You are on page 1of 37

Expletives as Arguments: Germanic Existential Sentences Revisited*

Claudia Felser & Laura Rupp

Abstract We propose an analysis for expletive-associate constructions within the minimalist framework outlined by Chomsky (1991, and later). Specifically, we argue that the expletive is an argument expression that must raise to the (most) external subject position in order to satisfy the Extended Projection Principle and check nominative case, while number agreement is checked by the associate. We present evidence against the assumption that the associate is assigned inherent or structural partitive by the verb, and instead suggest that it receives default case. In conjunction with the Kratzer/Diesing theory of indefinites, our analysis is shown to account for some well-known semantic restrictions on existential sentences, as well as for the relatively weaker definiteness effects observed in transitive expletive constructions.

1 Expletive-associate constructions: the problems With the development of the Minimalist Program (Chomsky 1991, and later), sentences containing overt subject expletives (henceforth, Exp) such as English there, German es, Dutch er, or Icelandic a have become the focus of much theoretical debate (see Chomsky 1995, 1998; Groat 1995; Jang 1997; Lasnik 1992, 1995; among others). Most of these studies assume that the expletive is a pure syntactic dummy, devoid of meaning and with a defective set of grammatical features, whose presence is motivated by the Extended Projection Principle (EPP) (Chomsky 1982: 10, 1995: 199, 284). The present article attempts to challenge this view, and reconsiders the syntactic status of the

* For constructive criticism and helpful comments on earlier drafts of this article, we are indebted to Martin Atkinson, Harald Clahsen, Jenny Dalalakis, Mike Jones, Roger Hawkins, Joan Maling, Andrew Radford, and several anonymous reviewers.

Summer 2000 (to appear in Linguistische Berichte, 2001)

Claudia Felser & Laura Rupp

expletive by comparing expletive-associate constructions of different types and from several languages, taking into account the effect that the presence of the expletive has on sentence interpretation. Given that morphosyntactically, expletives are pronouns, they are commonly categorised as (intransitive) determiners (see e.g. Radford 1997). The following examples (from Jenkins 1975: 11) show that expletive there occurs in positions typically occupied by argumental DPs or pronouns, rather than by the homophonous locative adverb (from which it is diachronically derived):1 (1) a. b. c. d. Is there any hope? Theres no hope, is there? There is believed to have been a revolution. There seems to be something brewing. (INTERROGATIVE) (TAG QUESTION) (PASSIVE) (RAISING)

"True" expletives such as English there are generally assumed to lack a -role. The following data confirm that there is excluded from positions associated with standard thematic roles such as AGENT or THEME, as well as from the subject position of those intransitive verbs that otherwise admit a quasi-argument such as weather-it (compare [2a-d]). (2) a. b. c. d. *There ate a biscuit. *We saw there. *She persuaded there to fund the party. It/*There rained.

True expletives usually occur with verbs that fail to assign an external -role such as existential be, unaccusative or raising verbs, and they always require the presence of a thematic associate (the logical subject). Expletive-associate constructions of the type exemplified by the (a) examples in (3)-(7) below are also known as existential sentences (henceforth, ES). As noted by Milsark (1974), they generally invoke an interpretation in which existence is predicated of the entity denoted by the associate, which must normally be a non-specific indefinite. This restriction on the semantic type of the associate is known as the definiteness effect, or definiteness restriction (DR) on the associate.2

1 See Breivik (1981) for further arguments that expletive there is grammatically distinct from the corresponding locative adverb. 2 The DR has been extensively discussed in the literature (see e.g. Reuland & ter Meulen 1987), and has alternately been described as a syntactic, semantic, or pragmatically motivated restriction. Observe, however, that definite DPs are not universally excluded from existential sentences (examples from Lumsden 1988): (i) There were the usual books outside the gates for sale. (ii) There was the top of a bottle on the table. (iii) There is the most remarkable woman in the room. According to Comorovski (1991), there sentences containing a definite associate illustrate the "presentative" use of ES: they serve the pragmatic function of introducing a new discourse referent

Expletives as Arguments: Germanic Existential Sentences Revisited

(3) (4) (5) (6) (7)

a. b. a. b. a. b. a. b. a. b.

There suddenly broke out a fight. *There suddenly broke out the fight. There arrived three new lecturers today. *There arrived every new lecturer today. There are many students waiting outside. *There are Marys students waiting outside. There appeared a ghost in the attic. *There appeared Elvis in the attic. There seemed to be someone in the room. *There seemed to be she/her in the room.

Definite descriptions, proper names or pronouns are usually excluded from existential sentences, as well as quantified noun phrases introduced by a universal quantifier such as every or most (Milsark 1974, 1977).3 DPs that are excluded from the coda of existential sentences are referred to by Milsark as "strong", whereas those which may serve as an associate of an expletive have been labelled "weak" DPs. Observe that the QP associate of expletive there in (8) has narrow scope only, as can be seen from the fact that it cannot take scope over the negation. (8) There are not many students in this class. i. 'It is not the case that many students are in this class.' 'Many students are such that they are not in this class.' ii.

rather than asserting existence. She further argues that presentative ES differ semantically from ordinary existentials in that the presence versus absence of the expletive does not affect the meaning of the sentence (compare section 4 below). The same appears to be true for the German example (iv), which has been suggested by a reviewer: (iv) Es war der Geist von Elvis im Schloss. there was the ghost of Elvis in.the castle The present article focuses on unequivocally "existential" ES though. 3 In view of the fact that the indefinite associate tends to favour a non-specific interpretation, En (1991) has suggested that the term "definiteness effect" should be replaced by the term "specificity effect". The following counterexamples (from Comorovski 1991 and Abbott 1994, 1995), however, show that certain types of specifics, including partitives, can also occur quite naturally in existential environments: (i) There is a certain young man at the door. (ii) Remember those bats that got loose last night? There was one of them in the fridge this morning! (iii) There are most/all/several of yesterdays exams left to correct. Abbott (1995) concludes that specificity (at least in the "partitive" sense of En 1991) cannot be the relevant factor triggering the DR either. According to Comorovski (1991), sentences like those above are not interpreted as genuine existentials, but again illustrate the presentative use of there (compare note 2). As a comprehensive discussion of all real or apparent exceptions to the DR is beyond the scope of the present study, we will leave this as a matter for future investigation.

Claudia Felser & Laura Rupp

Moreover, the predicate of existential sentences must have stage-level properties in the sense described by Carlson (1980), a fact previously noted by Milsark (1974). Predicates that strongly favour an individual-level reading such as be dead or be intelligent, on the other hand, are normally disallowed. (9) (10) a. b. a. b. There was a fly in my soup. *There was a fly dead. There were many people dancing. *There were many people intelligent.

Expletive-associate constructions involving true expletives are further characterised by the fact that the predicate shows overt number agreement with the associate rather than with the expletive:4 (11) There were/*was five students missing.

On the assumption that predicate-internal argument slots are reserved for marked expressions (compare Chomsky 1995: 312-316), it would seem that expletives must necessarily originate in a predicate-external position (see also Safir 1993). Chomsky (1995) proposes that in languages where T contains a strong [D] feature which forces overt substitution in its specifier, expletives are base-generated in (Spec,TP). In Chomsky (1998), the strength metaphor is abandoned, and the specifier of T is instead assumed to be licensed by an (uninterpretable) EPP feature. The associate itself remains in its base position, an option that is naturally available under the predicate-internal subject hypothesis (Koopman & Sportiche 1991, among others). True expletives are assumed to bear a defective set of -features. While according to Chomsky (1995), true expletives consist merely of the categorial feature [D], Chomsky (1998) suggests that their -set contains an uninterpretable [person] feature only. Their sole syntactic function is that of checking the EPP feature of T prior to Spell-Out, as required by the principle of Full Interpretation (FI). An existential sentence such as There were many people present can therefore be assigned the (simplified) representation in (12).5

4 In some languages, agreement on the verb is triggered by the expletive (e.g. French il; see Cardinaletti 1997 for a cross-linguistic comparison). Many speakers of English also permit a (colloquial) variant of (11), There's five students missing, in which the copula agrees with the expletive (for some relevant discussion, see Schtze 1999, and references cited there). The present study, however, focuses on "true" expletives that fail to trigger verbal agreement. 5 The label Pred(icate) P(hrase) has been chosen to remain neutral with respect to the precise categorial status of the "small clause" complement of T, which presumably is headed either by a lexical category (along the lines of Stowells 1983 analysis of small clauses), by a functional head of the kind postulated by Bowers (1993), or by some light verb v that forms part of an extended verb phrase in the spirit of Larson (1988) and Chomsky (1995).

Expletives as Arguments: Germanic Existential Sentences Revisited

(12)

TP ti Spec T g to there T[EPP] PredP g ep were many people present

This of course raises the question of how the associate checks case and agreement, given that the thematic subject is not in the checking domain of T at Spell-Out, and the external subject position is already occupied by the expletive. We shall discuss possible answers to this question in section 2 below. Full Interpretation requires that PF and LF, the only two levels of representation recognised within Minimalism, contain only legitimate objects. LF and PF objects are deemed legitimate only if they can be interpreted by the conceptual-intentional and articulatory-perceptual system respectively. Since true expletives do not seem to bear a thematic role or carry any interpretable features, their status at LF remains rather dubious. A successful analysis of expletive-associate constructions will have to provide answers to at least the following questions: i. ii. iii. iv. How is agreement triggered given that the associate is not in a checking configuration with the verb or auxiliary in T? How does the associate check its case? How can we account for the definiteness restriction, and for the fact that the predicate must have stage-level properties? What is the syntactic status of the expletive, and what (if any) is its contribution to semantic interpretation?

In what follows, we provide an analysis of expletive-associate constructions within the framework outlined by Chomsky (1991, and later) that attempts to answer the above questions, and further shows how the syntactic and semantic properties of ES are interrelated. The remainder of this article is structured as follows. Section 2 offers a brief critical overview of previous approaches to expletive-associate constructions within the principles-and-parameters/minimalist framework. In section 3, we provide arguments against the partitive case hypothesis initially put forward by Belletti (1988) and developed further by Lasnik (1992, 1995). Instead, it is argued that the associate, by virtue of being barred from checking either structural or inherent case, receives morphological default case. In section 4, we challenge the widely held assumption that true expletives lack a thematic role. Extending a suggestion made by Ramchand (1996), we propose that expletives are overt instantiations of the event or spatiotemporal argument postulated by Kratzer (1995). It is shown that a minimalist analysis of ES in the spirit of Chomsky (1995, 1998), combined with a

Claudia Felser & Laura Rupp

Kratzer/Diesing-style approach to the syntax of DPs, successfully accounts for the characteristic properties of existential sentences. Finally, in section 5, we examine transitive expletive constructions (TECs) which have been argued to be potentially problematic for Diesings (1992) theory of indefinites. It is shown that to the extent that these problems are real, they are not in fact incompatible with Diesings mapping hypothesis.

2 Previous analyses

2.1

Expletive raising

Based on the observation that ES share some of the properties of inverse copular sentences, Moro (1997) proposes an analysis of ES according to which the expletive originates as the predicate of the small clause complement of a copula (see Hoekstra & Mulder 1990 for a similar proposal). (13) [IP [IP therei [I' wask [VP tk [SC a fly ti ]]]] [PP in my soup ]] _______________________|

The expletive checks nominative case, which is shared with the subject DP via "case agreement", and also inherits the number properties of its associate (Moro 1997: 48, 98). The locative PP is analysed as an IP-adjunct, which, however, seems questionable on the grounds that its presence is often felt to be nearobligatory in copular ES. Expletive raising analyses also have some difficulty explaining why the alleged subject raising counterparts of ES such as (13) have a locative rather than an existential meaning. (14) A fly was there (in my soup).

Taking existential there to be identical to locative there seems problematic in view of the fact that the former can co-occur with locative expressions, including locative there as in (15) below, and given the various semantic, syntactic and phonological differences between the two noted by Breivik (1981) and others. (15) There were only a few students there.

What is more, given the Minimal Link Condition (Chomsky 1995: 311), we would - other things being equal - expect the small clause subject (i.e. a fly in [13]), rather than its predicate, to undergo raising in order to satisfy the EPP and check nominative case.

Expletives as Arguments: Germanic Existential Sentences Revisited

Finally, the inverse copular analysis would seem to have trouble accounting for transitive expletive constructions, which are discussed in some detail in section 5 below.

1.2

LF associate raising

Many earlier analyses of ES assumed that some kind of direct syntactic dependency relation holds between the expletive and its associate (see e.g. Safir 1985a). In view of the fact that existential sentences normally have a nonexpletive correlate, Chomsky (1986) suggested that the associate raises to matrix subject position at LF, replacing its expletive placeholder. (16) a. b. There was a fly in my soup. [ a fly ] was t in my soup _______|

If subject raising takes place overtly, insertion of there is unnecessary; if it is delayed until LF, the expletive is inserted as a "dummy" filler of the structural subject position in order to satisfy the EPP. Empirical evidence for covert associate raising appears to be provided by the associate's ability to behave like a matrix subject with respect to control, as illustrated by the examples in (17) (from Chomsky 1995: 274).6 (17) a. b. (18) There arrived three men last night [PRO without identifying themselves]. Three men arrived last night [PRO without identifying themselves]

*I met three men last night [PRO without identifying themselves].

Chomsky (1991, 1993) revises his earlier approach and instead analyses the expletive as an "LF affix" to which the associate covertly adjoins. The two together then form an "LF word", to the effect that the expletive becomes invisible at the LF interface (LF being blind to word-internal structure).7 However, the assumption that ES and their non-existential counterparts are isomorphic at LF proves to be problematic on both theoretical and empirical grounds. First, one might wonder why a semantically vacuous item should ever be lexically inserted in the first place - only to disappear, or become invisible, at LF - in a language where overt subject raising is the rule, and if such raising is

6 It should be noted, however, that some speakers consider (17a) less acceptable than (17b), or even reject the example. 7 See Groat (1995) for some criticism of this proposal.

Claudia Felser & Laura Rupp

ultimately required anyway.8 Secondly, both variants of the associate raising analysis wrongly predict (i) that (16a) and (16b) should be completely synonymous, (ii) that a QP associate should admit a wide scope reading (cf. [8] above; see Chomsky 1995: 157 for a brief discussion of this issue), and (iii) that every ES should have a non-expletive counterpart. The last prediction is proven wrong by the existence of bare existential sentences such as the following (cited by Jenkins 1975: 49).9 (19) (20) (21) a. b. a. b. a. b. There is a pain in my arm. *A pain is in my arm. Theres an answer to the question. *An answer is to the question. Theres a difference between X and Y. *A difference is between X and Y.

The issue of the non-synonymy of ES and their full sentence correlates will be addressed in detail in section 4 below. A more recent variant of the associate raising analysis is that of Groat (1995), who suggests that Exp itself carries a "null" interpretation that prevents any associate adjoined to it from being interpretable in (Spec,TP) at LF.10 This implies that the associate can only be interpreted in a predicate-internal position, which according to the Kratzer/Diesing theory of indefinites (see section 4.2 below) means that it can only receive a "weak" or existential reading. Though his analysis correctly predicts that strong DPs cannot serve as associates for expletives, Groats (1995: 361) characterisation of Exp as a maximal projection that is "neither an argument nor a nonargument" is not entirely satisfying, as it amounts to postulating a new type of LF object which is interpreted neither as an argument nor as a predicate (compare Rothsteins 1983 discussion of the predicate/argument dichotomy). In other words, it remains unclear what the presence of an expletive really signifies to the interpretive component, and why natural languages should make such elements available in the first place (see also Groats note 7).

8 Observe, however, that syntax regularly makes use of semantically vacuous elements such as uninterpretable case features, or dummy auxiliaries such as do that serve as carriers for otherwise stranded morphological affixes (see Ouhalla 1991; Radford 1997). By way of legitimatising the presence of a semantically vacuous expletive, one might argue that its semantic contribution is to ensure LF convergence by satisfying the EPP, or by providing a structural configuration which excludes certain interpretive possibilities from the range of options otherwise available (see section 4 for further discussion). 9 Further inadequacies of the expletive replacement analysis have been pointed out by den Dikken (1995) and Jang (1997). 10 More precisely, Groat (1995: 360) suggests that Exp provides an interpretable instruction to do nothing at the LF interface.

Expletives as Arguments: Germanic Existential Sentences Revisited 1.3 Feature raising / long-distance agreement

In Chomskys (1995) framework, LF associate raising can be conceived of as adjunction of the associates formal features to T (compare Chomsky 1995: 370, and Jang 1997 for a similar proposal), as indicated in the diagram below. (22) TP tp D T g rp there T PredP 2 ti [sg,NOM] T DP Pred g 4 6 was a fly in my soup [sg,NOM] Chomsky (1995)

On the assumption that the semantic features of the associate remain frozen in place, as suggested by Jang (1997), the above analysis accounts not only for the observed agreement facts, but also for the absence of wide scope readings for quantificational associates. In Chomsky (1998), the concept of feature raising is eliminated and replaced by the possibility of abstract long-distance agreement. For two elements to be linked through the operation Agree (which results in the erasure of uninterpretable matching features), both must contain uninterpretable features. Given that the associate's -features are semantically interpretable, it would seem that it can only be rendered active - and thus able to enter into an agreement relationship with the uninterpretable -set of T - by an uninterpretable case feature. An analysis along these lines is consistent with the observation that the associate is incapable of binding an anaphor (Chomsky 1995: 275): (23) *There seem to each other [ t to have been many linguists given good job offers].

What seems puzzling, however, is that the associate should be able to serve as a controller of PRO in many varieties of English (cf. [17a] above). According to Cardinaletti (1997), the observed difference between an associate's control and binding properties follow from the assumption that agreement and control are triggered by simple feature movement, whereas binding requires that the antecedent be the whole category (Cardinaletti 1997: 525). Her proposal remains essentially stipulative though, and is not obviously compatible with Chomsky's (1998) claim that feature raising does not exist. As noted by Landau (1999), the mechanisms that govern control of adjuncts are as yet poorly understood. Controller choice appears to be sensitive to semantic factors such as

10

Claudia Felser & Laura Rupp

agentivity or animacy, and closer examination of the data reveals that ccommand is not a necessary condition even for obligatory control (compare Landau 1999: 43f.). In other words, examples like (17a) do not provide particularly strong evidence in favour of covert associate or feature raising.

1.4

The partitive case hypothesis

Different from what Chomsky's (1986, and later) analyses would seem to predict, the associate does not universally carry nominative case. Lasnik (1992, 1995) and Groat (1995), among others, suggest that there is able to check the nominative case feature of T, while the associate is assigned (abstract) partitive, as proposed by Belletti (1988). Partitive case is morphologically realised on object plural count nouns in Finnish, including the THEME argument of copular or unaccusative verbs in existential sentences like the one below (Bellettis example): (24) Pydll on kirjoja. on the table is booksPART There are some books on the table.

By stipulation, partitive case is taken to be assigned to the complement of be, unaccusative and raising verbs.11 Unlike Groat, Lasnik adheres to Chomskys (1991, 1993) view that there is an LF affix which must be attached to an appropriate host. For Lasnik, then, the associate raises out of "enlightened selfinterest", in order to benefit there. Lasnik adds to the affixal requirement of there the condition that the associate host must be partitive (that is, it must be a non-specific indefinite). This condition successfully rules out examples like (25b), as definites do not carry partitive case and unaccusatives are unable to check objective case. (25) a. b. There arrived a man. *There arrived the man.

It further accounts for the unacceptability of (26a,b), where, according to Lasnik, the potential associates do not bear partitive, but nominative or objective case. (26) a. b. *There seems [(that) someone/he is here]. *There seems to [a strange man/him] [that it is raining outside].

11 A more recent variant of the partitive case hypothesis is de Hoop's (1996) proposal that the associate carries "weak" structural objective, a type of case that is licensed under V-government or by certain (transitive) verbs, and which is correlated with a weak reading of DPs (de Hoop 1996: 80).

Expletives as Arguments: Germanic Existential Sentences Revisited

11

Example (27) is ruled out by FI as well, since the associate is not in a partitive case position and therefore unable to adjoin to affixal there. (27) *There someone laughed.

Thus, under Lasniks approach, the crucial factor in the derivation of ES is morphological compatibility: the LF affix there requires an associate host with partitive case. Underlying Lasnik's analysis is the assumption that the checking of case and agreement features can be dissociated (as in Chomsky 1995). In more recent work, however, Chomsky has argued that the two are in fact inseparable, in the sense that structural case is a mere reflex of an abstract agreement relation (Chomsky 1998: 37).12 While we share with Lasnik and Groat the view that Exp carries case (contra Chomsky 1995, 1998), we provide arguments against the partitive case hypothesis in section 3 below.

1.5

Summary

While the expletive replacement analysis must be rejected on the grounds that ES and their non-expletive counterparts are not isomorphic at LF, several recent analyses are able to account for this observation, as well as for the fact that the associate determines number agreement. They differ, among other things, with respect to the case properties ascribed to there and its associate, an issue that will be discussed in more detail in the following section. An analysis of ES as variants of inverse copular sentences faces a number of empirical and theoretical problems, and thus seems difficult to maintain. Analyses that assume covert associate raising, or its correlates in more recent versions of Minimalism, on the other hand, manage to avoid most of these problems. With the exception of Groat (1995), existing analyses of ES tend to remain largely agnostic as to the expletive's contribution to interpretation, a question which we will return to in section 4.

12 In a similar vein, Cardinaletti (1997: 526) proposes that "[o]nly those expletives that are unambiguously marked as nominative trigger agreement on the verb." This generalisation, however, would appear to wrongly rule out the German es gibt construction, where agreement is determined by an expletive that is ambiguous between nominative and accusative (cf. also the colloquial English there's DP construction).

12

Claudia Felser & Laura Rupp

3 Against the partitive case hypothesis

3.1

Empirical and theoretical problems

In this section, we provide evidence against Bellettis (1988) claim that the associate is assigned partitive case by V. According to Belletti, Finnish transitive verbs check either accusative or partitive case, depending on the reading associated with their object: if the object has a definite/specific reading, it is marked for accusative; if it has an indefinite/nonspecific reading, it carries partitive case morphology. The existential sentence in (24) (repeated below), for instance, contains the verb be, which is unable to check accusative case and checks partitive case only. (24) Pydll on kirjoja. on the table is booksPART There are some books on the table.

Belletti generalises the existence of partitive case to other languages, and suggests that the definiteness restriction in ES follows naturally from assuming that "existential" predicates universally assign partitive case to the associate, and that only indefinites are able to carry partitive. In languages that lack morphological partitive, its only reflex will be in the interpretation of DPs. However, the role and distribution of partitive case in Finnish is not quite as straightforward as Belletti suggests. Heinmki (1984) has pointed out that the partitive/accusative alternation is correlated with aspectual distinctions, as can be seen from the difference between (28a) and (28b). (28) a. Maija luki kirjaa. M. read bookPART Maija was reading a book Maija luki kirjan. M. read bookACC Maija read (all) the book

b.

What is more, it appears that both partitive and accusative case can be realised on either definite or indefinite DPs (Heinmkis examples; see also de Hoop 1996: 67): (29) a. Metsstj ampui lehmn kuoliaaksi. hunter shot cowACC dead.to The hunter shot a cow dead.

Expletives as Arguments: Germanic Existential Sentences Revisited

13

b.

Manne kehui hevosta. M. praised horsePART Manne was praising the horse.

Vainikka & Maling (1996) also show that partitive case morphology in Finnish is not restricted to indefinites, and note that it is also compatible with universal quantification, or with the quantifier most. (30) Pekka kokeili useimpia reseptej. P. tried most recipesPART Pekka tried most (of the) recipes.

As we saw above, however, both definites and strong QPs are normally excluded from ES. Thus, in view of the fact that partitive case morphology in Finnish is actually dissociated from the strong/weak distinction, it seems that the distribution of morphologically partitive DPs in this language does not provide a very solid basis for accounting for the definiteness restriction in terms of partitive case-assignment to the associate. Moreover, as noted by de Hoop (1996: 68), Belletti's claim is questionable on semantic grounds as a partitive meaning is typically associated with definite rather than indefinite DPs. A theoretical drawback of Bellettis analysis is that one is forced to stipulate that partitive case is checked optionally by the verb in question, or else the subject DP in (31a) would be assigned case twice: nominative by T, and partitive by V. Examples like (31b), on the other hand, should not even be possible (as the partitive case feature of arrive remains unchecked). (31) a. b. [a man]i arrived ti [the man]i arrived ti

Further evidence against the partitive case hypothesis can be derived from the fact that many Germanic languages regularly make use of transitive expletive constructions, whose agentive associate can hardly be argued to be an object of the verb at any level (data from Bobaljik & Jonas 1996). (33) Es essen einige Muse Kse in der Kche. there eat some mice cheese in the kitchen There are some mice eating cheese in the kitchen. Er heeft iemand een appel gegeten. there has someone an apple eaten Someone has eaten an apple. a hafa margir jlasveinar bora bing. there have many Christmas.trolls eaten pudding Many Christmas trolls have eaten pudding. (German)

(34)

(Dutch)

(35)

(Icelandic)

English too marginally permits a type of transitive ES, as the following examples (from Chomsky 1995: 343) show.

14

Claudia Felser & Laura Rupp

(36)

a. b.

There entered the room a man from England. There hit the stands a new journal.

Finally, let us again consider ES in German, a language with a richer morphological case system than English. It has been argued that there are at least five different types of sentence involving expletive or impersonal es in German (compare Lenerz 1985: 102-104). German expletive constructions differ from their Dutch correlates in that the expletive is only ever spelled out in sentence-initial position (hence the terms "topic es" or "Vorfeld-es"). Where the verb second requirement does not hold or has been met otherwise, the expletive is phonetically null. Modern Icelandic patterns with German in this respect (see Lenerz 1985: 110-111). In the following, we shall focus on genuine existential sentences, which pattern with their English counterparts in that the associate must normally be a non-specific indefinite, the predicate involved must have stage-level properties, and agreement is triggered by the associate:13 (37) a. Es war ein Geist/*der Geist/*Elvis /*er im Schloss. there was a ghost /*the ghost/*Elvis/*he in.the castle 'There was a ghost in the castle.' Es war jemand betrunken/*intelligent. there was someone drunk /*intelligent 'There was someone drunk.' Es sind/*ist viele Mnner im Garten. there are /*is many men in.the garden 'There are many men in the garden.'

b.

c.

In German, the associate carries morphological nominative (rather than accusative, dative, or genitive - compare [38] below); the same is generally true for ES in Icelandic (compare Vangsnes 1995: 96).14

13 Reuland (1983) cites a few apparent exceptions to the DR in German, such as (i) below. However, the acceptability of presentative ES such as (i) declines considerably if, for instance, a locative PP such as in Berlin is added. (i) Es spielt das Londoner Symphonieorchester. there plays the London symphony.orchestra The London symphony orchestra is playing. (ii) ?*Es spielt das Londoner Symphonieorchester in Berlin. Since the presence of a PP or some other deictic expression usually improves the acceptability of ES (see Ramchand 1996: 189 for a possible explanation why), the ill-formedness of (ii) is rather puzzling. One might speculate that either the associate is not interpreted as definite or strong in the relevant sense, or that the expletive in sentences like (i) above is inserted as a pragmatic dummy, to aid presentational focus assignment (compare note 2). 14 This is not true for German Es gibt (lit. 'it gives') constructions, where the associate is marked accusative. This suggests that the verb geben 'give' is transitive in all its uses, and that the expletive - which also triggers verbal agreement - here carries a full set of -features.

Expletives as Arguments: Germanic Existential Sentences Revisited

15

(38)

Es war ein /*einen /*einem /*eines Geist(es) im Schloss. there was aNOM /*aACC /*aDAT /*aGEN ghost(GEN) in.the castle

Speakers of English, on the other hand, appear to show a preference for assigning a (normally illicit) pronominal associate objective rather than nominative case, as in There was (only) him in the room. To conclude, the hypothesis that the associate is assigned structural or inherent partitive by the verb lacks any convincing motivation.

3.2

An alternative proposal

Although the German and Icelandic facts noted above are consistent with Chomsky's (1993, 1995, 1998) assumption that a true expletive is case-free, and structural nominative is checked by the associate, even a very limited crosslinguistic comparison reveals that the associate is not universally marked nominative. Let us therefore re-examine the alternative possibility that it is the expletive that checks structural nominative (or objective, in non-finite contexts such as [39]), as has been suggested by Lasnik (1992, 1995) and others. (39) I found there to be no basis for the allegation.

The idea that the expletive checks nominative case is consistent with the fact that the German expletive pronoun es and Icelandic a are morphologically ambiguous between NOM and ACC.15 Given the objections to the partitive case hypothesis that were raised above, we suggest that the associate in existential sentences receives whatever happens to be the "default" case in a language.16 Schtze (1997: 44) defines default case as [a] case used by the morphological component to spell out a DP that checks neither structural nor inherent case in the syntax. Given the status of default case assignment as a last resort (or "elsewhere") rule, the default case hypothesis avoids the problems that arise

15 An anonymous reviewer asks why German es does not occur as the subject of small clauses in sentences such as (i): *Ich sah es einige Muse Kse in der Kche essen. (i) I saw there some mice cheese in the kitchen eat Note, however, that German es does not normally occur in embedded contexts anyway - except in a few quasi-idiomatic cases such as (iii). (ii) *Ich sah, dass es einige Muse Kse in der Kche gegessen haben. I saw that there some mice cheese in the kitchen eaten have (iii) Lass es Licht werden! let there light be That is, while it does not seem impossible in principle for expletive es to appear in accusativechecking environments, whatever constraint excludes it from embedded clauses is likely to be responsible to the ungrammaticality of (i) as well (see Felser 1998 for some relevant discussion). 16 We are grateful to Harald Clahsen for suggesting this possibility to us.

16

Claudia Felser & Laura Rupp

from the partitive case hypothesis, including the assumption that certain verbs assign partitive case optionally, and to indefinite DPs only (cf. [31] above). Of the languages under consideration, according to Schtze, German, Dutch and Icelandic all use nominative as a default case, whereas English uses objective. Schtze (1997: 53) lists twelve environments in which default case is used (in English). These include so-called "mad magazine" sentences (e.g. Her/*she cheat on you? Never!), ungoverned small clauses and gerunds (Him/*he liking beans, they bought some.), predicate nominals (The murderer is her/*she.), sentence fragments (Who did it? - Me/*I.), as well as certain types of left dislocation, ellipsis and conjunction structures. Following Vainikka & Maling (1996), the default case used in Finnish is partitive - the case borne by the logical subject of existential sentences in this language. Under this view, only the associate's -features will be checked against the corresponding features of T, whereas its - initially unspecified - case feature (if indeed it has one) will remain unchecked (or "unvalued"), and hence ultimately be spelled out as default case. How does this square with Chomsky's (1998) claim that structural case is inseparably tied to agreement? Since expletives can undergo raising, they must carry some uninterpretable feature that renders them active, i.e. accessible to Agree/Move. Chomsky (1998: 40) suggests that Exp carries an uninterpretable [person] feature only, which is checked against a matching (although equally uninterpretable) feature in T. Nominative case, on the other hand, is assigned to the associate DP as a reflex of abstract T-associate agreement. While a comprehensive discussion of Chomsky's (1998) views on checking and agreement is clearly beyond the scope of the present study, it seems to us that the above suggestions must be treated with some caution. For one thing, the observed cross-linguistic differences regarding the morphological case properties of the associate in languages that do not use nominative as the default are difficult to reconcile with the idea that the associate universally checks NOM. The latter assumption is also potentially problematic for expletive-associate constructions in languages such as French in which number agreement is triggered by the expletive (see also Chomsky 1998: 44fn.91). Independent evidence for a dissociation of case and agreement comes from the study of children suffering from Specific Language Impairment whose ability to handle agreement appears to be selectively impaired, while their case system is largely intact (Clahsen et al. 2000). As regards the featural content of the expletive, one may wonder why elements such as German es, which is morphologically an (inherently definite) 3rd person singular pronoun, should be assumed to carry an uninterpretable [person] feature. Recall that in Minimalism, -features on nominals are generally deemed to be interpretable (Chomsky 1995: 277).17 This suggestion is especially odd in the light of Lyons' (1999)

17

As Atkinson (1999) notes, the idea that two matching uninterpretable features (i.e., the [person] features carried by Exp and T) should enter into an agreement relation is rather odd.

Expletives as Arguments: Germanic Existential Sentences Revisited

17

observation that cross-linguistically, [person] (as opposed to the -features [number] and [gender]) can only be encoded on definite noun phrases, which would appear to eliminate the associate as a possible bearer of an interpretable [person] feature. Finally, from a theory-internal perspective, Atkinson (1999) draws attention to some technical problems and inconsistencies arising from Chomsky's (1998) treatment of expletive-associate constructions (in particular in the context of sentences in which the expletive itself undergoes raising). Given the above list of "caveats", consider instead the following alternative scenario, which is still broadly consistent with Chomsky's (1998) general approach to feature checking. The associate (which presumably is rendered active by the presence of an unvalued case feature) carries a [number] but no [person] feature. The associate's [number] feature checks, or "values", the corresponding feature on T (that is, T's [number] feature is assigned a singular or plural "value" by the associate, to be spelled out at PF). T, however, remains active due to the fact that the associate's -set is incomplete. Unable to "satisfy" T, which still hosts an uninterpretable, or unvalued, [person] feature, the associate's case feature cannot be checked/valued - in other words, the associate fails to be assigned structural nominative.18 Then Exp is merged into (Spec,TP), carrying both an interpretable [person] feature and an unvalued case feature. The latter activates the expletive, thus enabling it to enter into an agreement relation with T. Nominative case is assigned to the expletive as a consequence of the partial agreement relationship between Exp and T, which serves to erase the last remaining uninterpretable feature from T's -set. Under this view, the (simplified) representation of (37a) will look as follows (with interpretable features underlined).19

18 As has been pointed out to us by a reviewer, the associate appears to be able to undergo partial movement in passive structures such as (i): (i) There was believed to be a man killed. Although unexpected, partial associate raising in passive environments is not necessarily inconsistent with our analysis: the associate (rendered active by an unvalued case feature) may need to raise in order to check some uninterpretable feature(s) of v, or of an aspectual or Voice head. Chomsky (1998: 19fn.40) indicates that the unexpected verb-object ordering in examples like (i) may not result from any syntactic process, but instead represents a mere surface linearisation preference. 19 On the assumption that in declarative main clauses in German, C contains an EPP feature (forcing the projection of a specifier), the expletive will ultimately raise to (Spec,CP), an operation that may be driven by a force or topic feature (compare Chomsky 1998: 22) - but see note 27. Note that contrary to what our diagram in (40) suggests, TP in German and Dutch is standardly assumed to be head-final (but see Zwart 1994, 1997 for some arguments against this view). As the question of the relative ordering of T and its complement in V-final clauses is irrelevant to the present discussion, we have, for expository reasons, taken German and Dutch TPs to be structurally analogous to TP in English.

18

Claudia Felser & Laura Rupp

(40)

CP ti C ti C[EPP] TP tp D T g tp es T[EPP] PredP g to [3ps, Case] war DP Pred 6 [3ps, sg, Past] 5 ein Geist im Schloss [sg, Case]

NOM

"default" case

We shall return to the syntactic and semantic properties of true expletives in section 4 below. Potentially problematic for the default case hypothesis is the fact that in Icelandic, the associate sometimes carries quirky rather than nominative case. Consider the following example: (41) a fannst einhverjum hann vera duglegur. there found someoneDAT heNOM be clever 'Someone found him to be clever.'

It is commonly assumed that quirky case is inherent (i.e., lexically determined), which is consistent with the observation that it is not normally available for arguments associated with the AGENT role (compare Schtze 1993, and references cited there). Given the above definition of default case assignment as a "last resort" mechanism, the existence of examples such as (41) is not in fact in conflict with our proposal: while the expletive checks structural nominative, the experiencer subject is assigned inherent dative by the verb - so that the default case mechanism here need not be invoked at all.20

20 A reviewer wonders whether the use of inherent vs. default case may be subject to parametric variation. While the precise mechanisms governing quirky case assignment in Icelandic have been the matter of much debate (see Schtze 1993, and references cited there), we think that the answer to this question probably has to be "no". The two mechanisms of case assignment are qualitatively different (and the same language may use both of them): Inherent case is lexically determined, whereas default case assignment is assumed here to be a "last resort" or "elsewhere" mechanism only. There is no obvious way in which this type of variation could reduce to differences in the featural make-up of functional heads (Borer 1984, Chomsky 1995).

Expletives as Arguments: Germanic Existential Sentences Revisited 1.3 Summary

19

To summarise, we have presented evidence against the assumption that the associate is assigned partitive case by V, and instead argued that the associate receives default case, a claim that correctly predicts the observed crosslinguistic differences with respect to the morphological case properties of the associate. Given the modifications of checking theory suggested by Chomsky (1998), covert associate raising can now be understood as involving an abstract partial agreement relationship between the associate and T. We further proposed that the expletive carries an interpretable [person] feature, and that it is assigned nominative case by T. The "split-agreement" hypothesis outlined above accounts for the observation that in ES, number agreement appears to be dissociated from structural case assignment. The issues that remain to be considered are (i) why the associate must normally be a (non-specific) indefinite, if not for morphological reasons la Lasnik, (ii) the stage-level restriction on the predicate, and (iii) the origin and semantic function of the expletive. These questions will be addressed in the following section.

4 The expletive as a thematic expression

4.1

Interpretive restrictions on the associate

It has often been noted that ES and their non-expletive counterparts are not necessarily semantically equivalent (see e.g. Jenkins 1975: 45-50, and references cited therein). For illustration, compare the following pairs of sentences: (42) a. b. a. b. a. b. a. b. Minors must be in the dorm by midnight. There must be minors in the dorm by midnight. (Jenkins 1975) (43) Condors are in the Andes. There are condors in the Andes. (Kimball 1973, cited by Jenkins 1975) A man is in the room. There is a man in the room. Some girls were at the party. There were some girls at the party. (Groat 1995) (Jenkins 1975)

(44) (45)

20

Claudia Felser & Laura Rupp

Bare plurals such as minors or condors do, in principle, admit either a generic/universal or an existential interpretation. While the subject DP in (42a) is indeed ambiguous between these two readings (viz. all minors must be in the dorm by midnight versus the less natural reading the dorm must have some minors in it by midnight), the same DP can only be interpreted existentially in (42b). Essentially the same difference can be observed between (43a) and its ES correlate, (43b). Similarly, Groat (1995: 355) notes that while the indefinite DP a man in (44a) admits a specific reading, it only has a nonspecific reading in (44b). Finally, some girls in (45a) is ambiguous between a partitive, or presuppositional, reading (viz. some of the girls) and a cardinal reading (where some girls simply means more than one girl). In (45b), on the other hand, only the cardinal reading is available. In fact, we may assume that all the above examples illustrate basically the same phenomenon, namely that indefinites in non-existential sentences may be ambiguous between a strong (generic, specific, or presuppositional) and a weak (existential, non-specific, or cardinal) reading, to borrow the labels used by Milsark (1974, 1977), whereas as the associate of an expletive they favour the weak reading. How can this contrast be explained? Following Groat, it can be accounted for by appealing to Diesings (1992) theory of indefinites, which we will briefly sketch below. Taking Groat's suggestion as a starting point, we will argue in section 4.3 that true expletives are overt instantiations of the spatio-temporal argument proposed by Kratzer (1995), and that they originate within the extended predicate phrase.

4.2

The Kratzer/Diesing theory of indefinites

Diesing (1992) proposes that subjects of different types of predicates appear in different subject positions at LF. Observe, for instance, that bare plural subjects of stage-level predicates (SLPs) such as (46a) are ambiguous between a strong generic and a weak existential reading, whereas bare plural subjects of individual-level predicates (ILPs) as in 46b) admit a generic interpretation only. (46) a. b. Firemen are available. Firemen are intelligent. (existential or generic) (generic only)

Aiming to account for certain differences in the syntactic behaviour of definites and indefinites, Diesing applies the semantic framework of Heim (1982) to syntactic representations. According to Heim, the presence of certain quantifiers such as every causes the semantic representation of a sentence to be split into three parts: the quantifier, its restriction, and the nuclear scope. A default process of existential closure is assumed to bind all unbound variables (notably those introduced by indefinites) that remain within the nuclear scope in the semantic representation of a sentence. Diesing (1992) proposes that the nuclear scope of a sentence can be identified with VP, while IP (TP in our system)

Expletives as Arguments: Germanic Existential Sentences Revisited

21

provides the domain of the restrictive clause. This idea is expressed by the following mapping hypothesis (Diesing 1992: 15). (47) Material from VP is mapped into the nuclear scope. Material from IP is mapped into the restrictive clause.

Given that the internal subject hypothesis makes two subject positions available (compare [48] below), an internal and an external one, the interpretation of indefinites can now be correlated with the position they occupy at LF: DPs which are interpreted in (Spec,TP) receive a strong interpretation, whereas indefinites that reside within the predicate phrase at LF are assigned a weak reading. (48)
RESTRICTIVE CLAUSE TP 3 (DPsu) T 3 u T NUCLEAR SCOPE PredP 3 (DPsu) Pred 3 Pred

In line with observations made by Kratzer (1995), Diesing notes that different types of predicate impose different restrictions on the interpretation and distribution of subjects. She proposes that the LF position of indefinite subjects is restricted in the following way (Diesing 1992: 22): (49) Subjects of stage-level predicates can appear either in [Spec,IP] or in [Spec,VP]. Subjects of individual-level predicates can appear only in [Spec,IP].

According to (49), indefinite subjects of stage-level predicates may optionally be reconstructed in their original position within VP, which then renders them subject to existential closure (compare [50c]). (50) a. b. c. Firemen are available. [TP Firemeni [VP ti are available ]] |_ _ _ _ _ _ X [x is a fireman x is available] (ambiguous)

Within the minimalist framework, movement has been described as a copying process, with the possibility (or indeed, the necessity) of certain movement traces being deleted, so that they become invisible to LF (see Chomsky 1995:

22

Claudia Felser & Laura Rupp

300-304 for discussion). Pursuing this idea further, Hornstein (1995) claims that for an (A-)chain to be interpretable by the conceptual-intentional system, all but one of its links must delete (Hornstein 1995: 154, 186-192). The deletion of chain links can thus be regarded as the minimalist analogue to reconstruction (Hornstein 1995: 195).21 Under this view, a weak (or existential) reading for bare plurals results if the higher link is deleted (compare [51a]), whereas a strong (or generic) reading results from the deletion of the original link in (Spec,VP), as indicated in (51b). (51) a. b. [TP firemen [VP FIREMEN are available ]] [TP FIREMEN [VP firemen are available ]]

For an indefinite to be interpreted as strong it must reside outside the predicate phrase at LF in order to escape existential closure. Definite subjects differ from indefinites in that they may not appear inside the predicate phrase at LF (unless they are focused - compare notes 2 & 3) as they are incompatible with existential binding (Diesing 1996: 72). The same holds true for strong quantificational DPs, which must scope out of the predicate phrase in order to avoid a semantic type mismatch (Diesing 1996: 69). Kratzer (1995) maintains that only stage-level predicates provide an internal subject position. Subjects of individual-level predicates, by contrast, are always external subjects, and hence will automatically receive a strong interpretation. If we adopt Kratzers basic proposal, then stage-level and individual-level predicates will minimally project the structures in (52a) and (52b), respectively. (52) a. PredP 2 DPsu Pred 2 Pred b. XP 2 DPsu X 2 X PredP 2 Pred

The interpretive contrast illustrated by (42)-(45) above suggests that in ES, the associate is interpreted in its base (i.e., predicate-internal) position, an option which according to Kratzer is not made available by individual-level predicates.

21 While reconstruction appears to be the preferred option for A'-chains, reconstruction in Achains will sometimes give rise to binding-theoretic conflicts - see Chomsky (1993) for some discussion.

Expletives as Arguments: Germanic Existential Sentences Revisited 4.3 Expletives as arguments

23

In the following, we shall take a closer look at the syntactic status and semantic function of the expletive. Specifically, we will adopt and extend a suggestion by Ramchand (1996) to the effect that expletive there is an overt realisation of the argument of spatio-temporal location postulated by Kratzer (1995) (see Kiss 1996: 135 for a similar suggestion).22 There is independent evidence that the event or spatio-temporal argument is not only relevant to interpretation, but also has observable syntactic effects: Kratzer (1995) shows that, among other things, the presence versus absence of a spatio-temporal argument affects the grammaticality of sentences containing conditional clauses, is correlated with the possibility of subject-quantifier split and relative clause extraposition in German, and helps account for the (lack of) ambiguity of certain types of locative expression; Rapoport (1991) claims that event arguments play a crucial role in the licensing of adjunct predicates; Stowell (1991) analyses infinitival complements of so-called mental property adjectives as overt instantiations of the Kratzerian argument; Ramchand (1996) notes that the stage-level/individual-level distinction correlates with different positions for the subject DP in Scottish Gaelic, and Felser (1998) argues that the spatio-temporal argument is able to participate in anaphoric control relations. According to Ramchand, the spatio-temporal argument is to be distinguished from the lexically determined event variable originally proposed by Davidson (1967), which she suspects may be present in the argument structure of all predicates (compare e.g. Higginbotham 1985, 1989). She argues that the presence versus absence of the spatio-temporal argument is a matter of compositional structure-building rather than being determined by lexical properties of the predicate, and maintains that it is not correlated with the presence of the classical Davidsonian event variable (Ramchand 1996: 176-177). A smilar suggestion has been made by Heycock (1994: 234), who claims that lexical predicates are inherently individual-level. What allows them to be interpreted as stage-level is the possibility of them being predicated, not of the ostensible subject, but rather of an abstract Davidsonian event argument. Adding an event argument to an individual-level predicate then is analogous to adding an AGENT argument to an ergative predicate such as roll, yielding a ditransitive structure such as (53b) (compare Radford 1997: 367-370). (53) a. b. The ball rolled down the hill. John rolled the ball down the hill.

22 This idea is consistent with the observation that the element that functions as an expletive is often (though not necessarily) homophonous with a locative pronoun, indicating that they are diachronically related (cf. English there, Dutch er).

24

Claudia Felser & Laura Rupp

That is, addition of a spatio-temporal argument will turn an individual-level predicate like (54b) into a stage-level one as in (54b). (54) a. b. Mary was smart (i.e., she was a smart person). Mary was being smart (e.g. by making a "smart" remark).

Under this view, the spatio-temporal argument (henceforth labelled E) technically functions as the subject of stage-level predicates, as illustrated in (55) below (compare also Ramchand 1996: 178). (55) YP 2 E Y 2 Y PredP 2 DPsu Pred 2 Pred

For the sake of concreteness, and in line with suggestions made by Heycock (1994) and Felser (1998), let us identify Y in (54) with the functional head Asp(ect), which intervenes between T and V (or Pred), and which we take to be part of the extended predicate phrase. We further assume that it is AspP rather than VP or PredP that corresponds to the nuclear scope of the sentence, i.e. the domain to which existential closure applies (see Heycock 1994: 234 for a similar proposal). Our proposal is consistent with the observation that the spatio-temporal argument will normally be interpreted as being within the scope of sentential negation, and that it can be modified, for instance, by predicateinternal locative expressions.23 Observe that an ambiguous example like (56) below (from Kratzer 1995: 127) can be disambiguated in favour of the second reading - which results from associating the locative PP in diesem Bett 'in this bed' with the spatio-temporal argument rather than with the DP fast alle Flhe 'almost all fleas' - by inserting the affirmative particle ja before the PP (compare [57]). Particles of this type are taken to mark the left edge of the predicate phrase (see e.g. Diesing 1992: 31-33).

23 Alternatively, the functional category in question may be some light verb v, a Voice head of the type proposed by Borer (1993), or it may correspond to the head of Harley's (1995) Event Phrase. Sentential negation, which is syntactically represented by the functional head Neg, is generally taken to be located outside the predicate phrase, either above (Haegeman & Guron 1999) or immediately below T (Radford 1997, among others).

Expletives as Arguments: Germanic Existential Sentences Revisited

25

(56)

weil ihn fast alle Flhe in diesem Bett gebissen haben since him almost all fleas in this bed bitten have i. '...since almost all of the fleas in this bed bit him.' ii. '...since almost all the fleas bit him in this bed.' weil ihn fast alle Flhe ja in diesem Bett gebissen haben since him almost all fleas PRT in this bed bitten have '...since almost all the fleas bit him in this bed.'

(Ge)

(57)

(Ge)

Analysing Exp as a thematic or quasi-thematic subject correctly predicts the ungrammaticality of structures such as (58b), where the predicate nominal a solution is not saturated (in Rothsteins 1983 sense) by any subject. (58) a. b. There is a solution. *A solution is.

Compare also the grammatical A solution exists, where the subject a solution serves to saturate the one-place predicate exist. Our treatment of there differs from Groats (see section 2 above) in that the expletive is characterised as an argument expression. We further suggest that similar to the argument associated with the AGENT or CAUSER role in a layered VP system such as the one assumed by Chomsky (1995), it is -marked compositionally by the entire predicate phrase, including the logical subject. The spatio-temporal argument is like a quasi-argument in that it is thematic without referring to an actual participant in the action or event depicted - rather, in the spirit of Kratzer (1995), we may take it to be associated with an abstract location.24 If E lacks phonetic content, the thematic subject in EPP languages must raise overtly to the specifier of T, as indicated in (59) below.

24 We would normally expect a quasi-argument to be able to control PRO. This seems rather difficult to verify for expletives though. However, examples such as (ii) below (both [i] and [ii] have been suggested to us by Andrew Radford) could possibly be analysed as involving control by the expletive: (i) There do look like [ there are going to be casualties ]. (ii)There do look like [ being casualties ]. It may be that the expletive's incomplete -set renders it a poor controller generally.

26

Claudia Felser & Laura Rupp

(59)

TP ti DP T 5 ti a fly T AspP g ti was E AspP ti Asp PredP ti t Pred 6 in my soup

If, on the other hand, the spatio-temporal argument is spelled out as an expletive, the Minimal Link Condition requires that the expletive rather than the associate raises to (Spec,TP) to satisfy the EPP, as indicated in (60). (60) TP ti D T g ti there T AspP g ti was t Asp ti Asp PredP ti DP Pred 4 6 a fly in my soup

In this case, the presence of the expletive in (Spec,TP), and the fact that it checks both the nominative-assigning and the EPP feature of T, will block overt raising of the associate. Since the associate is quite literally "frozen in place" inside the predicate phrase, it is unable to escape existential binding at LF, and thus will obligatorily receive a weak reading. Number agreement will be checked via Agree, as suggested above. We may follow Kiss (1996) in assuming that overt expletives are inherently definite, which by Diesing's (1996) Scope Condition means that they must raise out of the predicate phrase. Definite or strong quantificational associates, on the other hand, are disallowed because

Expletives as Arguments: Germanic Existential Sentences Revisited

27

they may not normally remain inside the predicate phrase at LF, the domain to which existential closure applies.25

4.4

Summary

In sum, we have argued that the expletive is an overt realisation of the Kratzerian argument of spatio-temporal location, and as such contributes to sentence interpretation. The spatio-temporal argument serves to saturate, or "close off", stage-level predicates by virtue of being the last argument to be added. We have further assumed that Diesings (1992) mapping hypothesis holds, so that for the logical subject of stage-level predicates to receive a strong interpretation, it must appear outside the nuclear scope of the sentence (which we have identified with AspP rather than VP) at LF. If the spatio-temporal argument is realised overtly, as in existential sentences, it is interpreted as definite or "strong", and further affects semantic interpretation by precluding the possibility of the logical subject being interpreted in (Spec,TP), the external subject position. On the assumption that a predicate-internal subject position is made available by stage-level predicates only, and that only indefinite or weak DPs are eligible for existential binding, the proposed analysis accounts both for the definiteness restriction on the associate and for the stage-level restriction on the predicate.26

5 Are transitive expletive constructions a problem?

5.1

Transitive expletive constructions

Transitive expletive constructions (TECs) are normally ruled out in English and Mainland Scandinavian languages, but are attested in other Germanic languages

25 Under the minimalist copy theory of movement, the expletive should be able to optionally "reconstruct" in its original position, thereby permitting sentential negation or other types of sentential or aspectual operators to take scope over it. This would account, among other things, for the observation that the spatio-temporal argument appears to be within the scope of negation in sentences such as (i): (i) There wasn't a fly in my soup. 26 For lack of space, the question of why some languages should have overt realisations of the spatio-temporal arguments while others do not has not been addressed here. Chomsky (1995: 288-289) speculates that the presence versus absence of overt expletives may be determined by the phonological component, and correlated with the V2 requirement. As an anonymous reviewer points out, this correlation does not hold for all V2 languages though. Whether on the right track or not, the above hypothesis clearly requires further elaboration.

28

Claudia Felser & Laura Rupp

including German (Ge), Dutch (Du), and Icelandic (Ice) (examples from Bobaljik & Jonas 1996). (61) (62) *There has someone eaten an apple. *Der har nogen spist et ble. there has someone eaten an apple Someone has eaten an apple. Es essen einige Muse Kse in der Kche. there eat some mice cheese in the kitchen There are some mice eating cheese in the kitchen. Er heeft iemand een appel gegeten. there has someone an apple eaten Someone has eaten an apple. a hafa margir jlasveinar bora bing. there have many Christmas.trolls eaten pudding Many Christmas trolls have eaten pudding. (Danish)

(63)

(Ge) (= [33])

(64)

(Du) (= [34])

(65)

(Ice) (= [35])

Given the predicate-internal subject hypothesis, a natural assumption to make would be that the associate in TECs remains in its base position within VP. However, Vikner (1990), Zwart (1992), Vangsnes (1995) and Bobaljik & Jonas (1996) provide evidence that the thematic subject of TECs actually appears outside VP at Spell-Out. This can be seen from the fact that the raised subject may - and sometimes, must - precede shifted objects and certain aspectual or VP adverbs, as it does in the following examples. (66) a lauk einhver verkefninui alveg ti (Ice) there finished someone the.assignment completely Someone completely finished the assignment. (Jonas 1992, cited in Bobaljik & Jonas 1996) Es haben viele Leute das Buchi gestern ti gekauft. there have many people the book yesterday bought Many people bought the book yesterday. (Ge)

(67)

(68)

dat er veel mensen dat boeki gisteren ti gekocht hebben (Du) that there many people that book yesterday bought have that many people bought the book yesterday. (Zwart 1992)

Following Vangsnes (1995), we refer to the position occupied by the logical subject in (66)-(68) as the "intermediate" subject position. If we take the position in question to be a second specifier of T, as suggested by Chomsky (1995), then an example such as (69) below will have an internal structure along

Expletives as Arguments: Germanic Existential Sentences Revisited

29

the lines of (70) (note that the adverb gestern 'yesterday' is adjoined to AspP rather than VP, as AspP is the highest projection of the predicate in our system). (69) Es haben viele Leute gestern ein Buch gekauft. there have many people yesterday a book bought Many people bought a book yesterday. (Ge)

(70)

CP tu D C' g ti esi C TP g tu habenk ti' T ti DPj T' 6 tu viele Leute T AspP g tu tk ADV AspP g tu gestern ti Asp' tu Asp VP g rp tk tj ein Buch gekauft

Note that in V2 contexts, the expletive appears in the specifier of a functional head higher than T at Spell-Out, which we have identified with C.27 The ability to license two predicate-external subject positions is assumed to be a parameterised UG option (compare e.g. Bobaljik & Jonas 1996 "[Spec,TP] parameter", Thrinssons 1996 "split-IP parameter", and the discussion in Chomsky 1995: 371-376). The interpretive properties of the associate are discussed in the next section.

27 Languages that permit TECs differ with respect to the surface position of the expletive. It is traditionally assumed that German existential es resides in (Spec,CP) at Spell-Out (see Haider 1990, among many others). More recently, however, and within the context of Kayne's (1994) antisymmetry proposal, it has been argued that the expletive may actually be located in (Spec,TP) in German and Dutch ES (Zwart 1997). Although nothing crucial hinges on this question here, note that the latter assumption has trouble accounting for the fact that German DP-expletives (contrary to English there or Dutch er - see note 28) cannot appear embedded under a complementiser, or cooccur with topicalised or other Vorfeld constituents. See Vangsnes (1995), Bobaljik & Jonas (1996), or Thrinsson (1996) for further discussion and alternative proposals.

30 5.2

Claudia Felser & Laura Rupp TECs and the definiteness restriction

As noted by Bobaljik & Jonas (1996), the fact that the predicate-external associates in TECs are, apparently, subject to the definiteness restriction poses a problem for Diesings (1992) account for the distribution of (in)definites in the languages under consideration. However, the picture turns out to be more complex. Vangsnes (1995: 92) cites the following examples from Icelandic, which show that although definites are disallowed, other types of strong subjects such as partitives or universally quantified noun phrases are actually fine in TECs. (71) a. a hafa bir kettirnir/*kettirnir mnir ti msnar. there have both the.cats /*the.cats mine eaten the.mice Both cats/my cats have eaten the mice. a er srhver smokkur/*smokkurinn prfaur af RFSU. there is each condom / the.condom tested by RFSU Each condom/the condom is tested by RFSU.

b.

The situation is similar in German and Dutch:28 (72) Es haben beide/?alle/*meine Muse Kse gegessen. there have both / all / my mice cheese eaten All/both/my mice have eaten cheese. (Ge)

(73)

Er hebben ?twee van mijn vrienden/??alle studenten/*mijn ouders there have two of my friends / all students / my parents gisteren een boek gekocht. yesterday a book bought (Du)

Two of my friends/all students/my parents bought a book yesterday. Note that definite subjects of non-ES must appear in a predicate-external subject position (compare [74]-[76] below). The fact that definite subjects, though incompatible with expletives, are compatible with sentence-initial adverbials confirms that the expletive (or its trace) does indeed block the position that is normally targeted by a definite subject. Hence true expletives cannot, as has often been argued, originate in the specifier of CP or "topic" position (cf. Safir

28 In embedded (i.e., non-V2) contexts, Dutch er follows the complementiser, which we take to indicate that it occupies the outer specifier of T here: ? (i) dat er twee van mijn vrienden/??alle studenten/*mijn ouders that there two of my friends / all students / my parents

gisteren een boek hebben gekocht. yesterday a book have bought

Expletives as Arguments: Germanic Existential Sentences Revisited

31

1985b, among others) - even though they may appear in this position at SpellOut. (74) gr klruu (essar ms) sennilega (*essar ms) ostinn (Ice) yesterday finished (these mice) probably (these mice) the.cheese These mice probably finished the cheese yesterday. (Bobaljik & Jonas 1996) Hier essen die Muse den Kse gerade. here eat the mice the cheese just Here the mice are just eating the cheese. Gisteren hebben mijn ouders dat boek nog snel gekocht. yesterday have my parents the book PRT quickly bought Yesterday my parents quickly bought the book. (Ge)

(75)

(76)

(Du)

While predicate-internal associates in Dutch ES such as (77a) strongly favour a "weak" existential reading, a predicate-external associate (as in [77b]) will normally admit a partitive reading only (examples from De Hoop [1990: 285]). (77) a. b. Er waren gisteren enkele katten in de tuin. there were yesterday some cats in the garden Er waren enkele katten gisteren in de tuin. there were some cats yesterday in the garden

Similar interpretive differences can be found in German TECs: (78) a. b. Es haben gestern einige Muse Kse gegessen. there have yesterday some mice cheese eaten Es haben einige Muse gestern Kse gegessen. there have some mice yesterday cheese eaten

In (78a), the preferred reading for einige Muse some mice is the existential one, whereas in (78b) the associate favours a partitive reading. Notice further that bare indefinites are normally felt to be bad in the intermediate position (though for many speakers, the (b) examples below improve somewhat if the associate receives focal stress, according to Vangsnes 1995): (79) a. b. a hafa veri ms bakerinu. there have been mice in the.bath.tub
?

(Ice)

a hafa ms veri bakerinu. there have mice been in the.bath.tub (Vangsnes 1995: 93)

32

Claudia Felser & Laura Rupp

(80)

a.

Es haben oft Kinder auf der Strasse gespielt. there have often children on the street played Children often played in the street.
??

(Ge)

b. (81) a.

Es haben Kinder oft auf der Strasse gespielt. there have children often on the street played (Du)

Er hebben vaak kinderen op straat gespeeld. there have often children on.the street played Children often played in the street.
??

b.

Er hebben kinderen vaak op straat gespeeld. there have children often on.the street played

Moreover, observe that object DPs are also able to escape the definiteness restriction. (82) a hafa sennilega margir stdentar lesi bkina. (Ice) there have probably many students read the.book Many students have probably read the book (Bobaljik & Jonas 1996) a. Es hat ja eine Maus den/meinen Kse gegessen. (Ge) there has PRT a mouse the /my cheese eaten A mouse has eaten the/my cheese. Es hat ihn ja eine Maus (*ihn) gegessen. there has him PRT a mouse him eaten A mouse has eaten it/him. (Du)

(83)

b.

(84)

dat er veel mensen gisteren dat boek kochten. that there many people yesterday the book bought that many people bought the book yesterday.

If Diesing's (1996) prohibition against strong DPs remaining in the nuclear scope of a sentence at LF is correct, then we must assume that the definite objects in (82)-(84) are interpreted outside the predicate phrase. Object raising has taken place overtly in (83b), where the object pronoun ihn 'him' precedes both the thematic subject and the mood particle ja.29 In sum, closer examination of the data reveals that the definiteness effects found in TECs are considerably weaker than those found in intransitive or unaccusative ES. The above observations are summarised in (i)-(iv) below:

29 If Sabel (1999) is correct in claiming that covert object scrambling is not possible, there must be some other mechanism that allows objects to escape existential binding. See Diesing (1992, 1996) for some discussion of the interpretive properties of objects.

Expletives as Arguments: Germanic Existential Sentences Revisited

33

i. Definite subjects are normally excluded from TECs, whereas universally quantified NPs or strong indefinites are (marginally) permitted. ii. Strong associates preferably occur in the intermediate subject position (i.e., outside the predicate phrase). iii. Bare indefinites and other types of weak indefinite subject preferably appear predicate-internally. iv. Object DPs do not observe the definiteness restriction. Taken together, these observations appear to be essentially consistent with Diesings (1992) theory of indefinites, particularly if this is made more flexible by combining it with the minimalist copy theory of movement. DPs that are interpreted as weak are those which remain (or reconstruct) in a predicateinternal position, whereas indefinite or quantificational expressions that are interpreted as strong appear in a predicate-external position at LF, thereby escaping existential closure. Consider, for instance, the German example in (85) below, and the corresponding analysis in (86). (85) Es haben (einige Muse) ja doch (einige Muse) Kse gegessen. there have some mice PRT PRT some mice cheese eaten Some (of the) mice have indeed eaten cheese. [CP esi [C haben [TP ti' [T' (einige Muse) [T T [AspP ti ... PARTITIVE [VP ja doch [VP (einige Muse) Kse gegessen ]]]]]] EXISTENTIAL Depending on which copy of einige Muse some mice is interpreted at LF, the DP will receive an existential or a partitive reading. As noted by Diesing (1992, 1996), DPs in the languages under consideration show a fairly strong tendency to appear in their LF position at Spell-Out. Object DPs escape the DR because there is no "object" equivalent of a subject expletive that could prevent them from being interpreted in a predicate-external position. One important question, however, has not yet been answered: Given that TECs make two external subject positions available, why can the associate not be an ordinary (strong) definite? Chomsky (1995) suggests that in languages which allow TECs, the EPP feature carried by T may escape erasure once when checked, thus licensing a second specifier. This assumption does not, however, provide any mechanism for excluding definite subjects. As we saw above, the prohibition against definites appears to result from the fact that on the one hand, definites must raise out of the predicate phrase, while on the other hand, the position that normally hosts a definite subject is already occupied by the expletive (or its trace). Following on from the discussion in section 3, let us assume that definites differ from both indefinites and certain types of

(86)

34

Claudia Felser & Laura Rupp

quantificational DP in that they carry an interpretable [person] feature, which must be checked against a corresponding uninterpretable feature in an inflectional head.30 However, as the [person] feature of T is checked by the expletive, true expletives are incompatible with associates that carry a [person] feature themselves. Hence, definite subjects will normally be excluded from existential sentences, including TECs. Indefinites, as well as certain types of quantifier, by contrast, whilst capable of carrying [number] and [gender] features, are not specified for [person] (compare Lyons 1999). In TEC languages, then, associate raising is made possible by the presence of a "double" EPP feature, but as in non-TEC languages, the associate will only check the [number] feature of T.

6 Concluding Remarks We set out to review some previous accounts for the syntactic and semantic characteristics of existential sentences, including the fact that number agreement is with the associate rather than with the expletive, and the requirement that the associate must normally be a weak indefinite. We have presented evidence against the idea that the associate is assigned partitive case by the verb, and suggested instead that it receives default case. True expletives serve to check both the EPP and the uninterpretable [person] feature of T, and are assigned structural nominative. The associate checks number agreement by entering into a partial abstract agreement relationship with T. We have further shown that the presence of an expletive at LF does not pose a problem for Full Interpretation as the expletive has important contributions to make to semantic interpretation: It is a definite expression that constitutes an overt realisation of the spatio-temporal argument provided by stage-level predicates, and serves the function of preventing the associate from being interpreted in the external subject position (in non-TEC languages). In other words, use of an expletive is not "optional"; rather, by referring to some definite (albeit abstract) spatio-temporal location, overt subject expletives explicitly signal that existence is predicated of some DP (or of the event described) in the coda of an ES. Together with the Kratzer/Diesing theory of indefinites, our analysis accounts for the definiteness effects characteristically found in existential sentences, and for the fact that individual-level predicates are excluded from genuine existentials. We further argued that contrary to what has been suggested elsewhere, transitive expletive constructions do not provide any genuine counter-evidence to Diesings (1992) theory of indefinites. The

30 Under this view, checking of [person] features can be seen as the morphosyntactic correlate of Diesings (1996) Scope Condition, which forces referential definites to raise out of VP overtly in "scrambling" languages like German.

Expletives as Arguments: Germanic Existential Sentences Revisited

35

availability of a second predicate-external subject position in transitive expletive sentences makes it possible for the logical subject to escape existential binding. The fact that definite - but not strong indefinite or quantificational - associates are disallowed even in transitive expletive constructions has been attributed to a requirement on definite subjects to the effect that they must obligatorily check their [person] feature against a corresponding feature of T - which, however, is checked by the inherently definite expletive.

References
Abbott, B. (1994): "Referentiality, specificity, strength and individual concepts". Proceedings of the Twelfth West Coast Conference in Formal Linguistics, 473-484. Abbot, B. (1995): "Some remarks on specificity". Linguistic Inquiry 26, 341-347. Atkinson, M. (1999): "There seem to be likely to be some problems". Ms. University of Essex. Belletti, A. (1988): "The case of unaccusatives". Linguistic Inquiry 19, 1-34. Bobaljik, J. & D. Jonas (1996): "Subject positions and the roles of TP". Linguistic Inquiry 27, 195-236. Borer, H. (1984): Parametric Syntax. Dordrecht: Foris. Borer, H. (1993): "The projection of arguments". In: E. Benedicto and J. Runner, Eds., Functional Projections (UMass Occasional Papers 17). Amherst, Mass.: GLSA. Bowers, J. (1993): "The syntax of predication". Linguistic Inquiry 24, 591-656. Breivik, L. E. (1981): "On the interpretation of existential there". Language 57, 1-25. Cardinaletti, A. (1997): "Agreement and control in expletive constructions". Linguistic Inquiry 28, 521-533. Carlson, G. (1980): Reference to Kinds in English. New York: Garland. Chomsky, N. (1982): Some Concepts and Consequences of the Theory of Government and Binding. Cambridge, Mass.: MIT Press. Chomsky, N. (1986): Knowledge of Language. New York: Praeger. Chomsky, N. (1991): "Some notes on economy of derivation and representation". In: R. Freidin, Ed., Principles and Parameters in Comparative Grammar. Cambridge, Mass.: MIT Press, 417-454. Chomsky, N. (1993): "A minimalist program for linguistic theory". In K. Hale and S. Keyser, Eds., The View from Building 20. Cambridge, Mass.: MIT Press, 1-52. Chomsky, N. (1995): The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, N. (1998): Minimalist Inquiries: The Framework. MIT Occasional Papers in Linguistics 15. Cambridge, Mass.: MITWPL Clahsen, H., S. Bartke, and S. Eisenbeiss (2000): "Case marking and agreement in Germanspeaking SLI children: Evidence for a selective linguistic impairment". Ms. University of Essex. Comorovski, I. (1991): "Partitives and the definiteness effect". Proceedings of WCCFL 10, 91-102. Davidson, D. (1967): "The logical form of action sentences". In: N. Rescher, Ed., The Logic of Decision and Action. Pittsburgh: University of Pittsburgh Press, 85-95. Diesing, M. (1992): Indefinites. Cambridge, Mass.: MIT Press. Diesing, M. (1996): "Semantic variables and object shift". In: H. Thrinsson, S. Epstein and S. Peter, Eds., Studies in Comparative Germanic Syntax II. Dordrecht: Kluwer, 66-84. Dikken, M. den (1995): "Binding, expletives, and levels". Linguistic Inquiry 26, 347-354. En, M. (1991): "The semantics of specificity". Linguistic Inquiry 22, 1-26.

36

Claudia Felser & Laura Rupp

Felser, C. (1998): "Perception and control: A minimalist analysis of English direct perception constructions". Journal of Linguistics 34, 351-385. Groat, E. (1995): "English expletives: a minimalist approach". Linguistic Inquiry 26, 354-365. Haegeman, L. & J. Guron (1999): English Grammar. Oxford: Blackwell. Haider, H. (1990): "Null subjects and expletives in Romance and Germanic languages". In W. Abraham et al., Eds., Issues in Germanic Syntax. Tbingen: Narr, 49-66. Harley, H. (1995): Subjects, Events, and Licencing. MIT Dissertation, Cambridge, Mass. Heim, I. (1982): The Semantics of Definite and Indefinite Noun Phrases. PhD Dissertation, University of Massachusetts, Amherst. Heinmki, O. (1984) "Aspect in Finnish". In: C. de Groot and H. Tommola, Eds., Aspect Bound. A Voyage into the Realm of Germanic, Slavonic, and Finno-Ugrian Aspectology. Dordrecht: Foris, 153-178. Heycock, C. (1994): "The internal structure of small clauses: new evidence from inversion". Proceedings of NELS 25, 223-238. Higginbotham, J. (1985): "On semantics". Linguistic Inquiry 16, 547-593. Higginbotham, J. (1989): "Elucidations of meaning". Linguistics and Philosophy 12, 463-517. Hoekstra, T. &. R. Mulder (1990): "Unergatives as copular verbs; locational and existential predication". The Linguistic Review 7, 1-79 Hoop, H. de (1990): "Restrictions on existential sentences and object scrambling: some facts from Dutch". Proceedings of WCCFL 9, 277-288. Hoop, H. de (1996): Case Configuration and Noun Phrase Interpretation. New York & London: Garland. Hornstein, N. (1995): Logical Form. Oxford: Blackwell. Jang, Y. (1997): "Minimal feature-movement". Journal of Linguistics 33, 311-325. Jenkins, L. (1975): "The English Existential". Tbingen: Narr. Jonas, D. (1992): "Checking theory and nominative case in Icelandic". Harvard Working Papers in Linguistics 1. Department of Linguistics, Harvard University, Cambridge, Mass., 175-195. Jonas, D. (1996) "Clause structure, expletives and verb movement". In: W. Abraham et al., Eds., Minimal Ideas. Amsterdam: John Benjamins, 167-188. Kimball, J. (1973): "The grammar of existence". Papers from the Ninth Regional Meeting Chicago Linguistic Society, 262-270. Kiss, K. (1996): "Two subject positions in English". The Linguistic Review 13, 119-142. Koopman, H. & D. Sportiche (1991): "The position of subjects". Lingua 85, 211-258. Kratzer, A. (1995): "Stage-level and individual-level predicates". In: G. Carlson & F. J. Pelletier, Eds., The Generic Book. Chicago: University of Chicago Press., 125-175. Landau, I. (1999): Elements of Control. MIT Dissertation, Cambridge, Mass. Larson, R. (1988): "On the double object construction". Linguistic Inquiry 19, 335-392. Lasnik, H. (1992): "Case and expletives: notes toward a parametric account". Linguistic Inquiry 23, 381-405. Lasnik, H. (1995): "Case and expletives revisited: on Greed and other human failings". Linguistic Inquiry 26, 615-633. Lenerz, J. (1985): "Zur Theorie syntaktischen Wandels: das expletive es in der Geschichte des Deutschen". In: W. Abraham, Ed., Erklrende Syntax des Deutschen. Tbingen: Narr, 99-136. Lumsden, M. (1988): Existential Sentences: Their Structure and Meaning. London: Routledge. Lyons, C. (1999): Definiteness. Cambridge: Cambridge University Press. Milsark, G. (1974): Existential Sentences in English. Ph.D. Dissertation, Cambridge, Mass. [reproduced by the Indiana Linguistics Club, Bloomington/Indiana]. Milsark, G. (1977): "Peculiarities of the existential construction in English". Linguistic Analysis 3, 1-29.

Expletives as Arguments: Germanic Existential Sentences Revisited

37

Moro, A. (1997): The Raising of Predicates. Cambridge: Cambridge University Press. Ouhalla, J. (1991): Functional Categories and Parametric Variation. London: Routledge. Radford, A. (1997): Syntactic Theory and the Structure of English. Cambridge: Cambridge University Press Ramchand, G. (1996): "Two subject positions in Scottish Gaelic: the syntax-semantics interface". Natural Language Semantics 4, 165-191. Rapoport, T. (1991): "Adjunct-predicate licensing and D-structure". In: S. Rothstein, Ed., Perspectives on Phrase Structure: Heads and Licensing (Syntax and Semantics 25). San Diego: Academic Press, 159-187. Reuland, E. (1983): "On the subject of nonargument subjects". In: W. Abraham, Ed., On the Formal Syntax of the Westgermania. Amsterdam: John Benjamins, 1-46. Reuland, E. & A. ter Meulen, Eds. (1987): The Representation of (In)definiteness. Cambridge, Mass.: MIT Press. Rothstein, S. (1983): The Syntactic Forms of Predication. MIT dissertation, Cambridge, Mass. Sabel, J. (1999): "Das Passiv im Deutschen: Derivationale konomie versus optionale Bewegung". Linguistische Berichte 177, 87-112. Safir, K. (1985a): Syntactic Chains. Cambridge: Cambridge University Press. Safir, K. (1985b): "Missing subjects in German". In: J. Toman, Ed., Studies in German Grammar. Dordrecht: Foris, 193-229. Safir, K.(1993): "Perception, selection, and structural economy". Natural Language Semantics 2, 47-70. Schtze, C. (1993): "Towards a minimalist account of quirky case & licensing in Icelandic". In: C. Phillips, Ed., Papers on Case and Agreement II (MIT Working Papers in Linguistics 19). Cambridge, Mass.: MITWPL, 321-376. Schtze, C. (1997): INFL in Child and Adult Language: Agreement, Case, and Licensing. MIT dissertation, Cambridge, Mass. Schtze, C. (1999): "English expletive constructions are not infected". Linguistic Inquiry 30, 467-484. Stowell, T. ( 1983): "Subjects across categories". The Linguistic Review 2, 285-312. Stowell, T. (1991): "The alignment of arguments in adjectival phrases". In: S. Rothstein, Ed., Perspectives on Phrase Structure: Heads and Licensing (Syntax and Semantics 25). San Diego: Academic Press, 105-135 Thrinsson, H. (1996): "On the (non-)universality of functional categories". In: W. Abraham et al., Eds., Minimal Ideas. Amsterdam: John Benjamins, 253-281. Vainikka, A. & J. Maling (1996): "Is partitive case inherent or structural?" In: J. Hoeksema, Ed., Partitives: Studies on the Syntax and Semantics of Partitive and Related Constructions. Berlin: Mouton de Gruyter, 179-208. Vangsnes, . A. (1995): "Referentiality and argument positions in Icelandic". Working Papers in Scandinavian Syntax 55, 89-109. Vikner, S. (1990): Verb Movement and the Licensing of NP-Positions in the Germanic Languages. Ph.D. Dissertation, Universit de Genve. Zwart, C. J.-W. (1992): "Dutch expletives and small clause predicate raising". Proceedings of NELS 22, 477-491. Zwart, C. J.-W. (1994): "Dutch is head initial". The Linguistic Review 11, 377-406. Zwart, C. J.-W. (1997): Morphosyntax of Verb Movement: A Minimalist Approach to the Syntax of Dutch. Dordrecht: Kluwer.

Colchester

Claudia Felser & Laura Rupp

Department of Language and Linguistics, University of Essex, Colchester CO4 3SQ, U.K.

You might also like