the (b)-sentences occupies the head position of ToP. Thus the (b)-sentences violate (37).
Independent evidence for (37) comes from the fact that weak pronouns may also cliticise onto lexical material in the Spec of WhP, provided it is the first lexical element to the left of Agr, as in the examples below:

(38) Frisian
a. Ik wit net hoefolle boeken 'r lêzen het
I know not how many books he read has
b. Ik wit net hoefolle minskjen 'r telt het
I know not how many people he counted has
(39) Standard Dutch
a. Ik weet niet welke man-ie gezien heeft
I know not which man he seen has
b. Het gaat sneller dan-ie kan denken
It goes faster than he can think

These examples support (37), which in turn supports our analysis of embedded V-movement as movement to an ungoverned To-position.
We now expect that weak pronouns may occur in sentences in which embedded verb-movement takes place, provided that (37) is not violated. The first lexical item to the left of Ag is the verb in To. Hence we correctly predict that weak pronouns may cliticise onto the tensed verb, as shown below:

(40) Frisian
Hy sei dat syn buorman hie'r der net sitten sjoen
he said that his neighbour had he there not sit see
(41) Frisian Dutch
Zijn ogen zijn zo slecht dat zijn buurman had-ie niet zien zitten
His eyes are so bad Wh his neighbour had-he not see sit

As predicted, the weak pronoun cliticises onto the verb, providing evidence that the verb has been moved to To.
Dialectal variation involving embedded V/2 is thus described in term of the governing properties of WhP. Failure of government of To leads to V-movement to To, and topicalisation. Correspondingly, subject clitics cliticise onto the verb, and no longer onto the complementiser.

7. Variation involving CP

In many dialects, the C complementiser "al" or "als" may trigger V-movement. If it does, the verb is found adjacent to the complementiser. Thus, this type of complementiser triggers Verb-First rather than Verb-Second. To illustrate, consider the following sentences:

(42) North Holland (Langedijk 1963)
Al ben ik klaar, wat moet ik dan doen
when am I ready what must I then do?
(43) a. Standard Dutch (Brachin 1977):
al ren je nog zo hard, je komt te laat
C run you however fast, you come too late
"however fast you'll run, you'll be too late"
b. Hij rende als zat de duivel hem op de hielen
he ran as sat the devil him at his heels

(42) would be ungrammatical in Standard Dutch: "als" meaning "when" blocks V-movement, though it does not in the meaning "as if".
Expletives block V-movement by filling a head position and by binding lower head positions. I will assume the following statement is correct:

(44) Expletives are inserted into governed positions

Thus the expletive "of" is inserted into Wh, a position which is governed (but not bound) by C.
Cross-dialectal variation can now be analysed as parametric variation in the government properties of C. Clearly, these government properties are determined by the lexical item that is inserted into C. In (42-43), the C-elements do not govern Wh. Hence V-movement to Wh derives a Verb-First word order.
Some dialects exhibit the same phenomenon with the complementiser "of":

(45) Flemish, Frisian, Frisian Dutch (Overdiep 1949,
De Rooy 1965, Brachin 1977):
Het was of ging er een vrachtwagen voorbij
it was as if went there a truck past

However, this example involves the same semantic context als the earlier example with "als". Thus, "of" must be analysed as a C-element here, and the account goes through as before.
To sum, variation involves the governing properties of C with respect to Wh. V-First is analysed as V-movement into an ungoverned Wh-position.

It is a curious problem that some complementisers block V-movement in a first conjunct but not in a second conjunct, as shown below:

(46) Alsi [je [e]i te laat thuiskomt] en [je [hebt]k geen sleutel bij je [e]k] dan ...
if you too late home arrives and you have no key on you then

The problem cannot be denied by claiming that the second conjunct is outside the scope of the complementiser "als". This is clear from the interpretation of the sentence, which is such that the second conjunct is also interpreted as a conditional or temporal clause. In addition, the complementiser "als" can trigger negative polarity inside its complement clause. If the second conjunct were outside the c-command domain of "als", it would not be able to trigger negative polarity. The following sentences show that "als" can trigger negative polarity inside the second conjunct:

(47) Als je ook maar iemand hebt gezien of er is
if you anybody have seen or there is
ook maar iets ongewoons gebeurd, dan ...
anything unusual happened then
(48) Als er ook maar iets mislukt en ik zie
if there anything goes-wrong and I see
ook maar iemand in handen van de politie vallen, dan ...
anybody in hands of the police fall then

This makes it clear that both conjoined clauses occur in the c-command domain of the complementiser "als".
Not only the factual problem is to be solved. We have implicitly assumed that a binding relationship between two heads can be expressed in two ways: (i) V-movement to the higher head (ii) base-generation of an expletive in the higher head position. The conceptual question is whether these two ways of expressing a binding relationship are identical. In addition, there is a conceptual question concerning the interplay between the expletive strategy and the movement strategy. Expletives can be explained as an instance of Procrastinate (Chomsky 1992), the principle that discourages S-structure movement. But this raises the question why S-structure V-movement should occur at all, given that expletives are available to obviate S-structure V-movement.
An empty category coindexed with a moved verb can unproblematically be analysed as a variable. It is a different matter to coindex an empty category with an expletive. In the first place, expletives are not meaningful elements in the way verbs are. It is rather unlikely that a meaningless expletive can determine the range of a variable. In the second place, expletives are S-structure elements, which are assumed to be absent in LF. Specifically, Law (1987) proposes that the verb moves to C at LF, wiping out expletives on its way. V-movement at S-structure, on the other hand, is preserved at LF, and verbs are not wiped out.
Apart from these conceptual considerations, there are empirical arguments indicating that an expletive entertains a different relation with an empty head with which it is coindexed than the verb does. We saw above that the relation between an expletive and an empty head does not obey the coordinate structure constraint. The WH
variable relation obeys the Coordinate Structure Constraint, as shown below:

(49) a. * Wiei denk je dat [ik ti zag] en [Karel lachte]
who think you that I saw and Karel laughed
b. * Ik weet wiei [Jan ti zag] en [Karel lachte]
I know who Jan saw and Karel laughed
(50) a. Wiei denk je dat [ik ti zag] en [Karel ti fotografeerde]
who think you that I saw and Karel photographed
b. Ik weet wiei [Jan ti zag] en [Karel ti fotografeerde]
I know who Jan saw and Karel photographed

The verb-variable relation also obeys the coordinate structure constraint:

(51) a. * Wiei heeftj [hij ti gezien tj] en [zij ti voorbijgerend is]
who has he seen and she ran-past is
b. Wiei heeft [hij ti gezien tj] en [zij ti gefotografeerd tj]
(52) a. * Omdat zij en [de grap] en [Kees ti] ziet lachen
because she both the joke and Kees sees laugh
b. Omdat zij en [de grap] en [Kees lachen] ziet
because she and the joke and Kees laugh sees

In (51a), the Wh-phrase binds a variable in either conjunct. Thus it cannot be the Wh-phrase which causes ungrammaticality. The verb, however, only binds a variable in the first conjunct, causing ungrammaticality. The (b)-sentence shows that this account is correct. Once the verb also binds a variable in the other conjunct the sentence is grammatical. (52) shows that V-Raising also obeys the Coordinate Structure Constraint. In (52a), the raised verb binds a trace in one of the two conjuncts. In (52b), the verb is not raised but left in its base position (with two verbs, V-raising is optional), and, correspondingly, the Coordinate Structure Constraint is not violated.
The generalisation is that variable-binding obeys the Coordinate Structure Constraint. We saw that expletive-binding does not. This implies that expletives do not bind variables, a conclusion which we already reached on the basis of conceptual considerations.
If an empty category is bound without being a variable then it must be an anaphor. Indeed, anaphors, in general, do not conform to the Coordinate Structure Constraint:

(53) a. Jani heeft [zichzelfi getracteerd] en [Marie gefeliciteerd]
Jan has himself treated and Mary congratulated
b. Jani heeft [Marie gefeliciteerd en [zichzelfi getracteerd]
Jan has Mary congratulated and himself treated

Thus, it is not implausible to suppose that expletives bind anaphors.
There is also a difference between the anaphors in (53) and the anaphors bound by expletives. Anaphors must be governed by the expletives which bind them, which suggests that a generalisation of the Head Movement Constraint is at work.
The interaction between government and conjunction has been studied for Irish by McCloskey (1986). McCloskey proposes a definition of government which allows for the first member of a series of conjuncts to be governed. This accounts for the peculiar fact that pro-drop, in the presence of an emphatic particle (EMPH), may be found in the first conjunct but not in other conjuncts:

(54) a. Bhíos [ [e] féin agus Eoghan] i láthair
be-1S I EMPH and Owen present
b. * Bhíos [Eoghan agus [e] fein] i lathair
be-1S Owen and EMPH present

Even if one would like to analyse "féin" itself as a pronoun,8 then there is still the fact that the first conjunct determines the agreement on the verb. These facts indicate that the agreement relation may violate the Coordinate Structure Constraint.
The Dutch data are parallel to the Irish data, in that the expletive must bind an empty category in the first conjunct, just as agreement does in Irish. The similarity can be expressed if we assume that the first conjunct and its head are governed by an external governor. If this is correct, then we predict that weak pronouns which agree with the head of the first conjunct may cliticise onto the governor outside the conjunct, assuming that clitic movement does not leave variables as has been shown in Chomsky (1988). This prediction is borne out:

(55) Alsi-iej [[ej ei te laat thuiskomt] en [de deur is
if he too late home comes and the door is
al op slot gedaan]] dan moet-ie buiten slapen
already on the lock put tehn must he outside sleep

Notice that it is not crucial to analyse "ie" as a weak pronoun. It could also be analysed as agreement (cf. Van der Meer 1991 on Frisian). In the latter case, we have pro-drop inside the first conjunct.
There is cross-dialectal variation with respect to government of the first conjunct in case two NPs are conjoined. There are several speakers of Frisian, who accept an emphatic clitic pronoun only in the first conjunct (Van der Meer 1991):

(56) a. Wannearsto en Willem ...
when you and Willem
b. * Wannear do en Willem
when you and Willem

Interestingly, all speakers reject the (b)-sentence, in which the normal strong pronoun occurs in the first conjunct. Furthermore, the emphatic pronoun is barred from the second conjunct, whereas the normal strong pronoun is not:

(57) a. * Wannear Willem en stou ...
when Willem and you-EMPH
b. Wannear Willem en do ...
when Willem and you
c. Dou rinst
you walk
d. * Stou rinst

These facts are strongly remiscent of the Irish facts. There can apparently be a relation between an outside head and a first conjunct which cannot exist between an outside head and other conjuncts. Following McCloskey, we have supposed that this relation is government. Probably government in this sense must be construed as some PF licensing relation. This would fit in with the fact that both agreement and expletive heads are elements which are assumed not the exist at the level of LF. They derive their raison d'etre from the interplay between general principles of UG (e.g. Procrastinate) and PF licensing mechanisms.

9. Conclusion

The first conclusion of this paper is that there are three functional head positions above AgP. Each functional head is associated with its own semantic effect. Evidence for this claim was presented in the first half of this paper.
The second conclusion is that the binding relation between functional heads can be realised in two ways: either by V-movement or by coindexing with a base-generated expletive. The second option is a pure PF option, which cannot exist at LF. Parametric variation between dialects can be adequately described in terms of the choice for either option. Government plays an important role here. Lack of government entails lack of binding, which in turn triggers V-movement, much in the spirit of Safir & Pesetsky (1981).9 Where binding obtains, V-movement is ruled out by economy.
It turns out that the notion of government we are employing provides a principled view of parametric variation. The PF notion of government is presumably closely tied to morphological visibility, in contradistinction to the LF notion of government. A generalisation across PF government and LF government may be spurious, or the interpretation of locality might be different at either level. The advantage of a generalisation across PF and LF is that the relation between the two becomes clear, in the same way in which the principle of economy makes this relation more explicit.
Suppose now, following Chomsky (1992) that expletives are deleted at LF (or that they do not exist), because they are not meaningful elements. Thus deletion of expletives would follow from Full Interpretation. We now derive that the verb must move to the highest head position in LF, from where it turns all coindexed heads into variables.10 Suppose the verb would not move to the highest head position. Then the highest head position would be an expletive position, which in turn yields a violation of Full Interpretation. It is not surprising that cross-dialectal variation affects expletives much more than V-movement. Expletives exist only at PF, whereas V-movement (or Head-movement) belongs to LF. Thus generative grammar provides us with an analysis in which dialectal variation turns out to be less unsystematic than theory-free observation would lead us to believe.

References

Besten, H den (1989) Studies in West Germanic Syntax. Dissertation. U. of Amsterdam.
Bickerton, D. (1981) Roots of Language. Karoma, Ann Arbor.
Brachin, P. (1977) "Of + inversie (II)". De Nieuwe Taalgids 70: 422.
Chomsky, N. (1986) Barriers. MIT Press, Cambridge.
Chomsky, N. (1988) Language and Problems of Knowledge: the Managua Lectures. MIT Press, Cambridge (USA).
Chomsky, N. (1992) "A Minimalist Program for Linguistic Theory". MIT Occasional Papers in Linguistics 1. Massachusetts Institute of Technology, Cambridge (USA).
Debrabandere, F. (1976) "De SVf-woordorde in Zinnen met een Aanloop". In Handelingen van de Koninklijke Commissie voor Toponomie en Dialectologie, 87-97. Drukkerij G. Michiels, Tongeren.
Gerritsen, M. (1991) Atlas van de Nederlandse Dialectsyntaxis (AND). P.J. Meertens Instituut, Amsterdam.
Goeman, T. (1979) "COMP-Agreement?". In W. Zonneveld & F. Weerman (1979) Linguistics in the Netherlands, 291-306. Foris, Dordrecht.
Haan, G. de & F. Weerman (1986) "Finiteness and Verb-Fronting in Frisian". In H. Haider & M. Prinzhorn (eds) Verb-Second Phenomena in the Germanic Languages. Foris, Dordrecht.
Haan, G. de (1987) "De Syntacticus als Frisist". In S. Dyk & J. Hoekstra (1987) Ta de Fryske Syntaksis, 57-80. Fryske Academy, Ljouwert.
Haan, G. de (1988) "Hoofd- en Bijzinnen: Traditie en Progressie". Handelingen van het 40ste Nederlands Filologen Congres, 203-213.
Haeringen, C. van (1939) "Congruerende Voegwoorden". Tijdschrift voor Nederlandsche Taal- en Letterkunde 58, 161-176.
Hoekstra, E. (1991) Licensing Conditions on Phrase Structure. Dissertation, University of Groningen.
Hoekstra, J. (1987) "'Verb-Second' en de 'Imperativus Pro Infinitivo' in het Fries". Taalkundig Bulletin (TABU) 17: 96-121.
Hoekstra, J. & L. Marácz (1989) "Some Implications of I-to-C Movement in Frisian". In H. Bennis & A. van Kemenade (eds) Linguistics in the Netherlands 1989, 81-90. Foris, Dordrecht.
Kayne, R. (1992) "Word Order". GLOW Colloquium, Lisboa.
Koster, J. (1986) "The Relation between Pro-drop, Scambling and Verb Movements". Ms, University of Groningen.
Langedijk, H. (1963) Hé, is dat Westfries? Drukkerij 'West-Friesland', Hoorn.
Law, P. (1987) Effects of Head-Movement on Theories of Subjacency and Proper Government. Dissertation, MIT.
McCloskey, J. (1986) "Inflection and Conjunction in Modern Irish". Natural Language & Linguistic Theory 4: 245-281.
Meer, G. van der (1988) "Reported Speech and the Position of the Finite Verb". Leuvense Bijdragen 77: 301-324.
Meer, G. van der (1991) "The 'Conjugation' of Subclause Introducers: Frisian -ST". North-Western European Language Evolution (NOWELE): 63-84.
Overdiep, G.S. (1932) "Syntaxis en Dialectstudie I en II". Onze Taaltuin 1: 23-27, 46-52.
Overdiep, G.S. (1940) De Volkstaal van Katwijk-aan-Zee. Standaard Boekhandel, Antwerpen.
Overdiep, G.S. (1949) Stylistische Grammatica. Zwolle.
Pesetsky, D. (1982) Paths and Categories. Dissertation, MIT.
Platzack, C. & A. Holmberg (1989) "The Role of AGR and Finiteness in Germanic VO-languages". GLOW Newsletter 22, 53-54.
Pollock, J.-Y. (1989) "Verb-Movement, Universal Grammar and the Structure of IP". Linguistic Inquiry 20: 365-424.
Riemsdijk, H. van (1989) "Movement and Regeneration". In P. Benincá (ed.) Dialect Variation and the Theory of Grammar. Foris, Dordrecht.
RND (1930-1982) E. Blancquaert en W. Pée (eds) Reeks Nederlands Dialect-atlassen. Antwerpen.
Rooy, J. de (1965) Als-Of-Dat. Van Gorcum, Assen.
Rooy, J. de (1978) "Het ene 'of' is het andere niet". De Nieuwe Taalgids 71, 146-148.
Ross, J.R. (1967) Constraints on Variables in Syntax. Dissertation, MIT.
Safir, K. & D. Pesetsky (1981) "Inflection, Inversion and Subject Clitics". Proceedings of North-East Linguistic Society (NELS) 11: 331-344.
Vanacker, V.F. (1968) "Syntaktische Daten aus Französisch-Flämischen Tonband Aufnahmen". Verhandlungen des 2e Internationalen Dialektologencongresses, Beiheft 4 des Zeitschrift für Mundartforschung, 844-855. Franz Steiner Verlag, Wiesbaden..
Weijnen, A. (1966) Nederlandse Dialectkunde. Van Gorcum, Assen.
Williams, E. (1981) "On the notions 'lexically related' and 'head of a word'". Linguistic Inquiry 12: 245-274.
Winkler, J. (1874) Algemeen Nederduitsch en Friesch Dialecticon. Martinus Nijhoff, 's Gravenhage (reprinted (1972) Dr. Martin Sändig oHG, Wiesbaden).
Zwart, J.-W. (1992) "Verb-Movement and Complementizer Agreement". GLOW Newsletter26, 58-59. Ms, University of Groningen.

1 The lexical shape of the complementiser (abbreviated "C" in the gloss) is subject to dialectal variation. Most common are "dat", "of", and "a" (Weijnen 1966:325).

2 It has been claimed that all functional heads have the specifier to the left and the complement to the right (Kayne 1992). Cross-linguistic evidence for this claim comes from the observation that (i) subjects generally precede objects (ii) V-raising and VP-raising phenomena are typically found in head-final languages, but not in head-initial languages. In general, directionality effects suggest that Kayne's hypothesis is worth pursuing. Remember also that morphology is head-final (as laid down in Williams' (1981) Right Hand Head Rule), suggesting the idea that lexical categories are generally head-final. The English VP, for example, shows several head-final characteristics (Hoekstra 1991, and others).

3 At the conference "Dialektsyntax in den germanischen Sprachen", it became clear that the phenomenon of triple complementisers is also found in Bavarian.

4 In fact, if specifiers must occur internal to the projection of the licensing head (which is normally assumed to be the case), then it follows that at least C2, the Wh-complementiser, must project to a maximal projection. Below we will see that C1 and C3 may also have a specifier, which entails that they also project to a maximal projection.

5 The lexical content of the subordinator is often optional since it can be recovered from the filled Spec. Where an empty operator occupies Spec, the lexical content is obligatory, in e.g. Standard Dutch:

(i) Ik weet [wiei (of) [je gezien hebt]]
I know whom Wh you seen have
(ii) Ik weet [Oi *(of) [je Jan gezien hebt]]
I know whether you Jan seen have

6 "W" stands for "Winkler (1874)", a dialecticon in which dialect speakers retell the biblical story of the lost son. "RND" stands for "Reeks Nederlandse Dialectatlassen", a systematic investigation in which dialect speakers translate sentences from Dutch or French into their dialect, cf. Gerritsen (1991).

7 The richness of the language faculty is clear from the fact that creole languages like Sranan were developed in one generation and that a creole language may have syntactic rules which are unique to it, underived from the ancestors from which it derived its lexicon (Bickerton 1981).

8 The pro-drop analysis itself is well-supported. McCloskey presents strong arguments against the pronominal status of "féin". For example, the emphatic particle may not be accompanied by an overt pronoun in the first conjunct, indicating this is a pro-drop environment. In the second conjunct, the emphatic particle must be accompanied by an overt pronoun, indicating that this is not a pro-drop environment:

(i) a. * Bhíos mé féin agus Eoghan ...
were I EMPH and Owen
b. Bhíos Eoghan agus mé féin ...
were Owen and I EMPH

Of course, pro-drop in the first conjunct is expected once it is possible that agreement targets the first conjunct.

9 Thus it is not a feature like "finiteness" which triggers V-movement, cf. Platzack & Holmberg (1989). The same conclusion is drawn on independent grounds by J. Hoekstra (1987), who shows that infinitives (which are clearly non-finite) may undergo Verb-First in Frisian.

10 The challenge for this view concerns the conjunction facts. If an expletive entertains a relation with one conjunct, then so will the verb in LF, violating the Coordinate Structure Constraint. This reasoning, however, is too simple. It is clear from Antecedent Contained Deletion facts and sloppy interpretations that structure building is allowed in LF. For example, consider "John ate an apple, but Bill does not", where following the negation a VP consisting of a verb and an object must be built in LF. In the same vein, the problematic coordination can be saved at LF by letting the two conjuncts develop into full-fledged CP's (cf. Van Riemsdijk 1989). An example is given below:

(i) PF Als-ie [AgP te laat thuis komt] en [AgP zijn sleutel is weg]
if-he too late home comes and his key is away
LF [CP V ToP] en [CP V ToP]]

At LF, both conjuncts are grown into full-fledged CP's, and V-movement to C proceeds within each conjunct. On this view, S-structure is a structurally impoverished version of LF, with PF and LF government determining to what extent impoverishment is allowed.