Expletive Replacement, Verb-Second and Coordination1

ERIC HOEKSTRA (1993)

K.T. Taraldsen & O. Lorenz (eds) Comparative Germanic Syntax. A Selection of Papers from the 8th Workshop on Comparative Germanic Syntax. U. of Tromsø, Norway, 20-22/11/1992. The Linguistic Review 11, 285-297.

The purpose of this paper is to argue that superficial IP IP coordinations must be analysed as CP CP coordinations. Thus our paper provides support for the claim that “propositional reconstruction of coordinated structures is analytically valid” (Cremers 1993:61).To exemplify, a sentence like (1) will below will be analysed as involving an empty anaphoric element in the complementiser position of the second conjunct, as in (2):

(1) Loopt Jan naar huis en Piet naar school?
walks Jan to home and Piet to school
“Does Jan walk home and Piet to school”
(2) [CP Loopt [IP Jan naar huis]] en [CP [e] [IP Piet naar school]]
walks Jan to home and Piet to school

Section 2 below will present arguments against an analysis in which (1) is analysed as a conjunction of IPs, as in (3):

(3) [CP Loopt [IP Jan naar huis] en [IP Piet naar school]]
walks Jan to home and Piet to school

A challenge for our approach comes from asymmetric coordination, which may exhibit verb-second in the second conjunct but not in the first (first discussed in De Vries 1911-1912: 179ff, see also Höhle 1989). It turns out that our analysis elegantly accouns for asymmetric coordination, as shown in section 3, if a minimalist analysis of verb-second is adopted (cf. Chomsky 1992, Zwart 1992, this volume, and others). The precise analysis of forward gapping and verb-second is fleshed out in section 4. Remaining problems are discussed in section 5.

1. Superficial IP IP coordination is really CP CP coordination

In the Minimalist Program (Chomsky 1992), complementisers are typically syntactic elements, which are deleted, as expletives, by head movement in the mapping to LF.2 Law (1991) proposes that V moves in LF to C; Zwart (1991) proposes that an empty INFL moves to C in the syntax. Either of the minimalist proposals mentioned above faces the problem of accounting for coordination below the CP level, such as IP IP coordination or VP VP coordination. In that case, there are two verbs, one in each conjunct, and only one C-position, outside the coordinated structure, as in the example below:

(4) a. [[IP John laughed] and/but [IPPeter cried]]
b. Ik denk dat [IP Jan lachte] en [IP Piet huilde]
I think that Jan laughed and Piet cried
“I think that Jan laughed and Piet cried”

Movement of V to C, under an IP IP analysis would yield LF-structures like the following:

(5) a. [CP laughedi criedj [IP [IP John ti] and
[IP Peter tj]]]
b. ik denk [CP lachtei huildej [IP [IP Jan ti] and
[IP Peter tj]]]

Notice that the movement to C violates the Coordinate Structure Constraint (Ross 1967). In addition, there is a (less serious) landing-site problem since one C-position must hold two verbs.3
In fact, this problem shows up not only with LF-movement but also with clitic movement. Consider the following sentences:

(6) a. Omdattie zich vergist en zijn broer ziek is
because-he himself mistakes and his brother ill is
“Because he is mistaken and his brother is ill”
b. Omdat + iei [IP [IP ti zich vergist]] and
[IP zijn broer ziek is]]

Under an IP IP analysis as in (6b), cliticisation of the clitic ie onto the complementiser violates the Coordinate Structure Constraint.
The same applies to facts concerning complementiser agreement (indicated by subscripts) in substandard and dialectal Dutch. Consider the following sentences:

(7) a. omdatte Jan en Kees toffe jongens zijn en onse baas geen geld meer hep
because-PL Jan and Kees good guys are and our boss no money anymore has
“Because Jan and Kees are O.K. and our boss doesn’t have any more money”
b. omdattei [IP [IP [Jan en Kees]i toffe jongens zijn] en
[IP onse baas geen geld meer hep]]

It is hard to see how the complementiser can agree with the subject of the first conjunct of a coordinated structure, if the complementiser is itself outside this coordinated structure.
For each individual problem sketched above some ad-hoc solution can be thought up. However, these problems disappear if we adopt the idea that a conjunction of CPs is necessarily involved. Reconsider (4), repeated as (8), and its LF-structures (9) under a CP CP analysis:

(8) a. John laughed and Peter cried
b. Ik denk dat Jan lacht en Piet huilt
I think that Jan laughs and Piet cries
(9) a. [CP [CP laughedi [IP John ti]] and [CP criedj [IP Peter tj]]]
b. Ik denk [CP [CP lachtei [IP Jan ti]] en
[CP huildej [IP Piet tj]]]

Movement to C at LF proceeds within each conjunct inside the coordinated structure, without causing a violation of the Coordinate Structure Constraint. Furthermore, there is now for each verb a landing site. Clitisation and agreement similarly are unproblematic under a CP CP analysis,4 as shown below:

(10) a. Omdattie zich vergist en zijn broer ziek is
because-he himself mistakes and his brother ill is
b. [CP [CP Omdat + iei [IP ti zich vergist]] en
[CP [e] [IP zijn broer ziek is]]
(11) a. Omdatte Jan en Kees toffe jongens zijn en onse baas geen geld meer hep
because-PL Jan and Kees good goys are and our boss no money anymore has
b. [CP [CP Omdat -ei [IP [Jan en Kees]i toffe jongens zijn]] en [CP onse baas geen geld meer hep]]

Cliticisation and agreement now take place without violating the Coordinated Structure Constraint.
In addition to solving these empirical problems, the CP CP coordination analysis is more restrictive than its competitor. Whereas an IP IP coordination analysis must anyhow allow for the existence of CP CP coordination, the CP CP coordination analysis actually forbids IP IP coordination. This need not be stipulated: it follows from the Coordinate Structure Constraint in tandem with the assumption that verbs move to C in LF. We will present independent evidence for this analysis based on facts involving asymmetric coordination (section 1). Then we will consider the question what happens to the gapped (empty) complementiser position of the second conjunct.

2. Asymmetric Coordination

2.1. The problem

Some expletives block verb-second in the first conjunct but not in the second (De Vries 1911-1912, Höhle 1989). An example is given below, together with the bracketing it receives under an IP IP coordination analysis:

(12) a. Alsi [IP je [e]i te laat thuiskomt en [IP je [hebt]j geen geld bij je [t]j], dan ...
if you too late home come and you have no money with you then ...
“If you come home late and you don’t have any money with you,
then ...”
b. * Alsi [IP je [komt]i te laat thuis [t]i en [IP je [hebt]j geen geld bij je [t]j], dan ...
if you come too late home and you have no money with you then ...

These facts indicate that the complementiser blocks verb-second in the first conjunct but not in the second. Under the IP IP coordination analysis given in (12), it is a mystery why the complementiser als only blocks verb-second in the first conjunct. For example, Kathol (1990:252) stipulates that "the complementiser only identifies the first INFL", which is a rephrasal of the descriptive problem.
Several responses are possible now. One is to claim that the second conjunct is adjoined to the CP introduced by als. Thus being outside the scope (minimal c-command domain) of the complementiser, this complementiser would fail to block verb-second. The problem with this suggestion is that the second conjunct receives a conditional interpretation, like the first. Moreover, the complementiser als can trigger negative polarity inside its complement clause:

(13) a. Als je ook maar iemand ziet ...
if you anybody see
“If you see anybody ...”
b. Als er ook maar iemand lacht ...
if there anybody laughs
“If anybody laughs ...”

Now, it is known since Ladusaw (1980) that a negative polarity item must be in the scope of its trigger. If the second conjunct were truly outside the scope of als, then als should not to be able to trigger negative polarity inside the second conjunct. This prediction turns out to be incorrect:

(14) a. Als je ook maar iemand hebt gezien of er is ook maar iets ongewoons gebeurd ...
if you anybody have seen or there is anything unusual happened
“If you have seen anybody or anything unusual has happened ...”
b. Als onze onderneming mislukt en ik zie ook maar iemand in handen van de politie vallen ...
if our venture goes-wrong and I see anybody in hands of the police fall
“If our venture crashes and I see anybody fall in the hands of the police ...”

The fact that als can trigger negative polarity inside a second conjunct shows that the second conjunct must be inside the scope of als. Thus it does not seem possible to claim that the second conjunct is outside the scope of the complementiser, and that this is what makes verb-second possible.
Another possible response is to doubt the validity of the Coordinate Structure Constraint (Johannesen 1992). Although there are problems with this generalisation, discarding the Coordinate Structure Constraint does not provide any further insight into the empirical problem. Specifically, we still do not understand why only the second conjunct allows verb-second, and not the first.

2.2. Analysis

I would like to suggest that each conjunct must be analysed as a full-fledged CP, and that als occupies the C-position in the first conjunct of the coordinated structure. Thus (12a) should be analysed as in (15) below:

(15) [CP Alsi [IP je [e]i te laat thuiskomt]] en [CP _ [IP je [hebt]j geen geld bij je [t]j]
if you too late home come and you have no money with you

Notice that I assume that the Dutch IP is head-initial (following Zwart this volume). I will return to the analysis of verb-second in section 3.1. below. In (15) the head C of the second conjunct is empty as the result of forward gapping. This analysis is supported by an argument from clitisation.

2.3. Cliticisation

Note that the complementiser als occurs internal to the first conjunct. If this structure is correct then we expect that a pronominal subject may cliticise onto the complementiser. Conversely, if the complementiser occurs outside the coordination then we expect this to be impossible. The argument is identical to the one given in section 2. The following facts indicate that the pronoun may cliticise onto the complementiser in structures of asymmetric coordination:

(16) Standard Dutch
Als-ie te laat komt en hij heeft geen geld bij zich ...
if-he too late comes and he has no money with him
“If he is late and he doesn’t have any money with him ...”
(17) Standard Frisian
Ast te let komst en do hast gjin jild by dy ...
if-2SG too late comes and you have-2SG no money with you
“If you are late and you don’t have any money with you ...”

(16) is interesting because the pronoun ie only occurs if it is cliticised on the right side of the rightmost element in CP. If als were outside the coordinated structure (so if the structure were to be analysed as a coordination of IPs) then this cliticisation rule would violate the Coordinate Structure Constraint. No such violation occurs under the present analysis. (17) is interesting because Frisian allows inflected complementisers (Dyk & Hoekstra 1987:21-22). The relation between the inflected complementiser and the small pro in subject position again would violate the Coordinate Structure Constraint under a IP IP coordination analysis. Under a CP CP analysis, no violation arises.

2.4. Two questions

Reconsider the structure of asymmetric coordinations given above in (15). Two questions arise. (1) Why is Verb-Movement to the head of IP allowed in the second conjunct? (2) Is it possible to maintain the claim that the second complementiser is empty as the result of forward gapping (and what is the status of forward gapping)? We will discuss these interrelated questions in the next section.

3. Verb-Movement and Forward Gapping

3.1. Triggers for Verb-Movement

In main and embedded clauses, the strong features of C get eliminated by movement of the empty INFL to C (cf. Zwart 1992).5 Lexicalisation of either INFL or C is the PF-reflection of this process. Lexicalisation takes place either by base-generating a complementiser in CP or by moving the verb into INFL or C.
The locus of lexicalisation through V-movement is determined by ecnonomy and licensing requirements. In subject-initial main clauses, only the subject needs to be licensed (under Spec-Head agreement), so that V-to-INFL is sufficient and V-to-C ruled out by economy. In topic-initial sentences, the topic must be licensed, necessitating V-movement to C.6
The complementiser als (like V-movement) makes abstract head movement of INFL to C PF-visible. There is no need for embedded V-to-INFL, and hence this is ruled out by economy.
This account is supported by the fact that conditional clauses and counterfactuals exhibit both the option of base-generating a complementiser and the option of V-movement. Consider the following:

(18) a. Als je te laat komt dan moet je je melden
if you too late come then must you yourself report
“If you are late then you have to report yourself”
b. Kom je te laat dan moet je je melden
come you too late then must you yourself report
“If you are late then you have to report yourself”
(19) a. Als ik het geweten had dan had ik je opgebeld
if I it known had then had I you up-called
“If I had known then I would have called you up”
b. Had ik het geweten dan had ik je opgebeld
had I it known then had I you up-called
“If I had known then I would have called you up”

The (a)-sentences illustrate the option of base-generating the complementiser als, the (b)-sentences illustrate the option of embedded Verb-movement. Furthermore, this alternation also provides indirect evidence for the claim that complementisers (like als) are eliminated in LF, seeing that the (b)-sentences even exhibit this process in PF. This is a recurrent syntactic insight: what happens in LF may happen in PF. At the same time, the above alternation between V-movement and the complementiser als clearly supports our treatment of als as an expletive complementiser that is deleted in LF.

3.2. Forward Gapping

We will turn now to the interaction of verb-movement with forward gapping. Consider first forward gapping of a verb:

(20) a. Dan heeft Jan een auto, en [e] Piet een fiets
then has Jan a car and Piet a bike
“Then Jan has a car, and Piet a bike”
b. Jan heeft een auto, en Piet [e] een fiets
Jan has a car and Piet a bike
“Jan has a car, and Piet a bike”

The second C position in (20a), like the second INFL position in (20b), is empty through forward gapping. I will assume that such a position contains an anaphor, which is bound by its antecedent, following Koster 1978:104ff).7 In the above examples, heeft is the antecedent of the empty anaphor. Notice that, although INFL and C in the second clause are empty, no ungrammaticality ensues. The second INFL position, though empty, is perfectly well able to ensure nominative Case assignment to the subject.
Similarly, although Dutch is not a regular pro-drop language, subject positions may remain empty through forward gapping:

(21) a. Jan werkt te hard, en [e] lacht te weinig
Jan works too hard and laughs too little
“Jan works too hard, and doesn’t laugh enough”
b. * [e] Lacht te weinig
laughs too little

These facts make it clear that the empty anaphor has acces not only to the semantic content of its antecedent but also to its lexical content. If the anaphor did not have access to the lexical content of its antecedent then (21a) should be ungrammatical because Dutch is not a pro-drop language, as (21b) shows.
The anaphor makes use of the lexical content of its antecedent only if it has to, as in (21b) above, to obviate ungrammaticality. It is conceivable that there is no need for an anaphor to make use of its antecedent’s lexicality, if the anaphor’s lexicality can be ensured in a different way. We will explore this possibility in the next section in order to flesh out our analysis of asymmetric coordination.

3.3. Asymmetric Coordination Revisited

Reconsider a typical case of asymmetric coordination analysed as forward gapping, that is, (15), repeated below as (22):

(22) [CP Alsi [IP je [e]i te laat thuiskomt]] en [CP [e]i [IP je [INFL [e]i hebtj] geen geld bij je [t]j]
if you too late home come and you have no money with you

In (22), abstract INFL-to-C is represented by cosuperscripting.8 The empty complementiser of the second conjunct receives its semantic interpretation from the first complementiser, without sharing the latter’s lexical content. The reason for this is that V-movement to INFL in the second conjunct is sufficient for lexicalising C, because of abstract INFL-to-C. Seen in this light, asymmetric coordination is a case of forward gapping in which the anaphor does not make use of its antecedent’s lexical content.
Base-generation of a lexical complementiser in C will obviate the need for V-movement to IP, since in that case C already lexicalises INFL, after abstract INFL-to-C. This explains that the presence of the complementiser als is incompatible with V-to-INFL:

(23) a. * Als je komti te laat thuis ti
if you come too late home
b. Alsi je ei te laat thuis komt

(23a) violates economy: there is no trigger for V-to-INFL, INFL being lexicalised by the complementiser.
Of course, nothing prevents the anaphoric complementiser in a forward gapping construction from deriving its lexical content from its antecedent. In that case, however, the anaphoric complementiser will count as lexical, and block embedded V-to-INFL. Consider now the following sentence:

(24) a. Als je te laat thuis komt en je geen geld bij je hebt
if you too late home comes and you no money with you have
b. [CP Alsi [IP je [e]i te laat thuiskomt]] en [CP [e]i [IP je [e]i geen geld bij je hebt]
if you too late home come and you no money with you have

Comparison of (22) with (24) makes it clear that embedded V-to-INFL in the second conjunct is optional, which is derived from the optionality of the anaphor’s sharing the lexical content of its antecedent.
Finally, nothing in our analysis forbids V-movement to C in the second conjunct in asymmetric coordination (provided, of course, that the anaphoric complementiser takes the option of not sharing the lexical content of its antecedent). Thus we expect that the second conjunct may also be introduced by a non-subject topic, followed by the verb, as illustrated below:

(25) Als je vandaag te ziek bent en morgen heb je geen tijd
if you today too ill are and tomorrow have you no time
If today you are too ill, and tomorrow you don’t have time ...”

Here the verb lexicalises C, the verb licenses the topic, but C receives its semantic interpretation from als.

3.4. Remaining empirical problems

In fact, symmetric coordination as in (24) is the normal case. The question arises why asymmetric coordination is not generally found. Notice though that asymmetric coordination is more frequently found in spoken language than in written language, for example after dat “that”:

(26) a. Het irriteert ons dat je te laat thuiskomt en je hebt geen sleutel bij je
it irritates us that you too late home come and you have no key with you
“It irritates us that you are home late and you don’t have any key with you”
b. Het feit dat je te laat thuiskomt en je hebt geen sleutel bij je irriteert ons
the fact that you too late home come and you have no key with you irritates us
“The fact that you are home late and you don’t have any key with you irritates us”

The second conjunct must also have a rising intonation. Note further that the semantic environment is downward entailing.9 Reviewers point out that the phenomenon seems to be absent in Swedish (27) and Norwegian (28).

(27) a.?* Om du kommer för sent hem, och du har inte pengar på dej
if you come too late home and you have no money with you
b. Om du kommer för sent hem, och du inte har pengar på dej
(28) a. * Om du kommer for seint og du har ikke noen penger på deg
if you come too late and you have not any money with you
b. Om du kommer for seint og du ikke har noen penger på deg

The reason for the absence of asymmetric coordination in the Scandinavian languages might, speculating somewhat, be related to the absence of overt subject agreement on the verb (cf. Platzack & Homsberg 1989). We have assumed that V-movement to INFL, in tandem with abstract movement of an inflectional feature from INFL to COMP, suffices to lexicalise COMP. Suppose that the inflectional feature is Agr, subject agreement. Subject agreement is lexical in Dutch, empty in Scandinavian. This entails that V-movement to INFL in tandem with abstract movement of Agr to COMP suffices to lexicalise COMP in Dutch, but crucially not in Scandinavian.

4. Conclusion

The basic claim of this paper is that superficial IP IP coordination must be analysed as CP CP coordination, with forward gapping of the C position at the left periphery of the second conjunct. This claim allows us to maintain the idea that there is abstract head movement from INFL to C, without having to assume that this movement violates the Coordinate Structure Constraint. In addition, this claim leads to a transparant analysis of asymmetric coordination, in which forward gapping is analysed as an instance of binding. A succesful reduction of IP IP coordination to CP CP coordination provides support for reconstructing all coordination as propositional (=clausal) coordination.

References

Baker, M. (1988) Incorporation. A Theory of Grammatical Function Changing. Chicago University Press, Chicago.
Chomsky, N. (1992) "A Minimalist Program for Linguistic Theory". MIT Occasional Papers in Linguistics 1. Massachusetts Institute of Technology, Cambridge (USA).
Cremers, C. (1993) On Parsing Coordination Categorially. Dissertation, University of Leiden.
Evers, A. (1982) “Twee Functionele Principes voor de Regel ‘Verschuif het Werkwoord’”, GLOT 1, 11-30.
Hendriks, P. (1992) “Deletion in Coordinate Structures”. In M. Kas, E. Reuland & C. Vet (eds) Language and Cognition I, 99-110. University of Groningen, Groningen.
Hoekstra, E. (1993, in press) "Some Implications of Number Agreement on COMP". In F. Drijkoningen & R. van Hout (eds) Linguistics in the Netherlands 1993, Benjamins, Amsterdam.
Hoekstra, E. & J.-W. Zwart (1994) “De Structuur van de CP. Functionele Projecties voor Topics en Vraagwoorden in het Nederlands”. Spektator 23, 191-212.
Höhle, T. (1989) "Assumptions about Asymmetric Coordination in German". Ms. University of Tübingen.
Johannesen, J. (1992) "Coordinate-Alpha". Talk presented at Console 1, Utrecht. Ms. University of Oslo.
Kathol, A. (1990) "A Uniform Approach to V/2 in German". In J. Carter, R.-M. Déchaine, B. Philip & T. Sherer (eds) Proceedings of NELS 20. GLSA, Dept. of Linguistics, University of Amherst.
Kosmeyer, W. (1990) “Mood, Agreement, Tense and the Position of [+F]”. Proceedings of the Scandinavian Conference of Linguistics 12.
Koster, J. (1978) Locality Principles in Syntax. Foris, Dordrecht.
Ladusaw, W. (1980) Polarity Sensitivity as Inherent Scope Relations. IULC, Bloomington.
Law, P. (1991) Effects of Head-Movement on Theories of Subjacency and Proper Government. Dissertation, MIT.
Platzack, C. & A. Holmberg (1989) “The Role of Agr and Finiteness in Germanic VO-Languages”. Talk given at GLOW 1989, Utrecht. GLOW Newsletter 22, 53-54.
Ristad, E.-S. (1990) Computational Structure of Human Language. MIT, Dissertation.
Roberts, I. (1989) “Excorporation and Minimality”. Linguistic Inquiry 22, 209-218.
Ross, J.R. (1967) Constraints on Variables in Syntax. Dissertation. MIT.
Vries, G. de (1992) On Coordination and Ellipsis. Dissertation, University of Tilburg.
Zwart, J.-W. (1991) "Clitics in Dutch: Evidence for the Position of INFL". Groninger Arbeiten zur germanistischen Linguistik 33, 71-92.
Zwart, J.-W. (1992) "Verb-Movement and Complementizer Agreement". GLOW Newsletter 26, 58-59. Ms, University of Groningen.



Proof Corrections

I. Idiomatic English translations of example sentences
Idiomatic translation is only given once and not repeated.

II. Bibliographical references

III. Small corrections are in red on the proofs themselves. Comments between square brackets.

I. Idiomatic translation

(1) “Does Jan walk home and Piet to school”
(4) b. “I think that Jan laughed and Piet cried”
(6) a. “Because he is mistaken and his brother is ill”
(7) a. “Because Jan and Kees are O.K. and our boss doesn’t have
any more money”
(12) a. “If you come home late and you don’t have any money with you,
then ...”
(13) a. “If you see anybody ...”
b. “If anybody laughs ...”
(14) a. “If you have seen anybody or anything unusual has happened ...”
b. “If our venture crashes and I see anybody fall in the hands of
the police ...”
(16) “If he is late and he doesn’t have any money with him ...”
(17) “If you are late and you don’t have any money with you ...”
(18) a. “If you are late then you have to report yourself”
b. “If you are late then you have to report yourself”
(19) a. “If I had known then I would have called you up”
b. “If I had known then I would have called you up”
(20) a. “Then Jan has a car, and Piet a bike”
b. “Jan has a car, and Piet a bike”
(21) a. “Jan works too hard, and doesn’t laugh enough”
(25) “If today you are too ill, and tomorrow you don’t have time ...”
(26/5)a. “It irritates us that you are home late and you don’t have
any key with you”
b. “The fact that you are home late and you don’t have any key
with you irritates us”

II. Bibliography
Dyk, Siebren & Jarich Hoekstra (1987) [addition] “Oersjoch fan ‘e Stúdzje fan ‘e
Fryske Syntaksis” (“Overview of the Study of Frisian Syntax”).
In Siebren Dyk & Jarich Hoekstra (eds) Ta de Fryske Syntaksis.
(“Remarks on Frisian Syntax”). Frisian Academy, Ljouwert/Leeuwarden.
Evers [translation of title:] “Two functional principles for the rule ‘Move the Verb’”
Hoekstra & Zwart [translation and correction] “CP Structure. Functional Projections for topics and Wh-words in Dutch”. To appear in Spektator.
Kosmeyer [editor/publisher] Lars Hellan, University of Trondheim.
Vries, Wobbe de (1910-1911, 1911-1912) [addition, see also III.]Dysmelie.
Opmerkingen over Syntaxis.” (“Dysmelia. Remarks about Syntax”).
In Programma van het Stedelijk Gymnasium te Groningen.
(“Program of the Gymnasium of the City Groningen”)
(no editor). Stoomdrukkerij B. Jacobs, Groningen.
 
 
 

1 I would like to thank Hans den Besten, Marcel den Dikken, Petra Hendriks, Jarich Hoekstra, Janne Johannesen, Ken Safir, Jan-Wouter Zwart, two anonymous reviewers and the audience at CGS 8 for comments and discussion. This paper corresponds to the third part (on the handout) of the talk which I presented at CGS 8.

2 Expletives are base-generated functional elements, heads or specifiers (of functional projections), hence they are not moved out of the lexical domain (PP, VP, NP, AP). Thus the possibility is left open that expletives have semantic content.

3 Of course, one can abandon the assumption that there is abstract movement to C. However, this is not only conceptually unsatisfying since it leads to a less-uniform LF, but also empirically inadequate in view of the facts to be presented in the remainder of this paper. Nor is there, without abstract movement, a principled explanation of complementiser agreement (Zwart 1992, Hoekstra 1993).

4 Independent evidence for treating superficial IP-IP coordination and VP-VP coordination as CP-CP coordination is presented in Johannesen (1992).

5 In Zwart’s analysis, abstract INFL-to-C generally applies, except in subject-initial main clauses. In our analysis, it also applies in the latter clause-type. A discussion of this difference between Zwart’s account and our own is beyond the scope of this paper.

6 I am simplifying the analysis of verb-second considerably for the purposes of brevity and exposition. For a detailed analysis of topicalisation and verb-second, the reader is referred to Hoekstra & Zwart (1994).

7 Our treatment of the gapped element as a bound anaphor is supported by De Vries (1992:39-44) , who shows that the gapped element is not a deleted copy of its antecedent. Thus (ia) below is not equivalent with (ib) below but with (ic):

(ia) Some fishermen beckon and shout
(ib) Some fishermen beckon and some fishermen shout
(ic) Some fishermen beckon and they shout

De Vries’ argument against a deletion analysis of forward gapping is identical to the classical argument against a pronominalisation tranformation (the point of which was Everybody likes himself is not equivalent to Everybody likes everybody ). For further arguments against a copying/deletion analysis, see e.g. Ristad (1990), Hendriks (1992).

8 It might be objected that the verb adjoins to a trace (in INFL) in the representation in (22), violating constraints forbidding this (e.g. proposed by Baker 1988). This problem can be circumvented if the following more complex representation of INFL is adopted:

(i) [INFL [e]i [e] [hebt]j]

This representation entails that the abstract movement from INFL to C takes place from a subcategorised head position (cf. Roberts 1989). The subcategorised element in question, which moves to C, might be the Finiteness feature (cf. Evers 1982, cf. Platzack & Holmberg 1989), the Tense feature (cf. Kosmeyer 1990), the Aux feature (cf. Evers 1982) or the Mood feature (H. Kier p.c.). As it is unclear whether adjunction to a trace should be allowed or not, I will stick to the representation given in the text.

9 Both als and irritate are downward entailing, as shown below:

(i) a. If you dream, the machine will wake you
b. If you dream restlessly, the machine will wake you
(ii) a. It irritates me that you dream
b. It irritates me that you dream restlessly

The (a)-sentences entail the (b)-sentences but the (b)-sentences do not entail the (a)-sentences.