One of the basic premises of the theory of syntax is that clause structures can be minimally identified as containing a verb phrase, playing the role of predicate, and a noun phrase, playing the role of subject. In this study Andrea Moro identifies a new category of copular sentences, namely inverse copular sentences, where the noun phrase which co-occurs with the verb phrase plays the role of predicate, occupying the position which is canonically reserved for subjects, and the subject is embedded in the verb phrase. The consequences of such a discovery are pervasive. Four distinct areas of syntax are unified into a unique natural class. Along with inverse copular sentences, existential sentences, sentences with seem and unaccusative constructions are analysed as involving the raising of a predicative noun phrase to the most prominent position in the clause structure. In addition, new light is shed on some classical issues such as the distribution and nature of expletives, locality theory, cliticization phenomena, possessive constructions and the cross-linguistic variations of the Definiteness Effect.
CAMBRIDGE STUDIES IN LINGUISTICS General Editors: s. R. ANDERSON, J. BRESNAN, B. COMRIE, W. DRESSLER, C. EWEN, R. HUDDLESTON, R. LASS, D. LIGHTFOOT, J. LYONS, P. H. MATTHEWS, R. POSNER, S. ROMAINE, N. V. SMITH, N. VINCENT
The raising of predicates
In this series
58 MONIK CHARETTE: Conditions on phonological government 59 M. H. KLAIMAN: Grammatical voice 60 SARAH M. B. FAGAN: The syntax and semantics of middle constructions: a study with special reference to German 61 ANJUM P. SALEEMI: Universal Grammar and language learnability 62 STEPHEN R. ANDERSON: A-morphous morphology 63 LESLEY STIRLING: Switch reference and discourse representation 64 HENK J. VERKUYL: A theory of aspectuality: the interaction between temporal and atemporal structure 65 EVE V. CLARK: The lexicon in acquisition 66 ANTHONY R. WARNER: English auxiliaries: structure and history 67 P. H. MATTHEWS: Grammatical theory in the United States from Bloomfield to Chomsky 68 LJILJANA PROGOVAC: Negative and positive polarity: a binding approach 69 R. M. W. DIXON: Ergativity 70 YAN HUANG: The syntax and pragmatics of anaphora 71 KNUD LAMBRECHT: Information structure and sentence discourse referents 72 LUIGI BURZIO: Principles of English stress 73 JOHN A. HAWKINS: A performance theory of order and constituency 74 ALICE C. HARRIS and LYLE CAMPBELL: Historical syntax in cross-linguistic perspective 75 LILIANE HAEGEMAN: The syntax of negation 76 PAUL GORRELL: Syntax and parsing 77 GUGLIELMO CINQUE: Italian syntax and Universal Grammar 78 HENRY SMITH: Restrictiveness in case theory 79 D. ROBERT LADD: Intonational phonology 80 ANDREA MORO: The raising of predicates: predicative noun phrases and the theory of clause structure Supplementary volumes MICHAEL O SIADHAIL: Modern Irish: grammatical structure and dialectal variation ANNICK DE HOUWER: The acquisition of two languages from birth: a case study LILIANE HAEGEMAN: Theory and description in generative syntax: a case study in West Flemish A. E. BACKHOUSE: The lexical field of taste: a semantic study of Japanese taste terms NIKOLAUS RITT: Quantity adjustment: vowel lengthening and shortening in early Middle English Earlier issues not listed are also available
THE RAISING OF PREDICATES Predicative noun phrases and the theory of clause structure
ANDREA MORO Istituto Scientifico H. San Raffaele, Milan
CAMBRIDGE
UNIVERSITY PRESS
Published by the Press Syndicate of the University of Cambridge The Pitt Building, Trumpington Street, Cambridge CB2 1RP 40 West 20th Street, New York, NY 10011-4211, USA 10 Stamford Road, Oakleigh, Melbourne 3166, Australia © Cambridge University Press 1997 First published 1997 A catalogue record for this book is available from the British Library Library of Congress cataloguing in publication data Moro, Andrea The raising of predicates: predicative noun phrases and the theory of clause structure / Andrea Moro. p. cm. - (Cambridge studies in linguistics: 80) Includes bibliographical references and index. ISBN 0 521 56233 3 1. Grammar, comparative and general - clauses. 2. Grammar, comparative and general - syntax. 3. Grammar, comparative and general - verb phrase. I. Title. II. Series. p297.M67 1997 415-dc20 96-7773 CIP ISBN 0 521 56233 3 hardback Transferred to digital printing 2004
Contents
Acknowledgments
1 1.1 1.2 1.3 1.4 1.5 2 2.1 2.2 2.3 2.4 2.5 3 3.1 3.2 3.3 3.4 3.5
page ix
Introduction: four apparently unrelated empirical domains
1
The anomaly of copular sentences: the raising of predicates Subject-object asymmetries and the theory of clause structure The anomaly of copular sentences: the data The raising of predicates: a unified theory of copular sentences Exploring the syntax of predicative noun phrases: 6-role assignment and the local domain for binding Conclusions
17
The syntax of ci The standard analysis: on the notion of expletive Extraction from the embedded subject of c/-sentences and other unexpected phenomena An alternative analysis: ci as a raised predicate A difference between English there and Italian ci Conclusions Are there parameters in semantics? The defining properties of existential sentences Introduction: on the absence of the Definiteness Effect in Italian existential sentences On the so-called 'existential meaning': the role of syntax How to escape the Definiteness Effect: the role of the /?rodrop parameter A cursory view across languages: 'Jespersen's generalization' Conclusions
17 23 30 77 89 94 94 103 107 118 127
131 131 133 150 159 163 vii
viii
Contents
4. 4.1 4.2 4.3 4.4 4.5
The 'quasi-copula': the role of finite clauses in seem-sentences Finite clauses as arguments: passive vs seem Seem as the 'quasi-copula' On differentiating the role of it: empirical evidence Infinitival clauses with seem Conclusions
167 167 170 180 194 210
5 5.1 5.2 5.3
A view beyond: unaccusativity as an epiphenomenon On the notion of intransitive verbs A view beyond: unaccusative constructions as d-sentences Conclusions
214 214 222 245
i I1 I11
Appendix: a brief history of the copula The meaning(s) of 'copula' A brief survey of three major approaches 'Grammatical subjects': copular sentences in modern syntax
248 248 248 255
Notes References Index
262 301 310
Acknowledgments
Ideas are never developed in isolation; they always require the stimulation and the support of a community of people. In my case, the intellectual and emotional debt I have accumulated since I embarked on this project is enormous. Among the persons and institutions who have contributed to this work in different forms and at various times, I would like to thank: Giorgio Graffi, who allowed me to make myfirststeps in linguistics at the University of Pavia; Guglielmo Cinque and Giuseppe Longobardi at the University of Venice, who followed my Ph.D. programme; Noam Chomsky and James Higginbotham at MIT, where I spent a considerable time as a visiting scholar, partially supported by a Fulbright grant; Luigi Rizzi and Liliane Haegeman at the University of Geneva, where I took the Diplome d'Etudes Superieures. I would also like to express my gratitude to Luigi Burzio, Gennaro Chierchia, Marcel Den Dikken, Jacqueline Gueron, Teun Hoekstra, Richard Kayne, Anthony Kroch, Alessandra Tomaselli and Maria Rita Manzini for their generous attention to my work and to Peter Taylor for having translated my quasi-English into a natural language; special thanks to my (extended standard) family for having shared with me the catastrophic belief that copular sentences are the centre of the universe. The original idea of inverse copular sentences (discussed in the first three chapters of the present book) was presented in its first published form in 1988 (Moro 1988), and, during 1990, was presented at the Cambridge GLOW Conference in a talk entitled 'there-raising: principles across levels'; at the Syntax Lunch Seminar at the CUNY Graduate Center in New York City; and at the Penn Linguistics Club at the University of Pennsylvania in Philadelphia. The last part of chapter 1, where predicative nominals are analysed, was presented as a separate piece of research at a University of Geneva 'Seminaire de recherche' in 1993. The analysis set out in chapter 4 was presented in talks given at the Johns Hopkins University in Baltimore in 1991; at a University of Geneva 'Seminaire de recherche' in 1995; and at the 'Arbeitsgruppe Strukturelle Grammatik an der Humboldt-Universitat' at Max Planck Gesellschaft in Berlin in 1995. The content of the last chapter ix
x
Acknowledgments
was presented at the University of Lisbon GLOW Conference in 1992 in a talk entitled 'A syntactic decomposition of a lexical primitive: the unaccusativity effect'; at a University of Geneva 'Seminaire de recherche' in 1992; as well as at the first 'Langues et Grammaire' conference at the University of Paris 8 in 1994; and the section concerning possessive constructions was presented at the Societas Linguistica Europaea twentyeighth annual meeting in Leiden in 1995. It goes without saying that I also benefited by the various formal and informal presentations I had the opportunity of giving at the universities of Pavia, Venice and Padua, as well as MIT. Two preliminary versions of the theory presented here have been circulating as Moro (1991) and Moro (1993a). Finally, I would like to express my deep gratitude to Massimo Piattelli Palmarini, who encouraged me to undertake the final revision of this work while we were working together at the foundation of the Dipartimento di Scienze Cognitive at the Istituto Scientifico H. San Raffaele in Milan, and to Don Luigi Maria Verze for having believed in linguistics.
Introduction: four apparently unrelated empirical domains
In this work, we will explore four distinct empirical domains in detail and propose a unified treatment for them. I will begin by briefly illustrating some central aspects of these domains; for the moment the presentation will be kept at a rather informal level. 1 i
The anomaly of copular sentences
A well-established assumption concerning the structure of clauses is that in a sequence of the type noun phrase - verb - noun phrase such as the following, the two noun phrases have different syntactic properties: (1)
[NP a picture of the wall] revealed [NP the cause of the riot]
This assumption does not follow from any logical necessity; instead it is based on empirical evidence. To illustrate this we will consider the following simple case of extraction of an interrogative from a noun phrase: (2a) b
* [which wall]j did [NP a picture of tj] reveal [NP the cause of the riot]? [which riot]i did [NP a picture of the wall] reveal [NP the cause of tj]?
As we see, extraction from the preverbal noun phrase yields an ungrammatical sentence, while extraction from the postverbal noun phrase is successful. To prove that the ungrammaticality of (2a) is due to the position of the noun phrase from which extraction originates rather than to some other factor in the internal structure of the noun phrase involved, it is sufficient to observe that if [a picture of the wall] is placed in the same position as [the cause of the riot] in (2a), extraction becomes fully acceptable: (3a) b
[NP the inquiry] revealed [NP a picture of the wall] [which wall]i did [NP the inquiry] reveal [NP a picture of tj?
In the traditional representation that goes back to the first pioneering works in syntax this difference is implemented in terms of 'topological' asymmetry in the geometry of tree representation. The following simple formal representation will be sufficient for our purposes: 2 1
2 (4)
Introduction: four apparently unrelated empirical domains S
Using this structure, we can quite easily derive the phenomena we are considering here. Many independent facts have led to the generalization that extraction from a phrase is not possible if the phrase is in a left-branch position.3 Since only the preverbal noun phrase is in a left-branch position in this minimal clause structure, the contrast in (2) will follow. Of course, there are many further empirical data that support this asymmetry, such as, for example, verb agreement which reveals that inflection is sensitive only to the features of the higher noun phrase. Indeed one major contribution to the theory of clause structure stemming from this approach is that grammatical functions, like subject, predicate and the like, can be immediately derived from the configuration: thus, the subject of the predication is the 'most prominent' NP, i.e. the NP immediately dominated by the clausal node S; the object is the NP immediately dominated by the VP node, and the predicate would be the VP itself. Thus, the whole cluster of phenomena that yield differences between a preverbal and postverbal NP are generally called 'subject-object asymmetries'. The clause structure represented in (4) and the asymmetry that is associated with it is generally held to apply invariantly to all instances of clauses. In principle, there seems to be no reason to restrict it to a particular subclass. However, if we consider copular sentences, that is to say, those sentences where the main verb is be, several problems arise. Let us begin by considering the following case: (5)
[NP a picture of the wall] was [NP the cause of the riot]
It is easy to show that in this case, the usual subject-object asymmetries are fully preserved. Consider, for example, the following contrast: (6a) b
* [which wall]; wasj [NP a picture of ts] tj [NP the cause of the riot]? [which riotjj wasj [NP a picture of the wall] tj [NP the cause of tj]?
Extraction from the preverbal noun phrase (in a left-branch position) is impossible whereas extraction from the postverbal noun phrase is fully acceptable. So far, then, the standard asymmetric representation has been seen to hold when V is be:
The anomaly of copular sentences 3 (7)
But let us now examine another sentence which contains exactly the same lexical elements as a picture of the wall was the cause of the riot but where they are arranged in a different way: (8)
[Np the cause of the riot] was [NP a picture of the wall]
On the face of it, there seems to be no reason why we should not apply exactly the same analysis to this sentence and thus expect the same subjectobject asymmetries to show up. Surprisingly, however, if we apply the usual test used to detect subject-object asymmetry, we find that the results are bad in both cases: (9a) b
* [which riot]; wasj [NP the cause of tj] t, [NP a picture of the wall]? * [which wall]j wasj [NP the cause of theriot]tj [NP a picture of tj]?
Extraction is impossible not only from the preverbal noun phrase but from the postverbal noun phrase as well. More explicitly, the postverbal noun phrase of this sentence does not behave like an object. How can we explain this fact? Clearly, we cannot simply maintain the traditional representation given to the clause structure and stipulate that the copula does not allow extraction from the postverbal noun phrase. This assumption would go against the fact that there are cases where extraction is indeed possible from a postverbal noun phrase of a copular sentence, as in the example in (6b). A simple question then arises: why is there a difference between a picture of the wall was the cause of the riot and the cause of the riot was a picture of the wall? I will attempt to show that the answer to this question implies a revision of the standard theory of clause structure. In this book, a broad range of phenomena similar to those illustrated here will be analysed in detail and a principled theory will be proposed to account for them. I will now briefly summarize the other related issues that will be explored.
4
Introduction: four apparently unrelated empirical domains
ii
A case study in comparative linguistics: on the subject of existential sentences
A second empirical domain will be investigated in the present study. A common assumption is that, in both English and Italian, the following pairs of sentences share essentially the same structure: (10a) b c d
a man is in the room there is a man in the room un uomo e nella stanza c'e un uomo nella stanza
The elements in preverbal position in (10b) and (lOd), namely there and d, are regarded as semantically null entities. Their role is said to be that of place holders (technically 'expletives') which stand for the subject of the predication {a manlun uomo) when the latter is in postverbal position. The obligatoriness of such elements is traced back to a general principle of clause structure that requires subjects of the predication to be realized, namely the Extended Projection Principle. Thus, the standard schema applied to clause structure holds for these clauses as well, provided that the expletive element is inserted in the most prominent position: (11)
there c'
is a man in the room e un uomo nella stanza
The analogy between the two languages is only partial, though. A wellknown semantic restriction applies in English to the noun phrase following the copula in a sentence like (10b). Thus, this noun phrase cannot be 'definite', in the way that proper names or noun phrases introduced by a definite article are considered to be. The following exactly analogous examples illustrate a classic violation of this semantic restriction in English and show that the same restriction is not valid for Italian: (12a) * there is John in the room b c'e Gianni nella stanza This contrast immediately raises at least two conceptually distinct sets of questions. The first set centres on the comparative problem: what kind of evidence might be expected to trigger this cross-linguistic variation? Is a
The anomaly of seem
5
semantic restriction sensitive to parametric variation? How can we account for such a contrast in terms of language acquisition? A second set of questions can be formulated within the analysis of a single language: if there and ci are semantically null elements why is it that only those copular sentences in which these elements show up are considered to be 'existential sentences' as opposed to the other types? More radically, what are the defining properties of an existential sentence? To answer these questions we will explore the syntax of ci and there in detail, using original empirical tests. For the moment I will briefly illustrate some central data on which the argument will be based. Consider the following simple examples taken from Italian: (13a) b c d e
Gianni e uno scienziato (Gianni is a scientist) c'e uno scienziato (there is a scientist) Gianni IOJ e tj (Gianni clitic-is) * ce IOJ e tj (there clitic-is) ce n'j e uno tj (there of-them is one)
In the first sentence (13a), where the subject position is occupied by a lexical noun phrase, the postverbal noun phrase can freely undergo cliticization (13c). If ci occurs (13b), /o-cliticization is completely blocked and gives rise to a strong effect of ungrammaticality (13d). Why should a semantically null element block movement? This is puzzling; indeed, as the last example shows, one cannot simply stipulate that ci blocks cliticization tout court. In fact, ne-cliticization is perfectly possible in such types of sentence (13e). In this book, a new approach to the distribution of there and ci will be proposed, which, hopefully, will overcome many of the empirical problems such as the one just illustrated and provide answers to both sets of questions raised in this introduction.
HI
The anomaly of seem
A third kind of problem comes from those constructions which involve finite clauses as complements. It is a well established assumption that in passive constructions the object NP is raised to the subject position, while preserving the 'semantic' role it had in postverbal position:
6
Introduction: four apparently unrelated empirical domains
(14a) b
John read [the book] [the book]; was read tx (by John)
If the object is a finite clause, we find the following two options for passive constructions: (15a) b
[that John left]; was affirmed tj it was affirmed [that John left]
This paradigm is traditionally interpreted by assuming that either the embedded clause is raised to subject position (paralleling the case of object NPs) as in (15a), or it remains in situ as in (15b) and some kind of 'semantically null element' like preverbal there (i.e. it) is inserted in subject position. In formal terms, we obtain the following two structures: (16a)
was affirmed
was affirmed
t.
that John left
As in the case of there, the subject position cannot be left empty, because this would contradict the general principle that requires clauses to have a subject of the predication. There is one important exception to the general pattern that we have just illustrated; if the verb is seem no raising of the complement is possible on the analogy of passive structures. Thus, the only option is for it to be realized in subject position:
Remarks on unaccusatives 7 (17a)
seems
The very fact that it cannot be replaced seems somewhat paradoxical. Indeed, there has been a long-standing debate about why a semantically null element should be obligatory. What kind of explanation should we seek for this? In this book this problem will be reconsidered in the light of one of the central results of our analysis of copular sentences.
IV
Remarks on unaccusatives
In the final part of this study, we will examine an area of syntax that has recently yielded some important advances with implications for the entire field. In the late seventies, the traditional notion of 'intransitive verb' turned out to be inadequate from a descriptive point of view. In Italian, three purely empirical tests, namely ne-extraction from the postverbal subject, auxiliary selection and past participle agreement, show that intransitive verbs do not constitute a homogeneous class: specifically, if an intransitive verb allows extraction of the clitic ne from the postverbal subject then it selects the auxiliary essere (be) and its past participle agrees in gender and number with the subject. These three phenomena go together in the sense that if one property holds, the others will also hold. The following contrasts between arrivare ('arrive') and telefonare ('telephone') illustrate this distinction:
8
Introduction: four apparently unrelated empirical domains
(18a) b c
d e f
[molti ragazzi] sono/*hanno arrivati (many boys are/have arrived-plur.) 'many boys arrived' [molti ragazzi] sono arriva-ti/-*to (many boys are arrived-plur./-indef.) 'many boys arrived' nej sono arrivati [molti tj] (of-them are arrived many) 'many of them arrived' [molti ragazzi] hanno/*sono telefonato (many boys have/are telephoned-indef.) 'many boys phoned' [molti ragazzi] hanno telefona-to/*-ti (many boys have telephoned-indef./-plur.) 'many boys phoned' * nej hanno telefonato [molti t{] (of-them have phoned many)
This basic distinction, which, mutatis mutandis, has been confirmed across languages, could scarcely be considered as merely accidental: to account for it the so-called 'unaccusative hypothesis' was proposed; this states that the subject of a certain class of intransitives (represented here by arrivare) is in fact originally generated in the same position as the object of transitives and that the canonical subject position isfilledby a phonologically null pronoun (pro) as shown in the following representation: (19)
pro arrivano
molti ragazzi
Starting from this basic assumption, it became possible to trace the facts illustrated in (18a-f) back to independent and more general principles. Italian c/-sentences potentially undermine this explanation. We will now examine why. On the one hand, they react positively to all tests normally used to diagnose unaccusative constructions as in (20): (20a)
ci sono/*hanno stati [molti ragazzi] (there are/have been-plur. many boys) 'there have been many boys'
Plan of the work: towards a unified account b c
9
ci sono sta-ti/*-to [molti ragazzi] (there are been-plur./-indef. many boys) 'there have been many boys' ce nej sono stati [molti tj (there of-them are been-plur. many) 'there have been many of them'
On the other hand, the postverbal noun phrase of a d-sentence is currently analysed as an embedded subject, rather than as an object. Thus, it does not fall into line with the basic defining property of unaccusative constructions. We face a dilemma here: either we simply add esserci to the list of unaccusatives, and thus risk undermining the theory of unaccusativity, or alternatively we refine such a theory in such a way that both esserci and, say, arrivare will follow as instances of the same general structure. In this book, we will pursue the second possibility and explore some of its consequences.4
v
Plan of the work: towards a unified account
This brief overview has brought out a number of very different problems: an inconsistency with respect to the normal subject-object asymmetries (section i); a cross-linguistic difference with respect to an important semantic restriction (section n); a surprising obligatoriness of a semantically null element (section m); a potential problem for the current theory of unaccusatives (section iv). The central aim of this work is to show that all these problems can be solved in a unitary way once we abandon one single assumption which lies at the very heart of the theory of clause structure: that the noun phrase immediately dominated by the clausal node necessarily corresponds to the subject of the predication. The empirical support for this conclusion will be seen to come from copular sentences and this can hardly be considered accidental. In fact this is the only domain where the predicate (which is normally a verb phrase) can be realized by the same lexical category as the subject, i.e. a noun phrase. The crucial step will then be to allow a predicative noun phrase to occupy the most prominent position in clause structure, which is generally occupied by the subject of the predication. Compared to the standard 'rigid' structure, the only surviving requirement that applies to this 'flexible' structure is that the most prominent position should be occupied by a noun phrase. Thus, the syntax of clauses appears to be simpler than is usually assumed: it does not require that the most prominent noun phrase be the subject of predication; rather, it simply
10
Introduction: four apparently unrelated empirical domains
requires that a noun phrase be in such a position. The effect of this simple fact on the entire system is far-reaching and one that allows us to treat the four different sets of phenomena in a unified and natural way without exceptions. I will now briefly illustrate the plan of the entire work. First of all, dropping the central postulate about the most prominent position will allow us to explain away the failure of subject-object asymmetries in copular sentences. In order to see why, a somewhat more refined analysis of copular sentences than is nowadays adopted must first of all be developed. The current standard analysis for a copular sentence claims that the copula selects a 'small clause' complement.5 Thus, in formal terms a copular sentence would have the following representation: (21)
The subject is generated in the small clause and then raised to the higher position of the matrix clause. Notice that, although this structure is more complicated than the minimal one presented earlier in (7), the asymmetry between preverbal and postverbal noun phrases is fully maintained: the preverbal noun phrase is in a left-branch position whereas the postverbal one is in a right branch. This means that even in this more articulated structure the absence of subject-object asymmetry which we observed with certain copular sentences (cf. (9a-b)) still remains unexplained. All we have done here is to increase the distance between the two noun phrases. As in the simpler structure, the postverbal noun phrase is not in a left-branch position and thus extraction is expected to be possible. So far, then, there is nothing new. The original proposal that will be made here, and indeed the central one of this book, is that, along with the structure in (21), which I will call 'canonical', there exists another possibility which is the 'mirror image' of this standard account. In other words, the same lexical elements can be arranged differently by raising the predicative noun phrase to the most prominent position while leaving the subject noun phrase in situ:
Plan of the work: towards a unified account
11
(22)
NP-
VP
be
NP
This then is the formal correlate to what I earlier called the 'flexible' clause structure: aflexibleclause structure is one where the most prominent position allows the presence of either the subject or the predicate (provided that the latter is realized by a noun phrase). This structure, which I will call the inverse' copular sentence, allows us to account for the absence of subject-object asymmetries in a natural way. In fact, the postverbal noun phrase of an inverse sentence is now in a left-branch position exactly like a preverbal subject in English. Recall that normally it is not possible for anything to be extracted from a constituent occupying this position or indeed for any movement to take place from such a position, unless, that is, certain special conditions are met (which will be discussed in the text). Thus, the phenomena we observed earlier follow as consequences of the position of the postverbal noun phrase in inverse sentences. This major change in perspective will be extended to there-sentences and their counterparts in Italian, namely d-sentences. If we abandon the rigid structure in favour of the moreflexibleone, there no longer seems to be any principled reason for assuming that the grammatical function of elements like there and ci is to occupy the position of subject of the predication. In fact, it will be proposed that these elements are just the opposite of what they are usually assumed to be, i.e. that they are (place holders for) predicates rather than subjects and they are raised from a lower position where predicates are generated: (23)
there:
12
Introduction: four apparently unrelated empirical domains
In other words, I will propose that f/zere-sentences are particular instances of the much broader class of inverse copular sentences that involve raising from a small clause constituent rather than direct insertion into a higher position (cf. (22)). With these premises, it will be possible to furnish a natural interpretation of the cross-linguistic difference between English and Italian with respect to the so-called Definiteness Effect. We will work towards this in two steps: first, we will establish clearly what the basic syntactic correlate of the semantic notion of 'existential meaning' might be. Second, we will show that the simple fact that Italian ci is a clitic, unlike its English counterpart there, will ultimately enable us to relate the lack of the Definiteness Effect in Italian to an independent parameter, i.e. the pro-drop parameter. Moreover, the theoretical consequences of adopting there-rmsmg as an alternative to there-xnserixon will be discussed in detail. In particular, the expletive replacement hypothesis, which states that expletives must be replaced by their associate elements at LF (Logical Form), will be shown to be empirically inadequate, at least in this case. Theflexibletheory will also be seen to provide an account for the fact that it must be present when seem is followed by a finite clause. In this case, the traditional label for seem as the 'quasi-copula' will turn out to be very insightful. The main line of reasoning will be briefly illustrated. For independent reasons, it is widely assumed that certain seem sentences can be given the same analysis as the copular sentences in the sense that seem can take a small clause complement. For example, a sentence like John seems happy would be analysed in exactly the same way as John is happy, namely as follows: (24)
seem
The noun phrase in preverbal position is raised from the lower position within the small clause. My proposal would be to extend the small clause
Plan of the work: towards a unified account
13
analysis to cases where seem is followed by a finite clause. The crucial step involves adopting the flexible structure, that is to say, the one which allows the most prominent noun phrase to be a predicate. Looked at from this perspective, a sentence like it seems that John left could be analysed as an instance of the inverse structure I proposed for copular sentences, that is, one where the subject stays in situ and the predicative noun phrase is raised to the preverbal position (cf. (22)). In formal terms, we would have a structure like the following: (25)
seem
In this representation, the subordinate clause (S) following seem is not itself the complement of seem; instead, it is the subject of the small clause (SC) which is in fact the complement of seem. The pronominal element it here plays the role of a propredicative element raised from the small clause position, something which will be seen to be independently attested in inverse copular sentences such as it's that John left. The major question of why it cannot be replaced by the subordinate clause will be straightforwardly answered. Crucially, replacement of it by S is impossible as it would obliterate the chain of the propredicative element. Finally, we will approach the problem posed by esserci with respect to the theory of unaccusative structures. In essence, my proposal will be that unaccusative constructions should be analysed as instances of inverse structures in the sense that the verb selects a small clause, rather than a noun phrase, and that the predicative noun phrase contained in the small clause is raised to preverbal position (cf. (22)). Thus, unaccusatives will turn out to be parallel to d-sentences, with the only difference that ci is overt whereas the raised predicate of the unaccusative structure is not phonologically realized. In formal terms, a prototypical unaccusative sentence like arriva un uomo (arrives a man; 4a man arrives') will turn out to have the following structure:
14 (26)
Introduction: four apparently unrelated empirical domains S
arriva NP
Thus, the verb selects a small clause and the subject is left in situ while the most prominent position is filled by a phonologically null predicate. At this point, the new approach I am proposing as a way of unifying these apparently unrelated domains should be more or less transparent. In each and every case explored here, the most prominent position of the clause structure is occupied by a raised predicative noun phrase, while the subject remains in situ in a lower left-branch position. In formal terms, the following representation applies to all cases observed here: (27)
NP/S
Given these ideas, the apparently peculiar properties of the constructions analysed here will be traced back to the modular interaction of independently assumed principles and will be seen to lie within the range of the limited parametric variation allowed by grammar. This book is organized into afirstchapter that sets out the general position that will be adopted and into four other chapters, each of which deals with one of the themes briefly sketched out in this introduction. Chapter 2 is to be considered both as an exploration of the consequences of the theory presented in chapter 1 and as the logical premise for the problem treated in chapter 3. The analysis presented in these chapters is completed by an appendix illustrating the historical development of the theory of the copula from its origins right up to modern frameworks. It is my intention to show
Plan of the work: towards a unified account 15 that an analysis of the history of this notion is not just the reflex of some philological bent on the part of the author but rather that it actually provides the necessary background to understanding the complex literature on the copula. Whether or not I succeed in providing such a background is, of course, for the reader to judge. Indeed, the aim of the appendix is twofold, since I will also try to show that it is by no means accidental that a theory of the copula should play a role in the reformulation of the theory of clause structure. As we will see, the copula has always played a fundamental role in the development of this field ever since the Aristotelian model. Nevertheless, in spite of this continuity, the term 'copula' will turn out to have been used to refer to very different entities. Failure to notice this has been at the root of some major misunderstandings. In particular, it will be shown that considering the copula as expressing an identity relation has prevented grammar from reaching a simpler theory and has thus forced the introduction of complex mechanisms (such as the expletive replacement hypothesis). One last remark is in order concerning the potential relevance of this work for the general theory. As in all scientific fields, the forefront of research is often at odds with the accepted practice of everyday research. At the present stage of development, the entrenched generative framework often referred to as 'government and binding theory' is still producing a fruitful theoretical debate. On the other hand, a new approach is establishing itself in response to Noam Chomsky's recent reconceptualization of the field and from other extremely influential works such as Richard Kayne's attempts to derive the linear order of constituents from their hierarchical organization or Ken Hale and Jay Keyser's project to reduce 6-roles to configuration. It goes without saying that, however innovative, this new framework, which is generally labelled as 'minimalism', remainsfirmlywithin the bounds of the 'principles and parameters' approach which has been recognized as the signal breakthrough in the modern theory of syntax since the time of the Pisa Lectures. The work presented here reflects this transitional situation: on the one hand, certain notions from the 'government and binding' model that the new model tends to eliminate will be employed, discussed and perhaps even refined; on the other hand, in the light of the new perspective, the original analyses proposed here will also be offered as being congenial to the spirit of the new model still in the making. One area which is currently in the front line of research is what is referred to as 'locality theory', that is to say, an area of inquiry that seeks to establish the conditions that must be met by empty categories within a local domain. Specifically, two central notions of the theory of locality will be approached,
16
Introduction: four apparently unrelated empirical domains
namely the Empty Category Principle (ECP) and the Subjacency Condition.6 The need for two separate conditions has been generally related to the two conceptually distinct roles that they are called upon to play in syntax. As regards the ECP, Chomsky remarks that 'it is not unreasonable that Universal Grammar should require that the presence of an empty category be signalled in some manner by elements that are overtly present' (Chomsky 1981: 251). The Subjacency Condition, on the other hand, accounts for a quite different empirical phenomenon, namely the fact that direct extraction from more than one cyclic node is ungrammatical. Since Chomsky (1986b) there has been increasing debate about whether these two notions could be subsumed under a more general notion of 'barrier' and whether this could in turn be derived from the notion of lexical head'. More recently, Chomsky (1993) has proposed to reduce the theory of locality to the interaction of economy principles.7 In the first chapter of this book, evidence will be provided for maintaining the core content of the ECP and Subjacency as distinct, at least at this stage in the development of the theory. This approach will be made in two steps: first, original empirical data will be provided in favour of reducing proper government (hence, the ECP) to a spec-head relation with a local head containing agreement features;8 second, a number of problematic cases that resist an ECP account will be uncovered and reanalysed as Subjacency violations by pursuing the idea that this condition correlates with the selectional properties of a governing head (as proposed by Cinque 1990a and Rizzi 1990 among others) rather than with agreement. As a second line of research, which makes use of some core notions of the minimalist framework, I will offer support for the wide-ranging hypothesis that non-lexical elements are not visible to movement at LF and I will attempt to account for what appears to be a semantic variation (such as the lack of the Definiteness Effect in Italian) in terms of purely syntactic principles. Conversely, I will offer arguments against the expletive replacement hypothesis. Beyond the specific theory that will be proposed here, two independent facts should perhaps be underlined, which make any analysis of copular sentences interesting whatever syntactic model is being discussed: first, the fact that they involve predicative noun phrases which, as far as I know, are still far from being completely understood; second, it seems to me that, in spite of the extremely advanced theory of clause structure that has been developed lately through advanced research into largely morphosyntactic phenomena, the problem of predication remains virtually untouched and the copula, with its special status, still deserves our undivided attention. 9
1
The anomaly of copular sentences: the raising of predicates
1.1
Subject-object asymmetries and the theory of clause structure
A basic task for any theory of syntax is to account for the fact that, in a sequence of the type noun phrase - verb - noun phrase, the two noun phrases show very different properties. These different properties are evidenced by a wide range of empirical phenomena ranging from overt movement to verbal agreement patterns and Logical Form (LF) movement. Consider for example the following sentence: (1)
[Np a picture of the wall] [v revealed] [NP the cause of the riot]
A classic test for a structure of this kind - and one that was among the first to be discovered - involves wh-movement. This really amounts to two different tests since one can either attempt to extract each noun phrase as a whole or else attempt to extract a constituent embedded within either noun phrase. In each case, the results are extremely clear. First, an embedded noun phrase can be successfully extracted by whmovement from a larger noun phrase only if the latter is in postverbal position: 1 (2a) b
* [which wall]; do you think that [NP a picture of ts] revealed [NP the cause of the riot]? [which riot]i do you think that [NP a picture of the wall] revealed [NP the cause of tj]?
To show that the ungrammaticality is really due to the syntactic position of the noun phrase rather than to some factor having to do with the internal structure of the noun phrase, it is sufficient to notice that if the same noun phrase occurs in postverbal position extraction becomes fully acceptable: (3)
[which wall]i do you think that [NP the cause of the riot] revealed [NP a picture of tj]?
Second, wh-movement of a preverbal noun phrase over an overt complementizer is impossible, whereas this is not the case with a postverbal noun phrase: 17
18 (4a) b
The anomaly of copular sentences: the raising of predicates * [which picture]; do you think that tj revealed [NP the cause of the riot]? [which picture]} do you think that [NP the cause of theriot]revealed tj?
A traditional representation that goes back to the first pioneering works in syntax treats the difference between preverbal and postverbal noun phrase in terms of a 'topological' asymmetry in the geometry of tree representation. The following simple formal structure taken from Chomsky (1957: 27) serves to show the essence of this proposal: (5)
Once this representation has been adopted, two independent principles could then be used to derive the observed phenomena: the Empty Category Principle (which is responsible for the ungrammaticality of (4a)) and the Subjacency Condition (which explains (2a)). In general, following the pioneering work of Richard Kayne (see Kayne 1984), it is assumed that extraction from a constituent in a left-branch position is impossible, unless special conditions are met.2 Now, leaving technical details aside, what is immediately apparent is that the two principles in question, though conceptually distinct, both lead to a common conclusion: the postverbal noun phrase is in a 'privileged' position compared to its preverbal counterpart: movement is always possible whether this involves the noun phrase as a whole or extraction of a part of it. Of course, there are many further empirical data which support this asymmetry, such as the assignment of semantic roles such as 'agent', 'patient', etc., usually called thematic roles (6-roles). As a simple case, let us consider the following pair: (6a) b
John hit Mary Mary hit John
In such a case, the noun phrase which is adjacent to the verbal head is assigned the role of 'patient' (in relation to the action of hitting) while the noun phrase which is immediately dominated by the clausal node is assigned the role of 'agent' (in relation to the same action). Thus, inserting the two noun phrases in the two different positions would amount to associating them with two different 9-roles (see Chomsky 1986a and references cited there).
Subject-object asymmetries and the theory of clause structure 19 Indeed, one of the major contributions to the theory of clause structure that has come from this approach is the idea that grammatical functions, like subject, predicate, object and the like, can be straightforwardly defined in configurational terms (see Lepschy 1992: 88): thus, the subject of predication would be the 'most prominent' NP, i.e. the NP immediately dominated by the clausal node S; the object would be the NP immediately dominated by the VP node, and the predicate would be the VP itself: (7)
NP subj .
For this reason, the whole cluster of phenomena that show differences between a preverbal and postverbal NP are generally referred to as 'subjectobject asymmetries'.3 Such subject-object asymmetries hold cross-linguistically although parametric differences can sometimes obscure them. Let us illustrate the situation in Italian. The first class of phenomena is fully preserved. Extraction from within the noun phrase is sensitive to the position of the phrase in the clause structure: (8a) b
* [di quale murojj pensi che [s [NP una foto tj] rivelo [NP la causa della rivolta]]? (of which wall do you think that a picture revealed the cause of the riot) [di quale rivoltajj pensi che [s [NP una foto del muro] rivelo [NP la causa tj]]? (of which riot do you think that a picture of the wall revealed the cause)
As for the second class of phenomena, involving movement of the noun phrase as a whole, there is a sharp difference between the two languages. In Italian, a pre verbal subject can be extracted over an overt complementizer (che), something that we found not to be the case in English: (9a) b
[quale foto]j pensi che [s tj rivelo [NP la causa della rivolta]]? (which picture do you think that revealed the cause of the riot) [quale causajj pensi che [s [NP una foto del muro] rivelo tj]]? (which cause do you think that a picture of the wall revealed)
Given these data the question naturally arises of whether one should simply conclude that the asymmetric clause structure does not apply to Italian. The current account (which goes back to Rizzi 1982) indeed claims that the clause structure proposed for English holds invariantly across languages,
20
The anomaly ofcopular sentences: the raising of predicates
including Italian. Crucially, what distinguishes the latter language is the idea that movement of the subject does not take place from preverbal position. In fact, in a language of this type, the subject can freely occur in a postverbal subject position, where it is traditionally considered to be an adjunct to the VP, while the subject position is occupied by a non-phonologically realized pronoun, i.e. pro: (10)
Thus, the Italian examples are only apparently atypical, since movement in fact can take place only from a postverbal position, which is exactly parallel to what we find in English. The correct representation, then, for the sentence in question (i.e. (9a)) is the following: (11)
[quale foto]j pensi che [s pro [Vp [VP rivelo la causa della rivolta] tj]]? (which picture do you think that revealed the cause of the riot) Interestingly, this structural account of subject-object asymmetry has been a constant feature of the theory of clause structure throughout its development. We will now briefly review the various stages that have led to the most advanced model current today. A first crucial step (see Chomsky 1981) was the separation of the morphological component (verbal inflection) from the lexical part of the predicate (verbal root). Adapting the representation we have used so far (cf. (5)), we obtain the following enriched structure: (12a)
b
INFL = [± Agreement, ± Tense]
The inflectional elements, labelled as 'INFL', are essentially those agree-
Subject-object asymmetries and the theory of clause structure 21 ment and tense features that appear as suffixes on the verb. In this representation, they are represented syncretically as two features of the same node and simply added as a third branch between the subject and the predicate. It might be useful to recall that the inflectional elements could be realized autonomously as auxiliaries or combine with the verb via movement of the latter to the higher node. At a later stage, developing an idea that can be traced back to Chomsky (1981), Chomsky (1986b) brought the structure of the clause into line with the general theory of phrase structure, known as X-bar theory. Essentially, this theory holds that every constituent is the result of a lexical head (X°) projecting an intermediate one-bar projection (X') and a maximal two-bar projection (X" = XP). Every head can be adjacent to another maximal projection (the 'complement'); and every intermediate projection can be adjacent to another maximal projection (the 'specifier'). For every head X°, Y° and Z°, ranging over the set of heads, we thus have the following recursive representation: (13)
XP
X° ZP If we abstract away from the linear order of the adjacent elements, 4 we can assume that this structure is universal. The key step in the rethinking of the clause structure, then, was the assumption that INFL is the head of the clausal constituent and thus that the clause is organized in accordance with the principles of X-bar theory. This move led to the following structure, where '1°' represents (the head) of what was previously referred to as 4 INFL' (VP is simplified here): (14a)
V° NP P = [± Agreement, ± Tense]
22
The anomaly ofcopular sentences: the raising of predicates
Subsequently, for a number of different reasons, it was proposed that the syncretic components of the inflectional node 1° should be represented as heading their own separate projections (see the original work by Pollock 1989 and independently Moro 1988). Using 'Agr' and T ' as abbreviations for 'agreement' and 'tense', the structure was expanded in the following way (irrelevant intermediate spec-positions are omitted): (15)
AgrP
V°
NP
Finally, Chomsky (1988a), for Case-theoretical reasons, and Belletti (1989), on the basis of morphological evidence suggested by Baker's (1988) theory of incorporation, further enriched the structure by including an agreement node for the object (called 'Agr-o') which is distinct from the agreement node relating to the subject (called 'Agr-s'): (16)
Agr-s P
Agr-o°
V°
NP
All in all, throughout its entire evolution, the theory of clause structure has maintained at least two major assumptions: first, the grammatical functions
The anomaly of copular sentences: the data 23 (such as subject, object, etc.) can be defined exclusively in configurational terms; second, subject-object asymmetries can be derived from the asymmetric position of the NPs within the geometry of the structure. 5 We will now consider how this theory has been applied in the area of copular sentences and evaluate its empirical force in predicting their properties. 1.2
The anomaly of copular sentences: the data
As a first approximation, we will say that a copular sentence is a sentence where the main verb is be or its equivalent in languages other than English (see the appendix for a discussion of the notion of copula). Descriptively speaking, there are at least four different subtypes of copular sentences, defined in terms of the type of maximal projection that follows the copula: (17a) b c d
John is [NP the cause of a riot] John is [AP happy] John is [PP in the room] John is [Vp causing a riot]
In this work we will essentially concentrate on the first type {noun phrase copula - noun phrase) and will not deal with the syntax of the other types. In discussing copular sentences, the idea that sentences of this type display the same subject-object asymmetry that has been observed in other cases will be adopted as the null hypothesis. It will thus be assumed that, from the formal point of view, there is no a priori reason why a sentence of this type should not be assigned a representation like the following: (18)
In the course of the discussion in this section, we will see that this null hypothesis cannot be maintained. For expository convenience we will concentrate for the moment on Italian data. As a preliminary step the representation will be somewhat refined by the adoption of a more articulated theory of noun phrases as projections of
24
The anomaly ofcopular sentences: the raising of predicates
functional determiners (D°) selecting a fully lexical NP complement (see Abney 1987 and references cited there). The traditional representation for noun phrases will then be enriched as follows: (19)
spec
It is important to realize that this is not just a mere notational variant. In fact, it allows us to say, as a first approximation, that in Italian either bare NPs or full DPs can follow the copula, as in the following examples: (20a) b
Gianni e [NP causa della rivolta] (Gianni is cause of the riot) Gianni e [DP [D° la] [NP causa della rivolta]] (Gianni is the cause of the riot)
but that, on the other hand, only DPs can precede the copula (see Beninca 1980 and Longobardi 1991a, for a detailed discussion of the distribution of bare nouns in Italian): (21a) * [NP ragazze] sono la causa della rivolta (girls are the cause of the riot) b [DP le [NP ragazze]] sono la causa della rivolta (the girls are the cause of the riot) Henceforth, we will concentrate on sentences of the type DP - copula focussing on minimal pairs like the following: (22a) b
-DP,
[DP una foto del muro] fu [Dp la causa della rivolta] (a picture of the wall was the cause of the riot) [DP la causa della rivolta] fu [Dp una foto del muro] (the cause of the riot was a picture of the wall)
A first important fact that needs to be noticed is that these two sentences cannot be regarded as distinct with respect to 0-role assignment. 6 In general, as we noticed earlier, inserting the same noun phrases in preverbal vs postverbal position in sentences of the type DP- V- DP yields different interpretations with respect to the semantic role played by (the referent of) the two noun phrases. The situation here is different. In both cases we are saying that a picture of the wall has the property of having caused the riot. This fact has not generally been considered of much interest by
The anomaly ofcopular sentences: the data 25 syntacticians.7 Since in general it is assumed that the copula expresses identity, saying that x is identical to y is simply assumed not to be any different from saying that y is identical to x. We will see that this anomalous behaviour of the copula conceals more complex empirical properties and that it cannot be adequately accounted for without some alterations to the theory. In fact, once the usual subject-object asymmetry tests are applied to the pair of sentences under discussion, corresponding anomalous behaviour will be detected on a syntactic level. 1.2.1 Cross-linguistic anomalies: two types of A' -movement I will test here two different types of A'-movement, i.e. wh-movement and Quantifier Raising (Q-raising).8 Consider the following paradigm (for the sake of simplicity, preliminary rightward movement in the case of subject extraction will not be indicated):9 (23a) b
[quale foto del muro]j pensi che tj fu la causa della rivolta? (which picture of the wall do you think was the cause of the riot) * [quale foto del muro]j pensi che la causa della rivolta fu tj? (which picture of the wall do you think the cause of the riot was)
The anomaly here compared to what we have seen so far is clear. Bearing in mind the standard terminology for the clause structure, we can thus conclude that movement from the object position in the second sentence is totally ungrammatical. Extraction of an embedded constituent from the object position yields another unexpected result: (24a) b
[di quale rivoltajj pensi che una foto del muro fu [DP la causa tj]? (of which riot do you think that a picture of the wall was the cause) 'which riot do you think that a picture of the wall was the cause of?' * [di quale muro]j pensi che la causa della rivolta fu [DP una foto tj]? (of which wall do you think that the cause of the riot was a picture)
Why should wh-movement from the object position be blocked in (23b) and (24b)? Notice that any appeal to idiosyncratic properties of the copula or the noun phrase involved may be excluded in advance. Extraction from the postverbal noun phrase is in fact possible in a copular sentence like (24a) and una foto does not block extraction per se, as is shown by sentences like di quale muro Gianni vide una foto? (of which wall John saw a picture; 'which wall did John see a picture of?'). The same type of anomalies can be detected at LF, as is illustrated in the following cases:10
26
The anomaly ofcopular sentences: the raising of predicates
(25a) b
[DP ogni libro] fu [DP l'acquisto di [DP molti studenti]] (every book was many students' purchase) [DP l'acquisto di [Dp molti studenti]] fu [Dp ogni libro] (many students' purchase was every book)
The first sentence ((25a)) is ambiguous. It can either mean that every book is such that many students purchased it or that many students are such that they purchased every book. 11 In the second sentence ((25b)), instead, only the latter meaning is possible. Following usual assumptions stemming from May (1986), it is assumed that a quantifier a can be interpreted as being in the scope of quantifier (3 only if (3 has at least the same c-domain as a, that is only if the set of elements c-commanded by a is a subset of the set of elements c-commanded by (3. Bearing this in mind, we can conclude that in the first sentence the embedded quantifier molti (many) can either be interpreted in situ or be extracted so that it has scope over the universal quantifier ogni (every). 12 Movement of the quantifier molti has the effect of enlarging its c-domain so that the quantifier ogni is contained in it. From the formal point of view, the first sentence of the pair we are considering here (reproduced as (26a)) can be assigned the LF representation given in (26b): (26a) b
[s [DP °gni libro] fu [Dp l'acquisto di [Dp molti studenti]]] [s [DP molti studentiji [s [Dp ogni libro] fu [Dp l'acquisto di t;]]]
On the other hand, in the second sentence of the pair we are considering here (reproduced as (27a)) the universal quantifier cannot have scope over the other quantifier. In other words, the lower quantifier ogni must be interpreted in situ, i.e. it cannot be moved at LF in order to acquire a broader c-domain which includes the c-domain of molti, as indicated in (27b): (27a) [s [DP l'acquisto di [Dp molti studenti]] fu [Dp ogni libro]] b * [s [DP °gni libro]j [s [Dp l'acquisto di [molti studenti]] fu tj]] Once again, this fact runs counter to the general property of the object position of allowing movement in contrast to the subject position which does not. Thus, extraction of an embedded quantifier produces the same type of anomaly we detected in the corresponding case of wh-movement. We will now proceed to test the interaction between a quantifier and negation. In this case, the facts seem to be more complex, as the following pair of sentences shows: (28a) b
una foto del muro non fu [Dp la causa di [Dp molte rivolte]] (a picture of the wall not was the cause of many riots) la causa della rivolta non fu [Dp una foto di [Dp molti muri]] (the cause of the riot not was a picture of many walls)
The anomaly of copular sentences: the data 27 In the first sentence ((28a)), the quantified DP molte rivolte (many riots) can have scope over the negation (meaning: many riots are such that a picture of the wall was not the cause of them) or be interpreted within the scope of negation (meaning: a picture of the wall was the cause of not-many riots). A formal account of this would assume that the lower quantifier can be interpreted in situ (as in (29a)) or else raised to a higher position where it includes the negation within its c-domain (as in (29b)): (29a) b
[s una foto del muro non fu [DP la causa di [DP molte rivolte]]] [s [DP molte rivolte]j [s una foto del muro non fu [DP la causa di tj]]]
In the second sentence ((28b)), on the other hand, the quantified DP mold muri (many walls) cannot have scope over the negation (meaning: many walls are such that the cause of the riot was not a picture of them). The conclusion here seems less straightforward. We cannot simply conclude that mold muri is in the scope of the negation, at least not in the same way as molte rivolte is in the first sentence. This is due to the fact that the sentence can only mean that it is not the case that the cause of the riot was a picture of many walls and not that the cause of the riot was a picture of not-many walls. In other words, the negation in (28b) can only have a sentential reading.13 Although I have no explanation for the obligatoriness of the sentential reading, it is surely reasonable to think that mold muri cannot have scope over the negation in this sentence. Thus, bearing in mind the usual analysis in terms of movement and c-domain, we can conclude that mold muri (many walls) cannot be extracted from the postverbal DP in (28b) (reproduced here as (30a)) so as to have scope over the negation: (30a) [s la causa della rivolta non fu [DP una foto di [DP molti muri]]] b * [s [DP molti muri]i [s la causa della rivolta non fu [DP una foto di [DP tj]]]] Once again the crucial question these examples raise is why Q-raising cannot take place from the object position in a sentence like la causa della rivolta non fu una foto di molti muri? 1.2.2 Language-specific anomalies: ne, verbal agreement and lo The anomalies discussed in the previous section were based on Italian data but even a rather cursory survey of the glosses accompanying the Italian examples shows that they can easily be tested in English as well. We will not pursue this here; rather, we will keep our attentionfirmlyfixedon Italian and examine, in the light of the discussion so far, some well-known properties that this language shares with others of the Romance family. We will see that in each case the same anomalous pattern is confirmed with
28
The anomaly ofcopular sentences: the raising of predicates
respect to the usual subject-object asymmetries (the first and the last property have been discussed in detail by Longobardi 1985 for Italian and by Ruwet 1975 for French). In Italian, the pronominal clitic element ne can be extracted from the object position but not from the subject position: (31a) b c
[DP ilfigliodi Gianni] vide [DP una foto di Maria] (the son of Gianni saw a picture of Mary) [DP ilfigliodi Gianni] nej vide [DP una foto tj] (the son of Gianni of-her saw a picture) * [DP ilfigliotj] nej vide [Dp una foto di Maria] (the son of-him saw a picture of Mary)
This process is blocked in one of the two copular sentences under discussion: (32a) b
una foto del muro nej fu [Dp la causa tj] (a picture of the wall of-it was the cause) * la causa della rivolta nej fu [Dp una foto tj] (the cause of the riot of-it was a picture)
The generalization that obviously suggests itself is that in such cases the clitic ne can be extracted only from the same DP that allows wh-movement and Q-raising of an embedded DP (see the discussion in the previous section). A further contrast illustrates an important distinction between Italian and English. The following examples show that when the two DPs occurring with the copula do not match in number, the agreement pattern of the verb is different for the two languages: (33a) b c d
the pictures of the wall were/*was the cause of the riot the cause of the riot was/*were the pictures of the wall le foto del muro furono/*fu la causa della rivolta (the pictures of the wall were/was the cause of the riot) la causa della rivolta furono/*fu le foto del muro (the cause of the riot were/was the pictures of the wall)
In English the copula invariably agrees with the DP in subject position; in Italian, instead, it always agrees with the same DP, whether this precedes or follows the copula. Notice that prima facie this difference cannot be related to any independent properties of Italian syntax. In fact, apart from copular sentences, in Italian sentences of the DP- V- DP kind agreement is always with the subject position:14 (34a)
Gianni vede i ragazzi (Gianni see-sing, the boys)
The anomaly of copular sentences: the data 29 b * Gianni vedono i ragazzi (Gianni see-plur. the boys) c Gianni e Maria vedono il ragazzo (Gianni and Maria see-plur. the boy) d * Gianni e Maria vede il ragazzo (Gianni and Maria see-sing, the boy) A third anomaly can be detected through observing clitic movement. It is a well-known fact that in Italian the object of a verb can be cliticized onto it (by means of a full inflected range of clitics: la (fern, sing.), lo (masc. sing.), le (fern, plur.), etc.; see Burzio 1986). In copular sentences, on the other hand, there is a special clitic, namely lo, which is invariant in gender and number. The following paradigm based on the set of sentences we discussed above in relation to the agreement facts (cf. (33c-d)) shows that cliticization from the object position is successful in only one case: (35a)
le foto del muro k>j furono tj (the pictures of the wall lo were) 'the pictures were such and such' (e.g. the cause of the riot) b * la causa della rivolta IOJ furono tj (the cause of the riot lo were)
Once more, this asymmetry reproduces the pattern we obtained when examining wh-movement and Q-raising: in a copular sentence, a postverbal DP can be cliticized (by lo) only if it can be wh-moved and Q-raised; witness the examples given in the previous section. Summarizing the findings of section 1.2, it has been shown that copular sentences behave anomalously with respect to the classic subject-object asymmetry tests. We have analysed pairs of sentences of the type DP copula - DP, inverting the linear order of the two given DPs around the copula. In one of the two sentences both wh-movement and Q-raising from the object position fail to give grammatical results across languages in a consistent and systematic way (section 1.2.1). Some language-specific tests (Aze-cliticization, verbal agreement and /o-cliticization) from the current literature have also been seen to support these conclusions (section 1.2.2). What these facts appear to show, then, is that the standard clause structure cannot be assumed for copular sentences because it fails to predict their syntactic behaviour. From a formal point of view, we can then conclude that the following representation is (at best) insufficient to describe copular sentences:
30
The anomaly ofcopular sentences: the raising of predicates
(36)
S
More explicitly, if we were to adopt this representation, there would be no way to explain the strange behaviour of the postverbal DP in one sentence of the associated pair. In the next section, a revised structural representation for copular sentences will be proposed, from which the properties discussed will be seen to follow naturally. 1.3
The raising of predicates: a unified theory of copular sentences
The data reported in the previous sections indicate that a theory of copular sentences15 must be able to meet at least two minimal and contrasting requirements. On the one hand, it must capture the fundamental intuition that the pair of associated sentences such as those repeated here are not distinct from the point of view of 0-role assignment: (37a) b
[DP una foto del muro] fu [DP la causa della rivolta] (a picture of the wall was the cause of the riot) [DP la causa della rivolta] fu [DP una foto del muro] (the cause of the riot was a picture of the wall)
On the other, it must be able to explain the differences between these two sentences that were illustrated in section 1.2. In particular, it must explain why the postverbal noun phrase fails to behave like an object in the second sentence in contrast to the usual pattern displayed by the postverbal noun phrase in the first. Since the lexical elements construing these two sentences are the same, and since in particular they are both construed around the copula, any explanation based on lexical idiosyncrasies must be excluded in advance. In other words, the anomaly observed cannot possibly be due to the copula itself or to the DPs involved. The only possibility left is therefore that the asymmetries between these elements must be traced back to differences in the structure of the clause.
The raising of predicates: a unified theory of copular sentences 31 1.3.1 Canonical vs inverse sentences As a preliminary step we will adopt the more refined representation of copular sentences involving raising. For ease of exposition, we will shift our attention to English data. Stowell (1978) proposed that copular sentences should be regarded as expanded small clauses; his proposal completely overthrew earlier approaches which had regarded small clauses as reduced copular sentences. Prototypically, a small clause may be thought of as being the complement of a consider-typc verb (Williams 1975). The nature of the small clause and its structure are not completely understood yet (see Cardinaletti and Guasti 1995 and references cited there). For example, it is not clear whether it is an anomalous constituent, essentially an adjunct structure (cf. Stowell 1981 and Manzini 1983 and references cited there), or whether it is projected by a head (cf. Moro 1988 and Bowers 1993). This issue is not without importance for the argument being developed here and we will return to it at a later stage. For the moment, I will present a brief review of the current theory of copular sentences as involving small clauses. The small clause is normally considered to be a full phrase, i.e. a constituent that can occur where maximal projections occur. Empirically, this can be demonstrated by observing that it can be the complement of a head, or an adjunct to a maximal projection,16 and that it can occur in isolation, especially in informal contexts: (38a)
I [V' [v° consider] [Sc John the cause of the riot]]
b
[IP [IP Johtij arrived] [sc PROj drunk]]
c
[sc John the cause of the riot]? I can't believe it!
None of these three possibilities would be available if the small clause were not a phrase. The passive form of a sentence involving consider like (38a) is generally analysed as involving DP raising from the lower subject position within the small clause:
32
The anomaly ofcopular sentences: the raising of predicates
(39)
IP
John,
considered the cause of the riot
In the account that will be presented here, the copular sentence parallels this structure. Thus, it consists of a 'predicative nucleus' (the small clause) which is the complement of a verbal head (the copula), as given in the following representation:17 (40)
John;
IS;
the cause of the riot
Prima facie, this representation seems much more complex than the primitive one we are accustomed to:
The raising of predicates: a unified theory ofcopular sentences (41)
33
S
be
However, the complexity is only apparent, at least as far as the subjectobject asymmetry issue is concerned. In fact the two representations do not differ with respect to this matter. It is of course true that an intermediate position is absent in the simpler representation (i.e. the trace t t in the small clause) but this does not affect the essential asymmetry between the two noun phrases: the two elements are just 'more distant' in (40) but still asymmetrical in the same way as in (41). The problem posed by the anomaly of copular sentences then remains unsolved: the postverbal noun phrase is never in a left-branch position, so there is every reason to expect extraction to be possible. Notice also that the fact that V°raises to 1° isof no relevance: raising would have to be allowed for in all accounts of copular sentences. Thus, we cannot plausibly rely on it to explain why only one type of the pair of associated sentences we are examining here fails to show the usual subject-object asymmetry. All in all, adopting the more refined structure does not per se offer a solution to the problem raised here. Nevertheless, if we carefully rethink the raising analysis and the (implicit) postulates involved, a solution to the problem under discussion will be seen to emerge. We will now focus our attention on the noun phrase which shows up in spec-IP. This element is the subject of the predication. It is base generated in the small clause and then raised to spec-IP for at least two distinct reasons. The first reason is Case-theoretical: the copula, in virtue of Burzio's generalization, is unable to assign Accusative case (as opposed to say consider)^ thus, the only way for the subject of the small clause to acquire Case is to move to spec-IP, where it receives Nominative in a spec-head relation with 1°.The second reason for the subject of the small clause being raised to spec-IP relies on a general principle of clause structure. To explain the distribution of elements like there and it in sentences like there is a man in the room or it seems that John left, it is generally assumed that the position of subject of the predication (i.e. spec-IP) must be realized in all sentences (see Chomsky 1981: 116). This principle, called the 'Extended Projection
34
The anomaly ofcopular sentences: the raising of predicates
Principle' will be briefly discussed in the appendix. For the moment, it is sufficient to notice that the raising of the embedded subject from the small clause to spec-IP immediately satisfies this principle of clause structure. 19 A second question of a somewhat theory-internal nature, involving movement, is raised by the structure: traces must satisfy locality conditions. In the development of syntactic theory various attempts have been devoted to identifying such conditions. Since the end of the 1970s, the Empty Category Principle (ECP) has been adopted to account for the distribution of traces. Originally, the ECP was stated in a disjunctive formulation. A trace could either be antecedent-governed (i.e. governed by a phrase coindexed with it) or lexically governed (governed by a lexical head). 20 The most restrictive version of the ECP (proposed by Stowell 1981) requires that a trace satisfy both conditions. The question that arises from the point of view of the present study, then, is whether the trace of the subject raised from the small clause is able to meet this restrictive version. The answer is indeed in the affirmative. Both requirements of the ECP are in fact satisfied: antecedent government by the DP in spec-IP; head government by the inflected verbal head, i.e. is. The ECP has been constantly revised over the years. Apart from the conjunctive formulation proposed by Stowell (1981), efforts have been devoted to reducing the ECP to only one of the two components. For example, Chomsky (1986b) (following a suggestion by Richard Kayne) explored the possibility of reducing it to antecedent government via the notion of 'extended chain', while Rizzi (1990) pursued the opposite goal of reducing the ECP to head government.21 Chomsky's (1986b) version of the ECP turns out to be particularly well formulated from our point of view. The essential idea is that 'proper government can be reduced to antecedent government by chain coindexing' (Chomsky 1986b: 78), provided, that is, that the notion of chain is extended to include those heads which contain the agreement features of the moved element. 22 Now, in the structure under discussion, the trace of the subject is antecedent governed, thus properly governed, by the agreement features contained in the copula under a spechead relation with the subject DP. 23 With these issues in mind, the central proposal of this work can now be formulated. I propose to abandon the rigid clause structure in which the most prominent position (i.e. spec-IP) is always occupied by the noun phrase playing the role of subject of the predication and to explore the hypothesis that a predicative noun phrase can be raised to such a position in complementary distribution with the subject. Recall the pair of associated sentences of the type we are considering here:
The raising of predicates: a unified theory ofcopular sentences (42a) b
35
a picture of the wall was the cause of the riot the cause of the riot was a picture of the wall
To these sentences, I then propose to assign the following structural representations and labels: 24 (43)
a canonical
the cause of the riot b inverse
a picture of the wall Both the subject and the predicative noun phrase are uniformly generated in the same position in the two sentences and then raised to spec-IP in complementary distribution. 25 In other words, these two sentences share the same basic relation, as illustrated immediately below, but have entirely different structures after raising has taken place:
36 (44)
The anomaly of copular sentences: the raising of predicates IP
V°
be
SC
[a picture of the wall] [the cause of the riot]
As indicated, when the subject noun phrase is raised ((43a)), I propose to label the resulting sentence 'canonical (copular) sentence'; in the other case, where it is the predicative noun phrase that is raised ((43b)), I label the sentence 'inverse (copular) sentence'. How can this new structure explain what we have seen to be the anomaly of copular sentences, i.e. the absence of subject-object asymmetries in a subclass of copular sentences? Where the proposed analysis appears to be particularly promising is in the structure assigned to inverse copular sentences ((43b)), since, after raising has applied, the noun phrase that remains in postverbal position will be in a left-branch position. Thus, its peculiar properties will naturally follow. However, before developing this analysis in detail, we will consider some potential a priori objections to adopting such an inverse structure. The discussion will centre on three topics: the order of the two noun phrases in the small clause, the restrictions placed on movement by the locality conditions on traces and Case theory. First we will consider the question of the order of the two noun phrases in the small clause. The representation under discussion presupposes that the predicative noun phrase {the cause of the riot) can precede the subject {a picture of the wall) only if a landing site has been provided by the copula. If this were not the case, that is, if the predicative noun phrase could precede the subject noun phrase in the small clause, our theory could not be tenable. In such a case, in fact, the residual noun phrase in a small clause would not have a specific position and, in particular, the subject would not be necessarily left in situ in a left-branch position, undermining the explanation we want to give for the impossibility of extraction from it. In fact, there is independent evidence for the assumption that the subject and the predicate have a rigid order in the small clause. Consider the following contrast:26
The raising of predicates: a unified theory of copular sentences
37
(45a) John considers [s [DP a picture of the wall]; to be tx [DP the cause of the riot]] b John considers [s [DP the cause of the riotjj to be [DP a picture of the wall] tj] c John considers [s [DP a picture of the wall] [DP the cause of the riot]] d * John considers [s [DP the cause of the riot] [DP a picture of the wall]] The idea of the inverse structure seems to fit in with these data in a rather natural way. What we see is that in these contexts, the copula can be absent only if the two DPs display the base generated order, i.e. the subject precedes the predicate. Otherwise, we can now say, the copula is needed to allow raising of predicative DP. Nevertheless, the question arises why the order of the two DPs cannot simply be inverted in the small clause. 27 Of course, we cannot appeal to any restriction based on the category involved, since both the subject and the predicate here belong to the same category, namely DP. Nor can we say that consider must govern an argumental noun phrase as opposed to a predicative one, since a sentence like / consider the cause of the riot to be a picture of the wall is fully grammatical. In order to consider a possible solution to this problem, let us now take the two noun phrases which constitute the small clause and examine them in isolation: (46a) b
[DP a picture of the wall] [DP the cause of the riot]
These noun phrases are ultimately projections of the lexical head in question, namely picture and cause, which are capable of assigning 6-roles. These roles will be indicated in a straightforward way by adoption of the usual external-internal distinction that is used with verbs: (47a) b
picture: <0-external, 0-internal> cause: <6-external, 9-internal>
As is shown, both these two heads appear to assign external and internal 6-roles: thus, there is an agent of the action of taking a picture (6-external) and an object, which is the patient of this action (9-internal); similarly, there is an agent of a cause (6-external) and the object of the cause (9internal). This similarity breaks down for syntactic reasons, once we try to construe the full noun phrase projection and actually assign the external 9roles. Consider picture first: (48a) [DP his e.ext. picture of the wall e . int ] b * he B.ext is [DP a picture of the wall e.int ] Contrast now picture with cause: (49a) * [DP his H.ext. cause of the riot H.int] b he H_ext. is [DP the cause of the riot H.int ]
38
The anomaly ofcopular sentences: the raising of predicates
The difference is very clear: on the one hand, picture is able to assign the external 8-role within its maximal projection ((48a)) but not outside it ((48b)); cause, on the other hand, behaves in exactly the opposite way: it can assign its external 6-role outside its maximal projection ((49b)) but not within it ((49a)). I have no principled explanations for this contrast; however, I will make use of it to explain the rigid order of the two noun phrases within the small clause. We know independently from the analysis of verb phrases that V° assigns the external 6-role to the left. If we make the reasonable assumption that this property also applies to an N° like cause, the rigid order in the small clause between the two noun phrases under discussion would follow from two independent facts. First, the only element that can assign the external 6-role outside its maximal projection is cause; second, it can assign it only to its left. All in all, the contrast between / consider a picture of the wall the cause of the riot and * / consider the cause of the riot a picture of the wall is explained as a consequence of 6-role assignment and thus the fixed order of the two DPs in the basic structure in (44) makes (43b) tenable. The second issue that arises with inverse copular sentences concerns locality conditions on movement. In fact, what arises is a theoretical question of whether this structure is compatible with (any version of) locality theory. Let us once again consider first the more restrictive version of the ECP, which assumes that the trace must be both antecedent-governed and head-governed. Antecedent government is trivially satisfied by definition: the element in spec-IP is in the same position as the subject with respect to its trace in the small clause. Head government is also satisfied: a verbal head, namely the copula, governs the trace of the predicative element in the small clause. Nevertheless, it seems to me that these types of constructions are better understood if one adopts Chomsky's (1986b) version of the ECP, namely the one based on 'extended chains' (Chomsky 1986b: 78), provided that the notion of chain is extended to include those heads which contain the agreement features of the moved element. To establish, then, whether Chomsky's (1986b) version of the ECP is satisfied by inverse sentences like (43b), we must show that a predicative DP can independently trigger a spec-head relation with the I°-system. This can be accomplished in the following way. We will consider a nominal copular sentence where the two DPs mismatch with respect to number, as in the following: (50a) b
[DP John and Mary]+p,ur [DP the cause of the riot]+sing
The raising of predicates: a unified theory of copular sentences 39 The question now is: what kind of verb agreement will DP raising yield in either a canonical or an inverse sentence? Applying this test we obtain the following very clear result: (51a) b
[John and Mary]+plur are+plur [Sc t the cause of the riot] [the cause of the riot]+sing is+sing [sc John and Mary t]
Clearly, the two DPs in question preserve their number and determine verb agreement.28 If we fill in the slots of the I°-system with the DPs in question we will see rather better what is happening (to simplify the representation the usual V°/°label will be employed to indicate the complex formed by raising of the V° to F): (52)
John and Mary
V°jo
are
Since verb agreement is determined by a spec-head relation with (some head contained in) the T-system, the matching of the verb with the preverbal DP follows as a natural consequence. Interestingly, apart from the terminology, this type of proper government via head agreement, as proposed in Chomsky (1986b), is directly reminiscent of the mechanism of proper government proposed in Rizzi (1990) for preverbal subjects in spec-IP in English. In a simple sentence like who t left? the ECP would be satisfied in the same way as in the case of copular sentences:
40 (53)
The anomaly ofcopular sentences: the raising of predicates a
CP
Mutatis mutandis, the trace is licensed in both cases by activating the specposition of thefirstc-commanding head, provided of course that it is capable of bearing features. Coindexing the spec-position with the head indicates this special relation involving feature-sharing. Henceforth, I will adopt a simple locality condition on traces of DPs pushing the theories based on the notion of extended chains to the limit (see Moro 1993b for extensive discussion of this approach). For a first approximation, I will just assume that a trace satisfies locality (in the narrow sense of the ECP conditions) if there is a local head containing the features of the moved element. Technically, there might be several equivalent ways to implement this condition, depending on the way agreement is represented in syntax: for example, by means of index-sharing processes under a spec-head relation or by assuming incorporation of the relevant head with an abstract Agr°. In any case, the intuition underlying this proposal is clear: locality (again, in the narrow sense of the ECP) depends on morphology. More explicitly, the distribution of traces is inherently related to the morphological properties of the local environment where they happen to occur, agreement being the relevant morphological feature. Incidentally, this version of the ECP is particularly congenial to the spirit of the ECP as expressed in the Pisa Lectures: 'it is not unreasonable that Universal Grammar should require that the presence of an empty category
The raising of predicates: a unified theory of copular sentences 41 be signalled in some manner by elements that are overtly present' (Chomsky 1981: 251).29 All in all, then, the locality conditions assumed by (most versions of) the ECP module yield no obstacles to the flexible structure which is being proposed here: in inverse copular sentences like (43b), as far as locality is concerned, the chain established by the predicative DP is essentially no different from the chain of the subject DP in a canonical sentence: in both cases, the raised DP passes through the spec-position of the local ccommanding head activating agreement on it.30 The third preliminary problem which needs to be discussed involves Case theory. Arguments are assumed to need Case in order to fulfil the visibility requirement on 6-assignment (see Chomsky 1986a); thus, the subject must be raised to spec-IP, which is the only Case position available. Is it right to assume that a predicative DP can be raised to this position? As a first step, let us consider the following case: (54a) b
[*(for) [DP a picture of the wall]j to be [sc tj the cause of the riot]] is unusual [*(for) [DP the cause of the riotjj to be [Sc a picture of the wall tj]] is unusual
Traditionally, the first sentence is analysed by assuming that for is inserted to assign Case under government to the subject DP to overcome the lack of Case assignment on the part of the infinitival copula. Since whatever we say for thefirstsentence must also apply to the second one, we can conclude that a predicative DP can also be assigned Case on a par with the subject. On the face of it, this conclusion is not immediately compatible with the hypothesis that Case is assigned in fulfilment of the visibility requirement on 6-role assignment (Chomsky 1986a). In fact, predicates are normally assumed to assign 6-roles, rather than having 9-roles assigned to them. Moreover, the fact that the subject is left in situ within the small clause raises another problem: as we have already seen, there are no Case assigners within a small clause. Are we to conclude then that the subject simply does not receive Case? The answer to this is rather delicate. As a first approximation, it seems reasonable to adopt a 'conservative' position by assuming that the DP which remains in situ 'copies' the Case which is assigned to its raised counterpart. This solution is not ad hoc: in fact, instances of what is being proposed - which we might refer to as 'Case agreement' - are independently attested in languages with a rich case morphology, such as Latin: (55a)
Caesar dux/*ducem est (Caesar-nom leader-nom/*-acc is) 'Caesar is the leader'
42
The anomaly of copular sentences: the raising of predicates b
senatores dicunt Caesarem ducem/*dux esse (senators-nom say Caesar-acc leader-acc/*-nom to-be) 'the senators say Caesar is the leader'
c
Caesar dicitur dux/*ducem esse (Caesar-nom is-said leader-nom/*-acc to-be) 'Caesar is said to be the leader'
As these examples make very clear, the two DPs linked by the predicative relation, namely the subject Caesar and the predicate dux (leader), display exactly the same Case, the choice depending on context of assignment: thus, we have Nominative if the clause is finite, Accusative if the clause is infinitival. Alternatively, one could assume that the DP left in situ is simply assigned a 'default' Case, an idea which has been independently assumed to account for pronoun morphology in cases like the following: (56)
[IP [you and me] will never get along]
The pronoun me is not in a context where normally such a form is allowed (cf. *me will not go): since there is no reason why the coordinate structure should contain a Case assigner (for coordination see Kayne 1994 and references cited there), we can reasonably assume that me is displaying a default Case. Crucially, we may observe that in inverse copular sentences with a pronominal subject, the same Case may occur: (57)
[IP [DP the cause of the riot]j is [Sc me tj]]
A formal way of bringing the inverse structure into line with Case theory becomes possible if one adopts the minimalist approach to such an issue. In this framework, Case assignment is reduced to a process of spec-head relation between a DP and Agr° (at some point of the derivation; see Chomsky 1993, 1994). This amounts to adopting a very elaborate clause structure including the advanced version of the split-INFL hypothesis, specifically the one assuming two distinct Agr° heads (see section 1.1). Consider the enriched structure given in such a framework to clausal sequences of the type DP - V - DP (spec-TP omitted):
The raising of predicates: a unified theory ofcopular sentences 43 (58)
Agr-s-P
V°
DP
In this representation, there are two different positions where the two DPs might acquire Case: spec-Agr-s-P and spec-Agr-o-P. In the higher position Nominative Case is assigned (by [Agr° T°-Agr°]) while in the lower position Accusative will be assigned (by [Agr° V°-Agr°]). It is also assumed that while subjects raise overtly, objects generally raise only after spell out takes place, in the LF component.31 If we adopt this double Agr° structure for copular sentences the null hypothesis would be that preverbal DP raises to spec-Agr-s-P while the DP which is left in situ raises to spec-Agr-o-P at LF, paralleling the general situation. However, this would run counter to the fact that only one Case is assigned within copular sentences, as suggested by the Latin examples. Thus, if this theory of Case assignment involving Agr° projections is to be adopted, it seems reasonable to conclude that only one Agr° is present in such structures:
44 (59)
The anomaly ofcopular sentences: the raising of predicates AgrP
Furthermore, I think that there are good independent empirical reasons not to extend the double Agr-projection structures assigned to sentences of the type DP - V - DP ((58)) to copular sentences; this will be discussed in the next section. Let us now summarize the content of this long preliminary discussion. After the proposal of inverse copular sentences (in (43b)), three potential problems raised by these structures have been approached and solutions to them tentatively put forward: the rigid order of subject and predicative DP in the small clause, the locality condition on the chain of the raised predicate and Case assignment. Indeed, it turns out that there are no real obstacles to assuming the inverse structure. What we might ask now is whether there are any advantages. In the next section, we will go back to the main issue of this chapter, the unexpected absence of the subject-object asymmetries in copular sentences. For the sake of clarity, notice that the presence of a predicative DP is only one necessary condition for having an inverse sentence: in fact, for reasons that I will not pursue here, the predicative DP cannot be raised if the D° is an indefinite article, as in the following examples:32 (60a) [IP [DP John]j [v° is] [sc tj [DP a fool]]] b * [IP [DP a fool]j [v° is] [sc John tj]] More generally, it is clear that the theory presented here is not powerful enough to explain the syntax of all types of copular sentences. Nevertheless this limit does not undermine the essential proposal of this research, which aims at refining our understanding of the theory of clause structure.
The raising of predicates: a unified theory ofcopular sentences
45
1.3.2 Focus on inverse sentences: two interacting principles We will now reconsider the empirical facts set out in section 1.2, bearing in mind the new distinction between canonical and inverse copular sentences. 33 For ease of exposition, we will consider English cases first. The first major property which we have found in the inverse copular sentence is that the subject cannot undergo A'-movement (wh-movement and Q-raising) although it is in a postverbal position. The relevant examples, which involve wh-movement, are rephrased here. Consider the pair of sentences: (61a) b
(you think that) [IP [DP a picture of the wall]j was [Sc tj [DP the cause of the riot]]] (you think that) [iP [DP the cause of the riot]j was [Sc [DP a picture of the wall] tj]]
If we try to extract the subject a picture of the wall, a sharp contrast is produced: (62a) [which picture^ do you think [IP tj was [ sc tj [DP the cause of the riot]]]? b * [which picture]j do you think [rP [DP the cause of the riot]j was [Sc U tj]]? A similar effect occurs with Q-raising. Consider the two following sentences: (63a) b
(you think that) [IP [DP every bookjj was [Sc U [DP many students' purchase]]] (you think that) [IP [DP many student's purchase^ was [Sc [DP every book] til]
Only in the canonical sentences can the subject every book have wide scope over many students'. In the inverse copular sentence the subject cannot move. This is captured in the following representations: (64a) [IP every bookj [IP tj was tj [Dp many students' purchase]]] b * [IP every bookj [iP [Dp many students' purchase^ was [Sc tj tj]]] The proposal made in the present study allows us to account for these phenomena in a rather natural way. The advantage offered by the theory of copular sentences, and specifically by the hypothesis of inverse copular sentences, is clear. The postverbal noun phrase of an inverse sentence is seen to be in a left-branch position ((65a)), essentially in the same type of position as a preverbal subject ((65b)):
46 (65)
The anomaly of copular sentences: the raising of predicates a
V
Neither DP can be moved out unless it passes through the spec-position of the c-commanding head activating agreement on it. If this is already filled by another element, extraction is now expected to be blocked. Specifically, in inverse copular sentences the raising of the predicative DP is expected to block further extraction of the small clause DP subject, as is indeed confirmed by the facts observed above.34 All in all, we can conclude that our theory is successful on this score. In a copular structure only one of the two DPs can be extracted from the small clause because there are not enough heads to license two traces or, equivalently, not enough spec-positions to license the chains of two DPs. For the sake of clarity, from now on we will simply refer to this type of violation we have been discussing as violations of locality conditions, irrespective of the specific theoretical framework that is adopted. 35 Interestingly, these phenomena involving agreement shed further light on the special nature of the copula as a verb. Suppose one assigns to copular sentences of the type DP - V - DP the canonical two-Agr° projection structure assigned to DP - V - DP structures with verbs other than the copula:
The raising of predicates: a unified theory of copular sentences 47 (66)
Agr-s-P
be
DP
As far as I can see, it would not be possible to derive the absence of movement of the lower subject in inverse copular sentences. Let us see why. The predicative DP moves up to spec-Agr-s-P, activating agreement and thus satisfying locality conditions. After the predicative DP has raised to spec-Agr-s-P, there would always be a residual way for the subject DP left in situ to move out of the small clause, say by wh-movement. In fact, it could satisfy locality conditions by passing through the spec-position of the lower Agr-o projection, exactly as the object of a transitive verb does (see Chomsky 1993). Yet this is not reflected in our data. There seems to be no alternative, then, to concluding that the only possible structure for a copular sentence is the one that contains a single Agr°:
48 (67)
The anomaly of copular sentences: the raising of predicates AgrP
be
DP
In this representation, once the predicative DP has raised to spec-AgrP there is no way for the DP in situ to get out of the small clause. If it does, a sharp violation of locality conditions on movement is expected, and this is indeed reflected in our data. Notably, this conclusion fits in with what we concluded by considering Case theory as applied to copular sentences, where Case assignment is to be considered as a spec-head relation with an Agr head (see the end of section 1.3.1). In fact, we independently concluded that there cannot be two separate Case positions for the predicative and the subject DP. Rather we supposed that a process of 'case agreement' takes place, as happens in languages like Latin, where the two DPs overtly display the same Case. Locality theory thus brings us to the same conclusion as Case theory: in copular sentences there is only one Agr projection. In other words, the copula appears to be a defective verb: it is like transitive verbs, in that it occurs with two distinct DPs but it differs from them in that it contains only one agreement projection. The descriptive generalization we can formulate is that this can only happen with the copula because 'case agreement' appears to be instantiated only between two DPs that are linked by a predicative relation. Interestingly, this also suggests a generalization concerning the distribution of Agr° as a function of the distribution of argumental DPs: there must be an Agr° for every argumental DP, but not necessarily for every DP. All in all, we can conclude that, given our inverse structure, extraction from the subject of inverse copular sentences yields a violation of locality conditions on movement (in the narrow sense of the ECP-like violations)
The raising of predicates: a unified theory of copular sentences
49
and we can now examine the second major property that we observed in inverse copular sentences (in section 1.2). In inverse copular sentences, extraction from within the subject yields an ungrammatical result, although it is in a postverbal position. It was shown that this happens with both types of A'-movement, i.e. wh-movement and Q-raising. For ease of exposition, the pair of associated copular sentences given earlier are reproduced here: (68a) b
(you think that) [IP [DP a picture of the walljj was [ sc tj [Dp the cause of the riot]]] (you think that) [IP [DP the cause of the riotjj was [ sc [DP a picture of the wall] tj]]
A violation takes place if wh-movement extracts a DP embedded in the subject as opposed to one embedded in the predicative DP: (69a) b
[which riot]; do you think [DP a picture of the walljj was [ sc tj [DP the cause oft,]]? * [which walljj do you think [DP the cause of the riotjj was [ sc [DP a picture Oftjjtjj?
Similarly, consider the following sentences: (70a) b
[IP [DP a picture of the walljj wasn't [ sc tj [DP the cause of many riots]]] [iP [DP the cause of the riotjj wasn't [ sc [Dp a picture of many walls] tjjj
As we observed in section 1.2, the quantified DP embedded in the postverbal subject cannot be extracted, in contrast to that which is embedded in the predicative DP. This was captured by means of the usual LF representations showing the movement in question: (71a) b
[iP [DP many riots]; [jP [DP a picture of the walljj wasn't [ sc tj [DP the cause oft,]]]] * [iP [DP many wallsjj [iP [DP the cause of the riotjj wasn't [ sc [DP a picture of
What kind of principle can account for these types of violation? In this case, it is easy to conclude that (the various versions of) locality theory stemming from the original ECP proposal cannot furnish plausible explanations for the ungrammatical cases. In the previous cases ((62)-(63)), violations were due to the fact that raising of the predicate to spec-IP exhausted the only available strategy for the trace of a DP in a small clause to satisfy locality, i.e. activating agreement features on the first c-commanding head. In the present case, instead, the very fact that spec-IP is occupied by a DP does not create any interference; witness the grammatical examples ((69a) and (70a)). Thus, these violations cannot be treated on a par with the previous
50
The anomaly ofcopular sentences: the raising of predicates
cases. On the other hand, it would make no sense to assume that the trace in the DP headed by picture is not licensed for reasons related to the internal structure of the DP involved. Indeed, sentences like which wall do you think John saw a picture of? are perfectly grammatical. All in all, to explain these contrasts we have to appeal to a different principle. On the face of it, the paradigm under discussion seems to indicate that extraction from a DP in situ in the small clause depends on the grammatical function of the DP: the results are good if the DP is a predicate, and bad if the DP is a subject. This would, however, be misleading. What makes the difference is not in fact the grammatical function of the DP. Indeed it is easy to show that extraction from the predicative DP is just as bad if the predicative DP is raised to spec-IP: (72)
* [which riot]j do you think [iP [DP the cause of tj]j was [Sc [DP a picture of the wall] tj]]
In this case, the predicative DP behaves exactly like a subject in that it does not allow extraction. In general, violations of this type are not due to the ECP but rather to the Subjacency Condition. In what follows, an attempt will be made to account for the violation contained in the paradigm under discussion in terms of the Subjacency condition. The Subjacency condition was originally formulated as a generalization covering a series of violations that have features in common: they all involve an extraction that crosses at least two constituents of a certain kind (the socalled 'bounding nodes') without intermediate steps (see Chomsky 1977a). The most advanced formulations of this principle (stemming from Chomsky 1986b) attempt to give an intensional definition of what counts as a bounding node rather than a mere list.36 In general, it is assumed that a bounding node is a maximal projection that fails to be governed by a head that assigns a 6-role to it (in this case we have an 'inherent barrier') or one in such a position that the closest maximal projection it dominates is an inherent barrier (in this case we have a 'barrier by inheritance'). The only exception to this paradigm is IP: this maximal projection can only be a barrier by inheritance (otherwise it would always be a barrier, given that C° does not assign a 9-role to it). Subjacency is then violated when two barriers are crossed in a single step. For example, a case like (72) is straightforwardly ruled out: the DP in spec-IP fails to be governed by a head assigning it a 6-role (technically, it fails to be 'L-marked') so it is an inherent barrier; this makes IP a barrier by inheritance and the sentence is therefore ruled out:
The raising of predicates: a unified theory ofcopular sentences (73)
51
C
the cause of t
Cinque (1990a) refined the formulation of the Subjacency Condition by proposing two differences with respect to Chomsky's (1986b) version: first, a single barrier is sufficient to trigger a Subjacency violation; second, Lmarking is not defined in terms of 6-role assignment but in terms of 'selection' (see Cinque 1990a: 41-2 and references). In simple terms, the idea is that a maximal projection is L-marked only if it is selected by a ccommanding head which is not distinct from / + V/. This version of the Subjacency Condition offers a partial solution to the present problem.37 The impossibility of extracting from the embedded subject now follows without any recourse to the ECP, as required. Let us go back to the offending structures ((69b) and (71b)) repeated here: (74a) * [which wall]; do you think [DP the cause of the riot]j was [ sc [DP a picture Oftjtj]?
b
* [ip [DP many wallsjj [JP [DP the cause of the riot]j wasn't [Sc [DP a picture of
tj tjiii
As we noticed before, the two DPs where these extractions originate turn out to be in the same type of syntactic configuration as the preverbal subject; in other words, they are in a left-branch position immediately dominated by a clausal constituent: (75)
... t...
... t...
As has just been argued, in the first case extraction from the DP is banned because C° does not select the lower DP and this turns it into a barrier. As
52
The anomaly of copular sentences: the raising of predicates
for the second case, we simply have to reproduce the same argument. The copula does not select the lower subject DP and this turns it into a barrier. All in all, in both cases the embedded DP counts as a barrier and extraction from it is banned by the Subjacency Condition. This is a good result since, as we established earlier, this violation had to be distinct in principle from the violation of locality conditions examined earlier. The success is only partial, though. We must still explain why the extraction from the predicative DP is possible when this is left in situ within the small clause as in (69a) and (71a), repeated here as (76a) and (76b) respectively: (76a) b
[which riot]j do you think [DP a picture of the wall]j was [sc tj [DP the cause oftj]? [ip [DP many riotsjj [[P [DP a picture of the wall]j wasn't [sc tj [DP the cause Oft;]]]]
Why do these movements not involve a violation of the Subjacency Condition? The answer to this question is not straightforward. Indeed, potentially, we have at least two distinct lines of reasoning which both depend on the analysis we give for small clauses. In fact, there are at least two competing analyses for the small clause constituent; we will now briefly review them. In the early 1980s, it was proposed that small clauses should be analysed as a kind of adjunction structure, where a subject DP is adjoined to a maximal projection related to a lexical head (see Stowell 1981 and Manzini 1983 among others): (77a)
XP = SC
The intuitive idea behind this proposal was that a small clause consists solely of a predicative relation and that this relation is not mediated by any functional head. 38 In other words, a small clause is anomalous in that its distribution is that of a phrase, i.e. a full maximal projection (XP), but its inner structure is not the projection of a head. More recent proposals have suggested that the small clause can be regarded as a projection of some type of functional head (for a general discussion of small clauses see Cardinaletti and Guasti 1995, Bowers 1993 and Moro 1988). In Moro (1988), for example, it was originally proposed
The raising of predicates: a unified theory of copular sentences 53 that the small clause complement of the copula should be analysed as an AgrP (and the copula as T°; see fnn. 17, 58): (78a) b
[AgrP [Gianni] [Agr, Agr° [il mio miglior amico]]] (Gianni-masc.sing. the my best friend-masc.sing.) * [AgrP [Gianni] [Agr, Agr° [i miei migliori amici]]] (Gianni-masc.sing. the my best friends-masc.plur.)
Under this analysis, predicative DPs were in fact governed by a head, namely Agr°. The consequences of adopting one analysis rather than the other are of course relevant for the present discussion. Before showing what these consequences might be, I would like to illustrate briefly the complicated debate about small clauses by highlighting two separate issues that play a central role in this debate: the role of agreement in predication and the distribution of adverbials in small clauses. We will begin with the first of these two topics. In the conclusions of Moro (1988) it was observed that, while the representation of small clauses as AgrPs solves many problems, there remain a number of empirical facts which fail to support such a hypothesis. I will now present a number of considerations that were set out briefly in that paper (see also Longobardi 1988 for arguments against the hypothesis that small clauses contain heads).39 Consider two DPs which do not match either in gender or in number, as in the following case: (79a) b
[DP quest-i libr-i] (these masc.plur. books masc.plur.) [Dp 1-a caus-a della rivolta] (the fern.sing, cause fern.sing, of-the riot)
These two DPs can enter into a predicative relation, yielding a perfectly well-formed nominal small clause. If a small clause is an AgrP, then we expect them to agree. Let us now examine the data: (80)
Gianni ritiene [sc [DP questi libri] [DP la causa della rivolta]] (Gianni believes these books the cause of the riot)
Clearly, there is no agreement at all. The two DPs enter into a predicative relation without matching any component of their (^-features. Leaving aside the question of how to explain this important difference between the type of DP under discussion and the other categories (for DPs in general see for example Higginbotham 1990), we can simply conclude that there is at least one case where agreement between the subject and the predicative
54
The anomaly ofcopular sentences: the raising of predicates
DPs of a small clause cannot be realized. As usual, we will take the existence of a single counterexample to be sufficient reason to conclude that all other cases should be considered as epiphenomena. Applying such reasoning to the case in hand, we would naturally conclude that predication cannot always be identified with the projection of an Agr°. Interestingly, this conclusion appears to be reinforced by independent data. It is a well-known fact that in those languages where small clauses can constitute a full matrix sentence, like Russian, for example, agreement in the predicate is not only unnecessary but impossible. This provides a sharp contrast with those cases where the same element occurs as a modifier, in that when it occurs as a modifier agreement is obligatory. A classic example is the one given in Jespersen (1924: 120): (81a)
b
dom nov (house new -agreement) 'the house is new' dom nov-yj (house new + agreement) 'a/the new house'
The conclusion is clear: in Russian, only when agreement is absent do we have a predicative relation, and consequently a clause structure. 40 To sum up, even if the predicative relation sometimes involves a form of agreement this cannot be regarded as a necessary condition for this relation to be recognized. Put somewhat differently, on the basis of the simple fact that nominal small clauses contain an instance of predication without agreement we conclude that small clauses do not necessarily have to be analysed as AgrPs.41 They may be AgrPs, of course, as in the case of AP predicates; however, this must be regarded as an accidental fact that has nothing to do with predication but rather with the specific morphological requirements of adjectives. Thus, if small clauses are projected by a head, it cannot be that this head is an Agr°.42 A different empirical approach to the inner structure of small clauses comes from the analysis of the distribution of adverbial elements. Cinque (1994) proposes that adverbs can occur in spec-position of functional heads (see Lonzi 1991 for a different approach). Thus, if we can show that adverbs occur within a small clause we would have indirect evidence for assuming that this type of clause structure is actually projected by a functional head. In what follows we will explore this possibility. Let us begin by considering the following examples:
The raising of predicates: a unified theory of copular sentences (82a) b
55
Gianni e sicuramente il colpevole (Gianni is surely the culprit) Gianni non e piu il colpevole (Gianni not is any longer the culprit)
There are two adverbial elements in these sentences, namely sicuramente (surely) and piu (any longer) (see Zanuttini 1992 for a detailed discussion of the distribution of piu). The question arises of whether they are within the small clause or outside it, perhaps in the spec-position of some morphological head X° related to verbal inflection. If we restrict our inquiry to these cases, no answer is possible. Indeed, since the subject has been displaced both orders are possible and there is no way of distinguishing between them: (83a) b
. . . [Xp adv [ sc t . . . • • • [sc [XP adv. . . . t . . .
However, inverse copular sentences offer the crucial information on this point. We have established on independent grounds that the subject of an inverse sentence cannot be moved from the position it is generated in within the small clause. Thus, it can be of value in signalling the position of the adverb: if the adverb follows the subject of an inverse sentence, we can safely assume that it stands in a small clause internal inflectional specposition. Otherwise, we have indirect evidence against the hypothesis that small clauses are projected by a head. Let us now examine the data: (84a) * il colpevole e Gianni sicuramente (the culprit is Gianni surely) b il colpevole e sicuramente Gianni (the culprit is surely Gianni) c d
* il colpevole non e Gianni piu (the culprit not is Gianni any longer) il colpevole non e piu Gianni (the culprit not is any longer Gianni)
These examples all point to exactly the same conclusion: if we look at the linear order of constituents in inverse copular sentences, the subject must always be the last element to occur. Combining this observation with the independent assumption that the DP in question is actually in the subject position of the small clause, we must conclude that there is no spec-position available for adverbs in small clauses; this strongly diminishes the plausibility of analysing small clauses as projections of functional heads. 43 Conversely, when they occur as complements of believe-iype verbs, we should conclude that they are not in fact 'bare' small clauses, as they are
56
The anomaly ofcopular sentences: the raising of predicates
when occurring as complements of the copula. Rather, the complement of a believe-ty^ verb would be better analysed as an 'enriched' small clause, which includes some spec-positions of a higher inflectional head presumably related with verb morphology. All in all, the moral is that the distribution of adverbs cannot be used to prove that small clause complements of the copula are projected by a head. We now conclude this brief digression on small clauses. It seems that at the moment there are no independent and cogent reasons to adopt the one analysis or the other for small clauses. Indeed, if they were analysed as projections of an inflectional head, the possibility of extraction from the predicate but not from the subject would follow naturally, given that the predicate would be in the complement position of the head selecting it, while the subject would be in a left-branch position. 44 Turning now to the analysis of small clauses as adjunction structures (note that we are discussing small clauses with predicative nominals only), it seems to me that it would still be possible to explain the data under discussion even on the basis of this analysis. Notice, first, that although the X-bar module is currently undergoing a radical rethinking (see Kayne 1994, Chomsky 1994, Brody 1994), the possibility of adjunction structures is not completely ruled out. Let us now reflect for a moment on the nature of adjunction. The basic intuition underlying adjunction (of oc to (3) is that a certain constituent a can have the same distribution as P and can cooccur with it without being either the specifier or the complement of (3. In other words, the adjunction structure is a sort of 'enlarged p'. Technically, a small clause with a predicative DP would look like the following structure where the subject (DPS) is adjoined to the predicate (DP p ): (85)
DPp = SC
Now, if we take the hypothesis that small clauses are bare projections of predicates seriously, we can interpret the absence of Subjacency violations when extracting from within the predicate in a rather natural way. Since the copula selects a small clause and since the latter is nothing but the predicate with an adjunct, then we can reasonably assume that there is a selectional relation between the copula and the predicate. Thus, extraction from the predicative DP does not violate the Subjacency Condition because the predicative DP does not count as a barrier. Extraction from the subject, on the other hand, is indeed predicted to be blocked because there is no
The raising of predicates: a unified theory ofcopular sentences
57
selectional relation between the copula and the subject of the small clause. Notice that this argument would also be compatible with the explanation we gave for the impossibility of extracting an element across a predicative DP when this is raised to spec-IP. In this case, the Subjacency Condition would be violated because the first c-commanding head does not select the lower DP in spec-IP ((72)). To avoid potential objections, it should be pointed out that extraction of either DP from the small clause would not yield structural problems if we adopt the adjunction analysis. In fact there is independent evidence that extraction of both the adjoinee and the adjunct is possible from an adjunction structure. Consider for example the following cases: (86a) b c
John ate [DP the meat] John ate [DP [DP the meat] [AP raw]] [DP the meat]i was eaten [DP tj [AP raw]]
The DP to which the secondary predicate [AP raw] is adjoined (see Rothstein 1983 and references cited there) can be extracted. Similarly, one could consider examples like the following: (87a) b c d
John ate [DP the meat] John ate [DP [DP the meat] [AP raw]] [what kind of meat]j did John eat [DP tj [AP raw]]? [how]j did John eat [DP [DP the meat] tj]?
Thus, the assumption that small clauses are adjunction structures would not make them incompatible with movement of either DP. We can now conclude this discussion of the second fundamental property of the subject of inverse sentences. If the analyses of small clauses as projections of a head were to be adopted, we would have an immediate explanation of what makes extraction possible from predicates that remain in situ and yet impossible from the subject, since the latter would be in a left-branch position. On the other hand, even if we adopt the adjunction analysis, two important conclusions can nevertheless be drawn: (i) the impossibility of extraction from the subject of an inverse sentence should not be traced back to locality conditions but rather to the Subjacency condition; (ii) for some reason, arguably related to the nature of the small clause constituent, this restriction does not apply to the extraction of predicative DPs when they are left in situ. For our purposes, these two observations will enable us to understand a number of phenomena which are going to be discussed in the following sections, although a complete understanding of such phenomena will only be possible once the structure of small clauses has been thoroughly comprehended. Before investigating
58
The anomaly ofcopular sentences: the raising of predicates
the phenomena in question, however, we must solve some residual questions our analysis raises. A potential counterexample to the account given of the impossibility of moving the subject of an inverse sentence is provided by cases like the following: (88)
[what]j do you think [IP [DP a picture of the wall]j was [Sc tj tj]]?
Why is this sentence grammatical?45 If we do not want to abandon the idea that within a small clause one single DP trace at most can be licensed, we must find an alternative way of allowing the trace of what to fulfil locality conditions. One possibility is to reduce this case of movement out of a small clause to a case of extraction out of the predicative DP. This task can be accomplished by exploiting the articulated representation for noun phrases as projections of a functional head, i.e. D°. Bearing this in mind, the status of which and what can be distinguished as a first approximation by referring to this more articulated structure: (89) spec
D
which
spec
...
spec
N'
what
Thus, the potential offending sentence should be represented as follows: (90)
[NPwhat]j do you think [DP a picture of the walljj was [sc tj [Dp (tj) D°tj]]]?
If this proves tenable, the trace of what would now be properly governed by D° passing through spec-DP, and would thus overcome the impossibility of activating agreement on the copula.46 To support the hypothesis that what is extracted from the complement of D° consider the following independent facts: (91a) [DP what a girl] I met! b * [DP which a girl] I met! The contrast between these two sentences shows that what can be present
The raising of predicates: a unified theory of copular sentences 59 within the projection of the D°a, as opposed to which. The position of what in the first sentences can be derived by considering the following independent cases: (92a) b
[DP a [NP [AP very nice] girl]] [DP [AP how nice]j a [NP tj girl]] I met!
In the latter sentence, the wh-phrase is clearly raised to spec-DP from a lower position.47 Thus, we can reasonably assume that what also raises from a position lower than D°. In other words, we can conclude that what does not stand for the full DP but rather is extracted from it, arguably via spechead agreement. Thus the potential counterexample is effectively dealt with: there is no violation of locality conditions in the case under discussion. There is no extraction of a wh-phrase as a whole; rather the wh-element is subextracted from the predicative DP which remains in situ, something which is independently regarded as a proper instance of movement.48 The last residual observation concerns examples like the following: (93)
* [which causejj do you think [DP a picture of the wall]i was [sc tj tj]?
This case shows that the predicate of a canonical copular sentence cannot be extracted. Bearing in mind what we said about the impossibility of extracting the subject of inverse sentences, this result is hardly surprising. There is only one strategy for a DP to get out of a small clause without violating locality conditions on movement: it must pass through the spec-IP position. Since this position is already occupied by the subject of the predication, the trace of the predicate cannot be licensed: accidental agreement between the copula and the moved element simply would not count. Although this explanation is sufficient to rule out this sentence, it seems that this example violates something more than a grammatical principle. Mutatis mutandis, we can follow Geach (1962: 35; quoted in Higgins 1973): it is clearly nonsense to ask which cat cat stands for in Jemima is a caf.49 Whether this 'nonsense' can be entirely traced back to grammar is not a question that can be gone into here; nor do I have any decisive argument to offer. We have now reached the end of this section, the aim of which was to illustrate the advantages offered by assuming that predicative DPs can be raised to the most prominent DP position in clause structure. What we have shown is that the newflexiblestructure allows us to derive the anomalies of copular sentences with respect to the usual subject-object asymmetries. The postverbal DP of an inverse sentence is in a left-branch position and the
60
The anomaly of copular sentences: the raising of predicates
copula agrees with the raised predicate. This is sufficient to explain why the lower DP appears to be 'trapped' in that position as if it were a preverbal subject. Moving out the whole phrase would violate the locality condition on movement. Extracting from it would induce a Subjacency Condition violation, since the DP is in a left-branch position and counts as a barrier. So far we have discussed data taken from English. To complete the analysis, we can now turn to specific cases involving Italian, namely ne-extraction, /o-cliticization and verb agreement (cf. section 1.2.2) and see whether the inverse structure explains the data. The first phenomenon mentioned, Aie-extraction, can be dealt with immediately as follows: (94a)
[DP una foto del muro]j nej fu [sc tj [DP la causa t{]] (a picture of the wall of-it was the cause) b * [DP la causa della rivoltajj nej fu [Sc [DP u n a foto tj] tj] (the cause of the riot of-it was a picture)
The process of fte-cliticization is blocked from the postcopular DP of an inverse sentence ((94b)) by the same factor that blocks wh-movement and Q-raising, namely the Subjacency Condition. Such a DP is a subject (structurally, it is an adjunct to a predicative DP) and the only ccommanding head, i.e. the copula, does not select it. Thus it counts as a barrier. In a canonical sentence ((94a)), on the other hand, ne can be extracted from the postcopular DP because it crosses a predicate and the copula selects it, since it selects a projection of it, namely the small clause. Correspondingly, the latter case does not trigger a Subjacency violation when both wh-movement and Q-raising are tested.50 The two residual properties involving verbal agreement and /o-cliticization are less straightforward to explain and they will turn out to be closely related. For the moment, let us simply start by observing that /o-cliticization of a postcopular DP is impossible if the DP is a subject. Providing a principled account of these two phenomena will be the task of the next section. 1.3.3 Verbal agreement and cliticization in copular sentences In English copular sentences, when the subject and the predicative DPs fail to match in terms of number, the copula always agrees with the raised one. On the other hand, in Italian, it always agrees with the subject: (95a) b
[IP [DP le foto del murojj sono [sc tj la causa della rivolta]] (the pictures of the wall are the cause of the riot) [IP [DP la causa della rivolta]j sono [Sc le foto del muro tj]] (the cause of the riot are the pictures of the wall)
The raising of predicates: a unified theory of copular sentences
61
This paradigm shows that the copula is always plural when the subject of the predication (lefoto del muro) is plural, even in cases where this phrase follows the copula. This raises a non-trivial problem. Verbal agreement is by definition ultimately established as a spec-head relation between a DP and (the proper head contained in) F. 51 Indicating the complex formed by movement of the verbal head (V°) to the inflectional features (1°) by 'Vp', we can represent the relevant portion of the clause structure as follows: (96)
If we do not want to abandon this fundamental assumption, we must conclude that in the cases under discussion, the preverbal predicative DP is not in spec-IP. In other words, we must dismiss the second of the following two structures: (97a)
sono
62
The anomaly ofcopular sentences: the raising of predicates
Assuming it was in spec-IP, the DP of the second structure should trigger agreement on 1°, yielding the third singular e (is) not the third plural sono (are). What is there then in such a position? Assuming that the predicative DP is not in the spec-IP position, then we have to ask what is in that position. Similarly, we have to ask where the predicative DP has ended up. Before answering these questions, we will need to rethink the status of pro. This will occupy our attention in the next three subsections: in the first, we will combine the theory of pro with the unified theory of copular sentences; in the second, we will discuss a major consequence of the result obtained in the first one; and in the last section, discussion will centre on the residual idiosyncratic property of Italian copular sentences involving cliticization. 1.3.3.1 pro as a'null predicate' It is generally assumed that pro either plays the role of a non-phonologically realized argument or else is an expletive binding an argument (see Burzio 1986 and Rizzi 1986).52 In fact, the latter assumption was made to explain examples like the following: (98a) b c d
pro telefona (pro telephones) pro telefona Gianni (pro telephones Gianni) 'Gianni telephones' pro arriva Gianni (pro arrives Gianni) 'Gianni arrives' il pittore dipinge pro nudi (the painter paints pro naked) 'the painter paints people naked'
Notice that the underlying assumption is that the structure of the clause in Italian is basically the same as in English. When a preverbal DP is missing in the superficial realization, a non-phonologically-realized pronoun is anyway present, namely (99)
The raising of predicates: a unified theory of copular sentences 63 This element can either be an argument (as in pro telefona) or an expletive binding an argument (see Rizzi 1982, 1990, and Burzio 1986). Instances of argumental pro are simple cases where pro plays the role of the subject, like pro telefona, or else cases where pro plays the role of a direct object, like // pittore dipingepro nudi. Expletive uses of pro are assumed to be those when the subject is rightward-moved, like pro telefona Gianni, or when the subject is in the object position of unaccusative constructions, like pro arriva Gianni. The next question is what kind of pro we have in sentences like the following: (100)
pro sono io (pro am I)
First of all, let us observe that we are not facing one of the familiar cases like pro telefono io. In other words, this is not a case of rightward movement of the subject; witness the following contrast: (101a) pro telefono io (pro telefono io) b io telefono (I telephone) c
pro sono io (pro sono io) d * io sono (lam) Thus, whereas pro telefono io can be related to a sentence like io telefono, it is by no means the case that pro sono io is related to io sono, for the latter is totally ungrammatical. Let us now explore the possibility that in such structures pro is an argument, specifically the subject of the small clause selected by the copula:
64
The anomaly ofcopular sentences: the raising of predicates
(102)
IP
This representation is the one corresponding to canonical copular sentences, i.e. pro is the subject and io is the predicate left in situ. Is this representation the correct one? 53 To answer this we can try to determine whether it shares the properties of canonical sentences which we established in section 1.2 of the present chapter. Two tests will be sufficient. If this representation is correct, we predict agreement not to be sensitive to the features of the predicative DP. This is not in fact the case: (103a)
pro sono io (pro am I) b * pro sono noi (pro am we) c pro siamo noi (pro are-first plur. we)
That io cannot be considered a predicate in sono io is also indicated by the process of /o-cliticization. Judgments must be cautious in this case, since the following sentence is in fact grammatical: (104)
pro IOJ sono tj
Yet it is simply not possible to associate this sentence with sono io. Rather lo sono is the counterpart of a canonical sentence like sono il colpevole (pro am the culprit; 4I am the culprit'). Given this, we are forced to assign to pro sono io the following representation:
The raising of predicates: a unified theory of copular sentences (105)
65
IP
The sentence under discussion turns out to be an instance of the inverse copular sentence where the subject is left in situ and the predicative DP is raised to spec-IP. If this is correct, then we face a sharp departure from current terminology. In general, since the pioneering work of Rizzi (1982), the pro element is considered to play the role of 'null subject' (in fact, the pro-drop parameter is often called the 'null subject parameter'). The analysis just presented of sono io implies, however, that pro can also play the role of a 'null predicate' (in fact, a 'propredicate') linking the position indicated by tt. Of course, since playing the role of predicate is a general property of the class of DPs, we do not need any special assumptions to deal with this.54 Finally, to eliminate possible confusion, we should also notice that the following sentences cannot be associated: (106a) b
sono io (ami) chi sono? (who am)
In other words, chi is not the wh-counterpart of io in sono io. If it were its counterpart, this would be a problem for our analysis because the subject of inverse copular sentences cannot undergo A'-movement. However, apart from the interpretation of the two sentences, there are purely syntactic reasons for excluding this. In fact, it is sufficient to note that if chi were really the wh-moved counterpart of the subject io, copular sentences would violate a general restriction banning wh-movement of first persons:
66
The anomaly ofcopular sentences: the raising of predicates
(107a)
arrivo io (arrive-first sing. I) b * chi arrivo? (who arrive-first sing.)
Accordingly, we will analyse chi sono as an instance of the canonical sentence where chi is extracted from the DP, paralleling the case of what discussed earlier in the text (for an interesting discussion concerning who as a predicate see Jespersen 1924: 153, fn.l, reporting Sweet 1892: §215). Some special remarks must now be made with regard to the 'content' of pro. Although pro can normally stand for any consistent combination of cofeatures (as in: pro telefono (I phone), pro telefoni (you phone), pro telefona (he phones), etc.), in pro sono io, i.e. in an inverse copular sentence, it must have the same (^-features as the overt DP (cf. *pro e io (is I), *pro siamo io (are I), etc.). This phenomenon can be explained as the result of the interaction of two independent factors. On the one hand, it seems that an inherent property of propredicative elements is that they do not have any features of their own. In fact, there is only one overt clitic form for predicates, i.e. to, as opposed to the rich paradigm for argumental DPs, i.e. le (third-fem.-plur.), // (third-masc.-plur.) la (third-fem.-sing.), etc. On the other hand, pro must have some features (in the last resort, it is nothing but a collection of features). It is the predicative relation that explains this phenomenon: pro simply copies the (^-features of the referential element it is predicated of. Indeed, there is independent evidence for assuming that two DPs of a small clause can agree when they are linked by a predicative relation: (108)
Maria considera [sc Gianni amic-o/*-i] (Maria considers Gianni friend-masc.sing./masc.plur.)
Before we turn to the main issue, it is worth noting that licensing of pro as a null predicate does not call for any special assumption: (109a) * Maria considera [Sc pro il colpevole] (Maria considers pro the culprit) b * Maria considera [sc Gianni pro] (Maria considers Gianni pro) c prOj sono [Sc tj il colpevole] (pro am the culprit) d prOj e [sc Gianni tj] (pro is Gianni) Clearly, the grammaticality of the latter two sentences shows that there is no intrinsic prohibition on the subject or the predicate of a nominal small
The raising of predicates: a unified theory of copular sentences 67 clause being pro. Where pro is not possible, on the other hand, it seems reasonable to assume that it is the local environment within the small clause that prevents it from being licensed. In fact, as soon as the chain of pro is 'extended' to reach a richer inflectional context, the element is licensed.55 Bearing this in mind, we are now in a position to explain the original problem concerning verb agreement in inverse copular sentences with nominal small clauses without giving up the fundamental assumption that verbal agreement is ultimately established by a spec-head relation between a DP and the F-system. 1.3.3.2 A case study in parameter setting Bearing in mind the result of the previous sections, I would like to propose a solution for the apparent anomalous rightward agreement in Italian illustrated in (95). Specifically, I propose to assign the following structural representation to a sentence like la causa delta rivolta sono lefoto del muro (the cause of the riot are the pictures of the wall): (110)
sono
[le foto del muro] X{
As this representation indicates, there is no need to give up the idea that verbal agreement is the result of a spec-head agreement with the inflectional system. What must be abandoned is the idea that la causa della rivolta is raised to spec-IP. Instead, we must assume that it is an adjunct to IP56 and that what actually forms a chain with the predicative position contained in the nominal small clause is a propredicative pro.51 Notice that the analysis of la causa della rivolta as an adjunct to IP (and not, say, to CP) can be independently supported by showing that in 4Auxto-Comp' constructions la causa della rivolta is left in a lower position, as shown in the following representation:
68 (111)
The anomaly of copular sentences: the raising of predicates CP
C° essendoj la causa della rivolta i{ le foto del muro (being the cause of the riot the pictures of the wall)
In short, verb agreement in Italian inverse copular sentences appears to be rightward, running against the fundamental assumption that verb agreement is the result of a spec-head relation between a DP and (the proper Agr° contained in) the I°-system. In other words, the copula appears not to be sensitive to the features of the preverbal DP of an inverse sentence. Given the analysis proposed here, however, the usual fundamental assumption about verb agreement can be fully maintained for Italian.58 Even so, a residual question emerges: why is it impossible for the predicative DP la causa della rivolta to be raised to spec-IP, thus triggering agreement? In fact, it seems to me that there are no cogent reasons why a fully lexical predicative noun phrase should not appear in spec-IP and there trigger agreement with the verb. Indeed, this is exactly what happens in English. Thus, there would appear to be no principled reasons for this process to be blocked in Italian. I would like to suggest that there is no 'explanation' as such of why Italian is different in this respect: the difference is just parametric. Notice that I am not simply adopting the usual view that in Italian pro can be licensed while in English it cannot. My claim is much stronger and can be summarized as follows: (112)
if pro can be licensed in an inverse copular sentence then it must be licensed.
In principle, this conclusion has a wide range of consequences. In fact, although it is drawn from the analysis of nominal small clauses and copular sentences, it has much wider consequences for the distribution of pro in general. It is currently assumed that pro is an optional element, in the sense that a sentence does not necessarily have to contain it, provided that an overt subject occurs. So, along with pro corre (pro runs) we can have Gianni corre (Gianni runs) which would not contain any instance of pro. Inverse copular sentences, on the other hand, show that pro must obligatorily occur: both when a preverbal DP is absent (pro sono io) and when it is present (la
The raising of predicates: a unified theory of copular sentences
69
causa della rivoltapro sono io). The conclusion suggests itself. Any theory that assumes pro to be obligatory only in copular sentences, and, more to the point, only in the inverse subtype of such sentences, would be totally implausible. Thus, we are forced to conclude that pro is always in spec-IP with all sentences. As a corollary of this, we are obliged to conclude that preverbal fully lexical subjects in Italian, like Gianni in Gianni corre, always occur in a position which is higher than spec-IP. All in all, we can refine our generalization concerning the distribution of pro as follows:59 (113)
if pro can be licensed then it must be licensed.
If this analysis proves correct, (inverse) copular sentences assume considerable importance for the problem of what sets the pro-drop parameter. This is a central question and one that has produced major advances in the theory as a whole since it was formulated (the literature in the field is vast; for just a general survey, see among others: Chomsky 1981, Rizzi 1982, Hyams 1986, Manzini and Wexler 1987, Platzack 1987, Huang 1989, Jaeggli and Safir 1989, Haegeman 1990, Radford 1990 and all references cited in each case). In general, it is assumed that positive evidence is available to children in a pro-drop language when they hear sentences lacking the overt subject in the pre verbal position of fully inflected sentences. Thus, for Italian children to know that they are in a pro-drop language it will be sufficient for them to hear a sentence like the following: (114)
viene (comes) 'he/she/it comes'
The children thus would learn that when the subject position is not phonologically realized it is occupied by a matrix of features, namely pro. Again, recall that a common (albeit tacit) assumption is that pro is an optional element. In other words, the theory says that when the preverbal position is occupied by a full DP, the Italian children will not have recourse to pro. 60 I would like to propose that inverse copular sentences contribute to prodrop parameter setting in the following way. Universal Grammar provides the child with the fundamental principle that verbal agreement is the result of a spec-head relation between a DP and (the proper Agr° contained in) the I°-system. To preserve this principle, when the Italian child is exposed to sequences of the kind DP-copula - DP showing agreement to the right, he or she is forced to assume that spec-IP is occupied by pro, while the preverbal DP is an adjunct to IP. 6i If this analysis proves correct, then we
70
The anomaly ofcopular sentences: the raising of predicates
can assume that a piece of positive evidence is available to the child regarding the appropriate setting of the pro-drop parameter which is totally independent of the classical evidence provided by cases such as viene (comes; 'he/she/it comes'), where a phonologically realized preverbal noun phrase is simply missing. In an inverse copular sentence of the kind la causa siamo noi (the cause are we) a preverbal noun phrase is indeed overtly present, but it does not interact with the I°-system: crucially, the assumption that/?ro is occupying spec-IP is forced.62 As an independent piece of evidence supporting this analysis of inverse copular sentences based on propredicative elements, consider the following French paradigm which, of course, does not involve pro: (115a)
[IP [DP je]j suis [Sc tj la cause de la revoke]] (I am the cause of the riot) b * [IP [DP l a cause de la revoke]] est [ sc moi t{]] (the cause of the riot is me) c [IP [DP la cause de la revoke] [IP c'j est [ sc moi tj]]] (the cause of the riot ce is me) d * [IP [DP ce que je veux]j est [ sc ma femme tj]] (what that I want is my wife) e [IP [DP ce que je veux] [IP c'j est [ sc ma femme tj]]] (what that I want ce is my wife) f [IP [DP ce que je veux]j est [ sc tx incroyable]] (what that I want is incredible)
The distribution of ce can be derived once we assume that it can play the role of a propredicative element: ce intervenes obligatorily only in inverse constructions ((115b-e) vs (115a and f)), arguably with a propredicative role (note: agreement is to the left in French). Of course, this does not exclude the possibility of having ce in canonical sentences: (116a) b
[ip [DP la lecture^ est [ s c tj ma passion]] (the reading is my passion) [ip [DP 1^ lecture] [iP c'j est [ s c tj ma passion]]] (the reading ce is my passion)
In the latter case, French is entirely parallel to Italian in forcing the preverbal subject to a dislocated position, the only difference being that it employs an overt instead of a null pronoun. A residual problem to be explained is why ce is obligatory in inverse sentences as opposed to canonical ones; no attempt will be made here to find a solution. Let us now turn to the third and last idiosyncratic property of Italian, namely, locliticization.
The raising of predicates: a unified theory of copular sentences
71
1.3.3.3 The status of the propredicative clitic lo For the sake of convenience, the relevant data presented in 1.2.2 are reconstructed here: (117a)
le foto del muro sono la causa della rivolta (the pictures of the wall are the cause of the riot)
b * le foto del muro la sono (the pictures of the wall la are) c le foto del muro lo sono (the pictures of the wall lo are) d
la causa della rivolta sono le foto del muro (the cause of the riot are the pictures of the wall) e * la causa della rivolta lo sono (the cause of the riot lo are)
Given the unified theory of copular sentences, these facts can be captured descriptively as follows: a postverbal DP in a copular sentence can be cliticized only if it is the predicate and then only by the non-inflected clitic lo. The problem then is twofold: on the one hand, we must explain why the subject of an inverse sentence cannot be cliticized; on the other, we must explain why the usual inflected clitics are unsuitable for predicative noun phrases. The answer to the first question can draw on the independent hypothesis advanced above, to the effect that there is only one Agr° in copular sentences. What is the correlation between these two facts? It is well known that when inflected clitics move, they are obliged to activate intermediate agreement projections (see the seminal work of Kayne 1985).63 This can be shown for example by moving an object clitic: (118a) b
Gianni ha scritt-o le lettere (Gianni has written-neuter the letters-fem.plur.) Gianni tej ha scritt-e t} (Gianni clitic-fem.plur.has written-fem.plur.)
The past participle is neuter when the object is in situ but agrees in gender and number with the inflected clitic. Copular sentences show different behaviour: past participles cannot be neuter. In fact, in copular sentences a past participle always agrees with the subject. Consider for example the following inverse copular sentences: (119a)
(la causa) pro sono stat-i loro (the cause-fem.sing. pro have been-masc.plur. them-masc.plur.) b * (la causa) pro sono stat-o loro (the cause-fem.sing. pro have been-neuter them-masc.plur.)
72
The anomaly ofcopular sentences: the raising of predicates
With these premises, it seems reasonable to expect that inflected clitic movement would conflict with the structure of copular sentences. All Agr° heads appear to have already been activated for the chain of the raised DP. Moreover, we should not forget that there is independent evidence for assuming that there is no equivalent of object-agreement in copular sentences. In other words, inflected clitics are incompatible with copular sentences because there are not enough agreement projections. The answer to the second question requires a refinement of the notion of clitic that takes into account the more articulated structure proposed for noun phrases as projections of D°. In general, it is assumed that clitics, on a par with pronouns, are generated in D°: (120)
DP (D') D°
infl.cl.
As a first approximation, I would like to propose that the propredicative clitic lo should be analysed as being generated in N° rather than D°: (121)
uninfl.cl.
From the structural point of view, the contrast between 'inflected' clitics and lo can now be regarded as essentially equivalent to the contrast between which and what that was discussed earlier (section 1.3.2). There are two independent reasons for supporting this analysis. On the one hand, the very fact that lo cannot bear features suggests that it is the
The raising of predicates: a unified theory of copular sentences 73 counterpart of a head that is lower than any inflectional head contained in the noun phrase (see Cinque 1992). In other words, the idea is that D° and N° are the two extreme head positions within a noun phrase: in between them, we find all the various inflectional heads, i.e. Agr°, possibly split into Number0, Gender0, etc. The intuition underlying the representation of a propredicative lo as N°, then, is that this element is the lowest head of the noun phrase. It does not contain any agreement feature. 64 Fully inflected clitics, on the other hand, must have raised within the nominal structure so as to collect (the proper) inflectional heads, and are thus ultimately represented as D°. On the other hand, predicative DPs in canonical sentences clearly show that D° can be left empty: (122a) b
[IP proj e [ sc U [DP [D°il] [NP colpevole]]]] (pro is the culprit) [IP proj e [ sc t; [Dp [D°e] [NP colpevole]]]] (pro is culprit)
Bearing this in mind we can provide an explanation of the issue raised by fo-cliticization. In Italian, movement of the propredicative clitic lo departs from the general pattern of inflected clitics: (123a) Maria e stat-a la causa della rivolta (Maria is been-fem.sing. the cause of the riot) b
Maria lOj e stat-a tj (Maria clitic-neuter is been-fem.sing.) c * Maria lOj e stat-o tj (Maria clitic-neuter is been-neuter)
As is clearly shown by the second sentence of the paradigm in (123), clitic movement is unable to affect agreement on the past participle. The data show that the participle must show feminine singular features matching those of the subject which has been raised to preverbal position. Thus, we can reasonably conclude that movement of lo does not need to activate (intermediate) Agr° projections. By combining these independent assumptions we can now explain why predicates can be cliticized by lo in canonical sentences: (124a)
[IP pro; [loj e] [sc U [DP tj [Do e] tj]]] (pro lo e)
There is no conflict here between lo and the lack of object-agreement features in copular sentences.65 The propredicative clitic lo does not contain features, so its movement cannot interfere with agreement. 66 The analysis of the syntax of lo suggests a further question. Consider a sentence like the following:
74
The anomaly of copular sentences: the raising of predicates
(125)
lo e Gianni {lo is Gianni)
What structure can we assign to it? Potentially there are two options: (126a) b
[IP pro [lOj e] [ sc [DP Gianni] [DP D° [NP tj]]]] [IP [IP proj [lOj e] [ sc tj [Dp D° [NP tj]]]] [DP Gianni]j ]
In the first case, the subject Gianni stays in situ in the small clause; in the second case, instead, Gianni is rightward moved as in the classic intransitive examples such as telefona Gianni (pro telephones Gianni). The answer to this question can be offered by considering the following independent contrast: (127a) * [IP pro sono [Sc [DP molte foto] [DP la causa della rivolta]]] (pro are many pictures the cause of the riot) b [ip [DP molte foto] [IP prOj sono [ sc tj [DP la causa della rivolta]]]] c [IP [DP l a causa della rivolta] [|P prOj sono [Sc [DP molte foto] tj]]] Clearly, pro cannot be inserted in spec-IP position while DPs stay in situ within the small clause. If pro is licensed, it must be linked to the trace of one of the two DPs that build the small clause. Thus, since when lo moves the associated DP is left in situ, we can conclude that in a copular sentence of the kind lo - V - DP the postverbal DP is not in situ. Only the second structure of those potentially available in (126) can be consistently maintained, namely (126b). Interestingly, independent evidence that propredicative lo is linked to an NP projection as opposed to a functional DP projection comes from Clitic Left Dislocation Constructions (see Cinque 1990a). Consider contrasts like the ones between (128a, b) and (128a, c), where a non-predicative lo occurs (i.e. a third person masculine inflected clitic): (128a) * [DP il [NP colpevole]], Gianni lo e (the culprit, Gianni lo is) b [Np colpevole], Gianni lo e (the culprit, Gianni lo is) c [DP il [NP colpevole]], Gianni lo vide (the culprit, Gianni lo saw) We have so far considered lo as related to the predicative DP of copular sentences. However, in order to complete the analysis, a last question must be addressed: why is it impossible for the subject of inverse sentences to be cliticized by lo? In this case, we cannot correlate this phenomenon with the absence of object-agreement. Rather, we have to rely on the independent fact that the subject DP is the only element from which propredicative pro
The raising of predicates: a unified theory of copular sentences
75
can pick up features. Neutralizing it by means of the non-inflected lo which can only occur with an empty D° would amount to destroying the only possible source of features.67 It seems then that all language-specific data illustrated in section 1.2.2 which appear to go against the subject-object asymmetries have been explained: Aze-extraction, rightward verb agreement, and /o-cliticization can all be traced back to the peculiar status of inverse copular sentences. First, the absence of ne-cliticization for the subject of inverse sentences is due to the Subjacency Condition. Second, rightward verbal agreement in inverse copular sentences is explained: there is no need to abandon the fundamental hypothesis that it is the result of a spec-head relation between a DP and 1° for we assume that a propredicative/?ro is involved. Third, the fact that only the predicative DP can be cliticized and that the clitic is not inflected follows by assuming that lo is a pro-N° and that copular sentences do not contain ob j ect-agreement. The structure of Italian inverse sentences has turned out to be more complex than that of the English structures examined earlier on: on the one hand, the position where predicates are base generated within the small clause is coindexed with pro in spec-IP; on the other, the lexical content of the predicate is given by the adjunct DP so that in this sense pro should be considered a sort of pro-predicate. Nevertheless, despite this complexity, the structural representation given here is the only one that enables us to derive all the properties of inverse sentences and simultaneously maintain the fundamental hypothesis that verbal agreement is the result of a spechead relation involving 1°. Moreover, although I do not intend to repeat the whole analysis, the two fundamental properties of inverse copular sentences now follow for Italian: the first property (the inability of the subject to undergo A'-movement) can easily be explained by adapting the analysis given for the English counterparts. If the subject is moved out of the small clause in an inverse sentence, a locality violation is produced. The only way for it to get out of the small clause would involve it in passing through spec-IP but this position is already activated by the propredicative pro. The explanation for the second property (the inability of extraction from within the subject of an inverse sentence) entirely reflects what has been said for the English cases: extraction from the subject DP cannot take place because it would trigger a Subjacency violation. The major aim of the first chapter has thus been achieved. The canonical
76
The anomaly ofcopular sentences: the raising of predicates
structure involving raising of the subject to the most prominent position has turned out to be insufficient: (129)
The anomaly ofcopular sentences with respect to the normal subject-object asymmetries can be explained only by assuming that a predicative DP can be raised to spec-IP in complementary distribution with the subject: (130)
This opens up the possibility of analysing postverbal DPs as subjects, specifically as left-branch subjects in situ within the small clause. The syntactic environment where these elements occur is similar to that of a preverbal subject: they can neither be moved, because there is no way for locality conditions to be satisfied, nor do they allow extraction of an embedded element, because this would yield a Subjacency violation. It is worth emphasizing once more that the hypothesis of inverse copular sentences involves dropping a fundamental postulate of the theory of clause structure, i.e. that spec-IP position can be occupied only by subject DPs. Before exploring the consequences of this approach for other empirical domains, I would like to make some remarks on predicative nominals. The next section will accordingly be something of an excursus. Nevertheless, although it does not have direct consequences for the theory of clause structure, it develops certain issues that have arisen from the analysis of copular sentences presented which I therefore suggest deserve attention. In
Exploring the syntax of predicative noun phrases 11 particular it explores the syntax of noun phrases when they occur as predicative elements.
1.4
Exploring the syntax of predicative noun phrases: 6-role assignment and the local domain for binding
When we consider the development of generative grammar since the beginning one thing we notice is that the syntax of noun phrases, understood in the broadest sense as encompassing both their distribution and their internal architecture, has received a great deal of attention, but the study of the same constituents functioning as predicates has been more or less neglected. This is not only true for most introductory texts but also for a leading specific treatise: The Syntax of Noun Phrases by Giorgi and Longobardi (1991), where the term 'predicative nominal' is not even listed as an entry in the index. 68 Of course, such a major gap compared to traditional grammars, where predicative nominals played a central role, cannot be accidental. One could think of many reasons for this gap. Arguably, it is related at some deep level to a basic assumption of generative grammar (in fact, an assumption which is common to all distributionalist theories), namely that linguistics should avoid 'semantic' notions such as predication. 69 In other words, it may well be that the scarcity of studies in the field of predicative nominals is due to a desire to derive their status from more general properties (government, binding, control, etc.). However, although such an approach might eventually turn out to be correct, it seems to me that current models suffer from more serious shortcomings even at the very superficial level of descriptive adequacy. On the other hand, it is also true that for those systems where the notion of predication was taken into account, the predicative use of noun phrases has been explicitly denied. Consider, for example, the following citation taken from a recent systematic work: For concreteness, take categories to be as in (14), for nouns, verbs, adjectives, and pre-/post-positions, respectively (14)
(i) (ii) (Hi) (iv)
N= v = A= P =
[+ N, - V] [- N , + V] [+ N, + V] [- N, - V]
The feature [+N] is the traditional substantive; the feature [+V], predicate. (Chomsky and Lasnik 1993: 517)
78
The anomaly ofcopular sentences: the raising of predicates
The possibility of a noun phrase playing the role of predicate seems to be implicitly ruled out. Of course, this section will not even attempt to fill such a gap. Its very limited aim is simply to underline how two modules of grammar, 6-theory and Binding theory, fail to throw much light, as they stand, on the syntax of predicative nominals and therefore require a certain degree of modification. Anticipating a certain amount of important data, I will now outline how this section is organized: in the first part, we will see that there exists a striking similarity between passive structures and a class of copular sentences with predicative noun phrases: (131a)
Beatrice e [Vp desiderata da Dante] (Beatrice is desired by Dante)
b
Beatrice e [DP il desiderio di Dante] (Beatrice is the desire of Dante)
It will be shown that this similarity might only be apparent. The canonical treatment of passives as involving extraction of the object cannot plausibly be extended to the case of predicative nominals, unless, that is, one wants to go against a well-established generalization.70 In the second part, we will attempt to identify the local domain for binding. In particular, it will be shown that there are contexts where an anaphor contained in a predicative nominal can refer to the subject of the matrix clause, skipping the subject of the predication in the embedded clause in which it is contained, as in the following very clear case: (132)
Giannij ritiene [Sc questej [DP le migliori foto di se stessojjj] (Gianni considers these the best pictures of himself)
Here, the anaphor se stesso (himself) agrees and refers exclusively to Gianni, not to the subject of the embedded clausal constituent, namely queste (these). Again, the present work will not try to provide solutions to all these problems but will pursue the more modest aim of throwing light on the problems that have to be faced in the analysis of predicative nominals. 1.4.1 Pseudo-extraction from predicative noun phrases Traditionally, 0-role assignment has been studied in relation to VP predicates. Picking up from the lexicon the verbal head desider- (desire), we can obtain the following maximal projection:
Exploring the syntax of predicative noun phrases
79
(133a) Dante,
Beatricee
desiderk
6X = <experiencer>, 62 =
As indicated here, this verb is compatible with the presence of two distinct arguments: the DP adjacent to the head (the object), and the DP in specVP (the subject, as widely assumed nowadays following Koopman and Sportiche 1991). Thus, the lexical entry of desider- contains two distinct 6roles, which we can call the <experiencer> and the , labelled here as 6/ and 62. These 6-roles are rigidly assigned to specific structural positions (see references in chapter 5 for a condensed presentation of a new approach to 0-theory). 71 What happens to these two DPs and to their 9-roles when the VP is combined with other phrases to yield a clause structure is well known. For example, we can produce a fully inflected sentence like the following: 72 (134)
[IP [DP Dante]H, [VP tHI [v° desidera] D[ P Beatrice]e2]] (Dante desire-s Beatrice)
In this simple case, the subject DP is raised from spec-VP to spec-IP preserving its 6-role (i.e. the <experiencer> 0!); the other 6-role (i.e. the 62) is maintained by the object DP. Another common possibility is given by passive sentences where the internal argument is extracted and the external one is realized in a postverbal position where it occurs with da (by): (135)
dp [DP Beatrice]H2 e [Vp [v°desiderata] tH2 da [DP Dante]H1]] (Beatrice is desired by Dante)
Significantly, it is impossible for 6 2 to be assigned to the DP in preverbal position if the verb is not passivized, nor, conversely can Q} be assigned to the DP in preverbal position if the verb is passive. 73 Compare now these cases with the following: (136a) * [ip [DP Dante]H2 [Vp te2 [v° desidera] [DP Beatrice]el]] b * [IP [DP Beatrice]B1 e [Vp [v° desiderata] et, da [Dp Dante]H2]]
80
The anomaly ofcopular sentences: the raising of predicates
The next step is to see what happens when the maximal projection playing the role of a predicate is a DP rather than a VP. First of all, let us now construct an NP headed by the N° desiderio (desire), paralleling what we did with V° in (133) when constructing the VP: (137a)
a Dante,
desiderb
6j = <experiencer>, 6^ =
Then, let us embed such an NP within a proper DP, add Case assigners (i.e. per and di) and inflectional features: (138)
[DP il [NP [N° desiderio] di [DP Dante]Hi per [DP Beatrice]e2]] (the desire of Dante for Beatrice)
As indicated, the N° desider- preserves the same 9-role pattern as the corresponding V°, i.e. both the <experiencer> 6! and the 62 can be assigned to the arguments of the biargumental nominal head. 74 This DP can occur as an argument as in the following cases: (139a) b
[DP il [N° desiderio] di [DP Dante]ei per [DP Beatrice]H2] stupi Bonifacio (the desire of Dante for Beatrice astonished Bonifacio) Virgilio conobbe [DP il [N° desiderio] di [DP Dante]ej per [DP Beatrice]e2] (Virgilio knew the desire of Dante for Beatrice)
The central question that I would like to address here is the following: what happens to 9-role assignment when this DP occurs as a predicative element? There are two major contexts to consider here: small clause complements of a believe-type verb such as ritenere (believe) and the most typical context for predicative nominals, i.e. copular sentences. To avoid confusion during the discussion of copular sentences, it should be borne in mind that the linear order of a DP with respect to the copula is not sufficient to identify its grammatical function. In other words, the mere fact that a DP precedes or follows the copula is not a sufficient indication of whether this DP is a predicate or a subject. This is due to the existence
Exploring the syntax of predicative noun phrases
81
of a class of copular sentences which I have proposed calling 'inverse' (as opposed to the canonical variety): (140a) b
[IP [DP il desiderio di [DP Dante]el per [DP Beatrice]e2]i e [ sc \x [DP la vera novita]]] (the desire of Dante for Beatrice is the true novelty) [IP [DP l a v e r a novitajj e [Sc [DP il desiderio di [DP Dante] ei per [DP Beatrice]e2] tj] (the true novelty is the desire of Dante for Beatrice)
In the first example, a canonical sentence, the DP under discussion precedes the copula and plays the role of subject of the predication ((140a)). In the second example, an inverse sentence, the same DP still plays the role of a subject of the predication although in this case it no longer precedes the copula ((140b)). Let us now examine cases where the DP we are considering here plays the role of predicate. One thing is immediately clear: in both cases presented below, this DP cannot occur in a predicative position if a third argument, say Laura, occurs in the sentence: (141) * ritengo [ sc [DP Laura] [DP il desiderio di [DP Dante]el per [DP Beatrice]H2]] {pro believe Laura the desire of Dante for Beatrice) b * [iP [DP Laura]j e [Sc tj [Dp il desiderio di [DP Dante]e, per [DP Beatrice]H2]]] (Laura is the desire of Dante for Beatrice) Ought we then to conclude that this DP cannot play the role of predicate after all? The answer is not so straightforward, as becomes clear if we consider the following contrast: (142a) * ritengo [Sc [DP Dante]el [DP il desiderio per [DP Beatrice]H2]] (pro believe Dante the desire for Beatrice) b ritengo [ sc [DP Beatrice]e2 [DP il desiderio di [DP Dante]el]] (pro believe Beatrice the desire of Dante) c * [IP [DP Dante]ei e [ sc tei [DP il desiderio per [DP Beatrice]62]]] (Dante is the desire for Beatrice) d [iP [DP Beatrice]e2 e [ sc U2 [DP il desiderio di [DP Dante]el]]] (Beatrice is the desire of Dante) These examples show that if the predicative DP under discussion is deprived of one argument, specifically the one assigned 0 2 , it can play the role of predicative nominal; in this case, the subject of the predication is assigned 0 2 . 75 This is precisely what happens in the grammatical examples we are considering. A first major question therefore arises: why can only one 0role be assigned outside the predicative DP to the subject DP in spec-IP? The data presented here allow another interesting issue to be raised. Let us
82
The anomaly ofcopular sentences: the raising of predicates
concentrate our attention on the grammatical examples ((142b and d)). What emerges very strongly is that there is a certain resemblance, as far as 6-assignment is concerned, between sentences like Beatrice e il desiderio di Dante ((142d)) and passive sentences like Beatrice e desiderata da Dante ((135)). This resemblance lies in the fact that in both cases the subject argument is assigned what is generally regarded to be the internal 6-role of the predicative element, whether this is a nominal, as in (142d), or a transitive verb, as in ((135)). Clearly, this fact calls for some explanation. I am not in a position to offer a structural interpretation of this rather striking fact and will therefore leave it as an open question. What I would like to point out here is that one very appealing analysis that naturally springs to mind must be excluded. The analysis in question involves extending the classic analysis given for passive structures to cases such as the one under discussion involving a predicative nominal. In somewhat more explicit terms, we cannot hypothesize that the subject Beatrice is extracted from an internal DP position along the same lines as what is normally assumed for passive structures like (136b), reconstructed here as (143b):76 (143a) b
. . . [DP Beatrice]e2 • • • [NP desiderio te2 di [DP Dante]ei] . . . . . . [DP Beatrice]e2 . . . [VP desiderata te2 da [DP Dante]el] . . .
The reason why this analysis cannot be adopted lies in the fact that it involves extraction from a DP and would thus go against a rather well-established generalization, which owes its origin to an observation by Cinque (1980). This generalization states that a constituent can be extracted from a DP only if it can be possessivized (see Giorgi and Longobardi 1991 for an updated discussion and a new derivation of this phenomenon). Unfortunately, if we apply this to the DP under discussion, we actually obtain the opposite of the result we need to support an analysis involving extraction: (144a) * [DP il suoH2 desiderio tH2 di [DP Dante]ei] (the his desire of Dante) b [DP il suoei desiderio per [DP Beatrice]e2 tel] (the his desire for Beatrice) This contrast shows that the only argument that can undergo possessivization is Dante, which clearly bears Q{ role, and this of course is exactly the argument which cannot occur as a subject.77 Thus, the hypothesis that Beatrice is extracted from the predicative nominal in (143a) cannot be maintained. At this point, as I indicated earlier, I have no interpretation of this fact to offer. For the moment, then, I will simply suggest that such a class should be referred to as 'pseudo-extraction', a term which at least has
Exploring the syntax of predicative noun phrases 83 the merit of emphasizing - at one and the same time - the similarities and also the differences between this construction and the passives. In fact, before attempting any solution to the puzzle of pseudo-extraction, as a preliminary step, it is necessary to get some idea of the extension of such a phenomenon within the class of noun phrases. Naturally, an investigation of this kind would go far beyond our limits but we can perhaps conduct a brief survey, examining a number of noun heads which are compatible with two arguments (henceforth, biargumental nouns). Thus, it is easy to establish that, like desiderio, other nouns such as paura (fear), timore (fear), preoccupazione (worry), piacere (pleasure) allow pseudoextraction: (145a) b c d
l'inferno e la paura di Dante (the inferno is the fear of Dante) la citta di Dite e il timore di Dante (the city of Dite is the fear of Dante) Cerbero e la preoccupazione di Dante (Cerbero is the worry of Dante) la luce e il piacere di Dante (the light is the pleasure of Dante)
On the other hand, not all biargumental nouns, by any means, allow pseudoextraction: for example, descrizione (description), racconto (narration), fotografia (photograph) and apparizione (apparition) yield ungrammatical sentences: (146a) * l'inferno e la descrizione di Dante (the inferno is the description of Dante) b * la citta di Dite e il racconto di Dante (the city of Dite is the narration of Dante) c * Cerbero e la foto di Dante (Cerbero is the photograph of Dante) d * la luce e l'apparizione di Dante (the light is the apparition of Dante)
A first rough generalization may now be formulated. It seems that pseudoextraction is allowed only with those noun phrases which express psychological attitudes and not with those 'extensionaP verbs which denote activities like describing, narrating, etc. Thus, borrowing the terminology proposed for certain types of VP by Belletti-Rizzi (1988), it is tempting to call this class of noun phrases 4psych-nouns\ A provisional generalization can then be formulated as follows: (147)
Only psych-nouns allow pseudo-extraction
Keeping the discussion on a descriptive level, we can push this generalization
84
The anomaly ofcopular sentences: the raising of predicates
to the limit by asking whether it should be reformulated as an 'if-and-onlyif clause. In other words, we can try to establish whether all (and only) psych-nouns allow pseudo-extraction. It seems that the answer to this has to be in the negative. For example, there are cases like cruccio (worry) that are psych-nouns from a semantic point of view but show a different pattern compared to elements like desiderio (desire):78 (148a) Beatrice e il cruccio di Dante (Beatrice is the worry of Dante) b [DP il cruccio di Dante (*per Beatrice)] e noto a tutti (the worry of Dante for Beatrice is known to everybody) On the one hand, they allow constructions like (148a), namely Beatrice e il cruccio di Dante, which strongly resembles cases of pseudo-extraction (specifically, the 0-role assigned to Beatrice comes from the N°). On the other, they are clearly not biargumental, as (148b) shows, and so the very idea of extraction has to be excluded in principle. This (and other possible cases) suggests that the generalization presented above is too strong, if not totally on the wrong track. Summarizing, we have highlighted three major questions: first, why can only a specific 6-role among those compatible with a given N° be assigned to the subject of the predication when the predicate is a DP? Second, why is there a difference between psych-nouns and nouns indicating activities? Third, how can one define the subclass of psych-nouns that allows pseudoextraction?79 At this point our limited aim in this section of setting out the major questions regarding predicative nominals has been achieved. As we have seen here, as soon as such nominals are brought into the arena many specific problems are raised that do not seem to allow immediate solution; this clearly shows that much more careful examination is called for. That said, we can now shift to a different but closely related topic. So far, we have seen cases where for a DP to be used predicatively one of its arguments must remain unrealized within its maximal projection. The relevant minimal pair is represented by two previous examples in (141b) and (142d) which are reproduced here as (149): (149a) * [IP [DP Laurajj e [sc tj [DP il desiderio di [DP Dante]Hi per [DP Beatrice]H2]]] (Laura is the desire of Dante for Beatrice) b [IP [DP Beatrice]e2 e [sc tH2 [DP il desiderio di [DP Dante]H1]]] (Beatrice is the desire of Dante) We might ask whether it is right to conclude that a biargumental noun phrase must always be deprived of one of its arguments if it is to play the role of a predicate.80 On the face of it, this does indeed seem to be the natural
Exploring the syntax of predicative noun phrases
85
prediction. Whether or not the subject is literally extracted from the DP, it is clear that its 9-role belongs to the 0-grid of the nominal head N°. Thus, if both arguments are present within the maximal projection of N°, one should expect that any DP in spec-IP would fail to receive a 9-role since there are no other 9-role assigners around; consequently, a violation of the 9-criterion would be unavoidable ((149a)).81 Surprisingly enough, there are contexts where this prediction fails to hold: (150a) b
ritengo [ sc [DP questo] [DP il desiderio di [DP Dante]B1 per [DP Beatrice]e2]] (pro believe this the desire of Dante for Beatrice) [ip [DP questojj e [Sc tj [DP il desiderio di [Dp Dante]ej per [Dp Beatrice]e2]]] (this is the desire of Dante for Beatrice)
In both sentences, the DP under discussion can indeed play the role of a predicate without missing any argument. In other words, both Beatrice and Dante can stay within the DP without preventing it from assuming the role of predicative nominal. Notice that we have independent evidence to support the idea that the DP under discussion is actually functioning as a predicate: it is sufficient to apply the fo-cliticization test, which yields: (151)
[IP [DP questo]; [lOj e] [ sc tj tj]] (this lo is)
This shows quite clearly that the DP il desiderio di Dante per Beatrice is playing the role of predicative nominal. It has been established on independent grounds that this type of cliticization in a copular sentence is possible only if the corresponding DP plays the role of predicative nominal (see 1.3.3). Although I am not going to offer a full explanation of these phenomena, it is worth noting that the sentences under discussion have important implications. The only 9-role assigner here is the N° desiderio. Now, since the two arguments of the N° are Dante and Beatrice and since the role of subject of the predication is played by questo (this), the immediate conclusion is that in this case predication is not connected with 9-role assignment. It is interesting to notice that this state of affairs is not unique to copular sentences involving predicative DPs. In fact, certain unaccusative constructions in which the subject of the predication is an expletive share this same property. Consider, for example, the following simplified analysis of a classic unaccusative construction:82 (152)
[IP [DP there]_H [VP arrived [DP many girls]+e]]
86
The anomaly ofcopular sentences: the raising of predicates
Paralleling the case of predicative noun phrases we are discussing here, the head of the predicate, here the V° arrive, assigns its 6-role to its argument within its maximal projection (i.e. VP); the position of the subject of predication is occupied here by an expletive DP, namely there, which fails to receive any 6-role. The resemblance between the case of predicative noun phrases and unaccusatives stops here, though: by no stretch of the imagination could one say that questo (this) is an 'expletive' in the common sense of the term, that is an element which has no independent referential capacities. On the contrary, being a deictic, it is normally considered to be fully referential. However, in the example under discussion, we can only conclude by observing that questo seems to have an intermediate status between fully referential elements and expletives: it is like the former in that it can refer independently but it is like the latter in that it does not receive a 0-role from the head of the predicate. We will now shift our attention to a different but related topic involving predicative noun phrases. 1.4.2 Opacity phenomena for binding The possibility, observed above, of the subject of a predicative nominal occurring without being assigned a 0-role suggests a need for further investigation concerning Binding theory. Once again, the intention here is not to offer a solution, but rather to single out a potential problem for the current theory. In general, a minimal requirement that any version of Binding theory is expected to satisfy is that an anaphor can never escape the subject of the clause it occurs in. From an abstract point of view, what this amounts to is that for any anaphor a in a structure as in (153) :83 (153)
. . . DPj . . . [s DPj . . . a . . .
the index assigned to oc cannot be the same as the one assigned to the first DP.84 This is generally expressed by saying that a clause structure is a local domain for an anaphor and that an anaphor must be bound within its local domain (Condition A of the Binding theory). There have been several attempts to capture this fact. Efforts have been specifically devoted to deriving the identification of the minimal domain from independent structural properties. An approach that has become quite influential was proposed by Chomsky (1986a). The local domain for an anaphor or a pronoun oc is defined as the least Complete Functional Complex (CFC) containing a lexical governor for a (see Chomsky 1986a:
Exploring the syntax of predicative noun phrases 87 169). Let us reproduce the more refined definition of CFC given in Giorgi and Longobardi (1991: 54-5): (154)
(3 is a Complete Functional Complex if it meets at least one of the following requirements: a. it is the domain in which all the 0-roles pertaining to a lexical head are assigned b. it is the domain in which all the grammatical functions pertaining to that head are realized (where the R-relation counts as the structural subject of the NP)
The principal aim of this section is to show that predicative nominals challenge the theory of the CFC in a very direct way. To do so, let us consider the following example: (155)
Giannij ritiene [ sc [DP queste]f [DP le migliori (PROj j) foto di se stessOj/ *se stessef]] (Gianni-masc.sing believes these-fem.plur. the best PRO pictures of himself-masc.sing./themselves-fem.plur.)
This example is close to the ones we examined in (150): the subject of the predication in the embedded clause (queste (these)) does not receive any 6-role from the predicative nominal. In fact, the 9-grid of foto (including, say, and <patient>) is completely saturated within the maximal projection of the N° itself. The question now is: does the definition of CFC given here correctly identify the local domain for the anaphor? It would appear not to do so; we will now examine the reasons for this. The anaphor contained in the predicative nominal can in fact refer to the subject of the matrix clause, i.e. Gianni, as indicated by the agreement features of se stesso which is masculine singular. Notice that the subject of the predicate headed by foto (pictures) is queste (these). If we apply the definition of CFC given here, the prediction is that the subject of the predication, queste, will create opacity for the anaphoric linking. The anaphor cannot be bound by queste, as shown by the impossibility of having se stesse with its feminine and plural features. Assuming that DPs can contain a PRO subject will not be of much help here. Potentially, it could of course be that the anaphor is bound by PRO controlled by Gianni, but it is also true that the sentence can be interpreted as having an arbitrary PRO, as if someone else took the picture of Gianni. In this case, coreference between the anaphor and Gianni should be blocked, according to our definition of CFC, because PRO completes the functional complex of the head foto, being assigned the agent 6-role.85
88
The anomaly ofcopular sentences: the raising of predicates
Interestingly, if the subject of the predicative nominal is an overt DP (as opposed to PRO) there is no possibility for the anaphor to be bound by Gianni: (156)
Giannij ritiene [Sc [DP queste]f [DP le migliori foto di Mariaj di se stessa/ *se stessOj]] (Gianni-masc.sing. believes these the best pictures of Maria-fem.sing. of herself/himself)
In this case, the anaphor must be bound by Maria within the projection of the predicative nominal. This latter case deserves more attention. In general, Binding theory never makes essential distinctions between lexical DPs and pronominal empty categories. In particular, any version based on the notion of CFC in principle avoids any recourse to phonological features. The only relevant notions seem to be that of 6-role assignment and that of Grammatical Function; witness the version of CFC we are considering here.86 This seems to be implicit in the nature of Binding theory: this module is essentially devoted to accounting for (restrictions on) referential dependencies. Why should overt phonological realization be relevant, since pronominal empty categories can play the same role as fully referential elements? Again I do not see any immediate solution to this problem but, of course, it was not our initial aim to find one. Let us now summarize the content of this section. We have seen that when predicative nominals begin to be taken into consideration the current framework does not seem to be adequate as a basis for deriving the whole range of empirical facts. We have explored facts concerning 8-role assignment and Binding theory and we have shown why some traditional assumptions are challenged in a rather striking and puzzling way. The following two distinct topics have been explored here. We have singled out a class of predicative nominals we called 'psychnouns', like desiderio (desire), which can be so construed that one of the 8-roles assigned by the head N° is assigned to the subject of the predication with which such nominals occur. Three questions are raised by this fact: first, why can only one 6-role be assigned to the subject of predication? Second, why is it that other nouns expressing non-psychological activities, like descrizione (description), do not allow such a process? Third, what is the subclass of psych-nouns displaying such a phenomenon? Moreover, we have seen that in spite of the similarity with passive constructions an explanation involving extraction cannot be maintained without undermining Cinque's
Exploring the syntax of predicative noun phrases 89 (1980) generalization. We accordingly decided to refer to this phenomenon as 'pseudo-extraction' as a way of highlighting both the similarities and the differences between it and normal object extraction in passive constructions. It has been shown that predicative nominals challenge the Binding theory in a rather powerful way. In fact, whatever version of this theory one adopts, it is generally assumed that a clausal subject creates opacity for anaphoric binding. Predicative nominals challenge this assumption by showing a context where the anaphor contained in an embedded clause may be bound by the subj ect of a matrix clause, while skipping the subj ect of the embedded clause. This happens when the subject of the predicative nominal does not receive a 8-role from the head of the predicative nominal itself, or, to put it slightly differently, when the 6-grid of the nominal head is completely saturated within the maximal projection of the nominal itself. Interestingly, we have seen that this phenomenon occurs only if the internal subject of the predicative nominal is either absent or realized as PRO but not if it is realized as an overt DP. This seems even more puzzling since Binding theory should not in principle distinguish between null and overt lexical realization. Whether or not these facts can be explained without distinguishing predicative DPs from argumental DPs is a question which lies far beyond the scope of this section. It might turn out that the specific properties of predicative nominals are just the consequence of the particular structural position they occur in. However, even from this brief survey it is clear that this field of inquiry deserves a thorough investigation. 1.5
Conclusions
In this chapter we saw that sentences of the kind DP - copula - DP do not constitute a homogeneous class. We started by testing the common structure for the clausal constituent according to which there is an asymmetry between the preverbal DP (the subject) and the postverbal one (the object): (157)
90
The anomaly ofcopular sentences: the raising of predicates
Evidence was provided that, for a given pair of sentences in which there is alternation in the linear order of the two DPs around the copula, one sentence fails to show the usual subject-object asymmetries which are predicted by the structure. In particular, movement and extraction from the object position are blocked at all levels as if this were in fact a pre verbal subject: among the phenomena tested were wh-movement and Q-raising, as well as some language-specific phenomena from Italian syntax (neextraction, verbal agreement and /o-cliticization). It was then proposed to abandon the 'rigid' clause structure which supposes that the most prominent DP of the clause structure is always the subject of the predication, in favour of a more 'flexible' structure in which it is possible for a predicative DP to be raised to such a position. In this way, two different types of copular sentences have been derived from a unique base generated structure: along with the traditional type of 'canonical sentences', where the raised DP is the subject: (158)
S
it was proposed that it is necessary to recognize the existence of inverse sentences, where the raised DP is the predicate and the lower DP is the subject left in situ in a left-branch position: (159)
In this way it becomes possible to explain the whole cluster of anomalies in a unified way through the application of two independently assumed principles: the locality principles on movement (as stemming from the ECP)
Exploring the syntax of predicative noun phrases 91 and the Subjacency Condition. Crucially, it was shown that, if the most advanced theory of clause structure including distinct projections for the inflectional elements is adopted, the following structure including a single Agr° projection may be assigned to copular sentences: (160)
AgrP
be
DP
The chapter ended with an excursus on some challenging properties of predicative noun phrases which came to light during the investigation of copular sentences. In particular, we saw that 6-role assignment and the local domain for binding cannot immediately be understood on the basis of standard assumptions. At this point, a typical methodological issue arises. There are (at least) two distinct sources of support for a theory like the one being presented here: on the one hand, one could search for independent empirical evidence exploring new facts in many different languages; on the other, one could push the hypothesis to the limit and see if this will help in simplifying the grammar from a theoretical point of view by providing a new analysis of some well-known phenomena. Clearly, these two lines are not necessarily distinct. In this book we will essentially follow the second strategy: some well-known phenomena, such as the distribution of there and ci in English and Italian, and the syntax of seem and unaccusative verbs, will be reexamined in the light of the new clause structure. However, before embarking on this investigation, I would like to indicate some possible ways of developing the first line of research. An interesting example comes from the following citation from Jespersen's Philosophy of Grammar concerning a special pair of Danish copular sentences.
92
The anomaly ofcopular sentences: the raising of predicates Such words as ikke 'not' are placed before the predicative. Now we have two words spelt alike Moller, but if it is a proper name it is pronounced with the glottal stop in the /, while as a common name 'a miller' it has no glottal stop. The curious result is that Danes will never hesitate about the pronunciation of the four sentences: (1) (2) (3) (4)
Moller skal vaere Moller Moller skal Moller vaere Moller er ikke Moller Moller er Moller ikke
In (1) and (3) they will give the first Moller the glottal stop and thereby mark it out as the proper name, because the word order shows it to be the subject; inversely in (2) and (4). The English meaning of (1) and (2) is (Mr.) Miller is to be a miller, and of (3) and (4) Miller is not a miller, where the difference is shown by the indefinite article. (Jespersen 1924: 151)
The 'curious' paradigm presented by Jespersen can be immediately understood, once we assume that phrases move and that predicative noun phrases are visible to movement. The examples given in the citation may be completed by indicating the traces of the predicative noun phrases moved. The following simple representation is thus obtained: (161a) b c d
Mollerj skal vaere tj Moller Mollerj skal Moller vaere tj Mollerj er tj ikke Moller Moller, er Moller ikke tj
Of course, there are still many questions to be answered, such as those concerning verb-second phenomena, the type of DP involved and the like. However, this is just to show how apparently inexplicable phenomena ('curious', in Jespersen's own words) become perfectly understandable once the transformational approach to inverse copular sentences is adopted. Other interesting applications of this theory might involve pseudo-cleft and cleft sentences. Higgins' (1973) seminal work established that pseudocleft sentences like the following invented here have very different properties: (162a) what I don't like about mathematics is obvious b what I don't like about mathematics is prime numbers It would now seem reasonable to treat these cases as instances of canonical vs inverse sentences, respectively, and to assign them the following representations: (163a) b
[DP what I don't like about mathematicsjj is [ sc tj [AP obvious]] [DP what I don't like about mathematics]j is [ sc [DP prime numbers] tj]
Exploring the syntax of predicative noun phrases 93 In other words, the wh-element is to be associated with two different 'poles' of the small clause: the subject of predication in thefirstsentence, the predicate in the second one. Many interesting properties of these constructions such as the one illustrated in Higgins (1973) could be approached from this alternative point of view (see Heycock 1991 and references cited there, and Heycock 1994). In addition, it would also be possible to treat other facts hitherto not fully understood - such as the puzzling agreement pattern of the Italian counterparts of the English examples in pseudo-cleft sentences: (164a) cid che non mi piace della matematica e/*sono ovvio (what I don't like about mathematics is/are obvious) b cio che non mi piace della matematica sono/*e i numeri (what I don't like about mathematics are/is numbers) Since the wh-phrase in the second sentence is a raised predicate, agreement is to the right, for reasons that I indicated in this chapter. However, we will leave this line of research aside and turn instead to a re-examination of some well-known issues related to the theory of clause structure.
2
The syntax of ci
2.1
The standard analysis: on the notion of expletive
In chapter 1, we saw two constants in the theory of clause structure all through its development: first, clause structure must implement the subjectobject asymmetries in the geometry of tree representation; second, grammatical functions are defined on the basis of the configuration itself: (1)
S
V Then, we saw how the syntax of predicative noun phrases challenges these two properties. In particular, by exploring copular sentences, we have shown that raising to the most prominent DP position is possible for predicative noun phrases in exactly the same way as it is in the more standard case of argumental ones: (2a)
94
The standard analysis: on the notion of expletive 95
In this chapter, we will examine an empirical domain which is closely related to these cases and see how the structure proposed to account for the apparent absence of subject-object asymmetries in certain copular sentences will also allow us to improve our understanding of this other domain. Let us now illustrate the empirical data. Since Milsark (1974) the following two types of copular sentences have been considered as sharing the same base generated structure: (3a) b
there is a man in the room a man is in the room
In the first analyses proposed (see Milsark 1974, 1977),1 the two sentences were roughly considered as synonymous: in both cases, it is said that a man has the property of being in the room. In other words, according to these analyses, the element there, always regarded as a noun phrase, had no semantic role in a sentence like there is a man in the room. This intuition was strongly supported by the fact that this special noun phrase cannot discharge the function of argument, say subject or object, in cases like the following: (4a) b c
John likes books * there likes books * John likes there
The conclusion was clear enough: although there can occur where argumental DPs generally occur, namely in the subject position, it has no semantic content. The ungrammaticality of the above examples was then immediately explained as due to a violation of the reasonable assumption that semantically null elements are incompatible with 0-roles. Following on from this, an important question naturally arose: what forces there to show up in a sentence like there is a man in the room? In other words, what principle of grammar rules out a sentence like the following? (5)
* is a man in the room
96
The syntax of ci
The standard treatment of this puzzle came in the shape of what was called the Projection Principle, that is to say, a principle requiring lexical entries to be respected (at all levels of representation). A simple case like the following illustrates this principle: (6a) b c
John likes books * likes books * John likes
Only the first sentence is grammatical because only in the first sentence is the lexical entry of like respected. This verb, in fact, requires two arguments: in this case John and books. In the ungrammatical cases one of the two arguments is missing, and so the Projection Principle is violated. Clearly, the same principle could not immediately be applied to a sentence like * is a man in the room. In this case, there are no missing arguments. Immediate confirmation of this comes from the fact that a sentence containing exactly the same lexical elements arranged in a different way, like a man is in the room, is perfectly grammatical: what this means is that the ungrammaticality of * is a man in the room cannot be a question of the incomplete realization of the arguments of a lexical head. Although the Projection Principle was clearly not sufficient to account for such cases, there was little readiness to acknowledge the fact that it was simply irrelevant to the phenomena in question, thus the issue became one of refining the Projection Principle so that it would encompass the offending data. The crucial step was to extend it by exploiting an essential postulate of the theory of clause structure, namely that grammatical functions are defined on the basis of configuration. Thus, the Extended Projection Principle was formulated by Chomsky (1982: 10; see also Chomsky 1981: 28ff.); it required that all sentences have a subject of the predication (at least) at LF. Thus, to the question of why there is required in a sentence like there is a man in the room an immediate answer was now available: because there satisfies the Extended Projection Principle. The function of this semantically null element was then considered to be that of holding the place of the subject of the predication; consequently, it was termed 'expletive' (of the subject position). Stowell's (1978) analysis of be as a raising verb allowed the formalization of such a relation in the following way:
The standard analysis: on the notion of expletive
97
(7)
therej be
DP;
The basic relation of predication is contained in the small clause where the subject a man is linked to the predicate in the room. There is inserted at sstructure as an alternative to subject DP raising. Coindexing was introduced as a device for indicating that there and a man have no independent referential properties. This analysis of the distribution of there based on the assumption that the copula selects a small clause has met with general acceptance and has established itself as the standard analysis. It is nevertheless important to notice that an alternative theory has been discussed in the literature since its original formulation in Williams (1984) (see, for example, Higginbotham 1987, Reuland and ter Meulen 1987). This alternative is based on the idea that in a J/zere-sentence, the complement of the copula is a noun phrase, not a small clause, while there is still regarded as a semantically null place holder of the subject position: (8)
S
Crucially, in this analysis the idea that there plays the role of a subject is pushed to the limit. According to this analysis, there is not holding the place of a displaced subject: rather, there is the subject and, correspondingly, the DP following the copula is analysed as a predicate. In other words, in a f/zere-sentence like there is a prime number, the grammatical function of a prime number would be exactly the same as the one it has in seven is a prime number: in both cases a prime number is analysed as predicate. All in all, then, the analysis of the distribution of there has played a central role in the development of generative grammar although regarding it as an
98
The syntax of ci
expletive has not been without cost. Before proposing a new approach, the major stages in this development will be reviewed. Chomsky (1981) proposed that there-insertion should be regarded as totally free in noun phrase positions and that ungrammatical cases could be ruled out by the interaction of independent principles; this proposal was highlighted as a notable example of modular interaction: (9)
insert there anywhere (Chomsky 1981: 88)
Consider, for example, the following cases: (10a) * John likes there b * there is raining What are the independent principles ruling out these sentences? The first sentence is essentially a violation of the lexical entry of like, which requires it to have two arguments: given that there is inserted at s-structure, that would imply that at d-structure an argument is missing. As for the second case, it could be ruled out by assuming that the lexical entry of there requires it to inherit number from an associated DP; since there is no DP at all, the sentence has to be ruled out. Notice that the assumption that there has number appears to be independently necessary in view of cases like the following: (lla) b
there is a man in the room there are men in the room
Since verb agreement is established on the basis of a spec-head relation with the inflectional system (i.e. a relation between 1°and the element in specIP), there is the only element that can possibly trigger agreement. All in all, if we adopt the more refined theory of the clause as a projection of 1°,the structure of a there-sentence will turn out to be as follows: (12)
[a manjj [in the room]
The standard analysis: on the notion of expletive 99 A very different proposal for ruling out the ungrammatical occurrences of expletive there was suggested by Chomsky (1986a) who exploited the idea that there is a semantically vacuous element. His central argument concerns the interaction of there with the Principle of Full Interpretation which requires that at a certain level only legitimate objects be visible. Since by definition expletives do not have semantic content, they are not legitimate at LF, where grammatical structures undergo interpretation. For this reason, it was proposed that there be wiped out by replacing it with the DP it is coindexed with. Thus, at LF a f/iere-sentence would end up being identical to its raising counterpart: (13)
IP
in the room
This analysis, based on the expletive replacement hypothesis, had at least three major consequences. We will briefly deal with each of them. The most important consequence is that the distribution of there is seen to be determined by the independent principles governing chain formation. The associated DP must move to substitute there, thus creating a chain: if locality and Subjacency are not respected in each step of the chain, then the sentence will be ruled out. We can now consider once again the following cases (taken from Chomsky 1988a) from the new perspective: (14a) there is a man in the room b there are men in the room According to Chomsky (1988a) they are grammatical because a man and men can move to the subject position replacing there; witness the grammaticality of cases like: (15a) b
[DP a man]j is tj in the room [DP men]j are tj in the room
100
The syntax of ci
On the other hand, cases like the following are to be ruled out as instantiating illicit movement. (16a) * there seems that a man is in the room b * there seems that John saw a man c * there was thought that pictures of a man were on sale
In other words, the process of LF replacement of the associated elements yields either violations of ECP ((16a-b)) or of Subjacency ((16c)), as indicated by independent cases such as the following: (17a) * [DP a man]j seems that tj is in the room b * [DP a manji seems that John saw tj c * [DP a man]; was thought that [Dp pictures of tj] were on sale
Of course there would still need to inherit number from a DP, but otherwise it seems that its distribution can simply be regarded as the result of a more general phenomenon, i.e. chain formation. This was undoubtedly a considerable advance, but, as Chomsky himself noticed, a major exception survived (see Borer 1986). Consider a pair like the following: (18a) * there seems a man to be in the room b [DP a man]i seems tj to be in the room Apparently, there is no way to account for the ungrammaticality of the first sentence (18a), since movement of a man from this position to the matrix clause subject position is fully licensed, as the second sentence shows (18b). We will come back to this exception in the following sections. A second advantage implied by the expletive replacement analysis concerned Binding theory. It was soon realized that the hypothesis of thereinsertion yields a violation of Condition C of Binding theory, which requires referential expressions not to be bound by an element in an A-position.2 If we assume that expletives are to be replaced by their associated elements, the question concerning the violation of Binding theory Condition C by there-insertion is solved. The trace counts as an anaphor, and so at LF (where Binding theory checks the structure) there is no violation of Condition C; rather Condition A, which requires anaphors to be bound (in a local domain), is satisfied. A third major consequence of this analysis became apparent later, within the so-called 'minimalist' approach (Chomsky 1993). Indeed, expletive replacement has become a fundamental support to this model, as it illustrates the effect of the 'Principle of Greed' which is central to the new system based on 'economy principles'. I will briefly reproduce the essential intuition by considering sentences like the following (see Chomsky 1993: 32ff.):
The standard analysis: on the notion of expletive 101 (19)
there seems to [a a strange man] that it is raining outside
In this sentence, the replacement of there by a man 'would yield an intelligible interpretation (something like "There is a strange man to whom it seems that it is raining outside")'. However, 'derivations are driven by the narrow mechanical requirement of feature checking only, not by a "search for intelligibility" or the like, . . . benefiting other elements is not allowed' (Chomsky 1993a: 33). Thus, this process is not permitted and 'the derivation converges with an unintelligible interpretation'. The original hypothesis of 'expletive replacement' has been slightly refined in the most recent models. Following Williams' (1984) original observation, it was realized that there could not simply be wiped out at LF. Consider the following contrast: (20a) b
there aren't many men in the room many men aren't in the room
There appears to play the role of a scope marker blocking the wide scope reading of many men which is available in the other sentence. The theresentence cannot mean the same as the raising counterpart: in particular, it cannot mean that many men are not in the room. Thus, since scope is assigned at LF, there cannot be wiped out and the there-sentence cannot be exactly like the raising counterpart. Chomsky (1988a) then proposed to reduce the replacement of there by its associate DP to a process of thereaffixation which does not eliminate there from LF:3 (21)
[there];
tj [in the room]
This analysis was considered to solve the problem posed by Williams (1984) and the proposed structure has subsequently met with general acceptance. Even so, certain data remained unexplained, including the ungrammatical "there seems a man to be in the room, but these have generally been disregarded as minor problems. As a further puzzle, Chomsky (1988a)
102
The syntax of ci
reports an original observation by Kenneth Safir who discovered the following contrast: (22a) [how many men]k do you think that therej were tk in the room? b * [how many men]k do you think that tk were in the room? The second clause shows a typical that-trace effect: the subject of an embedded sentence cannot be extracted across an overt complementizer. The obvious question is why the presence of there makes any difference, especially since in this view it is affixed to the subject at LF. I will try to suggest an answer to this question in the next section. Generally speaking, this analysis for English there as an expletive for the subject position is also currently adopted for Italian ci (see Burzio 1986). Thus, the following pair: (23a) b are the the i.e.
molte copie del libro sono nello studio (many copies of the book are in the studio) ci sono molte copie del libro nello studio (there are many copies of the book in the studio)
generally analysed as stemming from the same d-structure. From formal point of view, then, there are two options: on the one hand, subject can be raised from the small clause to the subject position, spec-IP:
(24)
nello studio Alternatively, on a par with the analysis of there, the expletive ci would be inserted at s-structure as a place holder for the subject of the predication and coindexed with pro in spec-IP:4
Extraction from the embedded subject of ci-sentences 103 (25)
IP
[molte copie del librojj [nello studio]
The major difference between the two languages is that ci is a clitic, thus it is included within a segment of an X° category, namely 1°. There, of course, is a full XP and thus stands in spec-IP position. With this analysis in mind, we will now examine some unexpected phenomena yielded by J/zere-sentences and their Italian counterparts, cisentences. 2.2
Extraction from the embedded subject of ci-sentences and other unexpected phenomena
The following paradigm shows that extraction from the subject in a cisentence gives grammatical results: (26a) b c
[di quale librojj credi che ci fossero [ sc [DP molte copie tj] nello studio]? (of which book do you think there were many copies in the studio) 'which book do you think there were many copies of in the studio?' ce nej erano [Sc [DP molte copie tj] nello studio] (there of-it were many copies in the studio) non c'erano [Sc [DP copie di [molti Hbri]] nello studio] (there weren't copies of many books in the studio)
The complement of the head copie of the noun phrase can be successfully extracted by wh-movement as well as byrce-cliticizationor Q-raising. In fact, the last sentence ((26c)) is compatible with the following interpretation: there are many books such that copies of them are not in the studio. Following well-established ideas originally proposed by May (1985), we can conclude that the quantified noun phrase {molti Hbri: many books) embedded in the subject DP headed by copie can be moved at LF over the negation.5
104
The syntax of ci
This raises two related questions for the analysis of there as a semantically null place holder of the subject position. The first one is related to the expletive replacement hypothesis: if the subject DP is raised to a preverbal position, how is it that extraction from it is allowed? 6 In general, extraction from within a subject in preverbal position gives rise to strong ungrammatically. Consider, for example, the following raising counterpart to the first c/-sentence examined in this section (i.e. (26a)): (27)
* [di quale librojj credi che [IP [DP molte copie tj]j fossero [sc tj nello studio]]? (of which book do you think that many copies were in the studio)
One possible line of reasoning in the face of this problem would involve adopting a derivational approach as opposed to a representational one. Within d-sentences, extraction will not take place from the preverbal position, but from the postverbal position: in fact, expletive replacement would be activated only after spell out has taken place and so the problem of extraction from preverbal position would be eliminated (apart from intermediate traces, which can be deleted at LF). 7 Any commitment between a derivational and a representational approach is far beyond the scope of this book. Nevertheless, it is important to notice that whichever point of view is adopted, a second problem will arise for which no solution is immediately available: even if we assume that movement takes place from the postverbal position, it still has to be explained how extraction from within the subject is possible. Notice that the subject of a d-sentence is in the same syntactic context as the subject of an inverse sentence, namely it is immediately dominated by a small clause: (28)
. . . V°[sc
[DP
...
In chapter 1, it was shown that extraction from within the DP in such a configuration yields a Subjacency violation; witness sentences such as the following: (29)
* [which walljj do you think that [the cause of the riot]j was [Sc [DP a picture of t;] tj]?
The question, then, is how the Subjacency Condition can be deactivated in d-sentences, while remaining fully active in inverse sentences. In other words, why does a semantically null element make such a difference? In the rest of this section I will offer further independent evidence that appears to challenge the current analysis of there-sentences and I will propose an alternative. Let us begin by considering the following pair:
Extraction from the embedded subject of ci-sentences 105 (30a) b
[molte copie del librojj erano [Sc tj nello studio] (many copies of the book were in the studio) * [molte copie del Hbro]j erano [ sc tj [e]] (many copies of the book were)
Why is the second sentence ungrammatical? The answer is immediate: the PP contained in the small clause cannot be omitted. This would amount to omitting the predicate of the clause, and would thus be just as serious as omitting come from John has come, yielding * John has. What happens in the corresponding d-sentence? Clearly, no difference should be expected since the two sentences share essentially the same structure. This is not in fact the case: (31a) b
c'erano [ sc molte copie del libro nello studio] (there were many copies of the book in the studio) c'erano [ sc molte copie del libro [e]] (there were many copies of the book)
When ci is involved, the sentence that omits the PP is perfectly grammatical ((31b)). Yet why should the presence of the expletive subject make any difference?8 Consider now the following case: (32a) b
[molte copie del Hbro]j erano [ sc tj la causa della rivolta] (many copies of the book were the cause of the riot) * c'erano [ sc molte copie del libro la causa della rivolta] (there were many copies of the book the cause of the riot)
As we know, copular sentences can contain predicative noun phrases: indeed, this class of copular sentences is very productive, as was shown in chapter 1. Why then does d-insertion block the possibility of having a predicative noun phrase ((32b))? Is it reasonable to expect an expletive to interfere with the choice of the category of a predicate? A further sharp contrast can be added to the list if we look at cliticization phenomena. Consider the following sentences: (33a) b
Gianni e [DP uno scienziato] (Gianni is a scientist) c'e [DP uno scienziato] (there is a scientist)
If the expletive analysis is assumed, no difference is expected as far as movement of the postcopular material is concerned. This is not the case: (34a) b
Gianni IOJ e tj (Gianni lo is)
* ce loj e t{ (there lo is)
106
The syntax of ci
Cliticization appears to be blocked in a d-sentence ((34b)). Yet it is not clear what blocks this process. Notice that this problem cannot simply be solved by saying that cliticization is blocked tout court. In fact, fte-extraction is fully possible: (35a) b
c'e [DP uno scienziato] (there is a scientist) ce n'j e [DP uno tj] (there of-it is a)
From a theoretical point of view, this contrast is quite unexpected since it is standardly assumed that extraction from a DP and cliticization of the same DP depend on the same locality conditions. Here, locality appears to behave selectively allowing only Axe-cliticization. Why should ci make any difference? 9 The case we will examine now will serve to approach the problem we are discussing in the more general context of the pro-drop parameter. In Italian, the phonologically null pronoun pro is licensed as the expletive for the subject. Why, then, should the copula require an overt element as in (36a), in contrast with other verbs, such as arrivare (arrive) in (36b)? (36a) b
*(ci) sono [ sc [molte copie del libro] nello studio] (there are many copies of the book in the studio) (*ci) [Vp arrivarono molte copie del libro] (there arrived many copies of the book)
A further problem is posed by passive structures. The following case, in particular, is very puzzling: ci cannot occur with passivized VPs ((37a)) unless the object stands in an intermediate position in-between the auxiliary and the main verb ((37b)): (37a) b
(*ci) sono [Vp bruciate molte case dall'incendio] (there are burned many houses by the fire) *(ci) sono [molte case]; bruciate tj dall'incendio (there are many houses burned by the fire)
Why does object-displacement make ci compatible with passives? Where is the intermediate DP located in the second sentence? Notice that expletive replacement is of no help here. If it applies, the two sentences would both be reduced at LF to something like the following: (38)
molte case sono bruciate dall'incendio (many houses are burned by the fire)
However, this sentence is grammatical, and so the observed difference would be obscured.
An alternative analysis: ci as a raised predicate
107
To summarize for a moment, as far as I can see, if we maintain the hypothesis that ci is an expletive of the subject of the predication (along with there), many empirical and theoretical questions relating to issues that range from locality conditions on syntactic movement to the nature of the pro-drop phenomenon would go unanswered. The theory of copular sentences we developed in chapter 1 offers an alternative analysis that succeeds in explaining these facts in a natural way. 2.3
An alternative analysis: ci as a raised predicate
The unified theory of copular sentences showed that the position of specIP can be occupied by a predicate, provided that it is realized by the proper lexical category, namely a noun phrase. The question then is how this idea can be relevant for the theory of expletives. 2.3.1
First step: ci as a raised (pro)predicate
The specific proposal I would like to put forward here is that the element ci is not in fact inserted as the expletive of the subject position but rather is generated as the predicative element of the small clause and then raised. In other words, c/-sentences are to be analysed as instances of what I have called 'inverse copular sentences' (with the occurrence of an optional PP/ AP adjunct). From the formal point of view, I propose to abandon the current representation which is reproduced here in the first diagram ((39a)) in favour of that reproduced in the second one ((39b)): (39a)
DP
ci.
[molte copie del libro]i [nello studio]
108 b
The syntax of ci IP
[molte copie del libro] X{
Before illustrating the advantages of the new representation, two preliminary observations will be made. The idea that ci is a clitic does not call for any special assumption about the structure of small clauses: in particular, the analysis of ci as a clitic is neutral with respect to the major problem concerning the X'-theoretical status of this constituent. All we have to do is to stipulate that since ci is a clitic, the last step in its derivation will be an instance of head-to-head movement, as in the case of all other clitics (in fulfilment of (some version of) the Head Movement Constraint: see Travis 1984, Baker 1988, Kayne 1989a, Rizzi 1990 and Belletti 1994 among others). Abstracting away from the categorial status of d, we can indicate the small clause structure containing ci as follows: (40)
For the sake of simplicity, from now on the trace left by ci will be indicated by't\ without indicating systematically that it is a head. As for the categorial status of d, we will adhere to the traditional hypothesis that it has essentially
An alternative analysis: ci as a raised predicate
109
a nominal character; nevertheless, I would like to point out that within this framework there seems to be no direct empirical evidence for attributing it to a particular lexical category.10 Recall also that raising of the predicative head from the small clause is independently assumed on the basis of cases like the following (see section 1.3.3): (41)
Giannij [lOj e] [ sc tj [Dp [NP tj]]] (Gianni lo-is)
A theoretical problem, on the other hand, might be raised by the presence of an adjunct in (39b), specifically in the case of adjunction to the right by the PP, given in the alternative representation proposed. Recently, a debate has been opened on the question of adjunction as the result of the seminal work of Kayne (1994) (see Chomsky 1994, Brody 1994 and Manzini 1993 among others). Much effort has been expended on trying to eliminate this type of adjunction, i.e. right-adjunction. Naturally, if this turns out to be possible, we can expect pervasive consequences for the entire system. In the present study, I will not try to reduce the cases under discussion to some other independently existing configuration. For our purposes, then, adjunction will just be regarded as a convenient way of expressing the intuition that the postcopular AP/PP constituent of a d-sentence is not the predicate of the small clause. Thus, regarding it as an adjunct might eventually turn out to have been only a temporary step, but it nevertheless is the only analysis that can account for the data in question, as will be seen forthwith.11 Moreover, this option is costless, since there is independent evidence that AP and PP constituents can be considered as adjuncts, as in (42b): (42a) b
[IP [IP John gave me his book] [PP in the room]] [ip [ip John gave me his book] [AP drunk]]
We can now turn to the data discussed earlier and see how the new structure enables us to account for them without exception in a natural way. For the sake of simplicity, the syntactic material that follows the copula in a cisentence will be referred to as the 'coda'. With this new proposal in mind, the empirical facts illustrated above will now be re-examined. The first point that needs to be made concerns the optionality of the AP/ PP in a d-sentence as opposed to the corresponding sentence without ci ((30) and (31)). The AP/PP in the coda of a d-sentence can be omitted because it is not the predicate but rather an adjunct ((43c-d)). On the other hand, the AP/PP in a small clause of a canonical structure cannot be deleted.
110
The syntax of ci
since this would amount to cancelling the predicate of the clause ((43a-b)). The following (simplified) representations capture these facts: (43a) b c d
[molte copie del libro]j erano [ sc tj [PP nello studio]] (many copies of the book were in the studio) * [molte copie del librojj erano [ sc tj [e]] (many copies of the book were) [IP [IP Ci' erano [molte copie del libro tj]] [PP nello studio]] (there were many copies of the book in the studio) [IP (V erano [molte copie del libro tj]] (there were many copies of the book)
The analysis of AP/PP in the coda of d-sentences as adjuncts also explains why such sentences contrast with the canonical case in not allowing a DP to follow the subject (cf. (32)): (44a) * [IP [IP c^ erano [molte copie del libro tj]] [DP la causa della rivolta]] (there were many copies of the book the cause of the riot) b [molte copie del libro]j erano [ sc tj [DP la causa della rivolta]] (many copies of the book were the cause of the riot)
In fact, this simply correlates with a well-established generalization regarding adjuncts in general: DPs, as opposed to PPs and APs, cannot be adjuncts, as is shown by cases such as: (45)
* [IP [iP John came] [DP a friend of my wife]]
Notice that this must be due to purely structural factors, since a friend of my wife would be perfectly acceptable from the point of view of interpretability; witness a sentence like John came as a friend of my wife. All in all, the corresponding d-sentence with an adjunct DP ((44a)) is ruled out without it being necessary to appeal to any special assumptions. As for the process of /ocliticization, we do not need to stipulate any special condition (cf. (33) and (34)): as in all inverse sentences like (46a and c), lo cannot be employed because it bears no features and to do so would destroy the only context in which pro might acquire features, yielding the ungrammatical results of (46b and d) (see section 1.3.3.3): (46a)
[iP [la causa della rivolta] [|P proj sono [ sc [DP le foto del muro] tj]]] (the cause of the riot are the pictures of the wall) b * [IP [la causa della rivolta] [jP pro; loj sono [ sc [DP [NP tj]] tj]]] (the cause of the riot lo are) c [IP proj cij sono [Sc [DP l e f°to del muro] tj]] (there are the pictures of the wall) d * [,P proj cej lOj sono [ sc [DP [NP tj]] tj]] (there lo are)
An alternative analysis: ci as a raised predicate
111
Turning now to ne (cf. (26b)), the fact that it can be extracted without producing a Subjacency Condition violation does not appear to have a straightforward explanation: the issue will be approached directly in the next section. First, we will deal with the problems concerning unaccusatives ((36)) and passives ((37) and (38)). The problematic pair involving unaccusatives is the following: (47a) b
*(cij) sono [ sc [molte copie del libro] tj] (there are many copies of the book) (*ci) [Vp) arrivarono molte copie del libro] (there arrived many copies of the book)
The explanation of this contrast relies on the fact that ci is not simply inserted in the course of derivation to hold the place of the subject of the predication. As will be recalled, in our approach, it plays the role of a predicate and it is base generated. Thus, it follows naturally that it should be absent in unaccusative constructions, because they already contain a predicate, i.e. the verb, and it is required in a copular sentence because the copula is not itself a predicate. This allows us to avoid the rather unnatural assumption that a pro-drop language like Italian has an overt expletive (see the last chapter of this volume for a critical comparison between unaccusatives and d-sentences). Moving on now to the problem case involving passive verbs, the examples in question are reproduced with the traditional labelling: (48a) b
(*ci) sono [VP bruciate molte case dall'incendio] (there are burned many houses by the fire) *(ci) sono [molte case]j bruciate tj dall'incendio (there are many houses burned by the fire)
The analysis of ci as a raised propredicate sheds new light on the above data. The standard conclusion to the effect that ci does occur in passive constructions is simply misleading. In the case where ci is grammatical, I propose that the correct representation should treat the string molte case bruciate daWincendio as the DP subject of the small clause with an AP adjunct (to be represented as: [DP [DP molte case] [A P bruciate daWincendio]]
or perhaps as an embedded small clause; see fn. 11):
112 (49)
The syntax of ci IP
furono [[molte case][bruciate...]] t{ In fact, the same phrase can occur freely in subject position as in the following example: (50)
[DP molte case bruciate dall'incendio] vennero ricostruite (many houses burned by the fire were rebuilt)
On the other hand, the phrase [VP bruciate molte case daWincendio] cannot be analysed as a DP; witness the ungrammaticality of sentences like the following: (51)
* [DP bruciate molte case dall'incendio] vennero ricostruite (burned many houses by thefirewere rebuilt) Thus, the sentence containing this string must be a genuine passive sentence and thus there is no possibility for ci to be interpreted, because there isn't a small clause: (52)
[molte case]
To conclude this section, I will offer some independent evidence in support of the hypothesis that the PP/AP in a d-sentence is an adjunct. In general,
An alternative analysis: ci as a raised predicate
113
one would expect extraction of a phrase from within an adjunct to yield less acceptable results than extraction from predicates or argument positions. In our case, this prediction is confirmed by empirical data: extraction from the AP/PP in the coda is more problematic when ci occurs. Consider, for example, the following sentences involving an AP: (53a) b
a chij sembra che [iP [molte persone]; siano [Sc U [AP debitrici tj]]]? (to whom does it seem that many persons are indebted) * a chij sembra che [iP [iP cij siano [ sc molte persone tj]] [AP debitrici tj]]? (to whom does it seem that there are many persons indebted)
There is clearly a contrast between the two cases and this would go unexplained if we were to maintain the traditional analysis based on the expletive replacement hypothesis, since under this analysis the AP/PP would be in exactly the same position. Consider now the following pair involving a PP, leaving the source of perche (why) unspecified: (54a) b
perche sembra che [iP [molti italiani]j siano [Sc tj [PP in sciopero]]]? (why does it seem that many Italians are on strike) perche sembra che [iP [IP cij siano [ sc molti italiani tj] [PP in sciopero]]]? (why does it seem that there are many Italians on strike)
Thefirstsentence ((54a)) asks a question about the reason for the strike and about why such a large number of Italians are involved, whereas the second sentence ((54b)) only asks about why such a large number of Italians are involved.12 Although the contrast is sharply clear, it is not immediately obvious how to account for it. Nevertheless, it is clear that if the two sentences shared the same structure at LF (after expletive replacement) there would hardly be any way to account for this difference. One possibility is to adopt the more articulated structure which includes a small clause constituent for the adjunct PP as proposed by Chomsky (1981) (cf. fn. 16 of chapter 1). In this case, it becomes reasonable to analyse the clausal whelement why as being extracted from the adjunct clausal constituent and thus represent (54b) as follows: (55)
perchej sembra che [IP (jP cij siano [sc molti italiani tj]] [sc PRO [PP in sciopero] tj]]?
Given this structure, we can reasonably conclude that the trace of perche (why) cannot be extracted from the small clause since the latter is an adjunct, a kind of extraction which is generally assumed to be impossible. Further confirmation of the analysis of the PP in the coda of a d-sentence as an adjunct comes from the interpretation of quantifiers. Consider the following cases:
114
The syntax of ci
(56a)
[IP [due foto del muro]j sono [ sc tj [PP su tre riviste ciascuna]]] (two pictures of the wall are on three magazines each) * [IP [IP cij s o n o [sc due foto del muro tj ]] [PP su tre riviste ciascuna]] (there are two pictures of the wall on three magazines each)
b
The distributive reading of tre (three) on due (two) yielding six entities, as required by ciascuno (each), is not available when tre is contained in a phrase that is higher than the one containing due. In other words, the scope of tre can never be higher than that of due because due is generated in a lower position and cannot be raised to reach the same scopal domain as tre. If the two PPs basically shared the same domain, as the expletive analysis claims, this contrast would hardly be understandable. Taken together, these contrasts point to the conclusion that when ci is present, extraction from the PP or the AP gives the typical violations of extraction from an adjunct. Notice also that the presence of ci at LF now follows without stipulation since predicates are independently considered as legitimate objects at this level of representation. Moreover, considering ci as a predicate allows us to dismiss the problem concerning the violation of Condition C of Binding theory: this condition rules out the binding of a referential expression by another referential expression. Since there is a predicate in our analysis, the question does not even arise. To summarize, many empirical facts (including optionality of PP/AP in the coda, cliticization phenomena, wh-movement, contrasts with passives and unaccusatives) challenge the theory of ci as a subject expletive and support an alternative proposal based on the unified theory of copular sentences as developed in chapter 1. C/-sentences are to be analysed as inverse copular sentences, that is, as sentences where the predicative element contained in the small clause has been raised to the preverbal position. Of course, since ci is not regarded as a semantically null place holder occupying the subject position, there would be no need to replace it by its associate at LF. Notice, however, that the distribution of such an element would in any case continue to be determined by exactly the same locality conditions governing chain formation as were involved in the expletive replacement analysis. The only difference is that we do not need to have recourse to LF chains. Overt movement involving the raising of predicative noun phrases would be sufficient. Moreover, since expletive replacement would not take place, the problem of extraction from the preverbal subject would not arise (and there would be no need to choose between a derivational and a representational approach). Nevertheless, the second problem addressed in this chapter remains in
An alternative analysis: ci as a raised predicate 115 need of an answer: why can we extract from the subject of this type of inverse sentence without violating the Subjacency Condition? 2.3.2 Second step: the llexicalizatiori> of the copula In chapter 1, it was shown that there can be no extraction from within the subject of an inverse sentence. Let us take the contrast in ne-extraction as a reference and consider the following simplified representation (omitting pro): (57)
* la causa della rivoltaj [nej e] [Sc [DP u n a foto tj] tj] (the cause of the riot of-it is a picture)
Recall that the explanation for the ungrammaticality of this sentence relied on the Subjacency Condition: crucially, the problem is that the copula cannot 'remove' the barrierhood of the embedded DP. In fact, although the copula is in the proper local configuration it does not select the subject of the small clause: this prevents L-marking with the result that the embedded DP counts as a barrier. Since we are assuming that c/-sentences are inverse sentences, we expect extraction to be as bad from the subject position. As we already noticed in section 2.2, this is not the case: (58a) * la causa dellarivoltaj[nej e] [sc [DP una foto tj] tj] (the cause of the riot of-it is a picture) b [cej nj'e] [sc [DP una foto tj] tj] (there of-it is a picture) How can we explain this contrast? The shift from the analysis of ci as an expletive of the subject to that of ci as a predicate seems promising, once we take the empirical content of the Subjacency Condition seriously. In fact, if the Subjacency Condition is formulated in terms of selectional properties (as proposed by Cinque 1990a), then the very fact that ci is a predicate rather than a semantically null element turns out to be relevant. By relying on Chomsky's (1986a: 86-7) definition of 'selectional property', we now have a plausible way to interpret the data: clearly, predicates do select their subjects, expletives do not. This immediately suggests an explanation of the fact that extraction from the subject of an inverse sentence is possible. The incorporation of the head ci into the copula means that this verb inherits the selectional capacities of the predicate. Crucially, the copula now meets both requirements for L-marking: it locally ccommands the lower DP (inherently) and it selects it (derivatively). Thus, the lower DP is L-marked by the complex 'a+copula' and it does not count as a barrier any longer:13
116 (59)
The syntax of ci V
From a terminological point of view, then, we can conclude that the copula as a functional element is 'lexicalized' by ci in the sense that the clitic element makes the copula acquire the typical selectional property of an intransitive verb, namely that of selecting an argumental DP. The difference with respect to inverse sentences not involving ci is quite transparent: leaving aside the role of the propredicative/?ro, the head which has selectional properties here is the N° causa. This is embedded in a complex phrase: in fact the most prominent head in preverbal position is the D° la taking the NP projected by causa as a complement, as in the following simplified representation (VP omitted): (60) IP
We might reasonably assume that from this position the N° cannot affect the copula in the same way as ci does and thus it cannot endow the copula with selectional properties. Independent evidence for the hypothesis that ci 'lexicalizes' the copula can be provided as follows. It has been shown (see Beninca 1980 and Longobardi 1991a) that empty D°s are licensed in Italian only if the DP is governed by a lexical head (see also Diesing 1990 for a detailed discussion of this issue in German). For example, the distribution of empty D° is sensitive to the subject-object asymmetry:
An alternative analysis: ci as a raised predicate (61a) b
117
Gianni dice [c° che] [iP [DP [D° *(i)] [NP bambini]] pensano] (Gianni says that the kids think) Gianni dice che Maria non [Vp [v°vede] [DP [D° (i)] [NP bambini]]] (Gianni says that Maria not sees kids)
The contrast shows that when a lexical head (vede: sees) governs the DP, the D° can be empty: (62)
V
On the other hand, since in preverbal position there are no lexical heads governing the DP, an 'expletive' article must be inserted (see Longobardi 1991a for a detailed analysis of this phenomenon): (63)
Bearing this in mind, let us turn now to copular sentences and consider the following contrast: (64a)
la causa della rivoltaj [v° erano] [ sc [DP [D° i] [NP bambini]] tj (the cause of the riot were the kids)
118 b c d
The syntax of ci * la causa della rivoltaj [v° erano] [ s c [DP [D° e] [NP bambini]] t4] (the cause of the riot were kids) [v° Cj'erano] [ s c [DP [D° i] [NP bambini]] tj] (there were the kids) [v° Cj'erano] [ s c [DP [D° e] [NP bambini]] tj] (there were kids)
This shows that the copula cannot license an empty D° in the subject of an inverse sentence unless ci has been raised: the hypothesis that this process 'lexicalizes' the copula is then independently motivated. Notice that no theory of ci as an expletive of the subject of predication would be able to explain these facts, especially if it is assumed that a process of replacement or affixation is involved to raise the lower subject to a preverbal position where empty D° cannot be licensed. All in all, we appear to have reached the result we aimed at. We have provided an explanation for the fact that extraction can take place from within the subject of a d-sentence without yielding a violation of the Subjacency Condition, whereas this is not possible in inverse sentences with a full DP predicate. The crucial step was to consider ci as a predicate raised from the small clause, rather than as a semantically null place holder standing as subject of predication (to be wiped out at LF). Given this assumption, incorporation of ci into the copula allows the latter to inherit its selectional properties.l4 The resulting complex is now able to perform as an L-marker, and as such licenses extraction from within the DP. In the following section we will adopt a comparative approach, attempting to apply the present proposal to the English equivalent of Italian d-sentences. 2.4
A difference between English there and Italian ci
The aim of this section is twofold: on the one hand, I will sketch out an analysis of f/iere-sentences as inverse copular sentences, on a par with d-sentences examined in the previous section. On the other, a major difference between ci and there will be brought out; this will then be seen to take on a certain importance in chapter 3. My basic proposal is that the analysis of ci as a raised predicate can be extended to English there. It is not difficult to find empirical evidence for this idea and I will now briefly discuss the key data. First of all, f/iere-sentences contrast with the normal copular sentences in exactly the same way that we found to be the case in Italian (see section 2.3.1). Thus, the basic distributional facts regarding there are essentially the same as in Italian:
A difference between English there and Italian ci
119
(65a) many copies of the book were in the studio b * many copies of the book were c there were many copies of the book in the studio d there were many copies of the book e f
many copies of the book were the cause of the riot * there were many copies of the book the cause of the riot
g h
* there were burned many houses by the fire there were many houses burned by the fire
Similarly, extraction from the AP/PP following the copula is degraded in f/iere-sentences, showing that such constituents are adjuncts: (66a) to whom does it seem that many persons are indebted? b * to whom does it seem that there are many persons indebted? c why does it seem that many Italians are on strike? d why does it seem that there are many Italians on strike? e two pictures of the wall are on three magazines each f * there are two pictures of the wall on three magazines each In addition to this evidence, which essentially reproduces what we have already seen in Italian, English offers two further pieces of data to support the analysis of f/iere-sentences as inverse copular sentences; these data are not available in Italian for independent reasons. It is a well-known fact that in English infinitival copular sentences may occur as an alternative to small clauses, as complements of believe-type verbs: (67a) b
Mary believes John to be the cause of the riot Mary believes John the cause of the riot
A long-standing question about there has to do with the ungrammaticality of sentences such as the second one in the following paradigm: (68a) I believe [there to be a picture of the wall in the room] b * I believe [there a picture of the wall in the room] In fact, this has frequently been seen as a question of explaining why the copula cannot be omitted when the expletive there is inserted. On the other hand, if one adopts the new analysis, this fact will simply follow as an instance of a general property of inverse copular sentences. As was shown in the first chapter, 15 the predicate can precede the subject only if a landing site has been provided by the copula: (69a) Mary believes the cause of the riot to be John b * Mary believes the cause of the riot John
120
The syntax of ci
If there is now analysed as a predicate, this phenomenon will be immediately explained without any special assumptions having to be made: the predicative element there can precede the subject only if a landing site has been provided by the copula. The second fact regards Case assignment. Infinitival contexts show that there must be assigned Case; witness the obligatory insertion of for in C° position (see Safir 1985 and references cited there): (70)
[*(for) [there to be a picture of the wall]] is unusual
Under the hypothesis that there is an expletive, some specific assumptions obviously have to be made in order to account for the rather surprising fact that a non-argument has to receive Case. On the view being proposed here, on the other hand, this now follows without specific stipulations about there. Consider the following sentence reproducing example (54) in chapter 1: (71)
[*(for) [the cause of the riot]j to be [ sc [a picture of the wall] tj]] is unusual
Thus, Case is assigned to there as to all raised DPs in copular sentences irrespective of whether they are argumental or not. Assuming the analysis of there as a raised predicate, Case assignment to it is no longer surprising but rather exactly what we would now expect. 16 We may also notice that the remaining apparent counterexample to the theory of 'expletive replacement' (cf. (18a-b)) can also be dealt with on the basis of what is being proposed here. The crucial cases were the following: (72a) * there seems a man to be in the room b a man seems to be in the room Thus, the ungrammatical sentence shown here would appear to be problematic, since there is nothing to stop a man raising as far as the main clause subject position, as shown in the grammatical example. In our system, this case is not problematic. As a first step we simply have to assume that along with their argumental counterparts predicative DPs can raise a further step when the infinitival complement they occur in is embedded under seem. This is clearly shown in the first example of the following pair: (73a) b
[the cause of the riot]j seems [IP tj to be [Sc [a picture of the wall] tj]] [a picture of the wall]i seems [IP tj to be [Sc tj [the cause of the riot]]]
Once this assumption has been made all we need to do is appeal to locality conditions on movement in order to explain the offending case (see section 1.3.2). Thus, it is impossible for both DPs to raise out of the small clause, ending up in different positions as in the following cases:
A difference between English there and Italian ci
121
(74a) * [the cause of the riot]j seems [!P [a picture of the wall]j to be [ sc tj tj]] b * [a picture of the wall]; seems [IP [the cause of the riot]j to be [ sc U tj]] Similar considerations will therefore apply to there-sentences of the type in question: (75a) therej seems [iP tj to be [Sc [a picture of the wall] tj]] b * therej seems [iP [a picture of the wall]j to be [ sc tj tj]] Here, the offending sentence is ruled out by locality conditions on movement, exactly as in the cases involving raising of lexical DPs above. It would appear, then, that what remained as an unexplained exception under the theory of expletive replacement is straightforwardly derivable in our system, without any special assumption having to be made. All in all, then, //zere-sentences can be regarded as inverse copular sentences on a par with d-sentences in Italian: the subject remains in situ in the small clause and there is raised from the predicative position: 17 (76)
IP
there
[many copies of the book] t{ Nevertheless, although there-sentences and d-sentences are essentially similar from a predicational point of view they do differ substantially with respect to a property which will play a crucial role in the next chapter. As a preliminary we will consider some Italian data. The subject of a d-sentence can undergo wh-movement as opposed to the case of an inverse copular sentence containing a full predicative DP: (77a) * [quale foto]j credi che [IP [la causa della rivolta] [IP prOj fosse [ sc tj tj]]]? (which picture do you think that the cause of the riot was) b [quale foto]j credi che [IP pro [VP [cij fosse] [sc U tj]]]? (which picture do you think that there was)
122
The syntax of ci
Within our framework, this contrast can easily be explained. Recall that for a DP to be raised from the small clause it is necessary for it to pass through spec-IP where it will activate agreement on the local head: this will turn the inflected verb into a proper governor for the trace within the small clause (essentially as an 'extended chain' in the sense of Chomsky (1986b)). The contrast between (77a) and (77b) can now be seen as a consequence of the fact that ci is a clitic. When ci is absent, as in thefirstsentence, the spec-IP position is occupied by a propredicative/?ro, directly linked to the trace in the small clause (recall that la causa della rivolta is an adjunct). In this case, the subject qualefoto cannot be extracted. If the subject were to move through spec-IP it would inevitably obliterate the chain of the predicate generated in the small clause. Focussing on the relevant segment, we can capture the data as follows: (78)
una foto del muro
In the second sentence, on the other hand, the situation is quite different: here, the predicate is ci, a clitic element. This has an extremely important consequence for the structure: spec-IP is not directly related to the position where predicates are generated in the small clause. Instead, the direct link is between the predicate position and the clitic ci. In other words, the element pro occurring in Italian d-sentences is a genuine expletive:
A difference between English there and Italian ci
123
(79)
CI-
una foto del muro
This is sufficient to explain why the subject qualefoto can be extracted in a d-sentence. The trace of the subject can be properly licensed now because the subject can be raised to spec-IP first: obliterating pro in that position would not be of any consequence since the predicate, namely ci, has cliticized into the verbal head. Assuming this is correct, the traditional account of subject extraction in Italian proposed by Rizzi (1982) can now be applied. The subject is rightward-moved to an adjunct VP position and then extracted from a postverbal position.18 Let us now turn to English. In this language, where the equivalent of ci is not cliticized onto the verbal head but rather occupies the spec-IP position, the prediction is that movement of the subject of a there-sentence is as bad as movement of the subject of an inverse sentence with a full DP predicate. This is in fact the case: (80a) * [which wall]; do you think [IP [the cause of the riot]j was [sc tj tj]]? b * [which walljj do you think [1P [there]j was [Sc U tj]]? In both cases, moving out the subject yields a violation of the locality conditions on movement: as was extensively discussed in chapter 1, the chain of the subject cannot activate agreement on the copula because this process is already employed to properly govern the trace of the predicate. As a matter of fact, English there is not, of course, a clitic, as opposed to Italian ci:
124 (81)
The syntax of ci IP
there; I
a picture of the wall Thus, we observed the crucial difference between there-sentences and cisentences. Nevertheless, it is important to notice that there behaves just like ci in lexicalizing the copula. In fact, extraction from the subject of a theresentence gives well-formed results, unlike cases where a full predicative DP is raised to spec-IP position. Consider the following contrast: (82a) * [which wall]j do you think [DP the cause of the riot]j was [sc [DP a picture Oftjtj]? b [which walljj do you think [there]j was [sc [DP a picture of tj] tj]? Similarly, the same type of contrast between movement of the whole subject and extraction of a part of it can be detected at LF: (83a) b
[IP [the cause of the riot]j isn't [Sc [DP a picture of [many walls]] tj]] [IP [there]j isn't [sc [Dp a picture of [many walls]] tj]]
In the first sentence, the quantifier cannot be moved out of the scope of negation, i.e. it cannot mean 'many walls are such that the cause of the riot is not a picture of them'. In the second sentence, the quantifier can be extracted in which case it will yield the following meaning: 'many walls are such that there aren't pictures of them'. Again, these contrasts leave little doubt that the mechanism of lexicalization is also active in English, despite the different categorial status of ci and there. It remains less than clear, however, how we can account for it in English. For the Italian cases, we said that ci lexicalizes the copula by incorporating into it; this allows the copula to acquire the selectional properties required to L-mark the lower subject. Thus, extraction becomes possible without any violation of the Subjacency Condition. Exactly the
A difference between English there and Italian ci
125
same effect is visible in English even if there is not a clitic. What we need, then, is an explanation which will cover both there and ci. One possibility will be suggested in what follows. From a certain point of view, the difference between there and a full predicative DP is the same as that between ci and a full predicative DP. In the latter case it is the head N°, embedded in a DP, that has selectional properties: (84)
IP
Clearly, N° is too deeply embedded to affect V°, the copula. In the former case, instead, the element which has selectional properties is there, which is not embedded in a larger DP. Moreover, there has a twofold nature, from an X'-theoretical point of view: it behaves like a head, in that it does not contain intermediate projections, i.e. complements or specifiers, but it also behaves like a maximal projection, in that it occupies a spec-position. The following representation aims at capturing this fact: (85)
there
Thus, it seems not unreasonable to assume that the transmission of the selectional properties that make the copula become an L-marker is performed in English by a process of spec-head agreement holding between there in spec-IP and the verb. In contrast, when a full predicative DP is raised
126
The syntax of ci
to spec-IP, the spec-head relation is unproductive as far as L-marking is concerned: again the relevant head is the N° but it is too deeply embedded to have any possibility of transmitting its selectional properties to the inflected verb (Vio). Thus, when a full DP is raised, the copula does not Lmark the lower subject and a Subjacency Condition violation results whenever extraction takes place.19 Before concluding this section, I would like to highlight two consequences of the theory presented here. The first consequence is relevant to the analysis of there as a scope marker (Williams 1984). The hypothesis that there is a scope marker turns out to be simultaneously too weak and too strong. It is too weak, because extraction of a quantified subject at LF must be independently blocked by locality conditions just as it is from all inverse copular sentences; witness the following contrast: (86a) there aren't many books in the library b some students' purchase was many books Whether or not there is present, the quantified element cannot raise to take wide scope. As we know, subject extraction in inverse sentences is impossible for locality reasons because spec-IP is already occupied by the raised predicate: there and some students' purchase, respectively, in our examples. On the other hand, the analysis of there as a scope marker proves also to be too strong, once again for independent reasons, as is shown by this further contrast: (87a) there aren't many books in the library b there aren't pictures of many books in the library In fact, in (87b) extraction of the quantified element is successful, whereas in (87a) it is not. Thus, the second sentence can mean that many books are such that pictures of them are not in the library; it contrasts with the first one where many cannot have wide scope. It seems then that the status of there as a scope marker is in need of some qualification; perhaps it would be better to call it a 'selective scope marker', though it is far from clear what this might amount to. The alternative analysis being suggested here, where movement of the whole subject is blocked by locality conditions but subextraction is allowed, will immediately capture the contrast. The second of the two consequences mentioned above concerns the contrast discussed in Chomsky (1988a) which was presented in the introductory section of this chapter (cf. (22a-b)). The examples discussed there are repeated here: (88a) [how many men]k do you think that therej were tk in the room? b * [how many men]k do you think that tk were in the room?
Conclusions 127 As we noticed above, this contrast remains rather puzzling and problematic for the established theory of there: why should the presence of what is assumed to be an expletive allow a sentence to escape that-trace effects? Recall also that at LF the subject of a there-sentence should be affixed to there. Under our analysis, it becomes possible to provide a straightforward explanation for such phenomena. Various data show that extraction from the subject position of a there-sentence is fully grammatical both overtly and at LF. Thus, if we assume that at LF only how many is extracted from within the subject DP (see Moro 1991), the contrast can be explained in a rather natural way:20 (89a)
[how many]k do you think that therej were [sc [DP (tk) D° tk men] tj] in the room? b * [how many]k do you think that [oP tk menjj were [Sc U in the room]
Given this representation, the role of there is no longer mysterious. Extraction of the wh-phrase how many from the subject DP is possible only when there is present, because otherwise a Subjacency Condition violation will be yielded, as has been argued on independent grounds.21 In the same fashion, we can explain the following contrast involving which and what without any recourse to semantic conditions (see Heim 1987 for an extensive discussion on this matter): (90a) * [DP which one]; do you think therej was [sc tj tj]? b [what]k do you think therej was [sc [DP (tk) D°" tk] tj]? Through relying on the independent assumption that what is extracted from the DP as opposed to which (see section 1.3.2), the contrast can now be explained. The first sentence is ruled out as a violation of locality conditions on movement: once again, the only way for a trace within a small clause to be licensed is for it to be governed by the extended chain established by activating agreement features on the copula. Since this process is already employed to license the trace of the predicate there, the sentence is ruled out. On the other hand, the second sentence is fully grammatical because the trace of what is licensed by passing through spec-DP and extraction from the subject of a there-sentence is allowed thanks to the lexicalization of the copula.22 2.5
Conclusions
In this chapter we have exploited the proposal, originally made in the first chapter, to abandon the 'rigid' clause structure which assumes that spec-IP
128
The syntax of ci
is always the subject of predication in favour of a more 'flexible' structure which allows the raising of predicative DPs to such a position. This more flexible clause structure has been applied to f/iere-sentences and cisentences. At the outset, the traditional analysis was illustrated, according to which both there and ci are inserted at s-structure as semantically null elements holding the place of the subject of predication, technically 'expletives'. An analysis of this type, as we saw, allows the following structure to be assigned to a d-sentence like ci sono molte copie del libro nello studio (there are many copies of the book in the studio):
(91)
[molte copie del libro] {[nello studio]
A further refinement of this analysis involves Chomsky's (1986b) 'expletive replacement hypothesis'; as we saw, this is linked to the Principle of Full Interpretation, which forces the DP contained in the small clause to affix the expletive at LF. In the discussion that followed many empirical arguments were provided, both against the analysis of there and ci as subject expletives and against the expletive replacement hypothesis. Among the various phenomena examined were: optionality of PP/AP in the coda, cliticization phenomena, whmovement and contrasts with passives and unaccusatives. We then showed that the empirical data can easily be captured once we assume that cisentences are simply instances of a more general subtype of copular sentences, namely inverse copular sentences, and thus they involve raising rather than insertion:
Conclusions 129 (92)
Cl
[molte copie del libro] i{
The theoretical consequences of this alternative approach were then extensively discussed. First, we observed that this approach to the analysis of f/iere-sentences appears to retain all the considerable advantages of the analysis based on replacement, essentially the locality nature of the phenomenon. As was argued, the distribution of there can still be traced back to the independent principles that govern chain formation. The difference is that in our system the chain is overt whereas in the traditional system it is formed in the LF component (and is forced by the special hypothesis of 'expletive replacement'). Moreover, we observed that the analysis of there- and d-sentences is also to be regarded as offering indirect support for the analysis of inverse sentences based on the interaction of two distinct principles. In the first chapter (in section 1.3.2), indeed, we observed that inverse copular sentences have two major properties (the subject of inverse sentences can neither be moved nor allow a subpart of itself to be extracted) and we proposed to regard these two phenomena as the effect of the ECP and the Subjacency Condition, respectively. From a methodological point of view, the objection could be raised that such an analysis based on two distinct principles acting on the same DP could be regarded as redundant, not to say suspect: the conditions blocking movement o/a DP are expected to be the same as those blocking extraction from the same DP. TZ^re-sentences, nevertheless, show that the line of reasoning based on two distinct principles was indeed empirically correct because they actually provide evidence to the effect that the same DP can allow extraction of a constituent embedded
130
The syntax of ci
within it but cannot undergo movement itself. This was clear from examples like the following: (93a) b
there aren't [DP many books] in the library there aren't [DP pictures of many books] in the library
In the first sentence, many books cannot be moved as a whole at LF to take scope over negation; in the second one, instead, many books can be extracted. This indeed provides support for the validity of our analysis even from a purely theoretical point of view. The fact that the two instances of violation under discussion can occur separately is predicted by our approach, since the two principles involved have been related to two distinct empirical facts. The locality principles on trace licensing are related to the possibility of activating agreement on a local head (essentially following Chomsky's (1986b) ECP based on the notion of 'extended chain'): (94a) [which picture^ do you think [IP tj was [ sc tj the cause of the riot]]? b * [which picturejj do you think [IP [the cause of the riot]j was [ sc tj tj]]? The Subjacency Condition, instead, is related to the selectional property of a head. In general, we noticed that the possibility of extracting from within the subject of a there-sentence rather than from that of an inverse sentence with a full DP predicate can easily be understood in our framework: (95a) * [which wall]j do you think [DP the cause of the riot] was [DP a picture of tj] b [which wall]j do you think [DP there] was [DP a picture of tj] All in all, the shift away from the analysis of there as a semantically null place holder for the subject in favour of a new analysis where it is a raised predicate has turned out to be well motivated and this from a number of different perspectives. In the course of the discussion little attention has been paid to the semantics of there-sentences and to a particular phenomenon which distinguishes Italian from English, namely the so-called 'Definiteness Effect'. These two topics will be the core issues addressed in the next chapter.
3
Are there parameters in semantics? The defining properties of existential sentences
3.1
Introduction: on the absence of the Definiteness Effect in Italian existential sentences
In chapter 2, we explored the syntax of ci- and there-sentences and tried to highlight the similarities between the two structures in the two languages concerned. As we will now see, a sharp contrast can be found between the two languages if one turns one's attention most directly to the subject of this type of sentence. The following classic paradigm shows English and Italian in a word-for-word correspondence: (la) b c d
there are [DP many solutions] [PP in this book] ci sono [DP molte soluzioni] [PP in questo libro] * there is [DP John] [PP in this garden] c'e [DP Gianni] [PP in questo giardino]
In the same way, a DP headed by a definite article or by a universal quantifier or by a proper name cannot be the subject of a there-sentence whereas it may be the subject in d-sentences. Traditionally, the restriction which rules out the offending case in English has been formulated by appealing to semantics: (2)
sentences of the form there is definite noun phrase . . . are violations of the surface rules of semantic interpretation rather than of syntactic rules, it appears. (Chomsky 1973; reproduced from Chomsky 1977a: 52)
Since the notion of 'definiteness' plays the crucial role here, this restriction has been called the 'Definiteness Restriction' and the corresponding phenomenon the 'Definiteness Effect'.1 Why does the Definiteness Restriction fail to hold in Italian? Before attempting any explanation, let us begin by considering the issue from an abstract point of view and asking what kind of answer the theory might in principle allow. Comparative grammar shows that cross-linguistic variations are to be traced back to the interaction of a few independent parameters set by the child on the sole basis of overt evidence. The following citation from 131
132
Are there parameters in semantics?
Chomsky (1986a) illustrates the current theory: There may be general principles that determine how the switches [of a parameter] are set, for example, the subset principle discussed by Berwick (1982), which states that if a parameter has two values + and —, and the value - generates a proper subset of the grammatical sentences generated with the choice of value +, then - is the "unmarked value" selected in absence of evidence; this is a necessary and sufficient condition for learning from positive evidence only, insofar as parameters are independent' (Chomsky (1986a): 146). So for example, the pro-drop parameter is set by the occurrence of sentences which do not have a phonologically realized subject, by the presence of inverted subjects, etc. It may also be that in the last resort, the proper value of the parameter is decided by the presence of some overt (string of) elements.2 If we maintain that the restriction ruling out definite subjects in English is semantic, what kind of evidence should we expect to find that might overtly differentiate the two languages? In other words, how does the child know that definite noun phrases are prohibited in English and allowed in Italian in the context under discussion? Suppose we assume that the application of the Definiteness Restriction is parametric. The subset principle leaves only one option: the unmarked value is that this restriction does apply, for a grammar containing it would generate a proper subset of a grammar not containing it. Universal Grammar provides the child with the positive value of the parameter. The English child will never have to change it. The Italian child, instead, would eventually shift to the negative value (i.e. the restriction does not apply), since he hears sentences violating it. It seems then that we are obliged by this type of reasoning to conclude that natural languages do contain parametric variation concerning semantic restrictions, that is 'semantic parameters'. Whether or not this conclusion is tenable is hard to decide and definitely an empirical matter. More generally, the possibility of semantic parametrization has not been properly investigated yet, as far as I know, and the issue is still entirely open. Indeed, there may well be a priori reasons for not adopting this view. For example, in the recent minimalist framework stemming from Chomsky (1988a), (1993), it has been proposed that, apart from Saussurean arbitrariness, the field of language variation is limited to lexical differences, specifically, to differences in the features of the lexical elements that occupy the functional category nodes' (Marantz 1995: 372; see also Manzini and Wexler 1987). On the other hand, other theories have been put forward which purport to show that parametric variations in semantics can be considered as wellformed entities and indeed produce correct predictions (see for example Chierchia 1993, 1995).
On the so-called 'existential meaning': the role of syntax
133
All in all, if one adopts the point of view about the Definiteness Restriction which originally goes back to Chomsky (1973), it can hardly be doubted that there- and c/-sentences raise a serious theoretical and empirical problem concerning language variation. Nevertheless, although this seems to me a genuine issue, I will attempt to show that we do not in fact need to commit ourselves on the issue of the plausibility of what I called 'semantic parameters'. Indeed, the aim of this chapter is to show that regarding theresentences as inverse copular sentences (in the sense specified in the first chapter) will allow us to trace the cross-linguistic difference we have just exemplified back to an independent syntactic parameter, specifically the pro-drop parameter. This will be shown to be a highly desirable result, both in itself and for independent reasons, since, from a cross-linguistic point of view, the Definiteness Effect will be seen to correlate with such a parameter. Whether or not all apparent semantic variations between languages can be traced back to syntactic parameters in the same way is a question that cannot be examined here. 3.2
On the so-called 'existential meaning9: the role of syntax
A well-established practice ranging from linguistics (for example, Quirk and Greenbaum 1973) to logic (for example, Quine 1987) is that of calling theresentences 'existential sentences'. Although intuitively it is quite clear what this label means - i.e. that these sentences are used to assert or deny the existence of something - it remains difficult to give an explicit formal definition of what is meant by 'existential meaning' in natural languages. One possibility is to use artificial languages. Thus, a sentence like there are solutions could be translated into a first-order predicate formula of the type 'E x such that solution (x)'. Although this might well be the exact formalization of the content of an existential sentence, it is clearly not enough, on its own, to satisfy a linguist; indeed, if our aim is to construct a theory of natural languages, we clearly need to begin by singling out the basic syntactic properties that qualify a structure as an existential sentence and then attempt to explain the degree of variation which we encounter in each individual language, within the limits posed by this universal schema. Of course, this is not something specific to the study of existential sentences; rather, it is the natural consequence of adopting a principles and parameters approach. Take, for example, the analysis of passive sentences. We know that a cross-linguistic correlation holds between pairs of sentences of the following type:
134 (3a) b c d e f
Are there parameters in semantics? John kisses Mary Mary is kissed (by John) John hates Mary Mary is hated (by John) John hugs Mary Mary is hugged (by John)
The correlation in each pair is intuitively clear: on the one hand, the noun phrase following the verb in the first sentence plays the same role as the noun phrase preceding the verb in the second sentence; on the other, the noun phrase preceding the verb in the first sentence plays the same role as the one following the verb in the second sentence and we also know that the latter is immediately preceded by the preposition by and can be omitted, which is not the case with the corresponding noun phrase in the first sentence. Moreover, the two sentences differ also with respect to inflectional morphology: the two verbal suffixes {-(e)s and -ed respectively) differ and in the second case an auxiliary occurs, namely be. Since the very beginnings of grammatical study, the basic requirements of descriptive adequacy have been met by calling the second type of sentences 'passive' as opposed to 'active'. Modern linguistics, on the other hand, has approached a second level of adequacy, where the aim is to explain the phenomena and not just describe them, and radically different types of questions have been formulated such as: what do we know when we know the meaning of a passive sentence? Is this knowledge encoded in the lexicon or is it the output of a syntactic configuration? How can this vary across languages? Indeed, in the specific case of passives the answers to these types of questions are by now well known. A passive sentence is essentially a sentence involving a transformation which moves the object (more precisely, the 'internal argument') to subject position and makes the subject (again, the 'external argument') an optional element. Crucially, movement of an argument to the subject position is not driven by languagespecific rules nor even by rules specific to this configuration; rather, it is forced by a set of principles and by morphological requirements that apply to all languages in a modular fashion. Thus, the same type of movement is involved whenever the syntactic environment requires it, as it does, for example, in raising structures involving seem in sentences like John seems to kiss Mary. No one would any longer seriously conceive 'passives' as primitive entities. 'Passive' is now simply a sort of label that is assigned to an epiphenomenon, a taxonomic label that covers the interaction of more abstract primitives.3 Thus, knowing the meaning of a passive sentence, in the sense specified for example in Chomsky (1986a), is now thought to
On the so-called 'existential meaning': the role of syntax
135
amount simply to knowing that a particular type of movement is involved and that this movement is the output of a conspiracy of independent factors which may lead independent lives. In general, the set of properties which correlate with a certain construction is often technically called the set of the 'defining properties' of such a construction.4 Normally, the shift from construction-specific rules to the interaction of general principles has also applied to the cross-linguistic analyses. In this new perspective, on the other hand, the particular array of items has of course been less important. Consider, for example, the following pairs comparing English and Latin: (4a) b c d
John kisses Mary Mary is kissed (by John) Joannes Mariam basiavit (Joannes-nom Mary-acc kiss-third-pers.sing.active) 'John kissed Mary' Maria (a Joanni) basiatur (Mary-nom. by Joannes-abl. kiss-third.pers.sing.passive) 'Mary is kissed by John'
Whether or not the verb be occurs, the two sentences in which the object {MaryI Maria) is promoted to subject position and the subject becomes an optional element (by Johnla Joanni) are both considered as instances of the same general abstract 'passive' configuration. In other words, despite the different language-specific strategies, the defining properties of passive sentences are considered to be the same in each case. Turning back now to existential sentences all one has to do is adapt the questions we have just asked about passives, namely: what do we know when we know the meaning of an existential sentence? Is this knowledge encoded in the lexicon, say, in the lexical entry of there, or is it the output of a syntactic configuration? In this section it will be shown that existential meaning is exclusively the result of the syntactic structure. In other words, existential sentences will simply turn out to be taxonomic labels for an epiphenomenon, on a par with passives, interrogatives, relatives and the like. Let us start then from an apparently trivial observation: existential ^^resentences are copular sentences.5 More specifically, they are inverse copular sentences, i.e. the spec-IP is occupied by the predicative element raised from the small clause. There are at least three different reasons why it would be wrong to consider this a trivial observation. Even a rather cursory survey across
136
Are there parameters in semantics?
languages would immediately tell us that existential sentences are not copular sentences in all languages, including those languages which do not lack a copula. For example, in German we have es gibt (lit. 'it gives'), in Spanish hay (have-there), etc. In the last section of this chapter, this topic will be explicitly addressed. However, there is also a language-internal reason not to identify existential meaning with copular sentences. Along with £/*ere-sentences, there are other ways of asserting the existence of something in natural languages. The English lexicon, for example, contains the verb exist and the corresponding adjective existent. These can be used to construct a sentence that we would hardly claim was not a way of asserting the existence of something; take, for example: (5a) b
[DP many solutions] exist [DP many solutions] are existent
Henceforth, when I refer to 'existential meaning' I will be referring only to the interpretation of //iere-sentences. Moreover, we will not consider here issues related to the ontological implications of the notion of 'existence' (such as those concerning denotation of empty sets like unicorns) or to presupposition.6 A third reason why saying that existential sentences are copular sentences is not trivial is rather a theory-internal one. A r/iere-sentence is a copular sentence; specifically, it is an inverse copular sentence, i.e. the spec-IP is occupied by the predicative element raised from the small clause. However, it is clearly not sufficient for a given construction to be an inverse copular sentence in order for it to have an existential interpretation. Consider, for example the following two sentences: (6a)
the cause of the riot;
1°
j [a picture of the wall] tx
On the so-called 'existential meaning': the role of syntax
137
IP
there, I
tj [a picture of the wall] X{ Unless we want to deprive our terminology of any empirical force, we cannot possibly extend the class of existential sentences to include the first one as well as the f/iere-sentence. A further question is raised by pairs such as the following: (7a) b
there is a picture of the wall a picture of the wall is there
The latter sentence is associated with the following structure: (8)
IP
Within our framework, a sharp question immediately arises. If a theresentence is an inverse copular sentence, why has its canonical counterpart given in (8) such a different interpretation? Indeed it is clearly not the case that a picture of the wall is there is an existential sentence; rather, it has true locative meaning. Traditionally, the contrast under discussion has been dismissed by assuming that it is inherently related to the lexicon. Thus, two specific lexical entries are assumed: one for the existential there and one for a locative there.
138
Are there parameters in semantics?
Typically, the fact that these two different elements can cooccur - and that their phonological features are different - is assumed to make the issue very clear cut: (9)
there is a picture of the wall there
However, although the different phonological realization of the two theres does indeed seem to support the iexicalist' solution, comparative grammar shows that the problem is in fact quite independent from the issue regarding the two theres. We can anticipate the argument by reproducing the essential piece of data, taken from Italian. Consider the following pair: (10a) b
non ci sono molti numeri primi in questo insieme (there are not many primes in this set) non c'e Gianni in giardino (there is not Gianni in the garden)
Here, the intuitions are very clear: while the first sentence is an existential sentence, there can be little doubt that the second is truly locative. Indeed, it is simply not possible to trace these differences back to the supposed existence of two different c/s (unless, of course, we want to adopt an entirely ad hoc solution). Thus, if we adopt the 'Iexicalist' solution for the English case, the Italian cases would in any event have to be explained in some other way. In this chapter, I will try to suggest a unified theory that deals with the important problem of linguistic variation which has just been illustrated. 3.2.1 The logical form of an existential sentence The preliminary discussion in the previous sections has made one thing very clear: the level of representation we have chosen so far is not sufficient to distinguish between a there-sentence and any other inverse copular sentence, at least as far as existential meaning is concerned. Another significant fact that has emerged is that there-sentences are not just the only type of sentence that conveys true existential meaning but also the only type that shows a Definiteness Effect. This is clearly suggested by the following contrast: (lla) * there is John in the room b the cause of the riot is John We will see forthwith that the latter contrast is intimately connected with existential meaning. For the moment, it is sufficient to notice that any analysis linking the Definiteness Effect to the simple presence of the copula (for Case reasons, selection. . .) would not be able to derive this contrast.
On the so-called 'existential meaning': the role of syntax
139
Thus, we must enrich the structure by shifting to a more abstract level of representation: the specific property which characterizes there-sentences is thus to be sought at LF. Identifying it will be the goal of this section. As a preliminary step, recall the two main properties of //zere-sentences we discussed in section 2.4. On the one hand, the subject of a there-sentence cannot undergo A'-movement; on the other, an embedded constituent can be successfully extracted from it. Consider, for example, the following sentences involving the two basic types of A'-movement, namely whmovement ((12a-b)) and Q-raising ((12c-d)):7 (12a) * [DP which wall]; do you think [DP there]j was [ s c tj tj] (in the garden)? b [DP which wall]j do you think [DP therejj was [ s c [DP a picture of tj] tj] (in the garden)? c * [IP [DP many wallsjj [IP [DP there]j weren't [ s c tj tj] (in the garden)]] d [IP [DP many wallsjj [iP [Dp therejj weren't [Sc [DP pictures of tj] tj] ]] (in the garden)
In chapter 2, it was shown that, while the first property (which essentially follows from the ECP) is invariant for all inverse copular sentences, the second one (based on the Subjacency Condition) places //zere-sentences in sharp contrast with inverse sentences with a full predicative DP. Recall for example the following contrast: (13a) [which wall]j do you think therej was [sc [a picture of tj tj] (in the garden)? b * [which wall]j do you think [the cause of the riot] j was [Sc [a picture of tj] tj]? What this was held to indicate was that the potential barrierhood of the subject DP is neutralized in a there-sentence: the process of there-raising turns the copula into a lexical head, providing the structure with an Lmarker for the subject DP. If the raised predicate is a full DP this does not happen and the extraction yields a violation of the Subjacency Condition. Let us consider the effect of this difference at LF. 3.2.1.1 A principle of Logical Form The idea which will be implemented in this section is that at LF the link between predicates and their subjects is mediated by a syntactic operation. I will propose that syntax contains a principle requiring that a predicate should not apply directly to its subject but rather to a variable picked up within the range of entities denoted by the subject. Let us call this 'Principle Jt' and formulate it as follows: (14)
Principle Jt: 'predicates apply only to variables at LF'
Of course, there is no logical necessity here. As with all other empirical
140
Are there parameters in semantics?
principles, Principle jt is tenable only to the extent that it allows us to decompose complex empirical facts into simpler abstract entities. In this specific case, it will be shown that the interaction of this principle with independent assumptions allows us to formulate an explicit characterization of existential meaning and also to derive the Definiteness Effect. The obvious first step is to implement Principle jr. A variable is defined in syntax as a trace bound by a maximal projection from an A'-position; correspondingly, a maximal projection binding a trace from an A'-position is considered to play the role of an operator: (15)
XPi . . . tj
Traditionally, this configuration is associated with wh-movement and quantifier raising. Here we will go slightly further, assuming that, along with wh-phrases and quantifiers, subjects too are raised to an A'-position at LF from where they bind a variable, namely, the trace they have left. Let us illustrate the effect of Principle n on a very simple case: (16a) b
[s [DP John] [VP runs]] [s [DP Johnjj [s tj [VP runs]]]
Thus, in the sentence S, the VP predicate run is not applied to John but rather to a variable bound by John* This case is trivial: the application of Principle JT yields no relevant effect. So far, then, there is no need to complicate the structure. However, we may ask what happens if we apply this principle to inverse sentences. According to our analysis there are two subtypes of inverse copular sentences, namely those involving a full predicative DP and those involving there. Let us preliminarily consider the former case: (17)
(IP [tne cause of the riot]j is [Sc [a picture of the wall] tj]]
Potentially, there are two distinct ways of satisfying Principle Jt: the subject [a picture of the wall] must be moved to an A'-position in order to generate a variable; thus, in principle, it can be adjoined either to IP or to the small clause as in the following representation: (18a) b
[IP [DP a picture of the wall]j [lP [Dp the cause of the riot]j is [sc t; tj]]] [ip [DP the cause of the riot]j is [sc [Dp a picture of the wall]; [sc tj tj]]]
In fact, the first option ((18a)) is ruled out by locality conditions on movement. As we saw repeatedly in chapter 1, the subject of an inverse sentence cannot be extracted from IP. In order to satisfy the ECP, it should pass through spec-IP activating agreement on the copula and thus construing
On the so-called 'existential meaning': the role of syntax
141
an extended chain. This process is already exploited to license the trace of the raised predicate, which in fact occupies the spec-IP position. The second possibility, instead, is successful ((18b)): in the case of 'short movement', no locality condition is violated. Even the most restrictive versions of the ECP would allow it as an instance of antecedent government.9 This is sufficient to satisfy Principle jt: the trace of the subject counts as a variable and the predicate can be applied to it (via its trace). Let us turn now to the second class of inverse copular sentences, namely there-sentences: (19)
[IP [DP therejj are [sc [DP many girls] tj]] (in the garden)
Again, there are two potential ways of satisfying Principle JT: (20a) b
[IP [DP many girls]; [IP [DP there]j are [sc tj tj]]] (in the garden) [IP [there], are [sc [many girls], [sc tj tj]]] (in the garden)
The first possibility ((20a)) is ruled out for the same reason as in the case just discussed: the ECP blocks the extraction of the subject of a theresentence; once again, the raised predicate (i.e. there) exhausts the process of spec-head agreement with the copula which is necessary to satisfy the ECP. The remaining possibility is adjunction to the small clause, as represented in the second structure ((20b)). In this case, the difference between there-senXences and inverse copular sentences with a full DP becomes relevant. The crucial point here concerns adjunction. For independent reasons Chomsky (1986b) assumed that a maximal projection cannot be adjoined to the complement of a lexical head. The empirical reason for adopting such a hypothesis is hard to find. In the cited work, this hypothesis was the result of a highly complex system regarding adjunction to a 'barrier' as an escape hatch, in the technical sense. Simply allowing the element in question to adjoin to the complement of a lexical head would have given the wrong result. On the other hand, the same assumption has been carried over in the more recent frameworks proposed since Chomsky (1988a), where 'barriers' are not taken into account. Moreover, since adjunction is presently undergoing a radical rethinking, it is not clear yet what kind of empirical consequences this move has, nor is this the place to pursue such a complex matter in detail (see Cinque 1990a, Manzini 1992, Kayne 1994 and Chomsky 1994 among others). Nevertheless, it seems to me that there are indeed independent empirical facts that can be used to support the assumption that adjunction to the complement of a lexical head is not legitimate, at least at the LF interface. Consider, for example, the following well-known case study:
142 (21)
Are there parameters in semantics? [IP I consider [sc nobody a spy]]
Potentially, this sentence could generate two distinct meanings corresponding to the two possible scopal readings of the quantified subject nobody. (22a) b
[IP nobodyj [!P I consider [sc ix a spy]]] [IP I consider [sc nobodyj [Sc U a spy]]]
With wide scope as in the first structure ((22a)), the sentence should be paraphrased as 'nobody is such that I consider him a spy'. With narrow scope, as in the second one ((22b)), the sentence should rather be paraphrased as 'I consider that nobody is a spy'. In fact, only thefirstreading is available here. Although we are still far from a full understanding of such cases, we can at least register the fact that adjunction of nobody to the small clause is not possible. Technically, we can then assume that adjunction to the small clause is banned since the latter is the complement of a lexical head (consider). Bearing this in mind, let us go back to the there-sentence. We know on independent grounds that the copula behaves like a lexical head in theresentences, allowing extraction from within the subject DP embedded in the small clause. If we combine this independent fact with the assumption that adjunction to the complement of a lexical head is not possible, we must conclude that the second possibility of satisfying Principle JT given in (20b) is also to be ruled out. In short, then, the subject many girls can neither be an adjunct to IP for locality reasons, nor can it be an adjunct to the small clause, since this is the complement of a lexical head. Principle n seems to have a rather paradoxical effect here: there is apparently no way for a there-sentence to satisfy it. In the next section we will see that the solution to this apparent paradox becomes available if we formulate an explicit definition of existential meaning. 3.2.1.2 Existential meaning as DP splitting at LF The technical solution which is going to be proposed here is based on an informal intuition deriving from observing the following example: (23)
John hasn't met many girls
This sentence is ambiguous between two readings, which will be paraphrased as follows: (24a) b
many girls are such that John hasn't met them girls are such that John hasn't met many
On the so-called 'existential meaning': the role of syntax
143
Let us now make the difference between the two readings of (23) explicit by referring to the two paraphrases. In the first case ((24a)), although many girls have not been met by John he still could have met many of them (note that it is consistent to say many girls have been met by John and many girls have not been met by John). In the second one ((24b)), instead, the only possible situation is one where John has definitely met few girls. The intuition is quite clear: the two readings of (23) depend solely on the role that the element many plays in the sentence. Following a common practice (see the proposal in Higginbotham 1987 and references cited there), we can distinguish these two readings of many in (23) by calling the one associated to (24a) 'quantificationaP and the one associated to (24b) 'adjectival' and more generally say that many is a D° with an 'adjectival character'. 10 Of course, besides the intuitive force of this terminology, the specific property of a D° like many can be captured in an accurate formal way. For example, 'by adopting the standard algebraic picture in which restricted quantifiers over a domain D are interpreted as functions from ordered pairs of subsets of D to truth values, we can say that a quantifier q over D is of adjectival character if, for some functions f from the subset of D to truth values, q ( X , Y ) = f ( X n Y ) , for every pair X and Y of subsets of D' (Higginbotham 1987: 48). As for the structural position of many in the noun phrase, I will adopt the straightforward (perhaps simplistic) position that in a noun phrase there is only one head distinct from N°, namely D°. Thus, as a first approximation, many can always be regarded as a D°, although more subtle analyses including several different projections and empty D°s could be assumed (see Cinque 1992, Giusti 1992 and Longobardi 1991a among others): 11 (25)
DP
spec
many Bearing this in mind, let us turn now to f/iere-sentences and consider the following case: (26)
there aren't many girls
The intuition is quite clear: there is no ambiguity here. This sentence can
144
Are there parameters in semantics?
only mean that the set denoted by girls has the property of containing not many elements. For the sake of clarity, if we construct two sentences in (27a-b) on the model of those constructed in (24a-b) for the sentence in (23), John hasn't met many girls, we obtain the following result: (27a) girls are such that they aren't many b * many girls are such that there aren't them The first sentence is the proper paraphrase of the there-sentence under discussion. The second one, instead, is not only not a good paraphrase for it; it does not even make sense. In Higginbotham's terminology, then, we can straightforwardly capture this intuition by saying that many in the theresentence can only have the adjectival reading. This intuition will turn out to be important in our attempt to reach an explicit definition of existential meaning. The idea is that a sentence like there aren't many girls is interpreted as predicating the property denoted by many of the set denoted by girls (i.e. as 'girls are not many'). As a first approximation, let us define 'existential meaning' as follows:12 (28)
existential meaning is a function that maps a DP onto a predicative structure where the D°is the predicate of a set denoted by the NP
A rather surprising observation suggests itself here. The element there does not enter into this definition of existential meaning. Is this a welcome result? At this stage of the analysis it is hard to tell but as soon as cross-linguistic data are taken into consideration, it will turn out to be crucial (see section 3.4). All in all, notice that the meaning of a there-sentence is springing up from the DP: again, saying that there are not many girls is just equivalent to saying that girls are not many.13 The definition given for existential meaning, then, can be regarded as an explicit characterization of what we know when we know the meaning of a there-sentence. A there-sentence is nothing but a way to turn a DP into a sentence and, in fact, it is the most economical way. The problem of how to treat 'existential meaning' is now more perspicuous. In fact, it is reduced to indicating the syntactic conditions allowing such a result. An important preliminary observation must be spelt out here to avoid misunderstandings. Although it is true that the meaning of a there-sentence is springing up from the DP, by no means can we say that DPs can have a sentential interpretation per se or, equivalently, that DPs contain a predicative linking in the same sense as sentences. To prove this, it is sufficient to consider the following contrast:
On the so-called 'existential meaning': the role of syntax
145
(29a) * I think that [DP many girls] b I think that there are [DP many girls] If DPs could receive a sentential interpretation 'in isolation', then the first should be interpreted as 'I think that girls are many'. However, this is plainly false: DPs cannot be interpreted as sentences per se . Thus, a there-sentence can be regarded as the minimal syntactic environment allowing the sentential reading of a DP. Yet why do there-sentences force this? Principle Jt plays the crucial role here. Recall that this principle requires that predicates apply to the variable bound by the subject. Now, we independently know that there can be extraction from within the subject of a there-sentence as in which girls do you think there is a picture oft Thus, we can reasonably conclude that the variable required for predicative linking is provided in there-sentences by extracting the NP out of the DP as follows: (30)
[IP [NP girls]g [IP there; aren't [ sc [DP (tg) [D°many] tg] tj]]]
This representation now satisfies Principle JT and gives a formal equivalent of our definition of existential meaning as a last resort operation. A sentence needs a subject and a predicate and predicates apply only to variables: since the DP many girls cannot be moved as a whole, then the NP girls is extracted and many is applied to its trace tg, which counts as a variable. Let us focus on the role of the element there in this process of splitting. Indeed, its role would appear to be twofold. On the one hand, it excludes Q-raising, which would be necessary for the quantificational reading of many. In fact, by raising to spec-IP, there blocks the possibility for the trace of the subject DP to be licensed. On the other, it allows the process of NP extraction from the DP by turning the copula into a lexical head along the lines suggested in chapter 2. Since there does not add any 'lexical' content of its own, we might call it a 'propredicate'. 14 This offers us the possibility of remaining within the traditional terminology and considering there as an expletive, provided that the proper syntactic source of there is indicated. Specifically, there can be considered the expletive of the predicate of the small clause rather than the expletive of the subject of the predication. This, of course, corresponds to the proposal that has been made in the present work. Interestingly, the representation in (30) can be regarded as the LF counterpart of the following overt case of wh-extraction, which was discussed in the first chapter: (31)
[NP what]g do you think [DP there]; was [sc [DP (tg) [D«. e] tg] tj]?
146
Are there parameters in semantics?
Wh-movement is successful because it is a case of extraction from the DP subject of a there-sentence, not of the DP subject as a whole. In fact, extraction of a w/i/c/z-phrase would yield an ungrammatical sentence (see chapter 1, section 1.3.2). At this point, we can deal with the theory-internal problem raised by this analysis of there-sentences as inverse copular sentences. The fact that the canonical counterpart of a there-sentence has a locative interpretation now follows without it being necessary to hypothesize two different lexical entries for the locative and for the existential there: (32a) b
[,P [DP there]i are [sc [DP many [NP girls]] tj]] [IP [DP many [NP girls]]j are [sc U [DP there]]]
The process of DP splitting at LF which is possible (and indeed obligatory) in the first case is impossible in the second case, for obvious syntactic reasons. In fact, a well-established assumption says that there can be no extraction from within the preverbal subject position. This corresponds to a very clear intuition: the only way to interpret this sentence is to treat many as a quantifier and apply there to the trace of the quantified subject many girls. From the formal point of view, a sentence like many girls are there could not receive the following representation with the meaning 'girls are many and are there': (33)
* [IP [NP girlsjg [IP [DP many tj s are [sc t; there]]]
Thus, we do not need to assume two different lexical entries for there: there is no existential meaning in a sentence like many girls are there because there can simply be no DP splitting at LF. As we have already said, the fact that we are not obliged to assume two distinct lexical entries for there is a highly welcome result, since the contrast between locative and existential interpretation in Italian is yielded by using the same element, namely ci. This issue will be discussed in section 3.3. Summarizing, in this section an attempt has been made to give a formal characterization of the notion of 'existential meaning' as applied to copular there-sentences. It has been suggested that predicative linking is mediated by a syntactic operation requiring that predicates apply to a variable left by the subject (Principle Jt). The particular syntactic structure embodied by a there-sentence forces the subject DP to split at LF as a last-resort operation. In fact, since the subject DP cannot be moved, extraction of the embedded NP will be the only way to provide the structure with a variable to fulfil Principle jt. This formal characterization captures the intuition that underlies the pretheoretical notion of 'existential meaning'. An existential
On the so-called 'existential meaning': the role of syntax
147
sentence is nothing but the minimal clause structure that allows a sentential reading of a DP. What a speaker knows when he knows the meaning of an existential sentence is that in a sentence like there aren't many girls, the trace left by girls is interpreted as the subject of the residue of the DP, i.e. many. that is, the sentence there aren't many girls is to be interpreted as 'girls aren't many'. So far, we have concentrated on a relatively simple case. Clearly, an exhaustive theory of existential sentences must be able to account for a wider range of data. Although I do not in principle see any reason not to extend the theory to cover all possible cases, it remains an empirical question whether our theory of there-sentences as inverse copular sentences can actually account for all other cases and this aim clearly cannot be undertaken in the context of the present work. Nevertheless, I would like to indicate the line of reasoning that this theory might lead one to take. Consider, for example, the following crucial cases: (34a) b c
there are [DP [Do e] [NP girls]] there is [DP [Do a] [NP girl]] there are [DP no girls]
Can the D°s occurring in these three sentences be interpreted as predicates as required in our framework? Surely, this cannot be answered in a direct way; that is, we cannot construct corresponding copular sentences of the kind girls are D° as we did in the case of many. I think that we should distinguish two different aspects of this empirical problem. On the one hand, we must ask whether these D°s can be regarded as predicates from 'a logical point of view', that is, by referring to the semantic apparatus we have chosen (based on Higginbotham's 1987 approach). On the other hand, assuming that all types of D°s including those in (34) can indeed be regarded as predicates, we have to explain why they never occur overtly in copular sentences of the kind NP-copula - D°as opposed to a D°like many. These two different aspects must be kept in mind while briefly reviewing the three cases under discussion. Empty D°s occurring with plural NPs like the one in the first sentence ((34a)), for example, have been shown to count as existential predicates (roughly as null counterparts of some; see Diesing 1990, Longobardi 1991a and references cited there). Yet why is it impossible to have * girls are? Crucially, it has also been shown that empty D°smust meet specific syntactic requirements that can vary in a complex way across languages. For example, it has been argued that in English they can occur only if the DP they project is governed by a lexical head at some point in the derivation (see again
148
Are there parameters in semantics?
Longobardi 1991a). Now, the proper syntactic requirement is provided in (34a) by there-raising, as I have independently shown, but not in bare copular constructions (see section 2.3.2). Thus, we can straightforwardly assume that at LF the NP girls is extracted from the DP projected by an empty D° (/D° e]) paralleling the case of manygirls and that the existential predicate [D° e] is applied to it without further stipulations. The case with the indefinite singular a which we have in the second example ((34b)) parallels the case in (34a): following again Higginbotham (1987), we could consider a as a dummy element, which simply spells out the empty determiner for countable names (cf. there is water). If we abstract from number features, then, we can regard this case as analogous to the first one.15 The last sentence ((34c)) appears to require a more articulated analysis. The 'logical problem' could be quite straightforwardly solved by adopting Higginbotham's (1987) approach. In such a case, no is regarded as a member of the class of D°s with an 'adjectival character', so it is a proper candidate for //iere-sentences. Yet a residual question would nevertheless remain to be answered, i.e. why is it not permissible to have sentences like * girls are nolxty Assuming that the theory presented here is to be taken seriously, one possible account would be to analyse no as a complex form incorporating negation with an existential predicate. A sentence like there are no girls, then, would be the counterpart of there aren't (any) girls, where the negation appears syncretically on a null D° which is interpreted as an existential predicate. If this proves tenable, then predication could be assumed to occur between girls and the D° and negation would then apply to this link. The fact that a sentence like *girls are no is ungrammatical would ultimately be due to the fact that a null D° cannot be licensed in such a syntactic context, paralleling the line of reasoning we applied to the other cases discussed above.17 Again, what has been proposed here cannot be an exhaustive theory of the semantics of there-sentences', this is inevitable, given that no exhaustive theory of the semantics of noun phrases as yet exists. The limited aim of this work is to show that the new analysis suggested for there-sentences as inverse copular sentences opens up an alternative way of looking at some phenomena involved. Whether or not the analysis suggested here can derive the entire range of phenomena is an empirical matter that goes beyond the limits of the present work.18 In what follows I will nevertheless attempt to push this theory to the limit by examining the issue raised by the Definiteness Effect.
On the so-called 'existential meaning': the role of syntax
149
3.2.2 On deriving the Definiteness Effect Given the ideas that have been elaborated at length, we are now in a position to deal with the Definiteness Effect in a rather natural way. Indeed, it will be seen to follow straightforwardly as a consequence of independent principles. The basic consequence of our approach is that the class of D°s that can occur in there-sentences is expected to coincide with the class of D°s that can occur as predicates in copular sentences (provided that the line of reasoning indicated at the end of the previous section turns out to be correct). Referring again to Higginbotham (1987), we can simply say that the class of D°s that can occur in there-sentences coincides with the class of D°s with an adjectival character in the intended formal sense.19 For the sake of simplicity, let us consider the following pair of contrasting examples: (35a) there are many/few/three girls b * there are every/the/most girls The fact that this pair matches the following one not involving there no longer needs to be regarded as accidental: (36a) girls are many/few/three b * girls are every/the/most It follows from our theory that there should be a close link between the two sets of data. In a there-sentence the NP embedded in the postverbal DP is extracted and interpreted as the subject of the residue of the DP itself. Since the, every and most cannot play the role of predicates they cannot occur either in f/iere-sentences ((35b)) or in bare copular sentences ((36b)).2() Crucially, this result can be obtained only by assuming that theresentences are inverse copular sentences, that is, only if we abandon the analysis based on the replacement of the subject expletive there by the subject DP in favour of the analysis which treats this element as a raised predicate involving DP-splitting at LF. For the sake of clarity, we can compare this theory with the traditional one which regards there as the expletive for the subject of the predication. Let us consider the following contrast: (37a) there aren't many girls (in the room) b * there aren't the girls (in the room) If we follow the expletive replacement analysis, these two sentences are to be represented as in (38) where the entire DP has been raised: (38a) b
[IP [DP there [DP many girlsjj] aren't [Sc tj in the room]] [ip [DP there [Dp the girls];] aren't [Sc tj in the room]]
150
Are there parameters in semantics?
On the other hand, if we adopt the DP-splitting analysis, the same sentences receive the following representation: (39a) b
[IP [NP girls]g [IP [DP there]; aren't [sc [DP many tg] tj]]] (in the room) [IP [NP girlsjg [IP [DP there]; aren't [sc [DP the tg] tj]]] (in the room)
Choosing one theory rather than the other will naturally lead us to have very different a priori expectations. On the one hand, if the subject DP is to be raised at LF to affix there, then an ad hoc semantic restriction on the trace of the DP is really needed. Nothing would in fact prevent a preverbal subject from being definite as in: (40)
[DP the girlsjj are [sc tj in the room]
On the other hand, if the DP is split at LF forcing D° to be interpreted as a predicate, then independent knowledge relating to the contrast between, say, girls are many and * girls are the will be sufficient to rule out the ungrammatical cases. Furthermore, if we consider the issue for a moment from the point of view of language acquisition, we will see that the consequences of the analysis being proposed are far from trivial. In fact, the relevant piece of information the child needs to know about the Definiteness Effect is already independently available from bare copular sentences (cf. (35) and (36)): more explicitly, there is no independent semantic restriction that needs to be acquired in order for definite DPs to be banned from £/iere-sentences. To sum up, then, if an existential sentence is analysed as the minimal clause structure that forces the sentential reading of a DP, the Definiteness Effect will follow from independent facts. Not all DPs can occur as the subject of ^ere-sentences because not all D°s can be interpreted as predicates.21 Bearing this in mind, we can now turn to the main problem of this chapter: how can we account for the lack of the Definiteness Effect in Italian? 3.3
How to escape the Definiteness Effect: the role of the pro-drop parameter
We now know that the Definiteness Effect in English is not due to an independent semantic restriction: the speaker knows that certain D°s cannot be used in there-sentences because he knows that certain D°s cannot play the role of predicates. Nevertheless, the contrast among D°s in Italian coincides with the English one given earlier (cf. (36)):
How to escape the Definiteness Effect
151
(41a)
le ragazze sono molte/poche/tre (the girls are many/few/three) b * le ragazze sono ogni/le/la maggior parte (the girls are every/the/most)
Thus, the cross-linguistic variation under discussion must be traced back to some other property which distinguishes the two languages. At this point we can take advantage of an apparently marginal observation we made at the end of chapter 2 concerning a difference between there and ci (see section 2.4). These two elements differ structurally, because they end up occupying two different positions: there is in spec-IP while ci is a clitic. This simple fact will allow us to derive a full explanation of the cross-linguistic variation. 3.3.1 Embedded vs. rightward-moved subject in d-sentences Italian ci and English there have different X'-statuses: the first one is a head and the second one is a maximal projection. Let us represent them separately. On the one hand, the Italian sentence will be associated to the following structure:22 (42)
una foto del muro
On the other hand, the English sentence receives the following representation:
152 (43)
Are there parameters in semantics? IP
there, 1°
a picture of the wall This has an important consequence for the syntax of the two sentences. Unlike what happens in English, in a d-sentence the subject can either be left in situ within the small clause or raised to a preverbal position. These two options are represented in the following paradigm: (44a) b
Gianni dice che [!P pro [Cj'e] [Sc una foto del muro tj]] (Gianni says that there is a picture of the wall) Gianni dice che [IP [DP una foto del muro]; [Cj'e] [ sc U tj]] (Gianni says that a picture of the wall there is)
In English, it is not possible to have a corresponding pair, for there is not enough room for both there and a picture of the wall to be raised to preverbal position: (45a) John says that [iP [DP therejj is [ sc a picture of the wall tj]] b * John says that [iP [DP a picture of the wall]i [DP therejj is [Sc tj tj]] At the end of chapter 2, we also saw that this has an important consequence in differentiating the two languages: the subject of an Italian d-sentence can be wh-moved whereas this is not possible in English: (46a) * [DP which picture]} do you say that [!P [DP therejj is [ sc tj tj]]? b [DP quale foto]j dici che [,P proj [VP [c'j e] [ sc tj tj]]]? (which picture do you say that there is) It will be recalled that in essence the explanation given for this contrast did not differ from the traditional analysis of the apparent violation of the thattrace effect in Italian as explored by Rizzi (1982) (see section 1.1). In more traditional terms, one could regard the possibility of extracting the subject of a d-sentence as a case of rightward movement, which typically comes
How to escape the Definiteness Effect
153
into force in pro-drop languages in order to allow a preverbal subject to be extracted across an overt complementizer. 23 From this point of view, the structure of a sentence like quale foto did che c'e ((46b), repeated in the following pair as (47a)), would be entirely parallel to well-studied cases like quale ragazza credi che telefoni in (47b): (47a) b
[DP quale foto]i dici che [IP pro* [VP [Vp [c'j e] [ sc U tj]] t*]]? (which picture do you say that there is) [DP quale ragazzajj credi che pro [Vp [VP telefoni] tj? (which girl do you think that telephones)
Bearing this in mind, let us turn now to the crucial contrast under discussion in this chapter which is reproduced here: (48a) c'e Gianni (in questo giardino) b * there is John (in this garden) This contrast now follows without further stipulations. In English, the presence of a definite DP rules out the sentence because a definite DP does not allow the process of DP-splitting. Nevertheless, the same process is also banned in Italian, since there is no way even in a language of this type for the D° occurring in a phrase headed by a proper name to be interpreted adjectivally.24 The clitic nature of ci becomes crucial, now, for it creates a sort of escape hatch in Italian. Let us follow the derivation of a sentence like c'e Gianni step by step. The subject Gianni is generated in the small clause together with the propredicative ci and the small clause itself is the complement of the inflected copula (V° movement is omitted): (49)
[jp [DP e] [P e] [ sc Gianni [ci]]]
Subsequently, ci is cliticized onto the copula, while Gianni, which does not allow any splitting, is raised to spec-IP and then adjoined to IP at LF to satisfy Principle Jt (with no relevant effects, in this case): (50a) b
[IP Giannii [P Cj'e] [ sc U tj]] [IP Giannii [IP t{ [P Cj'e] [ sc U tj]]]
We can now make use of a well-established fact concerning pro-drop languages, namely, that subjects can alternate freely between pre verbal and postverbal position, as exemplified by such frequently studied cases as the following: (51a) b
Gianni telefona (Gianni telephones) pro telefona Gianni (pro telephones Gianni)
154
Are there parameters in semantics?
Thus, the idea to be implemented is that Gianni in c'e Gianni is not in situ within the small clause. Rather, it is in a dislocated postverbal position, paralleling the case of telefona Gianni. As a first approximation, the sentence c'e Gianni in giardino will receive the following representation (with in giardino as an adjunct itself): (52)
[IP [IP proj [,o cj'e] [sc t\ tj]] Giannij ] (in giardino)
Although the approach based on the adjunct position might be refined, it is worth emphasizing the underlying intuition it aims to capture. In a cisentence the postverbal subject does not necessarily have to be thought of as being in situ in the small clause; alternatively, it can be raised to a higher position outside the small clause constituent. In such a case, the fact that existential meaning is not available becomes trivial: simply, there is no DP splitting from a dislocated position. Instead, the propredicate ci is applied to the trace of DP subject (Gianni) yielding a locative interpretation: 'Gianni is there'. Notice that there is no need to assume two different cis. The locative and the existential meanings of a c/-sentence depend solely on the type of DP involved. In other words, the contrast between the existential interpretation of non ci sono molte ragazze (not there are many girls; meaning: girls are not many) and the locative one of c'e Gianni in giardino (there is Gianni in the garden; meaning: Gianni is in the garden) doesn't require a hypothesis assuming the existence of two cis. It seems likely that independent pieces of data can be offered to support the idea that Gianni has been adjoined to a higher postverbal position. Consider the following contrast: (53a) b
Gianni (*? c')e in questo giardino (Gianni there-is in this garden) *(c')e Gianni in questo giardino (there-is Gianni in this garden)
This fact can be explained as follows. Since the first sentence has a locative interpretation, due to the fact that ci is applied to the variable left by the subject, the addition of a PP creates a clitic doubling construction, as if the small clause contained two predicates, namely ci and the PP. Thus the sentence becomes awkward, since in Italian clitic doubling is generally avoided, as in the following case: (54)
Gianni (*? lo) legge il libro (Gianni lo reads the book)
As for the second sentence, on the other hand, the rightward-moved position of the subject forces the PP to be interpreted as an adjunct, rather
How to escape the Definiteness Effect
155
than as the predicate of the small clause, and the clitic doubling effect disappears. In fact, ci is now obligatory, since otherwise the sentence would lack a predicate. As confirmation of this idea, notice that if one inserts a negative adverb like mai (never) between the copula and the PP the sentence becomes fully acceptable, as opposed to (53a): (55)
Gianni non c'e mai in questo giardino (Gianni not there-is never in this garden) 'Gianni is never in this garden'
This can be explained by noting that the adverb mai forces the adjunct reading of the PP, thus avoiding any possible redundancy with respect to ci. In fact, mai cannot occur in a small clause; witness a contrast like the following: (56a)
Maria non ritiene mai [Sc Gianni il colpevole] (Maria not believes never Gianni the culprit) b *? Maria non ritiene [Sc Gianni mai il colpevole] (Maria not believes Gianni never the culprit)
In English, on the other hand, if the DP does not allow the DP splitting process at LF, there is no way to interpret the //iere-sentence, since there is no alternative escape hatch available as in Italian. This is what the Definiteness Effect turned out to be: the lack of a predicate in a sentence due to the impossibility of assigning a DP a sentential interpretation. The difference compared to the standard theory is at this point very clear. The Definiteness Effect was regarded as the violation of a specific semantic restriction banning a definite DP from occurring as the subject of a small clause if an expletive is inserted. To complete the picture, notice that the equivalent of ce Gianni in English would be something like the following: (57)
[IP Johnj is [sc U there]]
Here, given the premises concerning existential meaning, the only possible interpretation is a locative one. All in all, we can conclude this section by noting that, at least for the specific issue concerning cross-linguistic variation with respect to the Definiteness Effect, there is no need to assume that what we are dealing with is semantically based parametric variation, in other words, a sort of a 'semantic parameter'. Both in English and Italian, we find that the same class of DPs is incompatible with the existential meaning. The crucial difference is that Italian provides an escape hatch for the 'offending' class of DPs. This escape hatch is the result of a combination of two independent
156
Are there parameters in semantics?
properties: the fact that ci is a clitic25 and the pro-drop parameter.26 Thus, the Definiteness Effect turns out to be a mere label for the intersection of different independent conditions. Crucially, any theory that assumes the expletive replacement hypothesis could not treat such cross-linguistic variations involving the Definiteness Effect in a comparable way. Assuming that the DP subject is moved as a whole to preverbal position at LF in both languages would obscure the syntactic differences that in fact account for the cross-linguistic variation. As was said at the very start of this chapter, any analysis of existential sentences in natural languages should in principle be able to give us an abstract template which will allow us to account for the different ways in which each language instantiates this type of sentence. The theory proposed so far seems promising from this point of view. Existential sentences, and the corresponding conceptual correlate, i.e. existential meaning, have been broken down and reanalysed as the interaction of various different factors. Before trying to explore each of these factors and see whether any light can be thrown on cross-linguistic differences, I would like to make a brief departure from our main theme by exemplifying one application of the theory that has been proposed. 3.3.2 A note on partitive phrases and existential meaning This section is of a perhaps rather digressive nature. We will support the analysis of d-sentences as inverse copular sentences by applying it to a rather challenging contrast. Consider the following pair: (58a) b
non c'erano [DP molte ragazze] (not there-were many girls) non c'erano [DP molte delle ragazze] (not there-were many of the girls)
The interesting fact here is that the quantifier molte (many) cannot have narrow scope in the second sentence, as opposed to the first one, despite the fact that they have the same linear order. This can easily be tested in the following way. The narrow scope reading of a quantifier like mold within negation is preserved when replacing the pair non . . . mold (not . . . many) by pochi (few) as in the following case: (59a) b
Gianni non ha letto molti libri (Gianni not has read many books) Gianni ha letto pochi libri (Gianni has read few books)
How to escape the Definiteness Effect
157
Let us now apply this to the challenging contrast under discussion: (60a) b
c'erano [poche ragazze] (there-were few girls) c'erano [poche delle ragazze] (there-were few of the girls)
The result is quite clear: the first sentence ((60a)) has the same meaning as non c'erano molte ragazze ((58a)); the second one ((60b)), on the other hand, fails to have the same meaning as non c'erano molte delle ragazze ((58b)). Why should this be so? As a source of a possible answer to this question, the analysis of d-sentences as inverse copular sentences seems much more promising than the standard analysis based on expletive replacement. Let us first focus on the two DPs contained in the crucial pair. I will assume that the two DPs have different structures (see Moro, in press, and references cited there): (61a) b
[Dp molte [NP ragazze]] [DP molte [NP [N° e] [delle ragazze]]]
The phrase delle ragazze is not the complement of the D°: it should rather be regarded as an adjunct, playing the role of 'partitive restriction'. Consequently, the NP is instead headed by an empty category, i.e. [^ e]. This structure can be independently supported by considering the following paradigm: (62a) b c
[DP qualche [NP ragazza]] (some girl) * [DP qualche [NP delle ragazze]] (some of the girls) [DP qualc- [NP [una] [delle ragazze]]] (some-one of the girls)
In this case the complex structure is overt. The empty category is lexicalized by the pronominal form una (one-fem.sing.). 27 It is assumed that in molte delle ragazze the empty category is licensed by agreement with the quantifier. Let us now proceed a little further, by considering an important property of the partitive restriction. A constituent of this sort cannot be moved. A simple contrast of topicalization like the following illustrates the point: (63a) b
[DP molte [NP [e] delle ragazze]]i, Gianni vide tj (many of the girls, Gianni saw) * [NP [e] delle ragazze]g Gianni vide [Dp molte tg] (of the girls, Gianni saw many)
An explanation for this contrast is immediately available, once we adopt
158 Are there parameters in semantics? the structure containing an empty category. Crucially, an empty category cannot be licensed in such a configuration and thus yields an ECP-like violation.28 Let us turn now to the central contrast under discussion. What is the LF structure of the two sentences introduced at the beginning of the section? The minimal assumption is that DP splitting applies to them: (64a) b
[IP [NP ragazze]g [IP pro non [cj'erano] [sc [DP molte tg] tj]]] [IP [NP [e] [delle ragazze]]g [IP pro non [Cj'erano] [sc [DP molte tg] tj]]]
Consider the two structures separately. The first one ((64a)) is well formed: its existential interpretation (i.e. 'girls are not many') is accounted for along the lines indicated in the previous section. The second one ((64b)), on the other hand, contains an ECP-like violation: the empty category involved in the partitive restriction cannot be licensed in such a position. Since DP splitting cannot take place, there is only one possibility left here: that the entire DP molte delle ragazze has been rightward-moved to a higher position, paralleling the case of c'e Gianni. The two sentences under discussion here ((58a-b)), then, turn out to have quite different structures: (65a) b
[IP pro non [Cj'erano] [Sc [DP molte [NP ragazze]] tj] [iP [iP proj non [q'erano] [sc tj t,]] [DP molte [NP [e] delle ragazze]]j]
The asymmetric position of the two DPs can now be seen to explain the contrast we have focussed on in this section: molte cannot have narrow scope in the second case, as opposed to the first one, because it is adjoined to a position higher than the negation. In fact, the meaning of this sentence is locative rather than existential (i.e. 'many of the girls weren't there'). In the other case, in contrast, DP splitting at LF can occur giving the existential meaning as an output: crucially, in this case the quantifier remains in situ, within the scope of negation.29 As a corollary, notice that this analysis predicts that, in English, partitive phrases yield a Definiteness Effect violation in f/iere-sentences:30 (66a) there weren't many girls in the garden b * there weren't many of the girls in the garden In fact, the 'escape hatch' for those DPs that do not allow DP splitting at LF is not available in a non-pro-drop language. The DP can't be moved to a rightward position and the sentence is ruled out. This brief section has attempted to illustrate a fruitful application of the theory of existential meaning as DP splitting at LF. Without any doubt, a theory which crucially and solely relies on the type of D° occurring in the
A cursory view across languages: 'Jespersen's generalization'
159
subject of an existential sentence would not be able to derive this contrast, given that the D° is exactly the same in both cases - here molte (many). Let us now turn to a major question. How can the theory proposed here shed light on the general issue addressed in this chapter: what are the limits of variation between existential sentences allowed by Universal Grammar? 3.4
A cursory view across languages: 'Jespersen's generalization'
On a par with 'passive sentences' or interrogative sentences', 'existential sentences' have turned out to be just a taxonomic epiphenomenon. It has been shown that these constructions can be characterized as the interaction of independent factors which we might call the 'defining properties' of existential sentences. For ease of exposition, the basic interacting factors that yield an existential sentence are summarized here: (67)
The defining properties of existential sentences: (i) the subject cannot be moved as a whole (ii) extraction is possible from within the subject (iii) D° has an adjectival character
Each property has an independent life.31 We will now examine a few examples where these properties are instantiated individually. In English, for example, the preverbal subject of an embedded clause displays the first property when a wh-phrase occupies the adjacent spec-CP position: (68)
* Which student; do you wonder [CP which bookj C° I[P tj read tj]]?
Turning now to the second property, in Italian extraction is possible from the subject position of a small clause governed by a lexical head like ritenere (believe): (69)
Gianni nej ritiene [Sc [DP molti tj] intelligenti] (Gianni of-them believes many intelligent)
The third property, instead, is manifested in constructions like the following: (70)
John hasn't met [DP many [NP girls]]
where the adjectival character of the D° (in the sense of Higginbotham (1987); see section 3.2.1.2) can be captured by means of the following paraphrases: 'John has met few girls' and 'many girls are such that John hasn't met them'. Nevertheless, it is only when the defining properties in (67) are simultaneously present in a given structure that the output qualifies as an
160
Are there parameters in semantics?
'existential sentence'. In such a case, they interact with each other yielding the only possible structure where the D° is predicated of the (trace left by the) NP. For the sake of clarity, let us just repeat the standard case. A sentence like there aren't many girls (in the room) would ultimately be represented at the LF interface as follows: (71)
[IP [NP girls]g [IP [DP there]; aren't [sc [DP many tg] tj]]] (in the room)
The predicate many is applied to the trace left by the subject girls satisfying Principle n and the sentence is interpreted as 'girls are not many (in the room)'. It is easy to notice that this structure requires all the three properties to be met: if one of them were missing, the structure would be ruled out. It is perhaps worth pointing out once again that an existential sentence is nothing but the minimal structure satisfying these three properties. In other words, it is the minimal syntactic device that splits a DP in such a way that the D° is predicated of the NP. So far we have proposed an analysis for English and Italian. Even the most superficial analysis would have to take note of the fact that these two languages both have a locative expression (ci and there) in existential sentences. The question naturally arises of whether a similar locative expression is a necessary feature of existential sentences in all languages.32 A cursory survey shows that this is not the case: (72a) b c d e
there is (English), c'e (Italian) ily a (French), hay (Spanish) es gibt (German), si da (literary Italian) you (Chinese), echei (exei; modern Greek) eshte (Albanian), est (Romanian)
Indeed, even if we limit our aims to a descriptive level of adequacy, it nevertheless seems to be difficult to formulate a generalization. In fact, the partitions are substantially arbitrary and the examples can be grouped in several different ways. Let us start with the first two groups. Along with English and Italian a locative expression is also used in French and Spanish {there, ci, y, -y, respectively). Yet the parallelism between the four languages is far from complete. A first distinction is to be made between them once we look at the verb occurring with the locative element: we have be and essere in English and Italian respectively whereas we have avoir and haber in French and Spanish. A further difference which sets English apart from French, Italian and Spanish emerges when we look closely at the X-bar theoretical status of the locative elements involved. Italian ci, French y and the Spanish morpheme -y do not occupy full XP positions but rather clitic positions, and
A cursory view across languages: 'Jespersen's generalization
161
this distinguishes them from English there. Moreover, agreement phenomena would appear to be the basis for a further division which groups Italian and English (where the verb agrees with the subject)33 separately from French and Spanish (where the verb is not sensitive to subject features). Interestingly, the pro-drop parameter seems to produce no interesting divisions, at least as far as these data are concerned.34 If, on the other hand, we look at languages that construct existential sentences without a locative expression, as in the last three groups, the situation is equally complicated. In German and literary Italian, the verb used to form an existential sentence is a verb meaning 'give': geben and dare, respectively. It seems particularly significant that the subject clitic si must be present in Italian. In principle, since this is & pro-drop language, there is no need for the subject position to be overtly realized. However, if si is missing the corresponding sentence could not be read with an existential meaning. Thus, sentences like si danno due soluzioni and danno due soluzioni would be totally different, for only the first one qualifies as an existential sentence, while the second would just mean 'they give two solutions'. Again, the agreement pattern would also allow a further division to be made here: in Italian, the verb must agree with the subject, while in German it cannot.35 In Chinese and modern Greek, on the other hand (for Chinese see Huang 1987), the verb employed means 'have': you and echei (e/et), respectively. No locative elements are involved at all. Finally, in Albanian and Romanian existential sentences seem to be constructed with the 'bare copula'.36 A fully detailed treatment of these cases is obviously beyond the scope of this book. In fact, the empirical data could turn out to be even more complicated.37 Nevertheless, I would like to address here the potential lines of research that our theory of existential sentences would lead us to pursue. A major question which naturally arises from an exploration of existential sentences such as the one undertaken here is: what is the range of variability of such constructions across languages? Do these structures have anything in common? Seen in terms of the lexical items involved the situation appears to be anything but unitary. If, however, we concentrate on syntax, then a number of underlying common features come to light. A generalization concerning existential sentences was formulated by Jespersen (1924); I will refer to this as 'Jespersen's generalization': (73)
Jespersen's generalization: whether or not a word like there is used to introduce [existential sentences] the verb precedes the subject and the latter is hardly treated grammatically like a real subject. (Jespersen 1924: 155)
162
Are there parameters in semantics?
The content of this is quite clear: a comparative analysis shows that it is not the locative expression, or any other specific lexical item, that characterizes existential sentences but rather their peculiar syntactic structure. Two factors are singled out by Jespersen as crucial: the inverse order of the subject and verb and the anomalous behaviour of the subject. 38 Can this generalization be implemented in a modern framework? I think it would not be exaggerated to claim that the present theory actually comes very close to implementing it. The defining properties of existential sentences as listed above turn out to constitute a formal version of Jespersen's generalization. On the one hand, the fact that 'the verb precedes the subject' is implicit in our second property: extraction from the subject is possible. If the subject were in a preverbal position there could be no extraction of the NP from the DP. On the other hand, the intuition that the subject is hardly treated as a real subject' correlates with the third property: at the relevant level of interpretation (LF) the subject is split and a sentential interpretation of the DP is produced.39 This generalization (and the corresponding formal implementation) is able to capture the data reported here. Cross-linguistically, we might, as a first approximation, regard existential sentences as an instance of the following abstract schema: (74)
S
there; is pro c/e il Yi a es gibt i sijda pro pro; you eshte Proi pro; est
As we can see here, different languages may use different strategies for obtaining the proper configuration. In English there is employed with be, whereas in French we have expletive //, propredicative element y and avoir. Moreover, pro-drop languages do not seem to behave uniformly: on the one
Conclusions 163 hand, we have Italian, where an overt predicative element is employed, namely ci; on the other, we have Albanian and Romanian where a null propredicative element occurs, i.e. pro.40 In conclusion it seems not unreasonable to claim that Jespersen's generalization has been fully subsumed by our framework. All in all, as we have seen, we are still far from having a comprehensive and complete account of the syntax of existential sentences. Many major questions are still to be answered especially as regards the factors behind the distribution of be and have and their equivalents across languages. Nevertheless, the results obtained here would appear to constitute a significant preliminary step towards a theory of existential sentences: this particular type of structure has been broken down into its constituent elements - a number of interacting, independently given abstract factors. Thus, the analysis of existential sentences turns out to be entirely parallel to the analysis of other types of constructions such as, say, passives: no specific rules, nor specific lexical elements, are necessary to account for them. It is perhaps worth underlining that it only became possible to obtain this result when the theory of there as an expletive standing for the subject of the predication was abandoned in favour of the analysis in which existential sentences are seen as instances of the more general class of inverse copular sentences, whose preverbal position is occupied by a raised predicative element and whose subject remains in situ in the lower clausal constituent. Had we instead continued to regard there as an expletive standing for the subject of the predication and thus assumed that it is replaced at LF, we would simply have succeeded in rendering the data more opaque than they already are: crucially, we would have been saddled with the assumption that //iere-sentences are canonical subject-predicate constructions. 3.5
Conclusions
In this chapter, we have first addressed a problem of comparative grammar. The following contrast typically exemplifies a cross-linguistic variation: (75a) * there is [DP John] [PP in this garden] b c'e [DP Gianni] [PP in questo giardino] Thus, definite noun phrases cannot occur in the subject position of existential sentences in English, whereas this is perfectly normal in Italian. We began by considering the problem of what in principle can account for the absence of the Definiteness Effect in Italian existential sentences. We
164
Are there parameters in semantics?
examined and finally rejected the possibility of saying simply that there is a parametric variation with respect to the semantic restriction that has generally been assumed, since Chomsky (1973), to explain (75a). Our approach to the problem has been divided into two major steps: as a preliminary, we tried to formulate an explicit theory of existential meaning; then, we showed that the Definiteness Effect follows from this theory and thus reduced the apparent violation present in Italian to an independent syntactic parameter. I will summarize these two steps separately. We started by assuming that English //iere-sentences are inverse copular sentences. The argument was then guided by the fundamental intuition that a sentence like there aren't many girls in the room is interpreted as 'girls are not many in the room' rather than 'many girls aren't in the room'. Following Higginbotham's (1987) terminology, we concluded that in a //iere-sentence only the adjectival reading is possible for an item like many, as opposed to a quantificational reading. Technically, we implemented this intuition by means of a principle similar to ^-abstraction, called Principle jt, according to which at the LF interface predicates apply only to variables. Thus, after spell out the computational system will proceed further and reach the more complex structure where the NP is extracted from the DP: (76a) b
[IP [DP therejj aren't [sc [DP many [NP girls]] tj]] (in the room) [IP [Np girls]g [,P [DP there]j aren't [sc [DP many tg] tj]]] (in the room)
After DP-splitting the only surviving predicate in the small clause is many; this can now be applied to the variable, yielding the proper interpretation, i.e. 'girls are not many' (in the room). Thus, the so-called existential meaning turned out to be a mere epiphenomenon. In other words, theresentences could be treated in the same way as passives, interrogatives etc., as the output of a modular interaction. In particular, it turned out that an existential sentence is simply the minimal clause structure which forces the sentential interpretation of a DP. This led us to a first important conclusion: the Definiteness Effect is not the consequence of a specific semantic restriction; rather, it is the consequence of the independent fact that not all DP internal items are like many in having an adjectival character. Thus, the Definiteness Effect follows quite naturally: a 'definite subject', i.e. a subject containing a definite article, a universal quantifier, etc., cannot be the subject of a theresentence because a definite article, a universal quantifier, etc., cannot be interpreted adjectivally. In other words, if a definite DP occurs, the sentence would lack a predicate. Within the traditional analysis which regarded there as an expletive which stands for the subject of the predication, a conclusion
Conclusions
165
of this sort would not have been very helpful, because the role of the predicate was considered to be played by the PP in the room. In the present framework, where the PP is considered an adjunct, the absence of an adjectival item would amount to a violation of a general principle requiring sentences to have a subject and a predicate. Given this new perspective, it became possible to achieve the second objective. The absence of the Definiteness Effect in Italian could be traced back to a syntactic parameter, namely the/?ro-drop parameter. The fact that d, unlike there, is a clitic played the crucial role. The argument was based on two independent considerations. First, since ci is a clitic, the subject can be raised to occupy spec-IP position: (77)
Maria dice che [IP Giannij [ r Cj'e] [Sc tj tj]] (Maria says that Gianni there is)
Second, in pro-drop languages the subject can freely alternate between preverbal and postverbal adjunct position as in the classic example: (78a) b
Gianni telefona (Gianni telephones) pro telefona Gianni (pro telephones Gianni)
Combining the two observations we concluded that a sentence like c'e Gianni in questo giardino can be assigned the following representation: (79)
[IP [IP proj [jo Cj'e] [Sc tj tj]] Giannij ] (in questo giardino)
The subject Gianni has been moved from a pre verbal position to a postverbal adjunct position as in telefona Gianni. Crucially, there can be no DP-splitting process here. Thus, the sentence is simply interpreted locatively. In fact, the proper English translation of such an Italian sentence is something like John is thereIhere. Finally, this theory opened up the way for a wider-ranging analysis which aims at accounting for cross-linguistic differences in the field of existential sentences. Jespersen's (1924) generalization has guided the analysis and proved to be essentially correct. The defining properties that qualify a sentence as an existential sentence turned out to be essentially three. (80)
The defining properties of existential sentences are: (i) the subject cannot be moved as a whole (ii) extraction is possible from within the subject (iii) D°has an adjectival character
Thus, it is not the lexical elements involved that qualify a structure as an
166
Are there parameters in semantics?
existential sentence. Rather, it is the configuration in which these elements are finally placed that matters. However, over and above the various ideas just summarized, it is important not to lose sight of the more general aim pursued in the present work of abandoning the traditional rigid clause structure in favour of a more flexible one where predicative DPs are allowed to occupy the spec-IP position. In fact, the analysis of the Definiteness Effect as the violation of syntactic principles and the corresponding account that has been given for the lack of such an effect in Italian was made possible only by adopting this new flexible structure. Thus, the standard analysis which assumes that the subject expletive there must be replaced by the associated DP subject was compared with our proposal based on LF DP splitting. In the next two chapters, this general project will be further pursued. In the first three chapters, only sentences involving be have been discussed in detail. The theory of a flexible clause structure will now be tentatively applied to verbs other than be. In the next chapter, it will be applied to a verb that is roughly considered as a cognate of the copula, namely seem. In the final chapter, on the other hand, we will embark on a somewhat more ambitious undertaking which involves applying the same theory to unaccusative constructions.41
4
The 'quasi-copula': the role of finite clauses in seem-sentences
4.1
Finite clauses as arguments: passive vs seem
Since the first pioneering work (see Chomsky 1957), the clause structure proposed to account for subject-object asymmetries has always been considered as basically suited to the representation of passive constructions. If a simple sentence like John burned the book receives the following representation: (1)
Its passive counterpart, i.e. the book was burned by John, can easily be represented as follows (V being a complex verb): (2)
DP,
VP
(by DP)
The essential intuition this structure aims to represent is that the internal argument playing the role of the object in the active sentence is promoted to the subject position by movement transformation; on the other hand, the external argument, which plays the role of the subject in the active sentence, is now lowered' to the rank of an optional adjunct, here represented as an adjunct to VP. 167
168
The 'quasi-copula': the role of finite clauses in seem-sentences
If the internal argument of the verb is a finite sentence, a further possibility is available. Consider for example the following paradigm: (3a) b c
[DP Everyone] believes [s that John is intelligent] [s That John is intelligent] is believed (by [DP everyone]) it is believed [s that John is intelligent] (by [DP everyone])
In this case, the subject position can be held by a pronominal element which receives no semantic interpretation, namely it ((3c)). In fact, as far as 6role assignment is concerned, (3c) is entirely synonymous with (3a-b). These considerations lead to the conclusion that it is playing a role similar to there in copular sentences: i.e., it is the place holder of the subject of predication, technically an expletive. What is the structural position of the finite embedded clause then? The standard answer is that 'it can be left in place' (Lasnik and Uriagereka 1988: 19). Thus the passive sentence and the expletive insertion sentence would be analysed as follows: (4a)
(by DP)
VP
(by DP)
Why does /Jhave to occur in subject position? The answer is that an expletive is inserted in spec-IP to satisfy a general principle requiring sentences to realize the subject of the predication (see Chomsky 1981 and 1986a, following Rothstein 1983, and section m.ii in the Appendix). This analysis based on the idea of/^-insertion raises two further questions. Let us consider them separately. It is a well-known fact that the possibility of inserting the expletive it is blocked if the complement of the verb is a lexical noun phrase. In fact, the structure corresponding to *it was burned the book by John, turns out to be ungrammatical:
Finite clauses as arguments: passive vs seem (5)
b
169
S
VP
V
(by DP)
DP
Why is it impossible for the expletive to occur in such cases? The explanation for this phenomenon has generally been based on Case theory. Since the object position of passive constructions is not a Case position only clauses can remain there, because lexical noun phrases would need Case whereas clauses do not. Contrasts like the following justify this conclusion: (6a) b c
John is sure [s that Mary left] * John is sure [DP this] John is sure of [DP this]
The element of is interpreted as playing the role of a Case assigner. The contrast between the first and the second sentence shows that the finite clause can be licensed although no Case is assigned to it.1 The second problem raised by this analysis seems to be harder to explain and will therefore be the central topic of this chapter. The issue can be illustrated as follows. When the verb seem is followed by a finite clause, it is generally assumed that it shares the same underlying structure as the passive of believe (cf. (4a)). Focussing on the crucial part of the structure, the following representation would be assigned to a sentence of the type it seems that John left: (7)
S
seems
Thus, the finite clause is assumed to be in object position and the expletive it is inserted in subject position as a place holder for the subject of predication.
170
The 'quasi-copula': the role of finite clauses in seem-sentences
It is a well-known fact that the similarity between seem and the passive construction evaporates if the finite clause is moved to the subject position: (8)
*
S
seems Only the believe-type construction allows such an option, while seem yields a sharply ungrammatical result (cf. *that John left seems to everyone). The analysis based on the idea that it is an expletive offers no clear answer to this. In fact it seems quite paradoxical. 2 Why should a semantically null element, i.e. an expletive, be obligatory? The main goal of this chapter is to derive this contrast from independent assumptions. We will differentiate the role of it as occurring in passives and smn-clauses by appealing to the unified theory of copular sentences we developed in chapter 1. The first step will be to bring out the deep structural similarities between the copula and seem. 4.2
Seem as the 'quasi-copula'
The following pair shows that the copula and seem share a fundamental property: (9a) b
John is * (weird) [s that John left] seems * (weird)
The two verbs must combine with an independent predicational relation: in both cases, obliterating the predicate weird produces an ungrammatical sentence. This analogy is traditionally implemented by assuming that in both cases the verb must select a small clause complement whose subject is raised to spec-IP:
Seem as the 'quasi-copula' 171 (10a)
seems
weird
Following traditional terminology, we can call seem the 'quasi-copula'. In the next section we will push this analogy between seem and the copula to the limit and show that it will provide a solution to the main problem addressed in this chapter. 4.2.1 Inverse structures with clausal subjects We saw in the case of copular sentences that not only argumental DPs are allowed access to the spec-IP position. The crucial pair of examples are repeated here: (lla) b
John is the cause of the riot the cause of the riot is John
The sentence in (lla) is derived from the same base generated structure as the one in (lib): the difference between the two is that in the former the subject is raised while in the latter the predicate is raised. We distinguished these two structures by calling them respectively 'canonical' vs 'inverse' copular sentences and we assigned them the following representation:
172 (12a)
The 'quasi-copula': the role of finite clauses in seem-sentences S
With these premises, we can assign a sentence like it's John to the type of inverse copular sentences: (13)
[IP itj is [sc John tj]
The spec-IP position occupied by it is linked with the position where predicates are generated in the small clause while the subject stays in situ. Care is needed with judgments in such cases. Potentially, it's John could also be a canonical sentence, predicating the property of being John of the referent of it.3 In fact, copular sentences show that proper names can also play the role of a predicate. In Italian this claim receives clear empirical support from the following contrast: (14a) b
sono Gianni (am Gianni) e Gianni (is Gianni)
Verbal agreement reveals that in the first case the copular structure is canonical whereas the second one is ambiguous between a canonical and an inverse structure. Let us see why. Proper names are inherently assigned third person: thus, the very fact that in the first sentence a proper name can occur with a verb displaying first person shows that it is not able to trigger
Seem as the 'quasi-copula'
173
agreement. This amounts to saying that a fully argumental pro is present and that the proper name Gianni occurs as a predicate. In the second case, instead, the sentence is ambiguous, as in the corresponding English case, between a canonical and an inverse reading. Bearing these observations in mind, the syntactic properties of it's John will be seen to follow if the inverse structure is assigned to it. For example, consider the following case (in the proper reading): (15)
I thought [itj *(to be) [ sc John tj]]
As the asterisk indicates, the copula cannot be absent. In the first chapter, we saw that this is a typical diagnostic for inverse copular sentences. The copula provides a landing site for the predicate: if it is absent, inversion cannot take place. Notice that this particular type of inverse copular sentence involving a propredicative it has a peculiar property. The propredicate cannot be left in situ, as in the following case: 4 (16)
* [IP Johns is [sc ts it]]
Although this property might ultimately be traced back to independent facts, as will be proposed in the next section, for the present purpose we can simply observe that propredicative it is compatible only with the spec of a non-lexical head like 1°. Let us now turn to the 'quasi-copula', namely seem. The specific proposal I would like to make here is that the small clause analysis which is independently assumed for seem to derive a sentence like that John left seems weird ((10b)) should be extended to cover the case where this verb is followed by a finite clause, as in it seems that John left. The central idea should be transparent by now. I propose that, when occurring with seem, it plays the role of a propredicate instantiating an inverse construction. Thus, instead of the traditional analysis where it is inserted as the subject of the predication ((17a)), we will adopt a different representation where it is raised from the predicative position contained in the small clause ((17b)): (17a)
seems
174 b
The 'quasi-copula': the role of finite clauses in seem-sentences S
The new representation given in the second structure can be regarded as an instance of an inverse sentence with a clausal subject. The finite clause (S) following the quasi-copula is not its complement as is commonly assumed, but rather the subject of its complement, i.e. the subject of the small clause. Crucially, the spec-IP position is occupied here by a propredicative it raised from the small clause, not by a place holder which is standing in for the subject of the predication. Notice that the hypothesis that seem allows predicative DP raising is independently motivated. There are a number of different facts that point towards this conclusion. Recall first that 'full' predicative DPs as well as there can also be raised from infinitival complements ((18a-b)), on a par with argumental DPs ((18c)); this clearly parallels the case of the copula: (18a) b c
[ip [DP the cause of the riot]j seems [IP tj to be [ sc [DP a man] tj]]] [ip [ip [DP therejj seems [JP tj to be [Sc [DP a man] tj]]] (in the room)] [IP [DP a man]i seems [IP ts to be [Sc tj [DP the cause of the riot]]]]
Seem simply adds one more step to the chain of the raised DP (see section 2.4). Moreover, the analysis of it as a propredicate of a small clause whose subject is a finite clause will also be seen to follow on independent grounds if we consider copular sentences. Notice first that finite clauses can be the subject of inverse copular sentences as well as of canonical constructions: 5 (19a) b
[DP the worst hypothesisjj is [Sc [CP that John left] t{] [Dp that John leftjj is [ sc tj [CP the worst hypothesis]]
Consider now the following sentence: (20)
it's that John left
What structure should be attributed to this copular sentence? As usual, within our framework there are two options: either it can be regarded as a canonical copular sentence or else as an inverse one. We will now examine
Seem as the 'quasi-copula'
175
two different kinds of argument that support the conclusion that it should be regarded as an instance of the latter type. On the one hand, it appears to react positively to the empirical tests which can be used to detect inverse copular sentences. For example, the copula cannot be deleted in infinitival contexts of the following type (not to be confused with the perfectly grammatical / couldn't believe it that John had left which does not involve a small clause but rather extraposition): 6 (21)
I believed it *(to be) that John left
If, on the other hand, it were a canonical sentence instead, then this would actually be the only case where a finite clause plays the role of a predicate in a copular sentence; that such structures are normally ungrammatical is clearly shown in the following example: 7 (22)
* Peterj is [ sc tj [that John left]]
On the whole, then, it seems too ad hoc to assume that the finite clause in a sentence like it's that John left is the predicate. Comparative grammar also supports this analysis in terms of an inverse structure with a clausal subject. The Italian equivalent of a sentence like it's that John left is particularly revealing. There are two overt facts that signal the inverse nature of these constructions. Along with a pro subject ((23a)), we can have a d-sentence ((23b)), or a pronoun (e)gli which is attested in Fiorentino ((23c)): 8 (23a) b c
proj e [Sc [s che Gianni e partito] tj] (pro is that Gianni is left) Cj'e [sc [s c n e Gianni e partito] tj] (there is that Gianni is left) eglij e [sc [s che Gianni e partito] tj] (he is that Gianni is left)
Given the analysis of ci proposed in chapter 2, the very presence of a cisentence in this group strongly supports the hypothesis that we are dealing with instances of inverse copular sentences. Moreover, cliticization phenomena also confirm this analysis. In Italian, a postverbal finite clause can be cliticized by lo, in the proper context: (24a) b
Maria dice [s che Gianni e partito] (Maria says that Gianni is left) Maria lOj dice tj
(Maria lo says) The process of cliticization, however, is blocked for the finite clause in the constructions under discussion. Consider for example (23b), rephrased here as (25a):
176 (25a) b
The ' quasi-copula': the role of finite clauses in seem-sentences c' e [s che Gianni e partito] (ci is that Gianni left) * celOjetj
(ci lo is)
As we noticed in chapter 1, no explanation would be available for this fact if it were not considered to be an instance of an inverse construction. All in all, then, it would appear that the analysis of it as a raised propredicate which was suggested for sentences involving seem needs to be independently assumed for sentences like it's that John left and thus that the following representation should be assigned to such sentences: (26)
Summarizing, I have so far suggested that the small clause analysis given to sentences like John seems sad should be extended to clauses of the type it seems that John left. In these constructions, the clausal subject stays in situ, while a propredicative element is raised to the most prominent position. It has been shown that the hypothesis of inverse structures with clausal subjects can be independently motivated by analysing sentences of the type it's that John left. Additional evidence also came from the analysis of certain cliticization phenomena in Italian. 4.2.2
On the visibility of non-lexical elements at LF: a case study in the Principle of Full Interpretation The following contrast provides another piece of evidence in favour of the analysis of it as a propredicate: (27a) [IP itj is [ sc [CP that John left] tj]] b * [IP [CP that John left]; is [Sc tj it ]]
The element it cannot be left in situ. The only possibility is for it to be raised to spec-IP. This parallels the case of it's John. In neither case can the propredicative element be left in situ within the small clause. The same contrast also occurs with smn-clauses:
Seem as the 'quasi-copula' 111 (28a) [IP itj seems [sc [CP that John left] t{]] t> * [IP [CP that John left]j seems [Sc tj it]] Why are the canonical counterparts of these inverse structures ungrammatical? The descriptive generalization proposed would appear to be unsatisfactory: stating that it is only compatible with the spec-position of a non-functional head does not constitute an explanation. On the face of it, it seems to me that there are two independent ways of working towards the understanding of this fact: one based on Case theory and one based on the principle of Full Interpretation (as proposed in Chomsky 1988a). A priori, it is not clear whether we are dealing with two principles that simply happen to provide the same effect (exclusion of canonical sentences with propredicative it) and so are to that extent redundant, or whether there is some deep connection between them that for the moment escapes us. We will now investigate the empirical consequences of the two solutions separately. As we noticed above, finite clauses are generally claimed not to need Case; witness contrasts like the following: (29a) John is sure [s that Mary left] b * John is sure [DP this] c John is sure of [DP this] Assuming that this is right, an explanation of why propredicative it has to raise should be forthcoming. This element is a noun phrase; in chapter 1, we established that predicative noun phrases can acquire Case by raising to spec-IP (or by case agreement with the raised subject noun phrase). Clauses, on the other hand, do not need Case. If the finite clause were raised to specIP, there would be no way for it to acquire Case: in fact, there would be neither raising nor agreement. Thus, the only possibility is for it to raise and for the finite clause to remain in situ. This is exactly what is required. On the face of it, then, this would seem to be a plausible answer. However, it seems to me that the idea that clauses do not need Case is far from unproblematic. Consider for example the following sentence: (30)
[s that John left]; seems [sc tj [AP weird]]
If we maintain the idea that clauses do not need Case, the analysis of such a construction might appear puzzling. Neither of the two constituents involved in the small clause, namely the finite clause S and the AP, actually need Case. Why should the clause raise at all, then? We will not attempt to find a solution to this problem here. Instead, we will pursue the second line of reasoning.
178
The 'quasi-copula': the role offinite clauses in stem-sentences
Let us consider the second possibility based on the principle of Full Interpretation. Developing certain ideas that go back to the seminal work of Emonds (1985), Pollock (1989) suggested that the different distribution of adverbs in cases like the following could be captured transformationally: (31a) b
John has always kiss-ed Mary John always kiss-es Mary
What we notice immediately is that if the finite verb is an auxiliary, then the adverb follows it; on the other hand, if the finite verb is a full lexical verb, the adverb precedes it. Technically, this has been accounted for by assuming that the adverb is in the same position in both sentences. Thus, the contrast shows that auxiliaries are raised to acquire inflection whereas full lexical verbs require the inflection to be lowered. The following simplified representation will be sufficient for our purposes: (32a) b
John haSj-I° always tj kissed Mary John tj always kisses-I°j Mary
Let us now concentrate on the raising of the auxiliary, leaving the problems posed by lowering aside. Pollock's intuition has been regarded by Chomsky (1993) as the sign of a much more general principle of grammar: 'raising of the auxiliaries reflects their semantic vacuity; they are placeholders for certain constructions, at most, "very light" verbs. Adopting the intuition (but not the accompanying technology), let us assume that such elements, lacking semantically relevant features, are not visible to LF rules. If they have not raised overtly, they will not be able to raise by LF rules and the derivation will crash [fn. omitted]' (Chomsky 1993: 31). Theoretically, the idea underlying Chomsky's remarks on Pollock's intuition can be regarded as a strong version of the principle of Full Interpretation requiring only legitimate objects to appear at a given level (Chomsky 1988a). This principle was responsible, for example, for the process of 'expletive replacement', which was discussed in chapter 1. Expletives are not legitimate objects at LF, where interpretation takes place. Thus, they have to disappear, and consequently they end up being replaced (or in more recent versions, affixed to) by the element associated with them. The situation here is similar although this time it concerns operations rather than objects. Thus, raising may apply only to elements that are visible at the appropriate level. Since auxiliaries are considered as 'placeholders for certain constructions', they cannot raise at LF alone and must therefore raise in the overt syntax. What has just been said about auxiliaries has an obvious application to
Seem as the 'quasi-copula' 179 the domain we have been considering here. Although our analysis differs from the standard one (in that it is the place holder for the predicative noun phrase, rather than for the subject of the predication), we will still adhere to the traditional analysis in that it is not considered on a par with fully lexical elements but indeed as a place holder. Assuming that this is correct, the absence of canonical counterparts to sentences involving propredicates will simply follow from the principle of Full Interpretation. If the propredicative element it does not raise overtly, it would not be visible at all at LF, and the corresponding structure will simply be rejected as incomplete by the computational system: (33)
seem
t
To put it metaphorically, it is as if only the trajectory of place holders were visible to the computational system, not the place holders themselves: if their trajectory is null, they are invisible. If this analysis proves tenable, there would appear to be little need to appeal to Case reasons; instead we can simply regard the absence of canonical counterparts to sentences like it's that John left or it seems that John left as the result of a more pervasive principle of grammar, namely the principle of Full Interpretation. Whether or not this apparent redundancy will turn out to be avoidable, or perhaps just non-existent, is not a matter that can be pursued here. Some further remarks will be made on this question in the next section. To sum up, in this section we have pursued the analogy between the copula and seem, traditionally called the 'quasi-copula'. The small clause analysis independently assumed to account for certain constructions involving seem (such as: that John left seems weird) has been extended to cases where seem is followed by a finite clause (as, for example, it seems that John left). The central idea is that such cases should be analysed in the same way as inverse copular sentences, that is, as involving a predicative element raised to specIP. Specifically, we proposed to analyse the finite clause as the subject of
180
The 'quasi-copula': the role of finite clauses in seem-sentences
the small clause complement of seem and it as a propredicate raised from the small clause. The idea of a propredicative it with a clausal subject was shown to be independently motivated. 9 The next step is to evaluate the empirical advantages of such a theory and try to explore its range of application. This goal will be pursued from three different perspectives. As a start, we will attempt to deal with the major contrast addressed in the first section of this chapter, namely why seem does not behave like passives in allowing raising of the entire clause. Then, we will proceed to explore three distinct domains. First, we will apply this analysis to a long-standing question regarding Germanic languages. Second, we will examine extraction of adjunct phrases from seem-clauses. Third, we will see the consequences of this analysis for the treatment of Super Raising phenomena. 4.3
On differentiating the role of it: empirical evidence
Let us now turn to the contrast between seem and the passives of verbs like affirm. The base generated relations between the embedded S and V° in the two cases can now be distinguished as follows. For a sentence like it was affirmed that John left the following structure would hold: (34)
V
affirm On the other hand, for a sentence like it seems that John left, the following structure should be assigned: (35)
seem
In the case oiaffirm, we will not need to depart from the traditional analysis. The finite clause S is selected by V° and plays the role of the object as indicated in the first of the two structures above ((34)). In the case of seem, on the other hand, the situation is substantially different. There is no direct selection between V° and S. Rather, the verb selects a clause structure (the
On differentiating the role of it: empirical evidence
181
small clause, SC) whose subject is another clause structure (S) and whose predicate is the propredicate it; this is represented in the second of the two structures above ((35)). We can now turn back to the major contrast between passives and seemclauses, which was addressed in section 4.1 ((4a) and (8)). From this new perspective, the contrast will now be seen to follow. For the sake of simplicity, let us reproduce it here: (36a) that John left is believed b * that John left seems The base generated structure for seem allows only one possibility: (37)
S
seem
The propredicate it has to be raised. Crucially, it cannot be replaced by the finite clause: this would amount to deleting the predicate of the clause structure selected by seem. The alternative possibility of raising the finite clause, leaving in situ the propredicative it, is independently ruled out, as was shown in the previous section. The situation with passives, on the other hand, is entirely different. We have two options here. Let us consider them separately. The finite clause can be raised to spec-IP after passivization of V°, on a par with all objects: (38)
was affirmed
Alternatively, as we mentioned above and is generally assumed, the expletive it can be inserted into spec-IP in order to satisfy the Extended Projection Principle (see, for example, Lasnik and Uriagereka 1988: 19):
182 (39)
The 'quasi-copula': the role of finite clauses in seem-sentences S
was affirmed
It would appear then that we have accounted for the major contrast between seem and the passive construction: the basic relation between V° and the finite clause is totally different in the two cases, contrary to standard assumptions. In the case of the passive, we have a standard relation between a head and its object; with seem, on the other hand, we simply have a linear sequence of a head followed by another element, which in fact is the subject of a small clause which is itself the object of the original head. Before exploring certain consequences of this analysis for the syntax of seem, I will attempt to refine further the analysis of the distribution of it by considering copular sentences involving predicative APs and, again, passive constructions. We will begin with the following paradigm: (40a) [s that John left] is obvious b * it is [s that John left] obvious c it is obvious [s that John left] These contrasts raise a natural question. Why can't it be inserted in the specIP position of the sentence involving obvious, thus leaving the subordinate clause in situ, exactly as happens with passives? Let us consider the underlying relation between embedded S and the AP obvious. In these constructions, the copula selects the small clause containing a clausal subject (S) and a predicate (obvious). The subject is raised to spec-IP paralleling a case like that John left was affirmed. (41)
[obvious]
On differentiating the role ofit: empirical evidence 183 Clearly, there is no way to reanalyse (40a) as we did for seem. In particular, the element it cannot play the role of a predicate, because the predicate is indeed present, namely obvious. Thus, we might ask what accounts for the impossibility of it occurring in *it is that John left obvious? Leaving a Casetheoretical account aside, I would like to suggest that, in fact, this is also a consequence of the principle of Full Interpretation. In the previous section we concluded that those elements which are inserted as place holders for other elements are not visible unless they are moved overtly. From this premise, it follows that *it is that John left obvious is ungrammatical for the simple reason that it is not moved and consequently there is a violation of the principle of Full Interpretation. As for the position of the finite clause in grammatical examples of this type of construction, we can simply follow the standard account and assume that it is in an adjunct position (technically, 'extraposed'): (42)
is
tt
[obvious]
Let us now turn to passive constructions. The principle of Full Interpretation leads to a refinement of this type of construction. We cannot consistently maintain that it is simply inserted in subject position, because a violation of the principle of Full Interpretation would be expected since there would be no reason to assume that it ever moves: (43)
was affirmed
184
The 'quasi-copula': the role offinite clauses in seem-sentences
Thus, we are forced to assume that in passives it is moved from the object position and that the finite clause is in the same position as in the copular construction involving obvious: (44)
was affirmed
Notice that the idea that the finite clause is extraposed is independently required if we are to account for the fact that it can be omitted as in the following example: (45a) b
it was affirmed that John left it was affirmed
In fact, omission is generally impossible for complements: (46a) Mary affirmed that John left b * Mary affirmed All in all, we can conclude then that in each case the place holder it is moved from the position where it is originally generated. This apparentlyfitsin with the theoretical hypothesis stemming from the principle of Full Interpretation to the effect that place holders must move overtly. Thus, we have distinguished three distinct base generated positions for it: the object position (with the passive of affirm-type constructions: (47a)); the subject position of a small clause (with obvious-type constructions: (47b)); the predicative position of a small clause (with seem: (47c)): (47)
affirm
affirm
obvious
seem
S
On differentiating the role of it: empirical evidence 185 In the rest of this section, we will explore some implications of this analysis (with its specific assumption that it can be associated with these three different grammatical functions). Three cases will be discussed: an apparent exception to a common phenomenon in verb-second languages, an asymmetry in adverbial extraction and a possible refinement of the analysis of Super Raising. 4.3.1 A case study in verb-second languages The following paradigm taken from Bennis (1986: HOff.) shows a typical phenomenon of verb-second languages like Dutch: (48a) b
het is gebleken dat Jan ziek was (it is appeared that Jan ill was) gisteren is gebleken dat Jan ziek was (yesterday is appeared that Jan ill was)
This is generally interpreted in the following way. The spec-CP position in matrix clauses of these languages must be filled by some lexical material: it can either be a pronominal form like het (the equivalent of English it) or an adverb like gisteren (yesterday):10 (49) het/gisteren
C
is
This option is banned with the equivalent of seem, namely schijnen: (50a) b
het scheen dat Jan ziek was (it seemed that Jan ill was) * gisteren scheen dat Jan ziek was (yesterday seemed that Jan ill was)
In this case het is obligatory, as opposed to the more general pattern. There is no complementary distribution with, say, adverbs like gisteren. In our framework the explanation of this anomaly is straightforward: het does not play the same role with the two verbs. In the first case, het is just
186
The 'quasi-copula': the role of finite clauses in seem-sentences
required as a genuine expletive to fill the spec-CP position (possibly after extraposition as proposed by Bennis 1986) and thus satisfy the morphological requirements of C°. In the second case, it is the propredicate of the CP contained in the small clause complement of schijnen. Thus, omitting it would amount to omitting a fundamental piece of the clause structure selected by the matrix verb.11 4.3.2 Asymmetries in adverbial extractions A well-established hypothesis is that adverbial adjuncts can only be extracted in successive cycles (see for example Chomsky 1986b, Rizzi 1990, Cinque 1990a, Manzini 1992 among others, for a detailed discussion), i.e. passing through each intermediate spec-CP position (51)
Now, in a previous section, we saw that a unitary analysis for the affirmtype and the obvious-type constructions implies that when it occupies specIP the subordinate CP is in a left-adjunction position:
On differentiating the role of it: empirical evidence 187 (52)
IP
is tj obvious was affirmed i{
This contrasts with seera-sentences, where the CP is embedded in the subject position within the small clause: (53)
Does this asymmetry have any syntactic consequences? We will attempt to answer these questions by testing extraction of a sentential adjunct from the CP in the three distinct cases: (54a) b c
why does it seem that John left? why is it obvious that John left? why is it affirmed that John left?
The interpretation of these sentences is different. In the last two sentences why can hardly be interpreted as referring to John's leaving. These sentences rather ask why John's leaving is obvious ((54b)) and why it was the content of someone's affirmation ((54c)), respectively. In the first sentence, instead, why can easily be related to the subordinate clause ((54a)).12 Extraction of a manner adverbial gives an analogous pattern: (55a) b c
how does it seem that John left? how is it obvious that John left? how is it affirmed that John left?
188
The 'quasi-copula': the role of finite clauses in seem-sentences
A natural way to capture these asymmetries is to relate them to the different position of the CPs involved. One possibility is that the intermediate trace of the wh-chain passing through the spec-position of the subordinate CP violates locality conditions on the movement of adjuncts. For example, Rizzi's (1990) system which reduces the ECP module to head government will capture the data: since the CP is extraposed with obvious and affirm its spec position is not accessible to external government; crucially, this is different from the case of seem where the CP is in fact governed by the matrix verb: (56a) • • . [v° seem] [s c [CP tj that John left tj] . . . b * . . . obvious] [Cp tj that John left ts] . . . c * . . . affirmed] [CP tj that John left tj] . . .
Notice that the plausibility of an analysis that relies on the asymmetric position of the subordinate CPs involved, as opposed to one appealing to some idiosyncratic properties of seem, is supported by the fact that extraction of arguments from the embedded CP, as opposed to adjuncts, results in greater acceptability: (57a) b c
[which book]j does it seem [CP tj that John read tj]? [which book]i is it obvious [Cp tj that John read tj]? [which book]; is it affirmed [CP tj that John read tj]?
Any theory analysing seem as analogous to the passive of affirm would hardly be able to capture these asymmetries. In fact, the opposite state of affairs will be expected where passives are given the same analysis as seem and the obvious-type constructions are set apart as cases of extraposition. Rather, as I have attempted to show, passives behave like the obvious-type constructions. In the next section, we will test the capacity of the theory proposed here to predict a frequently examined class of violations, namely the Super Raising violations.13 4.3.3 On Super Raising The following contrast has played an important role in the study of locality conditions on movement: (58a) it seems that John appears to win b * John seems that it appears to win
Within the standard analysis given to seem, this contrast was explained as follows. The grammaticality of the first case ((58a)) was not considered surprising: on the one hand, John is raised from the embedded infinitival clause to the subject position of the embedded finite clause; on the other,
On differentiating the role of it: empirical evidence 189 the expletive it is inserted in the subject position of the matrix clause as an expletive: (59)
t to win
Further raising of John to the subject position of the matrix clause and insertion of an expletive in the intermediate spec-IP position (technically, Super Raising) are prohibited, showing that A-movement is a highly local process:
190 (60)
The ' quasi-copula': the role offinite clauses in seem-sentences *
John
IP
I'
t to win
A first attempt to reduce Super Raising to the Subjacency Condition (Chomsky 1981) was abandoned since it was soon realized that the extraction of wh-phrases can easily involve skipping an intermediate specIP position: (61)
whOj does it seem [CP that [iP Johrij appears tj to love tj]]
Chomsky (1986b: 22, fn. 18) also noticed that 'a binding-theoretic explanation is dubious, since pleonastics yield only "weak" violations'.14 Let us now examine two influential proposals. Chomsky (1986b) and Rizzi (1990) agree in regarding Super Raising as a violation of the ECP. These two systems start from different assumptions and converge to the same conclusion: the A-chain connecting John with the lower trace cannot skip spec-IP. In Chomsky (1986b), where the ECP is ultimately reduced to antecedent government (via the notion of 'extended chain': see chapter 1 and Moro 1993a), it blocks the only position for proper government. In Rizzi (1990) where the ECP is reduced to head government, a special requirement is added for A-chains on the basis of 0-theoretic reasons:15 movement cannot skip any intermediate A-positions.16 However, what happens if the raised element does move through the intermediate spec-IP position before reaching the spec-IP of the matrix verb? Consider the following example:
On differentiating the role of it: empirical evidence (62)
191
* Johrii seems tj appears tj to win
The sentence is still strongly ungrammatical: in fact, the level of deviance is comparable to the classic Super Raising violations. A priori, what kind of explanation might be expected to account for it?17 A Binding-theoretical violation is to be excluded in principle since it was excluded for Super Raising as too 'weak'. Thus, locality conditions on movement must be responsible for it. Leaving the technical details aside, current proposals essentially regard movement through the intermediate spec-CP position as the offending step. On the one hand, the chain is forced to pass through this position to satisfy locality conditions on movement. On the other, this A'-position is considered to be incompatible with the two A-positions it connects. These two independent facts thus conspire to rule the sentence out: (63)
* Johrij seems [CP tj C° [!P tj appears tj to win]]
In the rest of this section I will not attempt to criticize current proposals. Instead, I would like to illustrate the line of reasoning suggested by our approach to the syntax of seem; this would appear to be promising as a way of tackling Super Raising phenomena and the last case illustrated here (see Moro 1993a for a detailed analysis of current approaches). Standard accounts share the hypothesis that the embedded CP is the complement of seem: (64)
V
CP
seems
On the other hand, in our framework the following structure holds: (65)
V
seems CP
it
Assuming we adopt this more articulated structure, what are the consequences for the analysis of Super Raising? As a preliminary step, consider the following relation, where the copula selects a small clause containing two DPs:
192
The 'quasi-copula': the role of finite clauses in seem-sentences
(66)
V
a picture of John
the cause of the riot
Then, consider the following sentences: (67a) * [Johnjj is [Sc [DP a picture of tj] [DP the cause of the riot]] b * [the riot]j is [ sc [DP a picture of John] [DP the cause of tj]]
What accounts for the strong ungrammaticality of these structures? A rather natural explanation comes from Case theory. Spec-IP is the only Case position available in this structure. Moreover, either the subject or the predicative DP must raise to be assigned Case. Thus, if an argument embedded in either DP raises to spec-IP it will block Case assignment to the major DPs.18 The shift from this case of illicit movement to the one involving seem is quite obvious. Both Super Raising (*John seems that it appears to win) and the other structure under discussion (*John seems appears to win) can now be ruled out on a par in a rather natural way by Case-theoretical reasons. The line of reasoning is analogous to the one developed for the case of copular sentences ((67a-b)). Raising of an argument embedded in the subject of the small clause blocks raising of it to acquire Case and thus the sentence is immediately ruled out:19 (68) John
Notice that this explanation is in principle independent of whether the small
On differentiating the role of it: empirical evidence 193 clause contains a propredicative element or not. The same type of movement from the embedded CP is also prohibited in cases where the predicative element is fully lexical as in * John seems that it appears to win weird: (69)
weird
In fact, the only possible structure would rather be one where the entire CP is raised to the matrix subject position yielding that John appears to win seems weird. It would appear, then, that the principal aim of this chapter has been reached. It has been shown that, by abandoning the standard rigid conception of clause structure, what are considered exceptions in the standard framework follow from general principles in a rather natural way. After proposing a new analysis for the role of a finite tense CP with the quasicopula, an illustration of some applications of this analysis were given. First it was explained why sentences containing seem behave differently from passives; then, we explored three distinct empirical domains: a case study in verb-second languages, an asymmetry in the extraction of adverbial phrases and an alternative account of Super Raising. Although we are not going to pursue this here, adapting this analysis to different languages, as for example Italian, would be easy. The main difference in the case of Italian would be that a phonetically null propredicate (pro) is employed, as opposed to a phonetically overt one (it). Apart from these details all the other major assumptions would essentially appear to be valid for Italian as well (crucially, the assumption that a propredicative element raises from a small clause). It is worth emphasizing that there are now three independent domains for which the notion of inverse clause structure' gives good results: copular sentences, existential sentences and those seem-sentences where the main
194
The 'quasi-copula': the role of finite clauses in seem-sentences
verb is followed by a finite clause. It therefore seems that our proposal to abandon the rigid clause structure of traditional analyses is beginning to reach a certain degree of plausibility on empirical grounds. The rest of the chapter deals with a subject that is not crucial to the theory developed so far. Instead, whether or not the analysis that is to be developed there turns out to be plausible, it will not affect the theory proposed so far. Nevertheless, if the proposal that is to be made turns out to be well founded, this would constitute a significant simplification in the syntax of seem. Specifically, we will deal with cases where the quasi-copula is followed by an infinitival clause and argue in favour of a unitary analysis. In this case, we will see that the cross-linguistic differences are not trivial at all: it will in fact be shown that Italian allows a construction that is impossible in English. 4.4
Infinitival clauses with seem
We have so far considered cases where seem cooccurs with subordinate finite clauses. Can seem also occur with infinitival clauses? Three distinct cases will be discussed in this section; they are exemplified as follows: (70a) * it seems [Cp for John to be sad] b Johni seems [IP tj to be sad] c pro sembra [Cp di PRO volare] (pro seems di fly)
Let us briefly survey the problems related to these three cases. Why is a sentence like * it seems for John to be sad ungrammatical? Notice that a/orsentence is not necessarily incompatible with seem: (71)
[CP f°r John to be sad]j seems [sc tj unusual]
In order to account for the ungrammaticality of this sentence we will rely on a fundamental 'semantic' intuition formulated by Ruwet (1982). This will be dealt with in 4.4.1. As regards the second case in (70), John seems to be sad, this shows that an infinitival clause is perfectly acceptable with seem, provided that the subject of the embedded clause is raised to the matrix clause. However, this option involves a considerable complication. As indicated in (70b), a syntactic process called 'S'-deletion' is triggered (Chomsky 1981). For ECP reasons which will be illustrated in due course, the Comp periphery of the embedded clause has to be deleted and the clause is reduced to a bare IP. The aim of section 4.4.2 is to argue against this mechanism and in favour of a unitary analysis of seem with respect to its selectional capacities.
Infinitival clauses with seem
195
The third case in (70), sembra di volare {pro seems di PRO fly), is taken from Italian. This case appears to contrast with the English *it seems for John to be sad, in that an infinitival clause can occur with sembrare, the Italian equivalent of seem, without involving raising. The aim of 4.4.3 is to account for this cross-linguistic contrast. 4.4.1 for-clauses An infinitival subordinate clause is not compatible with seem unless the subject is raised to spec-IP of the matrix sentence or an independent predicate occurs, say obvious: (72a) * it seems for John to be sad b John seems to be sad c ? for John to be sad seems obvious The third case is trivial. The entire /or-clause is raised from the subject position of the small clause. We naturally expect the results to be grammatical (although some speakers find this sentence slightly degraded): (73)
? [Cp for John to be sad]j seems [Sc tj obvious]
The second case will be examined in the next section. For the moment we will concentrate on explaining why a /or-clause is banned in the first example. We will base our approach on a fact noticed by Ruwet (1982). The following two types of sentences are synonymous: (74a) b
it seems that John is sad it seems true that John is sad
The first sentence is interpreted as if the sentential predicate 'true' were applied to the embedded clause exactly as in the second sentence. We will explore the possibility that this coincidence is not accidental, but rather is codified in the syntax.20 More explicitly, I will suggest here that the configuration with the propredicate it (as (75)) is assigned by default the interpretation which we have in the case of the associated sentence that actually has the word true.21 That is, for every x ranging over propositions this construction is assigned the additional meaning 4x is true', 'x is the case', etc., which we might call a 'sentential predicate':
196 (75)
The 'quasi-copula': the role of finite clauses in seem-sentences IP
In fact, if we recall a piece of data we discussed in an earlier section, it will be easy to prove that this sentential predicative use of it can be independently attested. Recall for example the two following sentences: (76a) b
John is sad it's that John is sad
Saying it's that John is sad is not simply equivalent to John is sad. Its meaning is rather 'reinforced' as if it were: 'the fact is that John is sad'. This fits in with the hypothesis that the structure involving raising of a propredicative it is interpreted as assigning a sentential predicate to the subordinate clause. In fact, we independently know that the sentence it's that John is sad is an inverse copular sentence whose structure involves the raising of a propredicative it to spec-IP: (77)
[IP itj is [ sc [CP that John is sad] tj]]
With these premises, we can provide an independent reason to explain why a/or-clause cannot occur with seem. The fundamental observation is that a /br-clause, unlike a finite CP, cannot be the subject of a sentential predicate: 22 (78a) [Cp that John is sad] is true b * [Cp for John to be sad] is true Thus, since the construction involving the propredicate it is interpreted as involving a sentential predicate, a/or-clause is ruled out on a par with * for John to be sad is true. We will gain additional plausibility in favour of this explanation if we consider an independent contrast like the following: (79a) [IP [IP itj is [Sc U obvious]] [Cp whoj John saw tj]] b * [ip itj seems [ sc [CP whoj John saw tj] tj]]
Infinitival clauses with seem
197
Why can obvious occur with a CP marked [+wh] while seem cannot? Again, the reason is that a CP marked [+wh] can't be the subject of a sentential predicate. Let us apply the proper diagnostics:23 (80a) [Cp that John saw Mary] is true b * [Cp whoj John saw tj] is true Thus, both a sentence like * it seems for John to be sad and one like * it seems who John saw are ruled out for the same reason. The fact that obvious can occur with a CP[+wh] yielding it is obvious who John saw is due to the fact that the element it occurring with obvious is not a propredicate as it is in the case of it's that John left and it seems that John left. As further evidence, notice that obvious can also occur with a/br-clause, unlike seem: (81)
[ip [ip itj is [sc tj obvious]] [CP for John to be sad]]
Interestingly, if the wh-phrase is extracted from the embedded CP the sentence becomes grammatical, although both embedded clauses contain a variable: (82a) whoj doesI+wh| itj seem [sc [CP tj that(_wh| John saw tj] tj] b * [,P itj seems [sc [CP whoj C° [+wh] John saw tj] tj]] The meaning of thefirstsentence is 'for what x, that John saw x seems true?' The fact that the subordinate clause contains a variable does not matter. The first sentence is grammatical because the embedded CP is not marked [+wh-]; the [+wh-] clause is rather the matrix one where who enters into a spec-head relation with C°. All in all, we have explained the distribution of /br-clauses with seem. Crucially, it has been possible to maintain the inverse structure adopted for seem and this has allowed us to derive the difference between this and the obvious-type constructions. Let us turn now to the second case which needs to be discussed, namely sentences of the type John seems to be sad. 4.4.2 Against Sr-deletion: towards a unified theory o/seem There are at least three possible ways of combining a clause composed of a subject John and the predicate sad with the quasi-copula. Seem may select a small clause containing the subject John and the predicate sad ((83a)) or it may select a small clause containing the clausal subject that John is sad and the propredicate // ((83b)). Along with small clauses, a standard assumption is that seem may occur with an IP complement whose subject is raised to the matrix clause ((83c)). The three possibilities are represented as follows:
198 (83a) b c
The 'quasi-copula': the role of finite clauses in stem-sentences Johni seems [ sc tj sad] itj seems [ sc [that John is sad] tj] Johni seems [IP tj to be sad]
The main goal of this chapter was to reduce the representation of a sentence involving it in a sentence like it seems CP to the one of the type in ((83a)), namely to a structure where seems selects a small clause: the representation in (83b) satisfied such a goal. In this section we will rethink the analysis of sentences of the type in (83c) in an attempt to outline a unitary theory of seem. I would like to emphasize once again, however, that this further step is not crucial to the major aim of formulating an analysis for sentences like it seems that John left that treats them on a par with inverse copular sentences. Needless to say, of course, it would be a highly desirable result. We will begin by reviewing the analysis given to sentences of the type in ((83c)) in Chomsky (1981). A clause structure S with a complementizer periphery S' is base generated as follows: (84a) b
y [s NP INFL VP]] INFL (± Tense)
If INFL has the value — Tense, it is spelled out as to. When an S' containing a -Tense INFL is the complement of seem, the subject of the embedded clause is raised to the matrix subject position in order to acquire Case. Potentially, there are two possible paths: (85a) b
Joh^ seems [S' V [s tj to be sad]] Johni seems [S' [s tj to be sad]]
The trace in the subject position of the embedded clause violates the ECP in both cases. Antecedent government by t' in S' is not available because an A-chain (John, i) cannot include an A'-position (t')\ (86) John
Infinitival clauses with seem
199
Head government by seem is excluded as well because for independent reasons S' is generally assumed to block a government relation: (87)
t
The solution proposed by Chomsky (1981) was that if INFL has the value — Tense, then seem deletes S": (88)
Johiii seems [s tj to be sad]
The ECP is satisfied now because seem can head-govern the lower trace, S being a non-maximal projection: (89)
Mutatis mutandis, this explanation has essentially been adopted in most frameworks up until the minimalist programme (see for example Chomsky 1986b):24 (90)
Johrij seems [iP tj [v [i° to] [Vp be sad]]]
This analysis involving S'-deletion raises many questions. Two of them will be highlighted here. Notice first that S'-deletion must crucially depend on the±Tense value of INFL. If this process were available also for a +Tense INFL along with a —Tense INFL the following case would be a major problem: (91)
* Johrii seems [(P tj is sad]
Indeed, if S'-deletion were available for a +Tense INFL, this sentence could only be ruled out as a Binding Theory Condition A violation not as a violation of the locality conditions on movement. However, this would be contradictory since the degree of ungrammaticality of this sentence is the
200
The 'quasi-copula': the role offinite clauses in seem-sentences
same as that of a Super Raising violation, and yet a Binding-theoretical account is generally rejected when accounting for Super Raising violations as being too weak (see section 4.3.3). All in all, since there seems to be no principled way to make S'-deletion incompatible with the +Tense value, we should just assume this restriction to be ad hoc. Moreover, notice also that in the standard framework S'-deletion is obligatory for —Tense INFL clauses, otherwise PRO would be protected from government by seem and would then occur in the subject position of the embedded clause with an expletive inserted in spec-IP of the matrix clause: (92)
* [1P it seems [s- [s PRO to be sad]]]
All in all, these two independent restrictions associated with the deletion process appear to be quite ad hoc: the correlation between such a process and Tense features is not obvious at all. Although in principle it might not be impossible to trace these restrictions back to independent facts, nevertheless I would like to propose an alternative analysis that avoids any recourse to the deletion process. The idea that will be implemented here is that infinitival clauses simply do not contain inflectional heads, thus raising to the subject position does not need special assumptions. This idea will be developed in two steps: first, evidence will be provided against the assumption that to is the spell out of —Tense inflection; second, it will be shown that the clausal constituent involved can be traced back to another type of infinitival clausal constituent that is independently assumed, namely small clauses.25 Two properties of the INFL node diminish the plausibility of considering to as the spell out of 1°. First, Chomsky (1986b) (see also Roberts 1988a) provided evidence that 1° can properly govern its VP complement; witness the possibility of fronting it as in the following case: (93a) . . . and [Vp see Mary]; I believe John [v does tj] b * . . . and [Vp see Mary]j I believe John [r to tj] The ungrammaticality of the second sentence shows that to is unable to play such a role.26 Any explanation based on the ±Tense distinction of the two I°s would be very ad hoc here. Second, negation follows inflection in English finite clauses but the order is reversed when to occurs:27 (94a) b
I believe John [^ does] [Nego not] see Mary I believe John [Nego not] [r to] see Mary
Why should a —Tense inflection alter the situation? Again, assuming that
Infinitival clauses with seem
201
we set ad hoc solutions aside, the present contrast would appear to undermine further the assumption that to is an I°.28 Yet if to is not the spell out of inflection, what else could it be? What kind of clause structure should the complement of seem be considered to be in a sentence like John seems to be sad? An alternative analysis can be provided if we adopt the following line of reasoning. We will begin by considering the following example: (95)
I consider [sc [DP John] [DP my best friend]]
In this sentence a predicative relation with no inflectional specification holds between the two DPs in the small clause. Consider now the following case: (96)
I consider [a [DP John] as [DP my best friend]]
The same predicative relation with no inflectional specifications holds between the two DPs in this sentence, with an additional element, i.e. as. What is the role of as? What is the nature of the clausal constituent a? Clearly, there seem to be no semantic differences between the two sentences. Thus, it seems plausible to assume that as is present as the spell out of a syntactic function, rather than a full lexical element with typically lexical meaning.29 The hypothesis that I will adopt is that as plays the role of a predicative marker which is optionally inserted at a surface level. As a first approximation, the status of ^-insertion can be regarded as parallel to that of ofinsertion for DP complements of N°s as in the destruction of Rome. The underlying structure connecting destruction with its internal argument, i.e. Rome, is the following: (97)
destruction
Rome
In the course of the derivation, the element of is inserted to assign Case to the DP complement. On the face of it, then, it would appear that the following representation should be assigned to the phrase destruction of Rome:
202
The 'quasi-copula': the role of finite clauses in seem-sentences
(98)
N'
destruction
Rome
In general, however, it is widely assumed that o/-insertion is a very 'superficial' phenomenon. In fact, Rome is the argument of destruction rather than the argument of the P°: witness the fact that it is assigned the internal 6-role by destruction. As confirmation of this, notice that destruction is not able to have a fully lexical PP complement: witness the ungrammaticality of phrases like *the destruction in Rome, at least with the interpretation 'they destroyed Rome': (99)
*
N'
destruction
Rome
Thus, it seems that we do not lose any important information if we continue to regard the DP headed by Rome as the complement of the N° destruction. All in all, it seems to me that one can consider as and of as similar entities. They do not contribute to defining the grammatical functions involved; rather, they are present for purely morphological reasons such as those related to Case theory (see Chomsky 1986a). Notice that this hypothesis is confirmed by the fact that occurrence of as is impossible if Case assignment is performed in a different way, say by raising the subject to spec-IP within a copular sentence: (100)
[DP Johnjj is [ sc tj (*as) [DP my best friend]]
On the other hand, it would appear to be obligatory if the DP cannot acquire Case in any other way. Consider, for example, the following sentence: (101)
[IP [IP John telephoned] [DP *(as) the doctor]]
Infinitival clauses with seem
203
The DP cannot occur without the presence of as.30 Although we are still a long way from having a full understanding of DP adjuncts, these examples could all be dealt with by assuming that as plays the role of a Case assigner on a par with of. Finally, let us consider the following contrast: (102a) I consider [a [DP John] as [DP my best friend]] b * I consider [a [my best friend] as [DP John]] This pair shows that as can only be followed by a predicative DP. This excludes the possibility that as is playing a role similar to that of the copula, i.e. that it selects a small clause, otherwise we would expect the inverse order of predication to be possible exactly as in the cases discussed in chapter 1 (examples (45a-d)): (103a) b
I consider [[DP John]; to be [Sc U [DP m y best friend]]] I consider [[DP my best friend^ to be [Sc [DP John] tj]]
All in all, then, there seems to be no good reason for not assuming that the complements of consider in / consider John as my best friend and / consider John my best friend are exactly the same, namely small clauses: 31 (104) (as) DP Bearing this in mind, let us now go back to to. My specific proposal is that to plays the same role for VPs as as does for predicative DPs. In both cases these elements are inserted for morphological reasons: (105) DP
(to) VP
Apart from technical implementation, which is clearly very rudimentary, the underlying intuition that this proposal aims to capture is the following: from the point of view of its structure, a constituent like [John to run] is essentially the same as a small clause. The predicate is 'marked' (perhaps Case marked) by to when it is a VP and by as when it is a DP. Of course, no explanation for the distribution of to as opposed to as in such contexts is available but this does not seem to undermine crucially the line of reasoning we have pursued. 32 Furthermore, two independent facts support the analogy between to and as. As we noticed ((93a-b)), the complement of to cannot be moved, and this is true also of the complement of as:
204
The ' quasi-copula : the role of finite clauses in seem-sentences
(106a) * . . . and [see Mary]j I consider John to tj b * . . . and [my best friend]j I consider John as tj Moreover, we can expect the same ordering with respect to negation (cf. ((94a-b)): (107a) b
I believe John not to/*to not see Mary I consider John not as/*as not my best friend
Thus, assuming that this analysis of to proves tenable, there would appear to be considerable consequences for the theory of smn-clauses. In the first place we do not need to appeal to a process of deletion to account for the fact that the subject can raise to spec-IP from the lower position, as in John seems to run. (108)
Johrij seems [ sc tj to [VP run]]
We already know that this type of raising process is perfectly possible from a small clause and thus no further assumptions are required. Crucially, no S'-deletion process is needed any longer. However, there is an even more important advantage of this new analysis, over and above the fact that S'-deletion is no longer necessary. Bearing in mind the analysis of those cases where seem is followed by a finite CP, the lexical entry for the quasi-copula seem can now be reduced to the simple form given, which exactly parallels the case of the copula in the second of the two representations that follow:33 (109)
a
V
v°
b
sc
V
v°
sc
seem be This simple entry is now sufficient to yield the three possible ways of combining the clause composed of the subject John and the predicate sad with seem; these depend on how the small clause is organized: (110a) b c
[IP Johrij seems [ sc tj sad]] [IP Johrij seems [Sc U to [be sad]]] [IP itj seems [ sc [CP that John is sad] tj]]
From the point of view of language acquisition, what is particularly interesting about our suggestion is that the only piece of information the child needs to know about the lexical entry of seem is that it selects a small clause complement. There is no need to enrich the lexical entry by assuming
Infinitival clauses with seem
205
that seem is able to perform S'-deletion according to the relevant value of Tense features or any similar process. From the point of view of comparative grammar, on the other hand, the situation is more delicate. Some additional language-specific factors must be recognized to account for the type of predicates that can occur in the small clause in any given language. In fact, even a cursory survey across languages seems to suggest that if VPs are allowed across many languages as predicates of small clauses selected by seem, the other lexical categories allowed vary a lot within any given language and also across languages: (Ilia) * Jan schijnt [AP ziek] (Jan seems ill) b John seems [AP ill] c Gianni sembra [DP il re di Francia] (Gianni seems the King of France) d * John seems [DP the King of France] e * John seems [PP in the room] f John seems [PP in deep water] These data show that no generalization is within easy reach. In Dutch ((Ilia)), APs are banned while in English ((111b)) they are perfectly acceptable (see Bennis 1986: 114). In Italian ((111c)) DPs are grammatical whereas in English they are not ((Hid)). In English ((llle)) PPs are banned unless they have a metaphorical interpretation ((11 If)). Several factors, partially related to lexical categories and partially to semantics, seem to intervene here.34 Yet these rather scattered patterns suggest that they are all rather idiosyncratic and language-specific and would not therefore seem to undermine our analysis to a serious degree. In general, the analysis we have developed so far could easily be reproduced for Italian, as long as such parametric differences as distinguish the two languages (such as those related to pro licensing) are taken into account. In the next section, on the other hand, we will concentrate on a rather murky case where cross-linguistic differences should not be traced back to parametric variation. 4.4.3 A challenging case: Italian sembrare with di-clauses Let us now turn to the last case which will be investigated in the present section devoted to infinitival complements of seem: the relevant example ((70c)) will be repeated for convenience here: (112)
pro sembra [CP di PRO volare] (pro seems di fly)
206
The 'quasi-copula': the role of finite clauses in stem-sentences
The Italian equivalent of English seem, namely sembrare, occurs here with the infinitival clause di volare (di fly). This sentence represents a challenging problem. For the moment we will disregard the question of its meaning. Kayne (1984) showed that Italian di (and French de) should be analysed as complementizers. Thus, di is to be regarded as having the same status as English for, rather than as the equivalent of infinitival to. Adopting more up-to-date terminology di, de and for are to be analysed as C°: (113)
CP
Thus, the infinitival clause following sembrare in pro sembra di volare is a CP and cannot be reduced to a small clause as we proposed for the other cases. Kayne (1984) also showed that de and di protect the subject position from government from outside and that they are themselves inert to government: this forces PRO to occur in subject position (with an optional dative controller). A partial representation of pro sembra di volare now involving an optional dative controller is the following: (114)
pro (mij) sembra [CP [c°di] [iP PRO(j) volare]] (pro to-me seems di PRO fly)
Earlier (see section 4.4.1), we noticed that /br-clauses cannot occur with seem, though this is possible, for example, in obvious-type constructions: (115a) * it seems [Cp for John to fly] b it is obvious [Cp for John to fly] This contrast was explained by arguing that an infinitival clause cannot be the subject of a sentential predicate. Since the constructions involving propredicative it are interpreted by default as containing such a predicate, * it seems for John to fly was ruled out. If di is equivalent to for, why is the Italian ^/-clause grammatical with sembrare? The restriction which excludes /or-clauses as subjects of seem
Infinitival clauses with seem
207
cannot simply be relaxed in this case: all the empirical tests we used for English give exactly the same result in Italian. Consider for example the fundamental property of infinitival clauses, namely, that they cannot be the subject of sentential predicates. This also holds in Italian: (116a) * volare e vero (fly is true) b che voli e vero (that pro fly is true) Thus, to assume that this condition does not hold in Italian would not only be ad hoc, it would also simply end up undermining the whole basis of the explanation given for the English example. 35 The solution I would like to suggest here is that the role of the embedded clause in pro sembra di volare is entirely different from that of the embedded clauses occurring with seem we have analysed so far. Crucially, di volare does not appear as subject of the small clause selected by the quasi-copula. What role then does this CP play? To answer this let us consider the following empirical test. Italian come (like) can be used as a modifier of predicates. Consider, for example, the following paradigm: (117a) b
Gianni e (come) [DP il vero responsabile] (Gianni is like the actual responsible) Gianni e (come) [AP spaventato] (Gianni is like frightened)
As further confirmation of this observation notice that come cannot be used to modify subjects in inverse copular sentences: (118)
il vero responsabile e (*come) Gianni (the actual responsible is like Gianni)
Crucially, this modifier can also be used with infinitival sentences. The following contrast between a canonical and an inverse sentence, respectively, shows that come can only occur if the infinitival sentence has a predicative role: (119a)
[far questojj e [ sc tj [ (come) rovinare tutto]] (do this is like spoil everything) b * [la peggior cosa]j e [Sc [ (*come) rovinare tutto] tj] (the worst thing is like spoil everything)
Thus, we can use come as a diagnostic for the predicative use of infinitival sentences with sembrare. Consider the following two cases:
208
The 'quasi-copula': the role of finite clauses in seem-sentences
(120a) b
mi sembra (*come) che Gianni parti (to-me seems like that John left) mi sembra (come) di volare (to-me seems like di fly)
The resulting contrast is clear enough: come can be applied to a ^/-sentence but not to a finite CP. The ungrammaticality of mi sembra come che Gianni parti is not surprising in our framework, given that the modifier come is wrongly applied to a CP that has the role of subject. On the other hand, the acceptability of mi sembra come di volare would appear to indicate rather clearly that the ^/-sentence in this case has the function of predicate or, perhaps, that of an argument of a predicate headed by an abstract come. This conclusion surely calls for some care, since we earlier established that clauses cannot be predicates. Nevertheless, our explanation is not in contradiction with this generalization, so long as the appropriate distinction is made. The clause under discussion is infinitival and works like a DP. In fact it can also take a definite article: (121)
il volare (the fly-infinitival)
An overt case of the predicative use of an infinitival clause is given in the following (122)
partirej e [Sc U morire] (leave-infinitival is die-infinitival) leaving is like dying'
This sentence is a canonical copular sentence - witness for example the possibility of cliticizing the postverbal material by lo (partire lo e). 36 All in all, as a first approximation, we might refine the representation of mi sembra di volare by assigning it the following structure: (123)
[,P proj sembra [sc ts [CP di PRO volare]]]
Thus, the verb sembrare still selects a small clause. The crucial difference with respect to the case involving a finite CP is that the embedded clause is not playing the role of subject; it can be analysed as the complement of an abstract predicate corresponding to Italian come (cf. (120b)) or as a predicate itself. Interestingly, it appears that the English translation of this sentence would not involve seem but rather a copular construction containing the element like, which is commonly assumed to be the equivalent of Italian come:
Infinitival clauses with seem (124)
209
[IP it; looks [sc t; [like [CP PRO flying]]]]
Here the predicative role of the CP is obvious, although in fact mediated by like. The meaning of the two sentences in the two languages can then be paraphrased as follows: this situation resembles the act of flying. In fact, it seems reasonable to allow Italian sembrare two separate lexical entries, which are completely distinct in English, namely seem (proper) and resemble and we could then say that sembra di volare is covered by resemble as in it resembles flying. Failure to distinguish these two entries can obscure the data, suggesting that Italian sembrare (proper) behaves differently from seem in that it allows infinitival smn-clauses which do not involve subject raising {sembra di volare) unlike English (* it seems to fly). To support the conclusion that sembrare can also mean 'look like' or 'resemble' there is an independent piece of evidence from Binding theory (reproduced from Moro 1988). Copular sentences show that the local domain for binding is sensitive to the predicative vs argumental nature of a DP. Consider the following case: (125a) * Maria Elena; e [Sc tj [DP la sua; maestra]] (Maria Elena is her teacher) b Maria Elena; vide [DP la sua; maestra] (Maria Elena saw her teacher)
A possible way to interpret these facts comes from the notion of Complete Functional Complex (Chomsky 1986b; see also section 1.4.2 of this volume). Simplifying the definition, we can regard a Functional Complex of a head X° as complete if it contains all grammatical functions and 6-relations pertaining to that head. Crucially, a pronoun must be free within the Complete Functional Complex of its governing category. Let us now consider the two sentences. The relevant Functional Complex for the pronoun sua (her) is that of maestra (teacher), the governing category. If maestra is a predicate, as in the former sentence, the clausal subject will be included in the relevant Complete Functional Complex; if it is argumental, as in the latter case, the Complete Functional Complex will not include the clausal subject; thus, we correctly predict that the pronoun, which is contained in the DP headed by maestra, can be bound by the clausal subject only in the latter case. The following example clearly shows that the DP which is left in the small clause governed by sembrare is not a predicate: (126)
Maria Elena; sembra [Sc t; [DP la sua; maestra]] (Maria Elena sembra her teacher) 'Maria Elena looks like her teacher'
210
The 'quasi-copula': the role of finite clauses in seem-sentences
We must conclude that in this case the noun phrase headed by maestra is not a predicate, but rather an argument of the verb sembrare in the sense of 'resemble'. Thus, coreference is allowed. Clearly, this analysis is not exhaustive. Many questions concerning these constructions remain to be answered. For example, it seems that pro and it behave like quasi-arguments, in the sense of Chomsky (1981). In Italian, for example, a phonetically realized subject is always banned, exactly as in the case of meteorological verbs: (127a) b
(*esso) piove (it rains) (*esso) sembra di volare (it resembles di fly)
Moreover, it is not clear why fo-cliticization of the predicative element is impossible: (128a) mi sembra di volare (to me looks-like di fly) b * me lOj sembra tj
(to me it looks-like)
Nevertheless, the aim of this section has been achieved. We can conclude that the explanation we gave for the fact that infinitival clauses (/br-clauses) are excluded from subject position in English smn-constructions can be maintained. There is no contradiction between these cases and the Italian ones we have just examined, given that when sembrare is followed by an infinitival sentence it is simply not the counterpart of English seem. Rather, it is the equivalent of resemble/look like, which have very different properties and which call for further analysis. Thus, if this analysis proves tenable, we can assume that in all cases the complement of seem is a small clause. 4.5
Conclusions
In this chapter we have analysed the syntax of seem. All in all, the traditional terminology which treats it as the 'quasi-copula' has turned out to be very insightful. We started by considering the relation between this verb and the subordinate finite clause that can occur with seem in sentences like the following: (129)
it seems that John left
Traditionally, it is assumed that this type of sentence shares the same
Conclusions 211 structure as passives like it was affirmed that John left. More specifically, the postverbal clause is assumed to be the object of the matrix verb and a semantically null element, i.e. it, is supposed to be inserted in the position of subject of the predication, in order to satisfy a general requirement: (130)
seems
We then proceeded to deal with a long-standing question raised by this analysis, namely why it is that the embedded sentence can be raised to subject position (in complementary distribution with /^-insertion) with passives but not with seem: (131a) * [s that John left]; seems ix b [s that John left]} was affirmed tj We showed that the analogy with passives must be abandoned in favour of a more articulated analysis that reveals a deep analogy with the copula. Two independent facts played a central role. On the one hand, we made use of the well-known fact that seem can select a small clause, as in John seems sad: (132)
seems
On the other, we showed that the pronominal it can perform the role of a propredicative element in copular sentences with an embedded finite clausal constituent such as it's that John left, giving the typical configuration of inverse structures:
212 (133)
The ' quasi-copula': the role of finite clauses in seem-sentences S
IS
By combining these two facts, we first of all proposed that the small clause construction be extended to cover cases where seem is followed by a finite clause. The crucial idea is that in these cases spec-IP is occupied by a propredicative element which raises from the small clause exactly as happens in the independent case of inverse copular sentences: (134)
seems
The major aim of this chapter was thus realized. It was seen that the embedded clause cannot raise to take the place of it, as in the passive construction, because the pronominal element is playing the role of a propredicate. If it were deleted then this would amount to destroying the (predicate position in the) small clause. Subsequently, we explored some consequences of this analysis and showed that by adopting aflexibleclause structure (as proposed throughout the present work) a number of empirical facts can be treated more naturally and some theoretical assumptions can be simplified. We compared the syntax of seem with passives and obvious-type constructions and explored a number of different domains: an apparent exception to a verb-secondlanguage canonical pattern, an asymmetry in adverbial extraction and Super Raising phenomena. The last section of this chapter was presented as something supplementary - and not crucial - to our major goal. We investigated cases where the quasicopula occurs with infinitival clauses as in John seems to be sad and discussed
Conclusions 213 the traditional treatment involving the S'-deletion process. We concluded by proposing that the lexical entry of seem can ultimately be reduced to the selection of a small clause in all cases, thus removing the need for deletion processes. Crucially, we analysed infinitival IPs as small clauses and English to as a predicative marker for VPs paralleling the function of as for predicative nominals. Finally, following the analogy between English for and Italian di proposed by Kayne (1984), I explained why ^//-clauses can follow the quasi-copula in Italian while /br-clauses cannot in English. One conclusion, however, merits special attention. The very fact that there is independent evidence for assuming that predicative noun phrases raise to preverbal position is to be considered as indirect support for the theory of clause structure presented in this book. In fact, there are now three independent domains for which the notion of 'inverse clause structure' gives good results: copular sentences, existential sentences and those ^emsentences where the main verb is followed by a finite clause. In the last chapter of this book, this analysis will be extended to a further empirical domain.
5
A view beyond: unaccusativity as an epiphenomenon
5.1
On the notion of intransitive verbs
Traditionally, the class of verbs is divided into two subclasses: transitive verbs that can occur with a subject and an object; intransitive verbs that can only occur with a subject. Modern syntax has succeeded in refining this basic partition by showing that the class of intransitive verbs is not homogeneous. The main goal of this chapter is to consider what light might be thrown on this matter by the theory of clause structure presented in this book. As a preliminary to the discussion, we will summarize the empirical facts that led to a refinement of the notion of 'intransitive verb'. 5.1.1 Diagnostics for unaccusativity Perlmutter (1978) showed that the class of intransitive verbs in Italian needs to be split into two subclasses, and he did this by means of a number of purely empirical tests. In order to review these, we will refer to the following pair of examples: (la) b
molte ragazze telefonano (many girls phone) molte ragazze arrivano (many girls arrive)
The two intransitive verbs telefonare and arrivare are distinguished with respect to two properties. First, they select a different auxiliary (respectively, avere and essere: (2a-b), and second, only in one case does the past participle agree in gender and number with the subject ((2c-d)): (2a) b c d 214
molte ragazze hanno/*sono telefonato (many girls have/are phoned) molte ragazze sono/*hanno arrivate (many girls are/have arrived) molte ragazze hanno telefonat-o/*-e (many girls have phoned-neuter/-fem.plur.) molte ragazze sono arrivat-e/*-o (many girls are arrived-fem.plur./-neuter)
On the notion of intransitive verbs
215
A further test (generally considered as a straightforward diagnostic) is provided by /te-cliticization. A well-established generalization holds that the pronominal clitic ne can be extracted from a postverbal noun phrase only when that noun phrase is the object of the verb: (3a) b
i ragazzi videro [DP molte ragazze] (the boys saw many girls) i ragazzi ne; videro [DP molte tj] (the boys of-them saw many)
In fact when the noun phrase is in a preverbal subject position ((4b)) or plays the role of an adverbial adjunct ((4d)) the same process is not allowed: (4a) b c d
[DP molte ragazze] lessero il libro (many girls read the book) * [DP molte tj] nej lessero il libro (many of-them read the book) i ragazzi chiamarono [DP molte volte] (the boys called many times) * i ragazzi nej chiamarono [Dp molte tj] (the boys of-them called many)
This generalization can be derived from the configuration. Focussing on the relevant segments, the three types of extraction (from the object, the subject and the adjunct position, respectively) can be represented as follows: (5)
a
V1
b
IP
c
IP
V° (Dg) (Dg) I' IP (DF From this premise, the possibility of extracting ne from the object position follows on independent assumptions. Relying for example on Belletti and Rizzi (1981), one can assume that only in the first case will extraction meet the locality conditions on movement which require the presence of a lexical head in the proper configuration. In this case, only V° qualifies as such an element. Let us now turn back to intransitive verbs. In a pro-drop language such as Italian, the subject is free to precede or follow the verb: (6a) b c d
[DP molte ragazze] telefonano (many girls phone) [Dp molte ragazze] arrivano (many girls arrive) telefonano [Dp molte ragazze] (phone many girls) arrivano [Dp molte ragazze] (arrive many girls)
216
A view beyond: unaccusativity as an epiphenomenon
If we now apply the ne-extraction test, a further sharp distinction between arrivare and telefonare will be seen to emerge:1 (7a) b
* nej telefonano [DP molte tj] (of-them phone many) nej arrivano [DP molte tj] (of-them arrive many)
An interesting pattern emerged from these data: the verbs which allow necliticization from the postverbal subject coincide with those selecting the auxiliary essere and displaying past participle agreement. This could hardly be considered as purely accidental. In fact, further empirical tests (dating back to the first pioneering works in the field; see Perlmutter and Postal 1984, Rosen 1981 and references cited there) provided evidence that the empirical distinction between intransitives holds across languages.2 The question then arose of how to explain this regularity. This led to the formulation of the Unaccusative Hypothesis, which states that the subject of a certain class of intransitives has the same syntactic relation with the verb as the object of a transitive verb. Whenever an intransitive verb satisfies the three properties it will be called 'unaccusative'; otherwise, it will be called 'unergative'.3 5.1.2 A representation of unaccusative constructions The Unaccusative Hypothesis has been implemented in the principles and parameters framework by Burzio (1986). The two basic example verbs telefonare and arrivare can be straightforwardly distinguished as follows: (8a)
V°
telefonano
molte ragazze
On the notion of intransitive verbs 217
pro
VP
arrivano molte ragazze
Crucially, the postverbal DP is in a different position with the two intransitives. With a verb like arrivare it is base generated in the object position. On the other hand, with a verb like telefonare it is rather generated outside VP and then rightward-moved to an adjunct position.4 This minimal difference interacts with independent principles to capture the whole cluster of asymmetries that distinguish arrivare-typc verbs from the telefonare-type. The original formulation of the principles given in Burzio (1986: 30 ff.) is reproduced here as follows: (9i) ii iii
A past participle will agree (in gender and number) with an element holding a 'binding relation^ with its 'direct object'5 The auxiliary will be realized as essere whenever a 'binding relation/ exists between the subject and a 'nominal contiguous to the verb'6 ne-cliticization is possible with respect to all and only direct objects
Clearly, this theory does not admit 'intermediate situations'. The set of principles formulated here are crucially based on the fact that the subject of an intransitive is either in the object position or not, and those are the only two possibilities available. This was generally considered a welcome result. In fact, the theory predicts that either the three properties are simultaneously instantiated in a structure involving an intransitive verb or else none of them is. Again, there can be no intermediate situations. We will come back to this point in the next section. First, however, it should be highlighted that the representation standardly given to unaccusatives has important consequences for the theory of clause structure. Consider, for example, the following pair which represents a typical unaccusative verb in Italian and in English:
218
A view beyond: unaccusativity as an epiphenomenon
(10)
S
pro there
molte ragazze many girls
The intuitively unitary notion of subject appears to be split here into two distinct positions. The subject as a grammatical function appears to be realized in the most prominent position by inserting an expletive (pro, there). The subject as an element bearing a 6-role assigned by the verb, on the other hand, is realized in the object position.7 In fact, this split is simply the formal consequence of independent assumptions. Recall that in the current theory of clause structure grammatical functions are rigidly identified with specific positions (in particular, it is assumed that the subject of predication is always realized in the most prominent position). Thus, the postverbal noun phrase of an unaccusative structure simply does not qualify as the proper candidate for the grammatical function of subject of predication. The aim of this chapter is to see whether the flexible theory of the clause structure we have proposed to account for the syntax of copular, existential and quasi-copular sentences can be extended to the domain of unaccusative constructions. The rest of this chapter is organized as follows. In the next section (5.1.3) it will be seen that the current theory of unaccusativity is too restrictive. The diagnostics for unaccusatives give in fact positive results with d-sentences (but crucially not with all inverse copular sentences) even if the subject is not assumed to be in the object position. The last section (5.2), on the other hand, will be highly speculative. We will address some major questions related to unaccusativity. In particular, we will explore the possibility of giving a unifying account of d-sentences and unaccusatives by extending the analysis given for inverse copular sentences to the latter type of constructions. Finally, we will explore some conceptual consequences of our proposal which are related to the representation of 6-roles (5.2.1), to the difference between essere and avere (5.2.2) and to Case theory (5.2.3). 5.1.3 Esserci as an unaccusative verb In the first chapter, we analysed copular sentences and showed that, along
On the notion of intransitive verbs
219
with the subject DP, the predicative DP can also be raised to the most prominent position yielding what we called inverse copular sentences'. Thus a minimal pair like the following: (lla) b
tre foto del muro sono la causa della rivolta (three pictures of the wall are the cause of the riot) la causa della rivolta sono tre foto del muro (the cause of the riot are three pictures of the wall)
can be represented respectively as follows: (12a)
S
be
DP
Then, in the second chapter, it was shown that c/-sentences also belong to the class of inverse copular sentences: (13)
S
pro
VP
V° / ^ cij V° DP
SC
220
A view beyond: unaccusativity as an epiphenomenon
The fundamental property of this type of inverse structure is that the subject trefoto (three photos) is left in situ within the small clause constituent which is adjacent to the copula while the clitic propredicative element ci is raised to preverbal position from the position where predicates are generated within the small clause. Since ci is a clitic, the most prominent position is occupied by an expletive pro* We will now apply the tests normally used to diagnose unaccusative constructions to a construction involving esserci: (14a) b c
ci sono/*hanno state tre foto (there are/have been three pictures) ci sono stat-e/*-o tre foto (there are been-fem.plur./-neuter) ce nej sono state [DP tre tj] (there of-them are been three)
The result leaves little doubt. As far as present issues are concerned, there is no empirical difference between esserci and arrivare: they both take only one DP argument (i.e. they are 'intransitives'); they select the auxiliary essere; the past participle agrees in gender and number with the subject and ne-cliticization is possible from postverbal position. The theoretical problem at stake here is quite clear. If one of our basic aims is to use terminology consistently we have no choice but to consider esserci as an unaccusative verb. Nevertheless, we have established that the subject of esserci is not in the object position, where the subject of arrivare is assumed to be located, rather it is left in situ within the small clause.9 Focussing on the two relevant segments we can show the different position of the DP in an unaccusative and an esserci construction as follows: (15)
DP
How can we solve this problem? It seems that extending the class of unaccusatives to cover esserci on a par with arrivare amounts to undermining the Unaccusative Hypothesis itself. If a subject which is not in the object position yields a structure with the properties standardly thought to define unaccusatives, then there is no longer any need to assume that the subject of certain intransitives is in the object position.10 In reality, matters are even more complicated. What is crucially suggested
On the notion of intransitive verbs 221 by our theory of d-sentences is that the three properties under discussion are not the consequences of a single fact. Let us see why. In the first chapter it was shown that /^-extraction is not possible in inverse sentences with a full DP predicate as a consequence of the Subjacency Condition. Recall a simplified representation: (16)
* [IP [le cause della rivoltajj [nej sono] [sc [DP tre tj] tj]] (the causes of the riot of-them are three)
Nevertheless, these same inverse copular sentences with full DP predicates do select essere as an auxiliary and their past participle agrees in gender and number with the subject: (17a) b
le cause della rivolta sono/*hanno state tre foto (the causes of the riot are/have been three photos) le cause della rivolta sono stat-e/*stat-o tre foto (the causes of the riot are been-fem.plur./been-neuter three photos)
All in all, these tests force us to the conclusion that if esserci is indeed to be considered as an unaccusative this is in reality just epiphenomenal, the result of a conspiracy of different factors, rather than the consequence of one single structural fact. The position of the subject is not per se sufficient to yield the whole cluster of phenomena associated with unaccusativity, namely ne-extraction, past participle agreement and selection of essere as auxiliary. In fact, inverse copular sentences with full DP predicates show that ^-extraction from the postverbal subject is independent of the two other properties that are used as diagnostics of unaccusative constructions, again past participle agreement with the subject and essere auxiliary selection. This conclusion casts legitimate doubts on the current theory of unaccusatives. An exhaustive approach to this topic would of course take us too far. Our more limited aim in the present work will therefore be to indicate certain lines of reasoning that are suggested by our approach to clause structure. The next section then will be highly speculative. Four different topics will be addressed: first, a unitary account of esserci and arrivare will be proposed, which involves extending the small clause analysis to unaccusatives (5.2); then, we will explore some consequences of such a proposal. The small clause structure will be shown to be predicted if we follow the spirit of the Hale-Keyser (1991) framework (5.2.1); a unified theory of essere and avere as main verbs will be proposed assuming a small clause analysis for possessive constructions (5.2.2); Burzio's generalization will be reconsidered proposing a possible derivation (5.2.3).
222 5.2
A view beyond: unaccusativity as an epiphenomenon A view beyond: unaccusative constructions as ci-sentences
We have seen in the previous section that a particular subclass of inverse copular sentences, namely d-sentences, fulfil the diagnostics for unaccusativity. We concluded that extending the class of unaccusatives to include esserci undermines the theory itself for this is crucially based on the assumption that the subject of unaccusatives is in the object position and that all facts derive from one single cause. One possible way of getting round this is to reverse one normal assumption and propose that unaccusativity is always an epiphenomenon in exactly the same way we have suggested it is in d-sentences. This will clearly mean revising the entire analysis of unaccusatives. This is in fact the line of reasoning we are going to pursue here. For ease of exposition, let us formulate our hypothesis clearly: (18)
The structure of an Italian c/-sentence is not different from the structure projected by an unaccusative verb
First of all we will attempt to make it clear in what sense the two structures are 'not different'. Here, we come up against our first main problem. The peculiarity of inverse sentences (including d-sentences) is that the subject stays within the small clause and that the most prominent position is occupied by a predicative DP. Thus, the natural step to take here is that of analysing arrivare as involving a small clause. The most obvious possibility - that of assuming that arrivare selects a small clause containing d as a predicate - must be refuted. Consider the following contrast: (19a) * (ci) sono molte ragazze (there are many girls) b
(ci) arrivano molte ragazze (there arrive many girls)
This contrast shows that while ci is obligatory with the copula, it is pleonastic with arrivare (in the intended sense).11 Thus, the following analysis for arrivare as involving ci cannot plausibly be adopted:
A view beyond: unaccusative constructions as ci-sentences (20)
223
S
arnvano The theory of copular sentences, however, offers a further possibility. I will now reproduce the analysis of a particular type of inverse sentence like sono io (it is me): (21)
S
This sentence showed that pro can play the role of a predicate. This suggests a rather natural way of implementing the implicit content of our basic conjecture. As a first approximation, let us just take pro to be the predicate which raises from the small clause selected by arrivare: (22)
molte ragazze
224
A view beyond: unaccusativity as an epiphenomenon
Here, the relation between the V° and the DP is not the same as that between a V° and an object. Rather, it is mediated by an intervening clause structure, the small clause, paralleling the case oiesserci. The DP following the verb is not in the object position; it is in the subject position of the small clause complement selected by arrivare.12 Notice that the split between the grammatical subj ect (i. e. the subj ect of the predication) and the thematic subject is reabsorbed: the subject is uniquely base generated in a single position inside the small clause. We no longer need to assume that the most prominent position isfilledby an expletive standing for the subject of the predication: in a sentence like arrivano molte ragazze, this position can befilledby a predicative pro, which is required for independent reasons to explain cases like sono io (see again chapter 1). From this point of view, the two classes of intransitives can be distinguished in a new fashion with respect to the traditional representation. Of course, there is nothing to say here about unergatives (or 'true intransitives', as they are sometimes called). For such constructions, we will simply adhere to the traditional representation (possibly refined by adopting the more recent assumption that the subject is generated in spec-VP position; see Koopman and Sportiche 1991, Hale and Keyser 1991 and references cited there). Before evaluating the possible advantages of our new proposal, we mustfirstof all show that the new analysis is able to explain at least the same range of facts as Burzio's (1986) theory. From this point of view, it should be recalled that the presence of the subject in object position together with the assumed coindexation with expletive pro in specIP position was sufficient to derive the entire pattern. Can we still derive these same properties from independent facts if we adopt the structure involving the small clause? Let us consider the three properties separately. First, as we saw in chapter 2,rae-extractionis possible not only from the object position; in fact, it is also possible from an embedded subject in cisentences: (23a) b
[IP pro [cej nej sono] [sc [DP molte tj] t{]] [IP proj [nej arrivano] [sc [DP molte tj] tj]]
In a d-sentence, ne-extraction from the subject of a small clause is possible as a consequence of the incorporation of the predicate ci\ witness the fact that extraction is blocked in inverse sentences with full lexical DP predicates (cf. (16)). In the next section independent evidence will be provided for assuming that arrivare also results from the incorporation of a lower predicate. Assuming that the analysis will prove tenable, ne-extraction would be predicted along the lines suggested for ci (see section 2.3.2).13
A view beyond: unaccusative constructions as ci-sentences 225 Second, according to Burzio's (1986) theory, a past participle will agree (in gender and number) with an element holding a binding relation with its direct object (cf. (9i)). The central assumption made by Burzio (1986), in fact, was that in unaccusative constructions the DP in the most prominent position (pro) does not have autonomous reference. This was formally expressed by coindexing it with the DP in object position. Under our analysis, this follows without stipulation as a consequence of the predicative relation instantiated between the two nominals within the small clause (see Williams 1980 for a detailed proposal regarding coindexing the predicate with the subject). However, if we adopt the new representation, the verb is no longer 'contiguous to a nominal'. Thus, we must appeal to an alternative explanation which can be provided as follows. Recall that in a copular sentence raising to the most prominent position activates the agreement features contained in the copula (under a spec-head relation). This is very clear in English: (24a) b
[IP [DP these girls]j are; [sc tj [Dp the cause of the riot]]] [IP [DP the cause of the riot]j isj [sc [DP these girls] tj]]
Thus, since in our proposal a structure involving arrivare also involves extraction from a small clause, the obligatoriness of past participle agreement can be regarded as a consequence of a more general phenomenon related to raising: (25)
prOj sono [AgrP tj arrivat-ej [Sc [DP molte ragazze] tj]] (pro are arrived-fem.plur. many girls)
The predicative DP (i.e. pro) contained within the small clause is first raised to an intermediate spec-position where it triggers agreement on the past participle,14 and it is then further raised to spec-IP. Third, essere assignment was explained by assuming that this auxiliary is chosen 'whenever a binding relation exists between the subject and a nominal contiguous to the verb' ((9ii)). This fact can still be derived if we assume the small clause analysis. We know on independent grounds that essere is selected in raising constructions such as those involving sembrare or the passive of befove-type verbs, and not for instance in control structures, where avere is allowed: (26a) b c
Giannij e/*ha sembrato [tj essere [tj la causa della rivolta]] (Gianni is/has seemed to be the cause of the riot) Giannij e/*ha ritenuto [tj essere [tj la causa della rivolta]] (Gianni is/has believed the cause of the riot) Gianni e/ha dovuto [PROj essere [tj la causa della rivolta]] (Gianni is/has due to be the cause of the riot)
226
A view beyond: unaccusativity as an epiphenomenon
Thus, since we are assuming that arrivare-typc constructions also involve raising, essere assignment follows as a more general property. All in all, we can conclude that the new representation suggested here is (at least) empirically equivalent to the classic one given by Burzio (1986). The three fundamental empirical properties which distinguish the two classes of unaccusatives can be derived from independent facts in this framework as well. Notice that the explanation we gave for the three properties assumed with unaccusative constructions does not apply to the subject of telefonano molte ragazze: the postverbal subject is not L-marked (thus, fte-cliticization is impossible) and there is no raising (thus, the auxiliary is not essere and the past participle does not agree). Nevertheless, the advantages of assuming the new structure should be transparent: the theory can now explain in a unitary way why esserci and arrivare have analogous properties and why they are different from telefonare. From a technical point of view, then, the hypothesis we proposed would appear to be correct. The structure of an Italian d-sentence is no different from the structure projected by an unaccusative verb: in both cases the postverbal subject is left in situ in the small clause complement of the verb and the latter is fully lexical. However, one fundamental empirical question still remains unanswered: what is it that tells the child acquiring the language that a small clause construction cannot be assigned to telefonare, as opposed to arrivare? Naturally, it is not only in our framework that this question arises. Mutatis mutandis, a totally equivalent question is raised in the traditional framework as well: how does the child know that the subject of arrivare is in the object position and the postverbal subject of telefonare is not? In the next section we will attempt to answer this type of question. 5.2.1 Knowledge of 6-roles: the Hale-Keysef s project The crucial piece of information that makes it possible to distinguish between unaccusatives and unergatives is traditionally assumed to come from the mapping between ©-theory and X'-theory (see Marantz 1984, Burzio 1986, Baker 1988, Grimshaw 1990, Jackendoff 1990, Williams 1994 among others). Universal Grammar contains an inventory of 0-roles like , <patient>, , etc., expressing the prototypical relations linking the arguments of a verb. These 0-roles are not freely assigned, they are rather in a one-to-one relation with X'-positions according to the so-called Uniformity of Theta Assignment Hypothesis (UTAH) (see Baker 1988: 47ff.).15 Let us illustrate this with a simple case:
A view beyond: unaccusative constructions as ci-sentences (27)
227
S
i nemici V
distrussero molte navi A verb like distruggere (destroy) can assign two distinct 6-roles: call them and <patient>. These two 6-roles can only be assigned in one way: the is assigned to the external position (the 'logical subject' in Marantz's 1984 terminology) as opposed to the <patient> which is assigned to the object position: (28a) b
i nemici [VP distrussero [molte navi]<patient>] (the enemies destroyed many boats) * i nemici<patient> [VP distrussero [molte navi]] (the enemies destroyed many boats)
The UTAH, then, is a principle that determines the mapping between 6roles and syntactic positions. In the example above, we find that and <patient> are associated with the following two positions, respectively: (29)
Consider now the two verbs we have been using as examples in our discussion: (30a) b
pro telefonarono molte ragazze (pro phoned many girls) pro arrivarono molte ragazze (pro arrived many girls)
Here, 'the UTAH can be used to guide the construction of analyses - both by the linguist and by the child-in a non-trivial way' (Baker 1988: 47). Thus it is assumed that the fact that molte ragazze is assigned the 6role in the sentence involving telefonare and the 6-role in the sentence involving arrivare would be sufficient in itself for molte ragazze to be assigned to the two different structural positions that are consistent with the syntactic tests:
228
A view beyond: unaccusativity as an epiphenomenon
(31a)
S
arrivano molte ragazze
In the first case ((31a)), molte ragazze is moved from the external position where is normally assigned (occupied here by pro). Instead, in the second case ((31b)) molte ragazze is generated directly in the object position which we independently know to be the only position where can be assigned (on a par with <patient>). This is essentially the way that the UTAH works in 'guid[ing] the construction of [the] analyses'. Let us now turn to the analysis we proposed for unaccusative constructions: (32a)
pro
VP
V°
telefonano
molte ragazze
A view beyond: unaccusative constructions as ci-sentences 229 S
molte ragazze On the face of things, the argument which appeals to 0-theory in order to explain how we differentiate between the two classes of intransitives should not be available here. Indeed, it is by no means the case that the subject of a small clause is necessarily assigned the role of . Why then should a child choose such a complex structure for arrivare and not for telefonarel I would like to suggest two independent ways of arguing in favour of the new structure. The first is based on a framework stemming from the original proposal of Hale and Keyser (1991). The second is based on the analysis of d-sentences we proposed in chapter 2. The central idea of the Hale and Keyser (1991) project is that 'there are no thematic roles. Instead, there are just the relations determined by the categories and their projections, and these are limited by the small inventory of lexical categories and by the principle of unambiguous projection' (1991:40). For example, the label '' applied to a DP does not refer to an autonomous entity. Rather, it simply coincides with the syntactic relation that the DP bears to the other lexical elements as mediated by the X'-structure, specifically that of specifier of VP (see Hale and Keyser 1991: 39-40).16 Following the spirit of this new theory of 6-roles we can approach our problem from a rather different perspective. Although there is no general agreement on the terminology, it is commonly assumed that unaccusatives express a 'change of state', in a broad sense. More specifically, they refer either to 'change of location' {arrive, go, run, descend, come, etc.) or to 'change of condition' {improve, worsen, increase, diminish, fatten, cool, etc.). The semantic distinction between the object of transitives and the subject of unaccusatives is generally captured by using two distinct labels for the 0-roles: the 0-role assigned to the object of transitives is called '<patient>' as opposed to the one of unaccusatives which is called ''.
230
A view beyond: unaccusativity as an epiphenomenon
This semantic intuition can be used to support our theory. In the Hale and Keyser (1991) framework, the 6-role we refer to as 4 ' is regarded as associated with a particular configuration allowed by X'-theory which involves the incorporation of a lower predicate into a higher verb position. Although an accurate illustration of this theory would require a much more detailed treatment, it seems to me that one can grasp its essential content by considering the following simple case (see Hale and Keyser 1991: 44 for a detailed discussion): (33)
the cook thinned the gravy
In this sentence, there is an agent (the cook) and a theme (the gravy) which undergoes a change of status (it becomes thin). The corresponding base generated syntactic structure is represented as follows: (34)
VP the cook
thin
The verb thin which surfaces in the sentence the cook thinned the gravy is thus the result of a lexical incorporation of the A° thin, which is base generated in the lower VP. This element moves head-to-head up to the higher V° position as illustrated below:
A view beyond: unaccusative constructions as ci-sentences 231 (35)
VP
the cook
V
Here is the essence of Hale and Keyser's (1991) proposal: the gravy is interpreted as the theme in this structure but this is not a role assigned by a verb. Rather, its interpretation is associated with the specific configuration the gravy is in, namely the spec-VP position of a VP complement of a V°. The 9-role labelled is simply nothing but the following configuration: (36)
Similarly, the cook is interpreted as the '' in the cook thinned the gravy as a consequence of the fact that the structure [v> V° VP] expresses the 'causal relation'. If the higher spec-VP position is empty, the lower DP would ultimately be raised to spec-IP yielding a sentence like the following (Hale and Keyser 1991: 59): (37)
the gravy is thinning nicely
This sentence is essentially interpreted in the same way as the cook thinned the gravy, with the major difference that here the agent (the cook) is missing.
232
A view beyond: unaccusativity as an epiphenomenon
Although this framework cannot be immediately adopted here, 17 we can follow it as a guideline to answer the major question addressed. The idea that we will implement is that the role assigned to the subject of verbs which express change of state is a mere label for the structure involving the small clause. Thus, arrivare is to be analysed in the same way as thin, that is, as the result of an incorporation of a lower predicate into a higher verb. The line of reasoning is essentially clear, but the technical devices needed to implement such a proposal are somewhat harder to identify. However, as a first approximation, we might take arrivare to be the result of a lexical incorporation of a verb V° with the head of a lower locative predicate:18 (38) VP spec-VP
molte ragazze
loc.
Thus, V° selects a small clause containing a subject {molte ragazze) and a locative predicate (loc.) This raises to incorporate into V° yielding the overt form arrivare. A similar analysis can be given to those unaccusatives which express 'change of state': the only difference is that in this case the embedded predicate would not be a locative but rather the expression of a quality.19 We can leave the technical details aside: for our limited purposes it is sufficient to notice that within the guide lines suggested by Hale and Keyser (1991) the question addressed in this section can receive a plausible answer. The speaker knows that a small clause is involved in the case of arrivare, because he knows that the specific 6-role assigned to the argument of such a verb, namely the , is in fact associated with the incorporation of a lower predicate into a higher V° projection.20 As was said before, however, this analysis of unaccusatives as involving a small clause also suggests that some independent piece of evidence must be available to the child acquiring the language (and to the linguist) which allows him or her to detect the difference between this structure and unergatives. We can adopt the following line of reasoning. The role is clearly (perhaps prototypically) associated with the subject of a cisentence as in, say, cisono moltiproblemi (there are many problems). This
A view beyond: unaccusative constructions as ci-sentences 233 is particularly important because, as was proposed in chapter 2, d-sentences crucially involve raising of a predicate to a larger VP: (39)
Thus, we can simply conclude that the child has no alternative but to adopt the minimal assumption, namely that such a transparent configuration applies to all cases involving the role, including, of course, arrivare. If this were not the case, then this would mean that the same role is associated to two different structural positions, clearly an unwanted result. Thus, the only difference between esserci and arrivare turns out to be that, while the process of incorporation is morphologically transparent with esserci, it is not similarly transparent in the case of arrivare.21 All in all, the theory based on (a version of) the UTAH can still be maintained in our framework. Moreover, the hypothesis that unaccusatives involve an embedded clausal constituent has some non-trivial consequences also for the predicational architecture of clause structure. It is a well-known fact that certain unaccusatives have transitive counterparts. Burzio (1986) explains this alternation by assuming that the lexical entry of a verb contains the specification ± 6S (i.e. ± subject 6role). If the lexical element comes with the + 6S feature, then the external argument occurs; if it comes with the —0S feature, instead, the external argument is omitted. The following simple structure would be associated with a verb like affondare (sink): (40) S (DP)
V°
DP
If the value -6S is chosen, an unaccusative construction is produced. In such a case, a lexical DP could not occur in the external position, because it would receive no 9-role. Otherwise the verb is used as a transitive:22
234 (41a) b
A view beyond: unaccusativity as an epiphenomenon Gianni+es [VP ha affondato [DP molte navi]] (Gianni has sunk many boats) pro_Hv [Vp sono affondate [DP molte navi]] (pro are sunk many boats) 'many boats sank'
A relevant observation concerning this matter is made by Chierchia (1989) following a comment by Carol Rosen. He noticed that 'unaccusatives tend to be unstable. They tend to oscillate in valence from transitive to intransitive and vice versa, both diachronically and across dialects . . . one finds transitive uses even of verbs like morire (die), ribellare (uprise), suicidarsi (commit suicide) etc. In contrast unergative verbs like sudare (sweat), piangere (cry), mangiare (eat) do not seem to undergo these shifts in valence either diachronically or across dialects' (1989: 23).23 Within our approach a natural explanation of this phenomenon of 'unstable valency' is available. Unaccusative constructions involve an embedded small clause. The presence of a subject in this 'nuclear' clause structure is sufficient to fulfil the principle of Full Interpretation which requires sentences to contain a subject and a predicate (see Chomsky 1981, 1988a). Of course, this does not prevent the structure from including a second argument. However, this is now parallel to the following contrast: (42a) b
the cook thinned the gravy the gravy thinned
Thus, in certain geographically limited varieties of Italian a construction involving arrivare can include an external argument.24 This will be interpreted compositionally as the agent of a certain change of state. Consider, for example, the cases in (43), involving scendere (lower), and those in (45), involving prendere (take) and ridere (laugh): (43a) b
sono scese molte borse (are lowered many bags) 'many bags lowered' Gianni ha sceso molte borse (Gianni has lowered many bags)
(grammatical in many dialects)
When no external argument is inserted, as in (43a), the verb scendere yields an unaccusative construction with the meaning 'many bags lowered'. On the other hand, if a further argument is inserted as in (43b) which is grammatical in many southern Italian dialects, it is interpreted as the 'causer' of the change of state and the sentence would mean 'Gianni caused many bags to lower'. Crucially, for these cases involving incorporation of a lower predicate into the matrix verb, there is no need for the second argument;
A view beyond: unaccusative constructions as d-sentences
235
rather, there is full optionality. We can represent this by means of the following structure where pred. is the predicate to be incorporated into the matrix verb (cf. (38) and the related discussion): (44) (DP)
pred. With transitives and unergatives, instead, the situation is completely different. Consider now the case of prendere (take) and ridere (laugh): (45a) b
Gianni ha preso molte borse (Gianni has taken many bags) Gianni ha riso (Gianni has laughed)
Here, the external argument is required (possibly supplied by movement in passives) because these verbs do not select an embedded clausal constituent. Thus, omitting an external argument with transitives and unergatives would amount to omitting a subject of the predication (cf. 44): (46) •(DP) V° (DP) Although this issue cannot be pursued here, it should perhaps be noticed that this analysis implicitly assumes that a predicative relation can only hold between two maximal projections which c-command each other. In fact, this simply amounts to assuming (some version of) Williams' (1980) theory of predication. 25 All in all, we can conclude that the analysis given here of arrivare-type constructions as involving a small clause constituent not only covers the same empirical facts as the classical analysis but also seems to provide a rather natural explanation for two such entirely different phenomena as 'unstable valency' and the mapping between 0-roles and positions. In the
236
A view beyond: unaccusativity as an epiphenomenon
next section, we will approach a rather different phenomenon, attempting to relate essere (be) and avere (have) in a novel fashion. 5.2.2 Esser(ci) vs. aver(ci): a minimal difference It has so far been proposed that an unaccusative verb is to be analysed in the same way as esserci, that is, as involving a small clause. It has further been proposed that the external argument is optional because the principle of Full Interpretation is already satisfied by the subject contained within the small clause complement. The obvious question now is: can an external argument be inserted into a construction involving cil Or to put it in more formal terms, can a lexical DP be inserted into spec-VP in the following structure? (47)
essere DP ci On the face of it, the answer seems to be negative. Let us take for example a small clause containing a ci predicate and two arguments like [DP molti libri] (many books) and [DP i ragazzi] (the boys) and combine them as follows: (48a)
pro [cij sono] [sc [DP molti libri] t{] (pro there-are many books) b * i ragazzi [ci§ sono] [sc [DP molti libri] tj] (the boys there-are many books)
When the external argument is inserted as in the second case the sentence is completely ungrammatical. It would appear then that we should just conclude that an external argument cannot be inserted in a d-sentence.26 However, I propose to refine our understanding of d-sentences somewhat and of clause structure in general by examining cases like the following: (49)
i ragazzi [ci hanno] [DP molti libri] (the boys there-have many books)
This example deserves careful attention. The occurrence of ci with avere is totally unexpected in the present framework. If ci is regarded as the place holder for the subject position, how is it that it can occur with a subject?
A view beyond: unaccusative constructions as ox-sentences 237 Moreover, if avere is traditionally analysed as a transitive, why can it occur with d, in contrast with all other transitives? I would like to suggest here that the analysis we proposed for ci as a propredicate can be fruitfully exploited to refine our understanding of avere as well.27 As a preliminary step, notice that the distribution of ci with avere in Italian is pervasive indeed. Although ci can be omitted at a high stylistic level ((50a)), it becomes almost unavoidable when the object of avere is cliticized ((50b)), it is present in many Northern Italian dialects ((50c)); taken from the dialect spoken in Pavia), but crucially it cannot occur with transitives other than avere ((50d)) even if they are synonymous as in the case of possedere (possess):28 (50a) b c d
i professori (c')hanno molti libri (the professors there-have many books) 'professors have many books' i professori ??(ce) lij hanno tj (the professors there-them-have) 'professors have them' i prufesur a *(gh') ij han tj (the professors subj-cl.-there-them-have) i professori (*ci) possiedono molti libri (the professors there-have many books) 'professors have many books'
Basically, the hypothesis that I would like to propose here (essentially reproduced from Moro 1991) is that the two sentences like ci sono molti libri (there-are many books) and c'hanno molti libri (there-have many books) do have the same abstract underlying structure: in both cases the verb selects a small clause where the clitic ci figures as the predicate. In formal terms, the difference between a clause containing avere and one containing essere is the following: (51)
avere DP
This first step seems to be obligatory. How else could one account for the distribution of ci with avere if ci were not playing the role of a propredicate
238
A view beyond: unaccusativity as an epiphenomenon
raised from a small clause? A further and deeper question would be whether the information regarding these two structures is simply codified in the lexicon, as idiosyncratic properties of avere and essere, or whether the occurrence of these two verbs with ci can be traced back to some more principled difference. I would like to pursue the second line of reasoning and propose that the appearance in a given structure of either essere or avere does not depend on a lexical choice. Instead, the actual realization of V° depends on the presence of the external argument; thus V° is realized as essere if the external argument is absent, as avere otherwise. Seen in this light, the difference between essere and avere becomes very 'superficial'. This is hardly a new proposal in linguistics. The classic work of Benveniste (1966), for example, pursued the same hypothesis, from a different empirical perspective and towards different goals.29 The relevance of the contrast between, say, c'e un gatto (there-is a cat; 'there is a cat') and c'ha un gatto (there-have a cat; 'he has a cat') should be transparent. They show that the choice between essere and avere can be traced back to a minimal structural difference, namely the presence/absence of an external argument. In a sense, one can consider avere as the form the 'copula' displays when two arguments are inserted as opposed to only one. One interesting way of deriving this hypothesis is given by Chomsky's (1988a, 1993) theory of Case assignment. The essential idea is that Cases are not autonomous entities. A Case is a spec-head relation between a DP and an Agr° to be checked (at least) at the LF interface: (52)
AgrP
Furthermore, since all Agr°s are instances of the same entity, the difference between various Cases (traditionally, Nominative, Accusative, etc.) would have to be traced back to the elements that combine with Agr°. For example, Nominative would be associated with [Agr° T°-Agr°] and Accusative would rather be associated with [Agr° V°-Agr°]. The following representation is the one given to a clause containing a transitive active verb of the type DP - V - DP (irrelevant spec-positions omitted):
A view beyond: unaccusative constructions as d-sentences (53)
239
Agr-sP
V° DP On the other hand, we showed in the first chapter that in sequences of the kind DP - V - DP, where V° is the copula, only one Agr° occurs: (54)
AgrP
Recall that the latter assumption was made to explain three independent facts concerning copular sentences (see sections 1.3.2. and 1.3.3): the appearance of only one Case in languages with overt morphological realization of Cases (like Latin); the impossibility for the copula to cooccur
240
A view beyond: unaccusativity as an epiphenomenon
with inflected clitics in Italian; the apparent lack of subject-object asymmetries in inverse copular sentences. From an abstract point of view, then, we concluded that the number of Agr°s is a function of the number of argumental DPs. If two DPs are related by a predicative relation, as in fact happens in inverse copular sentences, only one Agr° may occur. Bearing this in mind we can turn back to the minimal difference between esserci and averci. If we are correct, we must assume that with avere two Agr°s are required, as opposed to the case of essere. This is the required endowment for Case-theoretical reasons, since in the first case two argumental DPs occur as opposed to what happens in the second one, where only one single argumental DP occurs. A straightforward way to test this hypothesis comes from cliticization phenomena. Recalling again the relation between cliticization and the presence of Agr° discussed in chapter 1, the immediate prediction is that cliticization is possible with averci as opposed to esserci. Facts show this prediction to be correct: (55a) b
c
d
c'e un gatto (there-is a cat) 'there is a cat' c'ha un gatto (there-has a cat) 'he/she/it has a cat' * ce l'j e tj
(there-it-is one) ce l'i ha tj (there-it-has one) 'he has it'
As further support for this analysis it should be recalled that the lack of cliticization cannot be explained by appealing to some specific locality condition applying to the DP and blocking movement tout court. In fact, neextraction, which involves subextraction from the DP, is fully acceptable in both cases: (56a) b
ce n'j e [uno tj] (there-of-them-is one) 'there is one of them' ce n'j ha [uno tj (there of-them-has one) 'I have one of them'
The general conclusion that I would like to draw here is that it is not implausible that the difference between avere and essere might not be
A view beyond: unaccusative constructions as ci-sentences 241 directly codified in the lexicon. Rather, it seems that these two verbs should be regarded as the lexicalization of the two possible structures involving the same functional element, Agr°. More explicitly, one can regard essere as the support for the V° which incorporates only one Agr° and avere as the support for the V° which incorporates two Agr°s.3() Interestingly, the proposal made here can be considered as an alternative line of research to the independent proposal of Kayne (1993), who relies on Szabolcsi (1983) and Freeze (1992). Both theories reach the same conclusion, namely, that avere and essere (and their equivalents across languages) should be regarded as epiphenomena. The path they take towards the conclusion, though, is different in each case. The essence of Freeze's idea is that avere should be analysed as an instance of essere into which an abstract preposition, P°, has been incorporated from a lower PP.31 In the theory proposed here, which relies on the unified theory of copular sentences, the difference is traced back to the presence of one vs two Agr° projections. Thus, the difference between essere and avere would simply turn out to be 'quantitative', as it correlates with the presence of one vs two instances of the same entity, namely Agr°. Assuming that the theory proposed here is well founded, an interesting line of research would then be to evaluate these two proposals. Although this goal cannot be pursued here, it is perhaps worth pointing out that diachronic syntax would offer a particularly interesting empirical domain for this evaluation.32 The key question that might be asked is whether the incorporation of a preposition is actually found in the development of languages. On the face of it, phenomena of 'desemanticization' involving transitive verbs which then begin to be used in possessive constructions - as occur in many languages, Spanish tener (bring) being an obvious example - stand in favour of the idea that what is needed is just a support for two Agr°s as opposed to the actual incorporation of a locative particle. Needless to say, any argument of this sort needs to be based on careful philological ground-work that cannot be undertaken here. As for the interpretation of the two sentences, esserci has the usual existential meaning. On the other hand, averci means possession (of the entity denoted by the inner DP by the entity denoted by the outer DP). If avere and essere are the spell out of functional heads, i.e. they are two 'copulas', where does the interpretation come from? With essere cisentences, we already know that the interpretation does not come from the 'lexical' contribution of the copula, but rather from a computational mechanism forcing the DP to split at LF (see chapter 3 of this volume). We can plausibly assume that, with avere, the situation is not essentially
242
A view beyond: unaccusativity as an epiphenomenon
different. One possibility would be to rely on the Hale and Keyser framework illustrated in the previous section. Recall that the essence of this wideranging programme is that 'there are no thematic roles. Instead, there are just the relations determined by the categories and their projections, and these are limited by the small inventory of lexical categories and by the principle of unambiguous projection' (Hale and Keyser 1991: 40). Thus, we can assume that the structural configuration combining the small clause containing ci with an external argument is interpreted as the most generic relation that can connect two entities (denoted here by the two DPs), namely possession. Again, a full theory would need to be based on detailed argumentation, which is beyond the scope of the present work.33 At this point, we can conclude this rather marginal section. We have seen that the theory of ci as a raised predicate can be indirectly supported by extending the domain of research from the syntax of essere to that of avere. Indeed, we went as far as to suggest that a careful analysis of the occurrences of ci might lead to a unified analysis of avere and essere. The crucial step was to rely on the theory of the distribution of Agr°s that makes them depend on the number of argumental DPs only and thereby show that essere and avere differ with respect to the number of Agr°s involved.34 Before concluding the chapter, I will discuss a different topic which has to do with Burzio's generalization. 5.2.3 Remarks on Burzio's generalization Burzio's generalization captures a universal relation between two independent modules of grammar, namely ©-theory and Case theory. It establishes that 'all and only the verbs that can assign 9-role to the subject can assign (Accusative) Case to an object' (see Burzio 1986:178ff.). This classic theory will be illustrated by a reference to a simple example. A verb like develop yields either a transitive construction or an unaccusative construction with there in preverbal position (see Burzio 1986: 159): (57)
John, developed many theories; there, developed many theories;
A view beyond: unaccusative constructions as ci-sentences 243 As has already been mentioned, when there is present no external 6-role is assigned to the subject position; when John occurs, the external 6-role is assigned to it. In the first example, moreover, the inflected verb, developed, assigns Nominative to the subject in spec-IP and Accusative to the object position. In the second example, instead, developed assigns no Accusative to the object position. Rather, the Case of many theories in the object position of the second example is Nominative. The hypothesis is that Nominative is directly assigned to there and then transmitted to the lower DP on the crucial assumption that the two DPs are coindexed.35 It is by no means clear what type of empirical phenomenon lies behind Burzio's generalization: why should the lack of a 6-role block the assignment of Accusative? In principle, there seems to be no reason for these two conceptually distinct modules of grammar to interact in this way. Clearly, it would be a major step forward if we could understand the inner logic that correlates Case with 6-role assignment.36 Before showing how the theory of clause structure developed here might shed some light on the matter, a further observation must be spelt out. To account for the phenomena related to Case theory in terms of Burzio's generalization, the implicit assumption must be made that Universal Grammar contains a principle of 'Case economy', namely, there is a hierarchy of Cases (Nom>Acc>Obl, etc.) and a principle requiring that an element should not pick up a given Case K unless all those Cases which can be assigned in a given structure and precede K in the hierarchy have already been assigned; otherwise, Accusative could be assigned to the subject position. Notice also that for this system to work a further crucial restriction must be added. One must stipulate that Case can be maintained only by an argument. Thus, since there is regarded as an expletive, not as an argument, Nominative case is transmitted to the noun phrase in object position. The hypothesis that unaccusatives involve the raising of a predicative DP from a small clause offers a new natural way to look at Burzio's generalization. There are two independent facts that can now be associated to explain the phenomenon under discussion. Let us consider them separately. First, we independently know that if two DPs are in a predicative relation they are assigned the same Case (see section 1.3.1). This is generally shown by languages with overt Case morphology. The examples discussed in chapter 1 are repeated here: (58a)
Caesar dux/*ducem est (Caesar-nom leader-nom/*-acc is) 'Caesar is the leader'
244 b c
A view beyond: unaccusativity as an epiphenomenon senatores dicunt Caesarem ducem/*dux esse (senators-nom say Caesar-acc leader-acc/*-nom to-be) 'the senators say Caesar is the leader' Caesar dicitur dux/*ducem esse (Caesar-nom is-said leader-nom/*-acc to-be) 'Caesar is said to be the leader'
The two DPs contained in these examples, Caesar and dux, are linked by a predicative relation. Throughout the paradigm they share their Case, which clearly depends on the context of assignment. Moreover, we know that elements like preverbal there are assigned Case in exactly the same way as other predicates in inverse copular constructions; witness the distribution of for in infinitival contexts (see once again section 1.3.1):37 (59a) b
[*(for) [DP there]; to be [ sc [DP a picture of the wall] tj]] is unusual j*(for) [DP the cause of the riot]j to be [Sc [DP a picture of the wall] tj]] is unusual
Thus, if unaccusatives are analysed as involving a small clause containing two DPs linked by a predicative relation, Burzio's generalization will now follow without stipulation. We can now replace the traditional representation by the one proposed here: (60) there,-
arrive DP
On a par with inverse copular constructions, Case is now assigned to the most prominent position as usual by the proper inflectional head combined with the verb. The lower subject DP copies the case of the other DP, due to the predicative relation. The advantage of this approach is clear. Apart from 'Case economy' which is still needed to ensure that Accusative case is not assigned to the most prominent position when Nominative can be assigned, Case transmission is no longer stipulative since we must independently assume it for inverse copular sentences. From a theoretical point of view, we can conclude that the missing piece of information that has obscured the
Conclusions 245 relationship between Case theory and 0-theory, as formulated in Burzio's generalization, comes from predication theory. What was missing was the idea that between the postverbal subject DP and the preverbal DP position a predicative relation holds which is exactly the same as in inverse copular sentences. One last remark is in order concerning Accusative case. It is generally assumed that Accusative case can be detected in Italian when we have occurrence of third person clitics such as le (them-fem.plur.) in the following examples (Burzio 1986): (61a) b c d
pro+Bs hanno affondat-o [DP tre navi]+acc (they have sunk-neuter three boats)
pro +es lej hanno affondate tj (pro them-fem.plur. have sunk-third fem.plur.) pro_es sono affondate [DP tre navi]_acc (pro are sunk three boats) * pro_es lej s o n o affondate tj (pro them-fem.plur. are sunk-third fem.plur.)
Thus, the last sentence is ungrammatical because Accusative cannot be assigned to the object position. Why should there be a relation between Cases and clitics? If we assume the more articulated structure involving raising from a small clause the phenomenon in (61) can be dealt with in a rather natural way, that is, by paralleling the analysis we gave for the impossibility of the subject of inverse copular sentences cliticizing onto the verb (see section 1.3.3.3). All in all, the absence of cliticization in unaccusatives is to be traced back to the lack of Agr-heads and the link with Case theory (in its more recent formulation stemming from Chomsky (1993)) thus becomes transparent. 5.3
Conclusions
To sum up, this chapter has been divided into two distinct parts. The first part attempted to trace the implications of a simple observation. Esserci behaves as an unaccusative verb like arrivare with respect to the classic diagnostics: it takes only one argument (i.e. it is intransitive, in the broad sense), it selects auxiliary essere, its past participle agrees in gender and number with the subject, ne-cliticization is possible from the postverbal subject. Nevertheless, whether or not the analysis of d-sentences proposed in chapter 2 is adopted, the subject of a d-sentence is not in the object position, as would be required in order to explain the properties normally associated with unaccusative constructions. Rather, esserci involves a small
246
A view beyond: unaccusativity as an epiphenomenon
clause with a postverbal DP in the subject position. Focussing on the relevant segment, we obtained the following representations: (62)
a
V
arrivare
b
essere DP
The first part of the chapter concluded by simply registering this apparent paradox and making the further crucial observation that the class of inverse copular sentences which do not involve ci meets only two of the three properties that characterize unaccusative constructions: they select essere as an auxiliary and display past participle agreement but do not allow neextraction. Again, this was regarded as highly unexpected, since in the standard theory of unaccusativity the three properties are normally regarded as traceable to a single cause, namely the presence of the subject DP in the object position. The second part of this chapter, on the other hand, had a highly speculative nature and appealed to a number of conceptually distinct arguments. First, we attempted to unify the account of esserci and arrivare. We proposed to analyse unaccusative constructions in the same way as cisentences and adopted the analysis of ci as a raised predicate proposed in chapter 2. Thus it was proposed that the structure for arrivare would also involve an embedded small clause construction and the raising of the lower predicate: (63a)
molte ragazze
Conclusions 247
molte ragazze
A twofold strategy was then adopted. On the one hand, we saw that the structure involving a small clause is just as capable of deriving the essential property of unaccusatives as the traditional structure: raising can be used to explain agreement patterns and auxiliary selection, while ne-extraction can be related to incorporation of a lower predicate. On the other, the advantages of this new approach were evaluated by applying it to three different empirical domains: 0-role assignment (section 5.2.1), the distinction between essere and avere as main verbs (section 5.2.2) and Burzio's generalization (section 5.2.3). In each domain, this theory turned out to capture the data in a rather natural way, suggesting that it is a good working hypothesis for future research. More specifically, it was suggested that d-sentences play a central role in 0-role assignment - provided that the theory stemming from Hale and Keyser (1991) is adopted - in that they overtly signal the structure associated with the role. Then, a unified theory of essere and avere was proposed that derives the distribution of the two verbs from the number of Agr°s required for the structure to converge. Finally, it was suggested that Burzio's generalization can be regarded as a consequence of the presence of a predicative nominal in the most prominent clausal position.
Appendix: a brief history of the copula
Mirrors and the copula are abominable because they multiply the number of people
i
J. L. Borges
The meaning(s) of 'copula'
There are terms in the history of science which, from the time that they were first adopted, have never been dropped. This is the case with such well-known terms as 'atom' or 'energy' or 'gene'. In reality, however, we all know that this continuity is only superficial; in the course of time terms of this type inevitably change their original meaning. The result of this is that, even though in most cases we are prepared to acknowledge that some basic core intuition, contained in the term in question when it was originally coined, has indeed survived, we are accustomed to treating such terms with great care: in particular, before they can be used in any sort of scientific treatise, exhaustive definitions have to be given. From this point of view, linguistics is no different from any other field of inquiry. Terms like 'verb' or 'subject' or 'negation' may well not mean the same thing now as they did at earlier stages in the study of language. There is therefore a risk of generating confusion, especially in studies that draw on more than one theoretical tradition. In this brief appendix my aim is to investigate the term 'copula': I will attempt to show that, in its common use, it conceals at least three distinct traditions which have been taken over by modern linguistics complete with their various inconsistencies and obscurities. ii
A brief survey of three major approaches
Three major approaches will be briefly illustrated here.1 In the first section, Aristotle's theory will be presented insofar as it can be reconstructed in its essential lines from the De Interpretatione. Aristotle's theory has of course played a fundamental role and practically every scholar who has approached the problem of the copula has felt obliged to refer to it, even though, in some cases, such references turn out to be little more than pro-forma tributes to Aristotle's auctoritas. In the second section, we will examine Abelard's theory. The term 'copula' first appeared in his Dialectica. This terminological innovation in fact brings with it a very deep change of perspective that ultimately influenced one of the landmarks of the entire field, i.e. the Port Royal Grammaire Raisonnee. 248
A brief survey of three major approaches 249 The third section is devoted to Russell's theory of the copula as presented in his Introduction to the Philosophy of Mathematics. It is hardly surprising that a theory
of the copula should be presented in a book of mathematics. This is in part due to the new approach to language problems that originated in nineteenth-century logical studies but also to the frequent use of the copula as a way to express mathematical formulae in natural languages. An attempt will be made to show that this logically driven approach to language is incompatible with a modern empirical approach and leads to a major misunderstanding. These three traditions all contribute to yield the meaning of 'the copula': failure to take a full account of them will mean that there is a risk of using the term inconsistently, referring now to one thing, now to another.2 //. /
The sign of tense
In Aristotle the copula is singled out and analysed as a special entity because of the central role it plays in the analysis of the declarative sentence (i.e. in 'logos apophantikos' (koyoq djtoa\aix6g)). It is a well-known fact that Aristotle's linguistic works (and much of his work in general) are not easy to interpret and contain a number of inconsistencies (see for example During 1966, Ackrill 1963 and references cited there).3 Nevertheless, his theory of clause structure has been reconstructed by scholars across the centuries and it is now rather well established. I will illustrate it very briefly by citing some major passages of his work.4 Let us refer to Aristotle's own words. First of all, what is a declarative sentence? '[N]ot every sentence is a statement-making sentence, but only those in which there is truth or falsity. There is not truth or falsity in all sentences: a prayer is a sentence but is neither true or false' {De Int. 4,17a, 3-6).5 Thus, what are in Aristotle's theory the necessary components of a declarative sentence? A single noun in isolation (a noun phrase, nowadays6) is not sufficient, because a noun cannot per se be true or false 'for falsity and truth have to do with combination and separation' {De Int. 1, 16a, 13).7 Thus, 'Every statement-making sentence must contain a verb or an inflection of a verb' {De Int. 5,17a, 12)8 because 'a verb is a sign of what holds, that is, holds of a subject' {De Int. 3,16b, 7).9 In the last citation, Ackrill's (1963) translation is potentially misleading: the term 'verb' used to translate the ancient Greek rhema (Qfjjxa), is a mistranslation - what is really meant is 'predicate'. In fact, there is a long-standing dispute about the translation of the term rhema (QTJUXX), as used by Aristotle. In general, commentators tend to assume that it can either mean 'predicate' (as in this quotation, see also the paragraph after the next one referring to De Mauro and Thornton 1985 and Graffi 1986) or 'verb' in the sense that this term normally has in modern frameworks as a part of speech. This second meaning of rhema (Qf](xa) is explicitly attested in many passages. In fact, verbs are recognized as having a further fundamental function in declarative sentences: 'A verb is what additionally signifies time[. . .^recovery is a name, but recovers is a verb because it additionally signifies something's holding now' {De Int. 3, 16b, 6-9).10 In short, then, a subject, a predicate and a tense specification are the conceptual primitives that can be regarded as constituting a declarative sentence. The role of the copula in this system is of extreme importance since it provides a
250
Appendix: a brief history of the copula
straightforward way to illustrate this triadic skeleton. The copula is analysed by Aristotle as the element which provides the affirmative sentence with the necessary tense specification when this is not realized in direct combination with the predicate. In fact, according to Aristotle, any sentence can be turned into a copular sentence: 'there is no difference in saying that a man walks and that a man is walking' (De Int. 12, 21b, 9).11 That the verbal inflection and the copula are the same thing is shown by the fact that they are in complementary distribution: 'In cases where is does not fit (e.g. with recovers or walks) the verbs have the same effect when so placed as if is were joined' (De Int. 10, 20a, 4).12 This fundamental idea that the copula does not participate in the predicative linking is also attested in several other passages. These tend to be quite explicit, as in the following case: 'For example, a man is just; here I say that the is is a third component' (De Int. 10,19b, 21).13 Yet they can also be indirect: 'If names and verbs are transposed they still signify the same thing, e.g. a man is white and white is a man' (De Int. 10,20b, 1).14 In this case there is only one thing that does not transpose and that is the copula; thus, it cannot be considered a 'verb'. It should not be surprising that the term 'verb' is not applied to the copula here. We may note en passant that this is another case where the Greek rhema (0fjuxx) is to be translated as 'predicate' rather than as 'verb' in the sense of a part of speech since in this case it clearly refers to white. Apart from the case cited in the previous paragraph, there is another crucial passage (see De Mauro and Thornton 1985, Graffi 1986 and references cited there) which is worth citing here. Unless rhema (£fj|ia) is translated as 'predicate' rather than 'verb', a proposition of the De Interpretatione becomes self-contradictory. The case in question is De Int. 10, 19b, 13.15 Many modern translators render it very misleadingly as: 'Without a verb there will be no affirmation or negation' (Ackrill 1963) or 'Unless there is also a verb, there is no affirmation or denial' (Cooke 1938). The ancient Greek original is a nominal sentence since the copula is missing; thus, we would simply have to regard this very sentence as contradicting itself, if we did not read rhema (£fj|u,a) as 'predicate' instead of 'verb'. Again, this shows that the copula is not to be considered as a predicate because, if it were, it could never be absent from a declarative sentence, as is the case, for instance, in nominal sentences. To complete the succinct reconstruction of Aristotle's system, which is still problematic today, we will examine a very much debated passage of his work. Aristotle also considers as declaratives those sentences made up of just a name and the copula, as in Philon estin (fyikiov eoxiv; 'Philo is') (see De Int. 2, 16b, 1, and De Int. 1, 16a, 17). Here, the text is very difficult to interpret for Aristotle does not indicate explicitly how this use of einai (eivai) can be reconciled with the theory of the copula as a 'third-component', as in a man is just (as in De Int. 10, 19b, 21, as we already saw). In general, many modern commentators simply assume that on this point Aristotle tacitly shifts from the copulative use to the existential use. Indeed, in the next section, we will see the reaction of the medieval commentators to this interpretation. Thus, the situation is anything but straightforward; nevertheless, a fairly clear conclusion can be seen to emerge: what we find in Aristotle is a theory of the declarative sentence as consisting of a predicational link plus tense specifications;
A brief survey of three major approaches 251 intimately connected with this is the theory of the copula as an expression of tense in complementary distribution with verbal inflection. Since then, this theory has never been completely abandoned, although it does not always exercise a direct influence on later models. Many references to Aristotle's theory in the subsequent literature are, it is true, little more than token tributes to his auctoritas, but it is also true that his theory has genuinely been considered as a model throughout the centuries. In thisfield,references to Aristotle reached modern times either explicitly (as in Benveniste, Jespersen and Chomsky) or implicitly (as in Vendryes and Meillet). Although the idea that the copula is not itself a predicate was also factually prompted by the discovery of nominal sentences in IndoEuropean, nevertheless the implicit theoretical background these authors refer to is quite transparent. For example, Vendryes says that Tintroduction de la copule dans la phrase nominale s'explique aisement. II y a en effet une notion que le simple rapprochement du sujet et du predicat ne suffit pas a exprimer: c'est la notion de temps . . . Une fois introduite dans la phrase nominale lorsqu'il y avait lieu d'exprimer le mode ou le temps, la copule s'y est parfois installee, meme lorsqu'elle n'ajoutait rien au sens. Le verbe, en tant que symbole de temps, devenait alors necessaire' (Vendryes 1921: 146). An analogous position is taken up by Meillet in various stages of his work; for example, 'Si le predicat, qui est l'element essentiel de la phrase, est un nom, la phrase est dite nominale; si le predicat est un verbe, ou du moins un verbe autre que le verbe "etre" ou copule, elle est dite verbale' (Meillet 1934: 356). Of course, Aristotle's theory of the copula has not survived to the present day without undergoing a number of important changes. In the next two sections are illustrations of two major contributions to the general development of the theory of the copula that are in some senses landmarks. //. // The sign of affirmation With Abelard, the term 'copula' enters into western thought. In fact, although widely attested, the use of the term 'copula' in reference to Aristotle's work is totally anachronistic.16 What led to this term? In his Dialectica, Abelard was mainly concerned with the way syllogisms can be construed. The interest of the copula was in fact derivative from this main concern. As Kneale and Kneale (1962: 206) put it, 'it is clear that for his [Aristotle's] theory of syllogism he assumes in every general proposition two terms of the same kind, that is to say, each capable of being a subject and each capable of being a predicate'. Thus, since the only linguistic entities that can play these two roles are nouns (in modern terms, noun phrases), it is easy to understand why the copula became central. Abelard pursued the Aristotelian theory by emphasizing the role of be as the element that can turn a noun into a predicate in a syllogism rather than as the element that provides the sentence with a time specification (see Dial. 161). It is this conceptual shift that underlies the invention of the term 'copula', which is cast on the Latin copulare meaning 'to link'. For example, in sentences like a man
252
Appendix: a brief history of the copula
is a mammal and Socrates is a man the copula allows the noun phrase a man to play the role of the subject, in the first, and that of the predicate, in the second. Clearly, in such a framework the assumption that the copula can be interpreted as a predicate meaning 'existence' cannot be maintained (see the end of the section on Aristotle). We have an explicit argument by Abelard to overcome this problem. Let us reproduce it synthetically as follows. A sentence like Socrates est (Socrates is), where est (is) just occurs with a subject, potentially undermines the theory of the 'copula' as link between a predicate and a subject like in Socrates est homo (Socrates is man). To avoid this inconsistency, Abelard appeals to a typical reductio ad absurdum (Dial. 137, 162). We interpret Socrates est as meaning Socrates est ens (Socrates is existing, where ens is (the Nominative, singular form of) the present participle of esse, be). Suppose that est means est ens, that is, est is a predicate meaning existence. Then nothing would prevent a sentence like Socrates est ens from being interpreted as Socrates est ens ens, and that from being interpreted as Socrates est ens ens ens, etc., ad infinitum without reaching a stable meaning.17 Thus, Abelard concludes, in the sentence Socrates est, the predicate of existence is just not expressed.18 This deep change of perspective concerning the copula has played an important role in the development of the theory of clause structure. Several centuries later, the idea of a 'copula' in the sense of Abelard was to be a crucial element in the thinking of the Port Royal school. The emphasis that this school put on this theory is by no means accidental. In fact, the grammatical distinction between a subject, a predicate and the copula was basic to their whole enterprise: this triadic partition (subject, predicate and the copula) was assumed to reflect the distinction between two fundamental activities of the 'spirit' (in the sense of the human intellect): i.e. conceiving (that is, establishing names for substances and accidents) and judging (that is, connecting names in a predicative relation by means of a copula).19 Of course, the influence of the Aristotelian model is still strong but the Port Royal grammar's departure from the simple idea of the copula as expressing tense is also very clear: 'le jugement que nous faisons des choses (comme quand je dis, la terre est ronde) enferme necessairement deux termes, l'un appelle sujet, qui est ce dont on affirme, comme, terre; et l'autre appelle attribut, qui est ce qu'on affirme, comme ronde: et de plus la liaison entre ces deux termes, qui est proprement Faction de nostre esprit qui affirme l'attribut du sujet . . . Et c'est proprement ce que c'est le verbe, un mot dont le principal usage est de signifier l'affirmation' (Grammaire: 94— 5; spelling is reproduced from Brekle 1966 edition). It is easy to see that the copula is still central within this framework, Ton peut dire que le Verbe de luy-mesme ne devroit point avoir d'autre usage, que de marquer la liaison que nous faisons dans nostre esprit des deux termes d'une proposition. Mais il n'y a que le verbe estre qu'on appelle substantif qui soit demeure dans cette simplicity . . . on n'auroit eu besoin dans chaque Langue que d'un seul Verbe, qui est celuy qu'on appelle substantif (Grammaire: 96-7). In fact, this triadic system is not always as perspicuous as in the example cited: 'Car, comme les hommes se portent naturellement a abreger leurs expressions, ils ont joint presque toujours a l'affirmation d'autres significations dans un mesme mot. Ils y ont joint celle de quelque attribut: de sorte qu'alors deux mots sont une proposition: comme quand je dis, Petrus vivit, Pierre vit: parce que le mot de vivit
A brief survey of three major approaches 253 enferme seul l'affirmation, et de plus l'attribut d'estre vivant; et ainsi c'est la mesme
chose de dire Pierre vit, que de dire, Pierre est vivanf (Grammaire: 96)2().
The principal departure from the Aristotelian model is explicitly stated: 'La diversite de ces significations jointes en mesme mot [i.e. predication, affirmation and tense in a verb as in vivit], est ce qui a empeche beaucoup de personnes . . . Ainsi Aristote s'estant arreste a la troisieme des significations adjoutees a celle qui est essentielle au Verbe, l'a definy: vox significans cum tempore: un mot qui signifie avec temps' {Grammaire: 98).
Although the linguistic data are treated in Abelard as direct support for a general theory of gnoseology, this approach has nevertheless had an enormous influence in linguistics itself. Up until modern times this system has been a key element of educational models in western civilization (probably, only Euclid's Elements had a comparable success). However, the long history of this theory of clause structure is by no means linear or easy to trace. Along with its incorporation in the Port Royal grammar, another key moment was its adoption by the German tradition. This started with the classic work of Wolf, Philosophia rationalis sive logica (1732), and continued at least until the work of linguists like Hermann and Bopp was developed.21 To sum up, in this second tradition, which stems from Abelard and extends as far as the nineteenth century, we see that the theory of the copula has already acquired a certain complexity. The debt to Aristotle is still evident in two persisting ideas: that all sentences are basically copular sentences and that the copula itself is not a predicate. However, the differences from the Aristotelian model are by no means small though they are anything but clear from the terminology (especially since the term 'copula' is applied retroactively to Aristotle's own framework). Thus, the copula is not only the element that provides the affirmative sentence with the necessary tense specifications; it is also the prototypical element that allows predicative linking to take place. A third major influential approach contributes to the meaning of the term 'copula' as it is generally used at the present time. //. ///
The sign of identity
A major change of perspective in the theory of the copula occurred in Russell's work. It is by no means accidental that there is such a large distance between the position of this author and those mentioned in the previous section. As Graff] (1991) has pointed out, by the nineteenth century the model stemming from Port Royal had already been dismissed. The very idea that all sentences are equivalent to copular sentences proved too abstract to maintain, especially in a framework that was grounded on the observation of historical and comparative data. Nevertheless, the discovery of nominal sentences in Indo-European, and the very existence of sentences without the copula in modern languages such as Russian, did in fact lead to a theory of the copula that roughly coincides with the old Aristotelian framework. As illustrated in the quotations from Vendryes and Meillet at the end of section n.i, the copula is 'once again' regarded as the element which provides a declarative sentence with inflectional specifications, when they are not realized
254
Appendix: a brief history of the copula
'within the predicate'. With such a background, it is rather natural that a new perspective on thefieldcould only come from outside linguistics, as it eventually did in this case with the work of the logician and philosopher Bertrand Russell. There is no better way of introducing Russell's theory of the copula than through a quotation from the Introduction to the Philosophy of Mathematics: The proposition Socrates is a man is no doubt "equivalent" to Socrates is human, but it is not the very same proposition. The is of Socrates is human expresses the relation of subject and predicate; the is of Socrates is a man expresses identity. It is a disgrace to the human race that it has chosen the same word is for those two such entirely different ideas as predication and identity - a disgrace which a symbolic logical language of course remedies' (Russell 1919: 172).22 Clearly, this theory is a drastic departure from the tradition: on the one hand the copula is still analysed in the traditional way (more precisely, the one which goes back to Abelard) as the linking element of predication in Socrates is human; on the other, and this is the major difference compared to previous models, the copula itself plays the role of predicate, specifically the predicate which, according to Russell, expresses identity in Socrates is a man. Nevertheless, as the final words of the quotation indicate, this departure is more than a iocal' change in the theory of the copula and it reflects the basic epistemological value of linguistics itself. Here, we see the interaction of many different forces. Russell's words reflect the effects of nineteenth-century logical thought (as in Boole and Frege) as well as the reaction to the British Hegelianism represented by Bradley. Nevertheless, a moral emerges from this approach: language is considered as imperfect, and its imperfection is to be remedied by squeezing the mass of linguistic data into a framework provided by a formal non-ambiguous language. Given an approach of this type, we can hardly expect to find much interest in actual linguistic data: in fact, the theory of the copula as the realization of an identity relation is not based on empirical investigation; it is simply imposed on the data. For this reason, the conclusion that Socrates is a man involves an identity is simply not open to falsification.23 The effect of this theory on modern linguistics has been enormous: the idea that the copula is ambiguous between predication and identity has been widely accepted and not only in the philosophical tradition (see for example the entry 'copula' in Quine 1987). Apart from illustrious representatives, like Jespersen (see m.i), 24 all major modern schools of grammar, such as generative grammar (see m.ii), relational grammar and Montague grammar,25 adopt it, as well as more traditional frameworks such as we can find in Benveniste (1966),26 Halliday (1967-8),27 Quirk and Greenbaum (1973),28 Kahn (1973) or Renzi and Salvi (1991). Interestingly, Russell's original passage has not been taken literally: it is worth emphasizing that, despite the fact that Russell considered the copula as a sign of identity if and only if it is followed by a noun phrase, in most frameworks the ambiguity of the copula between predication and identity is claimed to hold even when the copula is indeed followed by a noun phrase. To recapitulate, at least four distinct concepts come together in the term 'copula'. These can be subdivided into two main groups: those where the copula is not regarded as a predicate and those where the copula is recognized as having this
'Grammatical subjects': copular sentences in modern syntax
255
function. Thus, there is first a generally accepted conception of the copula according to which its function is to provide the sentence with inflectional elements required by declarative sentences (essentially following Aristotle's original thesis). Then, there is a second conception in which the copula is what can turn a term into a predicate and link it with the subject (this is the 'proper' copula as proposed by Abelard and followed by the School of Port Royal). Shifting now to those theories where the copula is itself regarded as a predicate, we find a third conception: in this case the copula is used to express identity (Russell's theory). Finally, there is a fourth possible meaning associated with the copula and that is existence, although attempts have been made to refute this idea since medieval times (starting with Abelard). We have also observed that the various different meanings of 'copula' which we have surveyed here have sometimes been combined together, as the citations of Benveniste and Halliday in particular clearly show. A first conclusion can be seen to emerge from this brief survey: all these analyses are independent of the syntactic structure of the clause. The different functions that the copula is assumed to be associated with (identity, existence, predication, etc.) are analysed as inherently dependent on the 'word' itself, i.e. on the lexical entry of be, einai (elvai), esse, etre, essere, etc.29 In the next section we will consider the impact of copular sentences on the theory of clause structure. HI
'Grammatical subjects': copular sentences in modern syntax
When Russell's theory first started circulating, the field of linguistics was going through a period of great change. Many problems which had received little or no attention in the heyday of historical linguistics (though without being abandoned completely; see Graffi 1991) were suddenly put back on the agenda with the development of Structuralism. As we will see, the copula is still central as inherently related to the problem of finding an adequate theory for clause structure itself. Again, we will not be able to draw a complete picture.30 Two major representative works among others will illustrate the trend, namely Jespersen's (1937) Analytic Syntax and Chomsky's (1981) Lectures on Government and Binding. At this stage in its development, linguistics appears to be completely deprived of any ontological, gnoseological or logical value: the problems it deals with tend to be conceived of in an entirely discipline-specific fashion, i.e. as exclusively the province of linguistics. Interestingly, we will see that although in these frameworks copular sentences have lost the central position they had in previous ones, they in fact function as a sort of 'Trojan horse', as they reintroduce into the theoretical debate a set of notions that it had been a basic aim of the theoretical model to dismiss. ///. / Jespersen ys framework: verbal agreement as a diagnostic for the subject
Right up to his last work, Analytic Syntax, one of Jespersen's major concerns was always to avoid using extra-linguistic concepts to account for linguistic phenomena.31 Even the terminology is highly unsatisfactory for Jespersen: the term 'subject' itself is regarded as too heavily compromised with logic to be straightforwardly adopted. Thus, in order to emphasize the fact that what he is speaking about is a linguistic concept he uses the term 'grammatical subject' (throughout his work).
256
Appendix: a brief history of the copula
Jespersen claims that 'the grammatical subject cannot be defined by means of such words as active or agent; this is excluded by the meaning of a great many verbs, e.g. suffer . . . The subject is the primary which is most intimately connected with the verb (predicate) in the form which it actually has in the sentence with which we are concerned; thus, Tom is the subject in (1) "Tom beats John", but not in (2) "John is beaten by Tom", though both sentences indicate the same action on the part of Tom; in the latter sentence John is the subject, because he is the person most intimately connected with the verb beat in the actual form employed: is beaten . . . subject as a grammatical term can thus be defined only in connection with the rest of the sentence in its actual form' (Jespersen 1937: def. # 34.1). Clearly, it is the 'actual', i.e. purely linguistic, relation between the parts of the sentence that Jespersen aims to capture, over and above any logical or semantic intuition. Within this conceptual framework, verbal agreement emerges as the major diagnostic for the subject. It is agreement that establishes a privileged relationship between the verb and one and only one of its arguments: this is for Jespersen the essential empirical correlate of the linguistic notion of 'subject'.32 Needless to say, copular sentences cannot be straightforwardly accounted for on the basis of this definition. Since the copula may occur with two noun phrases displaying the same gender and number, the question arises of how the subject and the predicate can be identified. Of course, for those who adopt Russell's logical point of view, when the copula occurs with two noun phrases (as in Russell's example Socrates is a man) the problem simply does not arise. The corresponding sentence would not be any different from a transitive construction: the copula would be treated as a predicate of identity and no other predicates would be around. However, Jespersen explicitly refutes the idea that the copula can express identity. 'The linguistic copula is does not mean or imply identity but subsumption in the sense of the old Aristotelian logic' (Jespersen 1924: 153). 'In the mathematical formula A = B we should not take the sign = as the copula and B as predicative, but insert the copula is before the predicative equal to B' (1924: 154; see also Jespersen 1937: 133).33 Jespersen does refer to 'identity' as in sentences like Beauty is truth; truth, beauty, but in this case he specifies that it is to be interpreted as 'coextension', which has nothing to do with the copula as expressing identity (Jespersen 1924: 153). The conclusion is that 'There are cases in which extension is equal, this means that we cannot decide which is the subject and which is the predicate' (Jespersen 1937:136). Interestingly, Jespersen makes an exception here to his fundamental principle of not appealing to logical concepts. In order to identify the subject in copular sentences, he proposes the following criterion: 'If one of the substantives is perfectly definite, and the other not, the former is the subject; this is the case with a proper name, Tom is a scoundrel (Jespersen 1924: 150). All in all, what is the real function of the copula in Jespersen's opinion? On the face of it, the answer is not very different from the one given by historical linguists (see citations from Vendryes and Meillet in section n.i): 'Sentences containing is probably have their origin . . . from "nominal sentences" in which two words were formlessly placed together as subject and predicative; later these were brought to the usual type by the addition of the least substantial verb . . ., in much the same way as other sentences were made to conform to the usual type by the addition of the colourless subject it (it rains, it pleases me to go, etc.)' (Jespersen 1937: 135).
'Grammatical subjects': copular sentences in modern syntax
257
Thus, for Jespersen the copula is what we might call a 'dummy verb', paralleling the term 'dummy subject' as used for the it of it pleases me to go etc. However, once it has been made, this proposal plays no further role in Jespersen's theory and indeed it is simply dismissed on the grounds that 'the questions of supposed original development should not determine our analysis of things as they are in our own day' (1937: 135). The following quotation also witnesses how delicate the issue of copular sentences was for Jespersen. As a marginal remark, Jespersen reports a dispute by 'Noreen and others' about 'which is the subject and which is the predicative, in some sentences where it is possible to transpose the two members, e.g. Miss Castelwood was the prettiest girl at the ball vs. The prettiest girl at the ball was Miss Castelwood'
(1924: 153).34 The question is immediately dismissed as 'not very important' and the conclusion is drawn that 'the two terms connected by is . . . may change places as subject and predicative' (1924: 153). How important these pairs in fact are should, I hope, be clear from the first chapter of this book. Nevertheless, it is sufficient here to recall that an English example involving two substantives which do not match in gender and number would be a major problem for Jespersen's approach. Take for example a case similar to those we discussed in the first chapter: books are my only desire and my only desire is books. Here, the copula unselectively agrees with the substantive that occurs on the left but one could hardly claim that the function of subject and predicate in these two sentences is not always held by the same elements. Moreover, and even more crucially, if one applies to these sentences the criterion based on definiteness we just mentioned, contradictory results would be obtained: on the one hand, the subject should always be the more definite noun phrase of the two (i.e. my only desire) and the predicate should be the less definite (i.e. books); on the other, verbal agreement (as a diagnostic for the subject) can go with either one. All in all, one of the major consequences of regarding the subject as being identified by verbal agreement is that in this way a link is established between this grammatical function and a specific structural position. In other words, since verbal agreement can only occur in a certain fixed position, then the subject is always associated with such a position. This fundamental assumption raises a major empirical question with respect to another type of copular sentence: what is the role of an element like there in there is a problem? Here, two contrasting intuitions seem to collide. 'In some respect (place in the sentence, etc.) this there behaves as an ordinary subject, and many grammarians therefore class it as a kind of subject' (Jespersen 1937: 129). On the other hand, the intuition is that the real subject has been displaced to 'an inferior position' (Jespersen 1924: 154). The issue at stake here is not marginal: given that within this framework the copula is not itself a predicate, the question naturally arises as to what the roles of the two elements that occur with it are. Jespersen notices that the terminology is not satisfactory: in his opinion, the best solution is simply to refer to this item as 'existential there'; as an alternative, he proposes the term 'lesser subject' (1937: 130). This analysis is extended to cases involving verbs other than the copula such as exist, stand, lie, come (1937:130) which are given the same explanations as the copula with the additional observation that only verbs with 'vague meaning' (sic) are included in this group. Finally, it should be noticed that Jespersen also refuses to consider the copula as
258
Appendix: a brief history of the copula
expressing so-called existential meaning. The issue here is tackled quite clearly: 'when philosophers form sentences like "God is", this is felt as a rather unnatural transference from the normal use of is as "copula". . . If we answer the question: "is he dead?" by saying: "Yes, he is" - does that mean: "he exists" or "his death exists"?' (Jespersen 1937: 133). Existential meaning, as we saw in the third chapter of this work, is rather regarded by Jespersen as the result of a particular syntactic structure where 'the verb precedes the subject and the latter is hardly treated grammatically like a real subject' (1924: 155).35 As we see, then, in a syntactic framework the explanation of copular sentences and perhaps even the problems that they raise are intimately related to a fundamental assumption of clause structure: that the predicational structure embodied in the clause is rigid, i.e. that the subject of predication coincides with the specific position where verbal agreement is triggered. Very similar problems and questions arise in the different empirical framework which has grown out of Noam Chomsky's work. ///. //
Frege revisited: subjects as saturators in generative grammar
The account of clause structure in generative grammar is grounded on two fundamental assumptions: first, the skeleton is given by the inflectional morphemes (essentially, verbal agreement and tense); second, grammatical functions such as subject, predicate, object, etc., can be entirely derived from syntactic configurations.36 How do copular sentences fit into this framework? Let us set about answering this question by considering the two assumptions separately. The first fundamental assumption is explicitly presented in Chomsky (1981: 51). The essential idea is that the inflectional morphemes do not belong to the verb phrase nor to any other specific element of the clause.37 Making crucial use of different levels of representation, Chomsky assumes that these inflectional morphemes are generated as autonomous entities that are then incorporated into the verb by the computational mechanisms.38 There have been various ways of implementing this hypothesis. Chomsky (1986b), for example, gives the following one. There is a general theory (called X'-theory) according to which every constituent is the result of a lexical head (X°) projecting an intermediate projection (X') and a maximal projection (X" = XP). Every head can be adjacent to another maximal projection (the 'complement'), and every intermediate projection can be adjacent to another maximal projection (the 'specifier'). For every head X, Yand Z, the usual bracketed notation is the following one: [Xp ZP [X' X°YP]]. If we abstract away from the order of adjacent elements, we can assume that this structure is universal.39 This schema has been subsequently extended to non-lexical constituents, including clauses. Thus, the verb phrase (VP) is the complement of a head (1°) containing the inflectional morphemes, i.e. agreement and tense.40 This complex, called I', has a specifier. Since by definition 1°contains agreement and assigns Nominative case, the specifier of 1°can only be a noun phrase (NP, disregarding here the DP hypothesis). This combination is represented in formal terms as follows: [IP NP [v 1°VP]]. Recursively, the Inflectional Phrase (IP) can be the complement of another head, namely C°, which is the label for what is referred to as 'complementizer', exemplified by English that, yielding: [CP C° I[P NP [,< 1°VP]]]. Intuitively, we can regard IP as the
'Grammatical subjects': copular sentences in modern syntax
259
'predicative nucleus' and the CP as the 'periphery' which is involved in carrying out syntactic operations like interrogative formation, subordination, scopal relations at LF, etc. (see Graffi 1994, Rizzi 1995 and references cited there). The second fundamental assumption about the way grammatical functions are identified can be regarded as a tribute to distributionalism, at least in the sense that any appeal to what can be intuitively called 'meaning' is avoided.41 It is easy to notice that the elements occurring within this skeleton have asymmetric relationships. Let us take a very simple sentence like [IPJohn [VP reads the book]]. The first NP is in spec-IP where it triggers verbal agreement on the verb; the second NP, on the other hand, is embedded in the VP and is in fact adjacent to V°. Thus, the two NPs have asymmetric positions with respect to V°. Since the very first approaches these asymmetries have been considered to be sufficient to identify the basic grammatical functions. The subject is the NP which is immediately contained by the clausal node, IP; the object is the NP which is immediately contained by the VP. Eventually, the predicate is the VP itself. Notice that this architecture, including the subject-object asymmetry, does not flow from any logical necessity. In principle, it would not have been impossible to have sentences where the subject and the object were totally symmetric: in many artificial languages this is in fact the case. This is simply not the case, however, in natural languages: 'human languages do not adopt the principles familiar in modern logic. Rather, they adhere to the classical Aristotelian conception that a sentence has a subject and a predicate, where the predicate may be complex' (Chomsky 1988b: 54).42 Let us now turn to the main question to be addressed here. How do copular sentences fit in this framework? Needless to say, no quick straightforward answer can be given. There are at least three distinct issues at stake here. Let us review them briefly by indicating the earliest bibliographical source in each case. 43 First, a morphological problem: the copula is anomalous with respect to other verbs in that it does not allow do support. Indeed, the impossibility of sentences like "John does be here was noticed in the earliest works, e.g. Chomsky (1957: 67). Second, a semantic problem: the class of copular sentences of the kind NP-copula - NP cannot be homogeneously described as containing a predicative NP in postcopular position. This would lead to the rather implausible conclusion that in the girl that helped us on Friday is Mary Gray, the noun phrase Mary Gray can only be interpreted as attributing a property to the entity denoted by the girl that helped us on Friday (see Higgins 1973). Third, a genuine syntactic problem: copular sentences like there is a man in the room are considered as variants of sentences such as a man is in the room, much in the way Jespersen analysed preverbal there (see Milsark 1974, 1977). Why is it so? The analyses given for these three cases are not going to be reproduced in detail. For our purpose it is sufficient to indicate the essential lines of reasoning. The first problem is not really relevant here. The absence of do support is analysed as a lexical idiosyncrasy of be and have (see Pollock 1989 for a recent discussion of this and related topics).44 On the other hand, the other two problems call for considerable attention. The second problem is generally solved by adapting Russell's analysis. The essential idea is that the copula is ambiguous: it can occur with a predicative NP but it can also occur with a non-predicative NP yielding an identity statement. In
260
Appendix: a brief history of the copula
this case, be is predicted not to be any different from a transitive verb, in that it can be followed by a fully referential object.45 The third problem is much more complex and brings back to the foreground many of the classic problems centring around the notion of predication. The two sentences in question are derived from the same base generated structure. If the subject does not end up in precopular position at a surface level, a dummy element {there) is assumed to be inserted in this position as an 'expletive'. What forces the appearance of the NP there? Usually, as we know, full NPs are required to satisfy the 'valency' of a verb as prescribed in the lexicon (Projection Principle). For example, the verb like requires two arguments, say John and movies. If one argument is missing at any level of representation, as in *John likes or * likes movies, the sentence is ruled out. Unfortunately, this principle cannot be applied to *is a man in the room because no NP is missing here: witness the grammaticality of a man is in the room. Clearly, the problem is of a purely syntactic order, in the sense that it is basically a problem of word order. The notion of grammatical functions, and specifically that of the subject of the predication, plays the crucial role here. In fact, Chomsky (1981: 28ff.) proposed analysing sentences like *is a man in the room as violating a fundamental requirement to the effect that all sentences should have a subject of the predication (at least) at LF. The element there would then be inserted in spec-IP as a null semantic element (an 'expletive') that realizes the subject of the predication. It is worth noting Chomsky's own words here: 'The projection principle requires that complements of heads must be represented at each syntactic level (D-structure, S-structure, LF), so that, in particular, objects must be represented, but it says nothing about subjects. Thus, it distinguishes between what Edwin Williams calls "internal" and "external" arguments, specifically, object and subject. The projection principle requires that the former be syntactically realized, but not the latter, although they are required as subjects of predication (either arguments or expletives). The two principles - the projection principle and the requirement that clauses have subjects- constitute what is called the extended projection principle (EPP) in Chomsky (1981)' (Chomsky 1986a: 116). From a methodological point of view, we see here that the notion of predication as derived from the Aristotelian model is still a central one. The theory cannot rely solely on configurations. A dramatic case supporting the persistence of such a model in this framework is given by the contrast between clauses and NPs: 'External arguments are required as subjects of VPs in clauses, as . . . [they destroyed the town], but not as subjects in corresponding nominalizations such as . . . [their destruction of the town] with a subject, and . . . without one [the destruction of the town]. The
reason is that the clause contains a VP predicate, but the nominalization does not contain a predicate' (Chomsky 1986a: 116-17). The debt to nineteenth-century thought is explicit in the direct reference a little later on to Frege's thinking on the subject: 'Rothstein (1983) proposes that the two clauses of the EPP are in fact closely related. We may think of a lexical head as a "lexical function" that is "unsaturated" (in roughly the Fregean sense) if it is not provided with appropriate arguments fulfilling the 9-roles it assigns, and we may correspondingly regard a maximal projection (apart from those that are quasireferential: NP and clause) as a "syntactic function" that is unsaturated if not
'Grammatical subjects': copular sentences in modern syntax
261
provided with a subject of which it is predicated. Then, the EPP is a particular way of expressing the general principle that all functions must be saturated' (Chomsky 1986a: 116). The last two quotations end this section on Chomsky's framework. It might well be the case that these references to Frege's thought do not in the end amount to anything substantial; nevertheless it is worth noticing that a theory of copular sentences still requires reference to such traditional notions as 'subject' and 'predicate'. Modern linguistics has succeeded in breaking many traditional notions down into more abstract entities. Thus, we no longer speak of 'active sentences', 'relativization', 'intransitive verb', except from a purely taxonomic point of view. With the notion of 'subject' and 'predicate', the situation is different: these two notions still appear to play a substantial part in syntactic explanation. Whether we will eventually replace them with more abstract notions or simply abandon them is a question that remains open and calls for considerable further study. I will now sum up and bring this last section of the appendix to an end. We have seen here that the copula is still at the very centre of the debate even in modern syntactic frameworks. Copular sentences challenge the basic syntactic notions on which various models of clause structure are grounded and show that the term 'copula' is not used in a uniform fashion. What we find is that widely differing theories are in circulation and that the theoretical underpinnings of the model that is adopted in any given case are not always made clear. This is true whether the model in question stems from the original Aristotelian theory, which treats the copula as a support for inflectional elements, or from the opposite extreme: Russell's theory of the copula as a predicate of identity, which is hardly compatible with modern epistemological assumptions. We have also seen that in both cases the data are crucially interpreted on the assumption that one postulate remains invariant: that from a predicational point of view the clause structure is rigid. Thus, the subject of the predication is always realized in preverbal position and no exception is allowed. In Jespersen's approach, this follows from the definition of subject as the element that triggers agreement on the verb. In Chomsky's framework, this follows from a configurational approach to grammatical functions. What is the cost of maintaining this assumption? Are there valid empirical reasons for abandoning the rigid structure? In the five chapters of this work I hope to have provided evidence to answer these questions.
Notes
Introduction: four apparently unrelated empirical domains 1 Credits and references will be omitted freely here. See the text and the appendix for a detailed discussion. 2 The text will shift freely from the more compact bracket representation to the tree representation when this is needed for the sake of clarity. In general, unless required as an essential part of the argument, non-crucial phrase markers will be omitted freely and the most simple representation will be used. 3 Of course, this is a very simplified version of the theory of extraction; for example it does not take adjuncts into consideration. We will approach this issue in detail in chapter 1. 4 I am referring to esserci and not simply to essere because the latter will be proved not to respond positively to all tests diagnosing unaccusative constructions. The reasons and the consequences of this will be discussed in chapter 2. 5 The constituent called 'small clause' was originally identified as the complement of consider-type verbs as [sc John a doctor] in Mary considers John a doctor. Copular constructions, then, are currently analysed as structurally equivalent to the passive form of consider-type verbs like John was considered a doctor. 6 In the literature, the ECP module in its various realizations is often referred to as 'locality theory' tout court, with the Subjacency Condition thus excluded. This practice will be followed in the text. 7 For a brief history of the ECP and the Subjacency Condition that goes from the origin to Chomsky (1986b) see Moro (1993b) and Roberts (1988b) respectively. For an overview of the Minimalist Programme see Marantz (1995) and Moro (1996). 8 Essentially, the line pioneered by Chomsky's (1986b) notion of 'extended chain' and Rizzi's (1990) proper head government by agreement will be followed here. The natural intersections with alternative models will be highlighted in the text. 9 Data will mainly be taken from Italian and English. A gloss will be given in brackets for each example together with a translation (in inverted commas) when the corresponding sentence is grammatical (and the gloss itself is insufficient). 1 The anomaly of copular sentences: the raising of predicates 1 To avoid interference from verb movement in interrogatives, I will use examples where the structure in question appears as an embedded clause. 2 More precisely, 'an empty category must not be unbound within a maximal 262
Notes to pages 19-26
263
projection that constitutes a left branch . . . there is no prohibition against a governed empty category itself constituting a left branch' (Kayne 1984: 168). Kayne also noticed that 'the idea that "left" vs "right" plays a role in the distribution of empty categories' was originally proposed by Cinque (1978: 347). See also Pesetsky (1982). 3 Notice that this structure with its architecture, including the subject-object asymmetry, is not dictated by any sort of logical necessity. In principle, it would not have been impossible for sentences to exist in which the subject and the object were totally symmetric: in many artificial languages this is in fact the case. But this is simply not true for the case of natural languages: 'human languages do not adopt the principles familiar in modern logic. Rather, they adhere to the classical Aristotelian conception that a sentence has a subject and a predicate, where the predicate may be complex' (Chomsky 1988b: 54). As for the so-called Comp(lementizer)-node (CP) which is also a part of clause structure, see the original work of Bresnan (1970), Chomsky (1986b), Rizzi (1995) and the appendix of this volume for a brief discussion; for our purposes here, we can simply concentrate on the inner part of clause structure disregarding complementizers. 4 See Graff! (1980) for a critical discussion of Greenberg's (1963) generalizations andX'-theory. For recent developments of X'-theory, instead, see Kayne (1994), Chomsky (1994) and Brody (1994). 5 Except where a detailed representation is necessary, the simplest representation of clause structure will be used. 6 This does not imply that the two sentences are synonymous in every respect. For example, they might have different informational properties. We will not explore this issue here. 7 For a brief history of the analysis of copular sentences see the appendix at the end of this book. 8 We will consider here 'A'-movement' in the traditional sense, as a movement from an argument position (specifically, subject and object position) to a nonargumental position (spec-CP, IP adjunction). Although the very notion of 'A-' vs 'A'-position' is now being re-examined (see Rizzi 1991a, Mahajan 1990, Chomsky 1993 and Moro 1993a), this distinction will be fully maintained in this section in view of its descriptive utility. 9 To avoid interferences with movement of the verb to Complementizer (C°) in Italian (see Rizzi 1991b), embedded copular sentences will be tested. 10 This supports the so-called 'correspondence hypothesis' to the effect that Qraising and wh-movement are subject to the same licensing conditions. See Longobardi (1991b) for a detailed discussion of this matter. 11 Notably, the two readings appear to be the same as in passives: ogni libro fu acquistato da moltistudenti (every book was purchased by many students). In the last section of this chapter 0-role assignment in predicative nominals will be analysed. 12 For the sake of clarity, the version of c-command relation we will use here is as follows: a node oc c-commands a node p if oc does not dominate (3 and the first node dominating a also dominates (3.
264
Notes to pages 27-34
13 Another perhaps clearer example of the sentential reading is given by yes/no questions: q a b
John hasn't manyy books [hasn't]; John tj many many bo books?
Although this is not the place for a full explanation of this contrast, we may notice that the interpretation of the two sentences differs sharply in spite of the fact that the c-domain relations between the negation and the quantifier are the same in each case (as in the pair discussed in the text). In the first case, the sentence is ambiguous as expected, but in the second one it can only mean 'isn't it the case that John has many books?' 14 Excluding, of course, cases where V is complex, such as those involving past participial agreement triggered by clitic movement, as below where part of the complex verb agrees with the object: Gianni lej ha vist-e tj (Gianni them-fem.plur. has-sing.third-person seen-fem.plur.) For these cases see La Fauci (1984) and Kayne (1985), (1989b). 15 Henceforth, I will refer simply to a 'theory of copular sentences' as a way of avoiding the longer formulation 'theory of nominal copular sentences of the kind DP-copula-DF. 16 The analysis of adjunct APs/PPs as small clauses was proposed in Chomsky (1981) and it reflects a minority view. In chapter 4 some independent evidence in favour of this analysis will be offered. 17 In Moro (1988) it was proposed that the copula is not a V° but rather the direct spell out of an inflectional component, namely T° (and that it selects an AgrP). As far as the subject-object asymmetry is concerned, there seems to be no empirical reason to distinguish between the two alternative theories. Thus, the copula will be indicated simply as a V° in the text. 18 It will be recalled that Burzio's generalization states that a verb assigns Accusative case if and only if it assigns the external 6-role to the subject position (see Burzio 1986). 19 The alternative way to satisfy the Case assignment and the Extended Projection Principles, namely by inserting an expletive in spec-IP, will be discussed in the next chapter. 20 Government is essentially a restrictive version of c-command (see fn. 12 in this chapter). A head or a maximal projection governs everything that it c-commands within the minimal maximal projection it is contained in. In other words, government is c-command between two elements that are not separated by a maximal projection. However, this version is too restrictive. In general, there is no overall agreement on the structural limits on the government relation. A certain number of exceptions must be allowed for, the core cases involving IP and X°-movement. Leaving aside for the moment the case of head movement (which is generally dealt with by some version of the Government Transparency Corollary; see Baker 1988 and references cited there), one can follow Chomsky (1986b) in assuming that IP is 'defective', that is, it can only be a barrier to government by inheritance, or else Cinque (1990a) in assuming that a maximal
Notes to pages 34-41
265
projection is a barrier only if it fails to be selected by a head non-distinct from [+V] and provided that C° is included in the proper class of heads. The fact that the small clause is transparent to the government relation can be explained using the same kind of reasoning. 21 In the most recent version of the generative framework, known as the 'minimalist programme', the ECP has disappeared (as a consequence of the fact that government itself has disappeared). The government condition on traces is replaced by general principles of economy that compare derivations and choose the chains where the number of steps is highest or, equivalently, the distance between each step is smallest (see Chomsky 1988a, 1993 and 1994). 22 We will see that as far as extraction from a small clause is concerned the difference between Rizzi's and Chomsky's accounts is mainly terminological. For a critical account of the development of the ECP see Moro (1993b). 23 We will follow Chomsky (1986b) in assuming that the agreement features necessary to satisfy the ECP can also be abstract. This is needed to explain raising from uninflected IPs like: Johnj seems [iP tj to be [Sc tj the culprit]]
24
25
26 27 28
29
30
The ECP is fulfilled by the intermediate trace in the same way as it is in cases where the copula is inflected. The hypothesis that predicative nominals can raise to the subject position, and the corresponding terminology introducing inverse sentences for the first time, was originally presented in Moro (1987) and first published in Moro (1988). Indices here are to be considered as mere notational devices indicating a collection of ^-features. In particular, no commitment is implied on the question of their 'referential' values as assumed by Rizzi (1990) and Cinque (1990a); indeed, predicates cannot be referential by definition (for a criticism of the notion of referentiality as applied to noun phrases see Moro 1988, essentially following Geach's 1962 principle known as Buridan's law, cited in Higgins 1973 and Den Dikken 1995). For a different account of these data originally presented in Moro (1988) see also Moro (1991). I am indebted to Gennaro Chierchia for a stimulating discussion on this topic. As far as I know, if the unified theory of copular sentences adopted here proves correct, this will be the only case where verb agreement is determined by a predicate. For a discussion of the relevance of verb agreement in general see Graffi (1991) and the appendix at the end of the present volume. To put it metaphorically, agreement would be the 'particle' mediating the 'long distance action' between a trace and its antecedent. I will not examine the special status of intermediate traces (see Moro 1993b). I will simply adopt the standard view that intermediate traces are deleted at the LF component. Notice that this structure appears to be compatible with the minimalist account of locality theory based on economy principles. Although a detailed discussion would require a much deeper reflection on the structure of small clauses, the basic lines of reasoning can be sketched here. In fact, the chain established by the predicative DP crossing over the subject is essentially not distinct from the chain established by LF movement of the object to spec-Agr-o-P crossing over
266
31
32
33
34
Notes to pages 43-6
spec-VP. This would be the shortest possible path to such a position, in fact the lower subject position and spec-IP position would be 'equidistant' in the technical sense specified by Chomsky (1993). This could either be regarded as a consequence of the fact that the subject is an adjunct (if the small clause is considered as an adjunction structure) or as a consequence of the fact that the head of the small clause is incorporated into the copula (if the small clause is the projection of a head). But, again, a full treatment would require a deeper understanding of the internal structure of the small clause (see Cardinaletti and Guasti 1995). I am referring here to English and Italian. Chomsky assumes that languages may differ with respect to the point of the derivation where movement takes place (see Chomsky 1993, 1994). Perhaps the subject-in-spec-VP hypothesis could shed light on this matter. At least at the descriptive level, we could assume that the indefinite article can be licensed only if it is governed by a lexical head at some point in the derivation. Since the copula does not meet this condition, i.e. it is not lexical, the article cannot in fact be licensed (cf. a fool came). Henceforth, we will refer to the basic paradigm in (43). To avoid confusion, then, the reader is reminded that when I refer to the 'subject' or 'predicate' of a certain copular sentence, I will be referring to the grammatical functions assigned to the (chains of the) two DPs contained in the small clause. For the sake of clarity, I am referring to the classical violations yielded in examples like: a b
* whOj do you wonder [CP what carj [IP tt fixed tj]]? you wonder [CP whOj [IP ts fixed this car]]
For a critical discussion see Manzini (1992) and references cited there. 35 As we have already observed, locality conditions can be expressed in various ways according to the particular framework that one chooses. But whatever the chosen framework, it remains clear that the offending element is the predicative DP which blocks the path of the subject towards a higher position. We can briefly review the possible different implementations. If we adopt the version of the ECP suggested by Chomsky (1986b) this could easily be captured by saying that the extended chain would be blocked. Agreement is already established with the raised predicate, thus the trace of the other DP cannot be licensed. Manzini's (1992) system, on the other hand, would treat these same facts as a violation of her principle of addressing: the intervening 1° would have the wrong address and block the chain. Interestingly, it does not seem that Rizzi's (1990) version of the ECP can be immediately applied to the case we are discussing. In this system, a trace satisfies the ECP if it is governed within the first projection of a lexical head (V° being lexical). Since the copula is a V°, the wrong prediction is made. As an alternative to this line of reasoning, one could assume that a Relativized Minimality effect is responsible for these violations, either assuming that spec-IP is regarded as an A'-position (which I rejected in Moro 1993a contra Moro 1988) or reformulating head-government along the lines suggested in Moro (1993b). In all probability our hypothesis of inverse copular sentences would allow correct predictions to be made in the minimalist framework as well. Thus the extraction
Notes to pages 50-5
36 37
38
39
40
267
of the subject from an inverse sentence would give a derivation that does not correspond to the smallest possible length of successive steps of a chain; indeed a simpler alternative is theoretically available: this would involve the element in question moving through spec-IP, exactly as happens in the derivation of canonical copular sentences. I use 'intensional' here in the sense of Roberts (1988b): a definition is intensional to the extent that the objects that satisfy it are not just put together as a mere list. The crucial empirical reason for adopting this version of the Subjacency Condition is based on the possibility of extracting phrases from the subject in ciand there-sentences (see section 2.2). Following current assumptions to the effect that noun phrases can be considered as projections of determiner phrases, we will update the list of X°s by adding D°. Clearly, this functional head cannot be considered to realize a functional projection of the SC. The same representation for small clauses as AgrPs is adopted in Chomsky (1988a), (1993). Notice that, in this case, the idea is only related to Pollock's (1989) insofar as theory and terminology are concerned. Similar cases occur in other languages. For example, it is a very well-known fact that in German there is a similar phenomenon. Take for example: a b
das Haus ist neu (the house is new — agr./null agr.) das neue Haus (the new +agr. house)
Although the extension of this phenomenon across languages and its correlation with respect to Case assignment are matters that call for careful study, we can just limit ourselves to this brief note (I am indebted to Giorgio Graff], Alessandra Tomaselli and an anonymous reader for the suggestion about the German case). 41 Of course, technically it is always possible to assume that an Agr° is present even in cases where we do not see any overt agreement. In such cases we would speak of a 'null' Agr°. I will not adopt this solution, though, because it seems to me to be totally ad hoc and, what is perhaps more important, incompatible with cases where agreement is actually overt, as in the Russian examples from Jespersen (1924) quoted in the text. 42 Moro (1988) concluded, by observing, that small clauses could be regarded as projections of an abstract predicative head, and so as predicative phrases. I will not pursue this line of reasoning here. For a similar proposal and a much more detailed discussion see Bowers (1993). 43 Similar arguments can be construed by looking at floating quantifiers. These have been analysed as involving movement of the inner NP to the spec of a higher functional projection, leaving the quantifier in situ (see Giusti 1992 and references cited there). Examination of inverse copular sentences shows that this process is barred in these constructions: a b
la causa sono [DP tutti loro] (the cause are all them) * la causa sono loroj [DP tutti tj] (the cause are them all)
268
Notes to pages 56-9
Once again, this supports the idea that small clauses do not contain inflectional heads. 44 In Cinque's (1990a) system this would give the correct result if one could assume that this head is non-distinct from [+V\. this head would plausibly select the predicate, neutralizing its barrierhood and making extraction possible. Again, it would be very ad hoc simply to assume that this head bears the [+V\ features, unless some independent evidence were found. 45 One could object that the wh-element under discussion in (88) is linked to the subject position rather than to the predicative one. In other words, (88) could be regarded as the counterpart of a sentence like a picture of the wall is this one, where a picture of the wall is the raised predicate. This would also be problematic for our approach since we observed that the subject of an inverse sentence cannot be extracted. To prove that this is not the case it is sufficient to construct a relative clause of this kind: a
* John found whatj you think [a picture of the wall]j was [Sc tj tj]
In such a case, what must be analysed as a referential noun phrase (as opposed to a predicative one) and the sentence is ungrammatical. Notice that this is the result of the syntactic construction involved and has nothing to do with the nature of what. In fact, in the associated canonical sentence what can be associated to the subject position: b
John found whatj you think was [Sc U a picture of the wall]
Similar considerations would apply to who. Consider the following sentence: c
who do you think the culprit to be?
Potentially, one could analyse this sentence as a case of wh-extraction of the subject of an inverse sentence. That this is not the case can be proved again by constructing an example involving a relative clause like the following: d
* I met John whOj I consider the culpritj to be [Sc tj tj]
As in the case of what, the wh-element cannot be interpreted as a referential element (in the intended sense). 46 I assume that empty D°s can contain abstract agreement features paralleling the usual assumptions concerning empty C°s (see Chomsky 1986a, Rizzi 1990). For similar considerations involving P°, see Riemsdijk (1978a). 47 For the sake of simplicity, the AP has been indicated as internal to the NP. A more accurate representation would include functional heads for noun phrases (like Agr°; see Cinque 1992 and references cited there) and one could indicate the AP as higher than the NP but still lower than the D°. Notice that what can cooccur with part of the NP it is related to as in what book did you read? and the more structured what kind of a book which is the interrogative counterpart of this kind of a book/a book of this kind. This implies that what can occur either with an empty N°([NP what [No e]]) or with an overt one {[NP what [N° book]]). In Italian, the alternation between che (what) and che cosa (what thing) supports this conclusion, showing that there can be a sort of a dummy head, cosa, which fills the N° position.
Notes to pages 59-62
269
48 Sylvain Bromberger (p.c.) has pointed out to me that this analysis of what as whcounterpart of a predicate provides a straightforward way of accounting for the very different interpretation of two superficially similar sentences such as the following: a b
[which (one)]i is [ sc tj [water]]? whatj is [ sc [water] tj]?
In (a) (a canonical sentence) we ask which element has the property of being water. In (b) (an inverse sentence), on the other hand, we ask what are the properties of the element named 'water'. 49 Of course, this fact is reproduced in Italian: a
* [quale causa della rivoltajj Gianni crede che una foto del muro fu tj? (which cause of the riot Gianni thinks that a picture of the wall was)
Care must be taken when considering cases like the following: b
la causa della rivolta, quale Gianni sarebbe . . . (the cause of the riot, such as Gianni would be)
In this case, we are not moving a predicative DP out of the small clause. Rather, we are extracting quale from it as a pro-NP paralleling the case of what (see the next section for an analogy with lo). We will not follow Longobardi (1985; reproduced also in Salvi 1990) in considering the following case as grammatical: c
*e il suo migliore amico che Mario e sempre stato (is the his best friend that Mario is always been)
50 The analysis here is limited to cases where ne is the complement of the head of the NP (see Cinque 1991, Giusti 1992 and Moro, in press, among others, for discussion on the categorial status of ne). In Moro (1988) it was observed that there are cases where ne-cliticization appears to be blocked even from a predicative DP. Consider the following contrast: a b
di avvocati, Gianni nej conosce [DP uno tj buono] (of lawyers Gianni of-them knows one good) * di avvocati, Giannij nej e tj [DP uno tj buono] (of lawyers Gianni of-them is one good)
These violations were treated as analogous to Rizzi's (1985) cases like: c
* LorOj sij sono t4 [simili tj] (they to-each other are similar)
where the crossing of two coindexed chains is assumed to yield a violation of the 0-criterion. 51 The neutral term 'establish' is used here in a wide sense to subsume both incorporation and checking as proposed in Chomsky (1993). Notice also that the split-INFL hypothesis (as in Pollock 1989 and Moro 1988) is irrelevant to the present discussion. 52 I will disregard the quasi-argumental occurrence of this element (see Chomsky 1981, 1986a), such as in pro piove (pro rains; 'it rains').
270
Notes to pages 64-9
53 Mutatis mutandis, this is essentially the analysis given in Chomsky (1981: 88, 280), following suggestions by Luigi Rizzi and Giuseppe Longobardi respectively. 54 Notice that there can be canonical copular constructions of the kind pro copulathird sing, pronoun-first sing, but crucially the pronoun in such a case does not show Nominative: pro e me/*io (pro is me/*I) 55 Incidentally, notice that copular sentences of the type 'copula - proper names' with first person singular copula are fully grammatical: prOj sono [tj Gianni] (pro am Gianni) 'I am Gianni' This is only compatible with the assumption that a proper name can play the role of a predicate. See also section 4.2.1. 56 Alternatively, it seems not unreasonable to assume that the preverbal DP in an inverse copular sentence is in a spec-position of a functional projection (i.e. Focus Phrase or Top Phrase, following the terminology of Rizzi 1995). 57 The identification of the lexical content of pro by an adjunct DP can be regarded as a case of secondary predication. This is not the only case, since the same assumption must be made independently for all cases of rightward movement, as in: . . . [IP pro telefona a Maria] . . . Gianni . . . Also, in this case, the adjunct DP Gianni must necessarily specify the lexical content of pro. 58 Incidentally, rightward agreement was at the root of the theory of inverse copular sentences and the proposal to split INFL into two independent nodes suggested in Moro (1988), independently from Pollock's (1989) theory. The idea was to consider the copula as a pure spell out of the INFL node (as in the traditional Aristotelian sense: see the appendix at the end of the volume) and INFL as composed of Tense and Agreement (as proposed in Chomsky 1981 among others). Thus the innovative step was to split INFL and consider grammatical functions as linked to the Agr features: [TpT°[A grP DPAgr°DP]] This move let me propose that a predicative element could occupy the most prominent position of the clause structure, thus instantiating inverse copular sentences. The reason why I now adopt a different proposal involving pro is due to the impossibility of dealing with cross-linguistic differences by relying on the structure above. 59 The view is restricted to subject agreement. Whether or not this claim can be extended to other instances of pro is an empirical matter that cannot be pursued here. For the original proposal of this analysis of preverbal subject in Italian as higher than spec-IP see Moro (1991), (1993a).
Notes to pages 69-73 271 60 As a theme for future research, it would be interesting to set up an experiment to test whether the pro-drop parameter is set before, after, or at the same time as, inverse structures begin to appear in the child's syntax. In fact, since inverse sentences require raising of the predicate to spec-IP a natural prediction is that they do not show up before the inflectional system is mature. 61 Of course, this assumes that this fundamental principle is actually part of the grammar of the child. Whether or not this is true can only be tested empirically and I am not now in a position to offer any relevant data. 62 In fact, this new type of evidence could perhaps be regarded as stronger. Potentially, with examples like chiama (calls), the child could just skip the subject position altogether and produce an incomplete structure. With inverse copular sentences, on the other hand, he will be forced to introduce pro, to avoid having to have agreement of the verb with the overt preverbal DP. 63 For the sake of clarity, a minimal portion of clitic movement can be represented as follows: ...[x-cliXo]...[AgrPtiAgro...ti... As indicated here, the first step of the clitic chain involves a maximal projection (hence, a spec-head relation with an agreement head). The last step, on the other hand, involves a head-chain yielding incorporation (leaving aside the distinction between adjunction-affixation and substitution). For a detailed discussion see Belletti (1994) and references cited there. 64 Incidentally, notice that lo is also a suitable clitic for AP predicates. As for PP predicates, see chapter 2. 65 Cocchi (1994) and Friedemann and Siloni (1994) have argued that participial AgrP cannot be considered as an Agr-oP. Along the same lines, it is interesting to notice that even if the past participle occurs in an inverse copular sentence, cliticization of the postverbal subject DP is still blocked, whatever the clitic employed: a b
la causa della rivolta sono (state) le foto del muro (the cause of the riot are been the pictures of-the wall) * la causa della rivolta lej/lOj sono (state) t; (the cause of the riot lello are been t)
This result is expected only if Agr-o is not identified with past participle AgrP. In fact, it would be easy to show that the anomalous behaviour of inverse copular sentences can be detected even if the structure has been 'enriched' by a further Agr° projection correlated with the past participle. 66 Interestingly, this would also allow us to understand why the predicate of the complement of ritenere-type verbs in Italian cannot be cliticized by lo: a b
Gianni ritiene Mario il colpevole (Gianni believes Mario the culprit) * Gianni IOJ ritiene Mario t; (Gianni lo believes Mario)
Following Hale and Keyser's (1991) framework, ritenere should be considered the result of a lexical incorporation of a lower predicate or, equivalently, the
272
Notes to pages 75-82
complement of ritenere should not be considered a 'bare small clause' (as in the case of the copula) but rather a richer constituent projected by an abstract head. The ungrammaticality of the propredicative lo in (b) can then be regarded as a violation of the Head Movement Constraint. The head of the predicative noun phrase cannot be cliticized because it would cross the head of the predicate which has been incorporated into the matrix verb. 67 Notice that ne has an intermediate status. It is not inflected (like lo) but it can activate intermediate agreement (like le, la, li etc.): a b
i ragazzi hanno vist-o tre ragazze (the boys have seen-neuter three girls) i ragazzi nej hanno vist-e tre tj (the boys of-them have seen-fem.plur. three)
We can reasonably assume that ne is simply compatible with agreement features which are not assigned if they are incompatible. Thus, when ne is involved in copular sentences, cliticization is fully licensed (pace the Subjacency Condition): c
68
69 70
71 72 73
74
75 76
delle rivolte in Milano, Gianni nej e stat-o il responsabile tj (as for the riots-fem.plur. in Milan, Gianni of-them was been-masc.sing. the responsible)
That agreement features are not inherent to ne is shown by past participle agreement with the subject. Of course, this is not to say that there are no references to predicative nominals in this and other texts. All I am saying is that the class of predicative nominals is at best regarded as of marginal importance, when not simply treated as a mere taxonomic label. For a historical survey of the relevance of the notion of predication within generative grammar see the appendix. I am referring in particular to what has become known as 'Cinque's generalization' since the original paper by Guglielmo Cinque (here, Cinque 1980) was published. See also Giorgi and Longobardi (1991) for a detailed discussion of this issue. For a detailed critical discussion of the theory of argument structure, see Marantz (1984), Jackendoff (1990), Grimshaw (1990), and references cited there. For the sake of simplicity, verb movement is not indicated. In fact, the rigid assignment of 0-roles is generally assumed to hold already at the more primitive level of d-structure. For a more advanced approach to 0-role assignment see the last chapter of this book and references cited there. Both arguments within the NP have been shown to the right of the head N°.This contrasts with the case of VPs, where the subject is on the left (in spec-VP) while the object is on the right (adjacent to V°). For the linear and hierarchical order of arguments within NPs see Giorgi and Longobardi (1991); for our purpose, this difference will have no empirical effects. Notice also that the linear order of di Dante and per Beatrice appears to be free to most speakers. As for 0-role assigned across the copula see Grimshaw (1990). For the sake of simplicity, verb movement to the I°-system is not indicated. As for the linear order of Dante and Beatrice, see fn. 74.
Notes to pages 82-7
273
77 To avoid confusion, it should be noticed that the string il suo desiderio di Dante is grammatical, but with the reading 'he or she desires Dante', that is, with Dante bearing 62 and suo Q\. 78 I am indebted to Guglielmo Cinque for this example (and a never-ending discussion on the matter). 79 Although I will not pursue this idea here, one could explore the possibility that the class of nouns allowing pseudo-extraction is made up of those nouns which correspond to verbs allowing the internal object construction. Thus, we could distinguish desiderio (desire) from foto (photograph) because the corresponding verbs behave differently: a b
desidero un desiderio impossibile (pro desire a desire impossible) * fotografo una foto impossibile (pro photograph a photograph impossible)
I am grateful to Liliane Haegeman and Luigi Rizzi for a helpful discussion on this topic. 80 I say 'deprived' in order to maintain a neutral position with respect to the hypothesis of extraction. In a certain sense, the idea that a noun phrase should be deprived of one argument to play the role of a predicate could be considered as indirect evidence in favour of the theory of predication as saturation. The core assumption of this theory is that predication is in fact established whenever an argument saturates an empty slot contained in a maximal projection. This theory, originally developed in Fregean-style analysis, has been adopted in modern grammatical models: in Montague grammar (see for example Montague 1973) as well as in generative grammar (see Williams 1980, Rothstein 1983 and Chomsky 1986; for a critical approach to this topic see also Moro 1991 and the appendix of this book). 81 I am maintaining the assumption that the copula cannot assign any 9-role, it being just the support (or the spell out) of the I°-system (Moro 1988). Were this not the case, a sentence like the following: Beatrice e il desiderio di Dante (Beatrice is the desire of Dante) would now violate the 0-criterion, for Beatrice would receive two 9-roles, one from the N° and one from the copula. 82 For a different approach to unaccusativity see the last chapter of this volume. 83 For the sake of simplicity, the discussion here will be restricted to clausal constituents and the two DPs will be considered as c-commanding the anaphor in (153); a more appropriate terminology, originating in Chomsky (1981), would refer to governing categories. 84 It is implicitly assumed that i is different from/. That is, cases like: John; thinks that hef loves himselfj where / can be interpreted as equal to i, are excluded. 85 For an example that does not involve PRO consider the following English case:
274
Notes to pages 88-93 [John and Mary]; considered [ sc [these]k [[each other's best performances]]
The anaphor each other is bound by John and Mary crossing over the subject of predication (I am indebted to David Pesetsky for having suggested this example to me). 86 That predicative nominals behave differently from argumental DPs is a rather well-known fact. In Moro (1988), for example, the following asymmetries were reported (see also Giorgi and Longobardi 1991 among others for references and discussion): a b c d e
Gianni; conosce [DP un suo; ammiratore] (Gianni knows a his admirer) * Giannij e [DP un suOj ammiratore] (Gianni is a his admirer) * Gianni; conosce [DP un ammiratore di se stessOj] (Gianni knows an admirer of himself) * Maria conosce [DF un ammiratore di se stessa] (Maria knows an admirer of herself) Gianni; e [DP un ammiratore di se stesso;] (Gianni is an admirer of himself)
These contrasts show that the predicative nominal extends its CFC to include the subject of the matrix clause (in contrast to what happens with the argumental DP). Unfortunately, these asymmetries do not seem to be of much help in understanding the specific problem regarding Binding theory presented here. For similar considerations see also fn. 33 of the appendix. 87 It goes without saying that we have not exhausted the empirical domain of copular sentences. Many interesting facts still deserve an explanation; for example, the so-called 'backward reflexivization' discovered by Ruwet (1975). The phenomenon can be simply illustrated by the following examples: a b c d
le sue; lettere sono l'alibi di Gianni; (the his letters are the alibi of Gianni) l'alibi di Gianni; sono le sue; lettere (the alibi of Gianni are the his letters) le lettere di Gianni; sono il suo; alibi (the letters of Gianni are the his alibi) * il suo; alibi sono le lettere di Gianni; (the his alibi are the letters of Gianni)
Only in one case is it impossible for the pronoun to be bound by Gianni and, as far as I know, no obvious reason has been given to explain this fact. The only improvement offered by the theory of copular sentences presented here is that we can now refine the description of such a receding phenomenon by indicating that the problem arises for a pronoun contained in the predicative noun phrase of an inverse copular sentence. Whether or not this can be related to the particular focus properties of such a structure is a question that cannot be addressed here.
Notes to pages 95-109
275
A possible line of research stemming from the theory of copular sentences presented here has been explored by Kayne (1994), Den Dikken (1995) and Zamparelli (1995) who independently proposed that instances of predicative raising across a left-branch subject (as in inverse copular sentences) can be found in the domain of noun phrase structure such as in the following example taken from Kayne (1994: 106-10): e
that [DP [NP idiot]] of [[DP a doctor] tj]]
2 The syntax of ci 1 See the appendix of this book for a brief history of the analysis of this empirical domain and for references cited there. 2 The following is a simple definition of 'binding': the node a binds the node (3 if and only if a c-commands (3 and they are coindexed. 3 The representation given here essentially reproduces the original one given by Chomsky (1988a). Later (Chomsky 1993), the DP was placed on the left oi there, i.e. [DP a man [DP there]], on a par with all other instances of syntactic movement. 4 More precisely, Burzio (1986) suggested that ci is inserted directly at d-structure in Italian, in contrast to the later s-structure insertion oi there in English. In any case, replacement would take place at the same level, namely LF. 5 The glosses show that extraction is also possible in English: we will consider the English examples later. The 'semantics' of existential sentences will also be discussed separately in the next chapter. 6 The difference between replacement and affixation will not affect the argument here. Thus, I will simply adhere to the most common practice of referring to the process straightforwardly as 'expletive replacement'. 7 For the sake of simplicity, recall that 'spell out' is the point where derivations undergo phonological interpretation. After spell out takes place, computation continues but no further access to the lexicon is assumed to be possible (see Chomsky 1993, 1994). 8 Williams' (1984) hypothesis that the copula selects a bare noun phrase (playing the role of a predicate) in a there-sentence would solve this specific problem. However, as is easy to deduce from the discussion, this idea would go against many other facts, such as those involving cliticization of the postverbal DP. Thus, it will not be adopted here. 9 Incidentally, the fact that only ne-cliticization is possible would appear to undermine the idea that the postcopular DP of a there-sentence is a predicative noun phrase as proposed by Williams (1984). 10 Indeed, ci overtly occurs as a (nominal) propredicative element of a small clause in Italian substandard constructions like: a b
Gianni Papa, io non ce lo vedo/faccio (Gianni Pope, I not ci him see/make) 'Gianni Pope? I cannot consider him to be like that' ci sei o ci fai? (ci are or ci make) 'are you really like that or are you pretending to be like that?'
276
Notes to pages 109-26
11 An alternative analysis (see Moro 1991: fn. 24) would be one where the AP/PP adjunct is in fact an adjunct to the subject (a) or the 'head' of a further embedded small clause (b): a pro [cij sono] [ sc [DP DP AP/PP] tj] b pro [cij sono] [ sc [ sc DP AP/PP] tj] Notably, (b) would now be compatible with Kayne's (1994) antisymmetry. Since at the moment I do not have any cogent empirical reason to choose among these analyses, I will leave the option open (see also the discussion concerning example (49)). What is crucial here is that ci is raised from the predicative position and that the AP/PP is not playing the role of the predicate in the small clause. 12 I am disregarding here the interpretation in which perche (why) relates specifically to sembrare, as a question on the reason for thinking that /?, rather than the reasons for/?. This interpretation is in fact available with both sentences. 13 Notice that this process is essentially not different from V°-to-P raising. This process must be kept theoretically distinct from VV-to-C0 raising, otherwise in verb-second constructions or in aux-to-Comp, the lower DPs are always Lmarked voiding barrierhood of the lower IP. 14 An alternative way of looking at this process is to think of it as a consequence of the Government Transparency Corollary (see Baker 1988). The analysis will not be developed here. 15 These data were originally discussed in Moro (1990) and (1991). 16 Case assignment to there is not a problem for expletive replacement theories. Case is assigned to there because at LF the subject is replaced by there. Nevertheless, inverse copular sentences involving full predicative DPs show that Case must be assigned to the subject position although the subject does not actually display it as an affix. 17 I am aware of the fact that this theory raises an immediate theory-internal question: are there canonical counterparts to there-sentences? If so, what is their interpretation? Both questions will be directly addressed in the next chapter. 18 If one assumes the so-called 'subject-in-VP' hypothesis, there would no longer be a need to assume rightward movement as assumed in pioneering work by Rizzi (1982). Extraction could take place directly from this position, which would not be possible in non-pro-drop languages, given the usual assumptions on parametric variation. 19 This note concerns contrasts like the following: a * [which girls]; do you think that there are tj in the room? b [which girls]; do you think that there are [pictures of tj] in the room? This type of contrast seems to me to be theoretically relevant. It shows that the conditions blocking movement of a DP (a) are distinct from those blocking extraction from a DP (b). Since it has been shown that movement is blocked by the ECP and extraction is blocked by the Subjacency Condition, the conclusion is that these two principles must also be kept distinct. This is a welcome result in our framework, since the two principles involved have been related to two distinct empirical facts: the ECP is related to the possibility of activating agreement under the spec-head relation; the Subjacency Condition, instead, is related to the selectional property of a head.
Notes to pages 127-40
277
20 I simply indicate the trace of how many as 't' without any commitment with respect to the internal structure of the DP from which the wh-element is extracted. 21 Notice that this type of extraction (disconnecting the actual interrogative element, namely how many, from the rest of the DP) can be overt in French: a b
[DP combien de livresjj a-t-il lus tj? (how many of books has he read) combieni a-t-il lu [DP tj de livres]? (how many has he read of books)
In general, given the new system proposed by Chomsky (1993), it seems to me that these examples involving how many could easily be dealt with by assuming a copy theory for the trace, provided that the analysis of there as an expletive subject is abandoned. 22 I am aware of the fact that the first sentence can be ruled out by appealing to the Definiteness Restriction, banning definite DPs from the subject position of a there-sentence. I will not adopt the Definiteness Effect as a primitive. In the next chapter I will suggest a way to derive it from independent principles. 3 Are there parameters in semantics? The defining properties of existential sentences 1 For a general survey of the issue see Reuland and ter Meulen (1987) and references cited there and the seminal works by Milsark (1974) and Barwise and Cooper (1981). 2 For a critical discussion on parameter setting see Lightfoot (1991). 3 The original formulation of such a radical departure from traditional grammar can be found in Chomsky (1981: 7). 4 Similar considerations, of course, apply to constructions other than passives. See, for example, Chomsky (1977b) and Riemsdijk (1978a) where wh-constructions are discussed, and Cinque (1990a), where clitic left dislocation (CLLD) constructions in Italian are discussed. 5 There may occur with verbs other than be as in there arrived many girls. In this case, these sentences are generally called 'presentational sentences' to distinguish them from existential sentences. In the last chapter, a possible unified account of these structures will be suggested. 6 Noam Chomsky has pointed out to me that in the following sentence existence of a book is entailed only if the content of the items in brackets is present: there is a book (that is) missing on the table I do not have any explanation for this contrast. In general, as noticed in the text, I will not discuss the ontological implications of 'existential meaning' but only attempt a preliminary syntactic rethinking of the notion of 'existential meaning'. 7 Recall that the theoretical premise of this chapter is that the Definiteness Effect should be derived from independent principles. 8 This principle is directly reminiscent of the ^.-calculus: a well-formed formula y can be turned into a predicate by an operation of ^-abstraction which abstracts
278
Notes to pages 141-8
a variable x from y yielding kx[y] (see Chierchia and McConnell Ginet (1990) for references and a critical discussion). The operation of ^-abstraction can be roughly regarded as the equivalent of the operation of subject raising proposed here. 9 In fact, a similar analysis allowing 'short movement' is to be independently assumed for cases like: a b
John thinks that [iP everyone{ [IP ts left]] John wonders [CP what bookj [IP everyonej [iP tj read tj]]]
If short movement were blocked by locality conditions, an instance of the thattrace effect would be expected. Similar considerations apply to the subject of a small clause in an inverse copular sentence with a full DP predicate: c [many students' purchase^ is [ sc every bookj [ sc tj tj]] d * [IP [every book]} [IP [many students' purchase^ is [Sc U tj]]] The subject every book cannot undergo long movement yielding a wide scope reading (d) but, since it must be raised anyway, we may conclude that short movement is allowed by locality conditions (e). 10 Note that here the use of the term 'adjectival' is a technical one, in that it does not coincide with the possibility of occurring in an overt predicative position; as we shall see, for example, no will count as adjectival in this sense (cf. the discussion below). 11 In fact, if empty D°s are assumed to exist, a further possibility arises to the effect that adjectival many is extracted together with the NP from a lower position. Since the issue of the distribution of empty D°s across languages is still under debate (see Longobardi 1991a and references cited there), I will not pursue this possibility here. 12 I would like to emphasize here that by 'existential meaning' I only refer to the interpretation stemming from a specific syntactic structure, namely theresentences, and not to a primitive semantic meaning. 13 Of course, this is not the only case where many has an adjectival reading. In fact, we saw that a sentence like John hasn't met many girls can also be interpreted as predicating many of girls, i.e. John has met girls and they were not many; but this sentence contains a further predicate (i.e. headed by meet). These two predicates (meet and many) combine to yield a more complex property. Indeed, it seems to me not to be unreasonable to conclude that John hasn't met many girls 'contains' an existential sentence, in the intended sense. 14 Its content is to be derived from the discourse: by default, it denotes the whole world, as in there aren't many ideas to discuss; alternatively, it can be restricted to a specific domain, when an adjunct PP is added as in there aren't many ideas in this book. 15 The obligatoriness of the overt spell out with the singular as opposed to the plural remains a mystery as far as I am concerned. 16 See also fn. 10 in this chapter. Incidentally, with the alternative semantic framework that has been generally adopted for existential sentences (since at least Milsark 1974), the occurrence of no would constitute a serious puzzle. 17 See also fn. 32 to chapter 1. To support this conclusion, notice that no seems to
Notes to pages 148-53
279
be unacceptable in the DP subject of inverse copular sentences with full lexical predicative nominals which we have seen do not lexicalize the copula: a there; are [ sc [DP no girls] tj b? * [DP the cause of the riot]; is [ sc [DP n o girls] t,] The latter example contrasts also with the following case where I assume, following Longobardi (1991a), that N° raises to D°: c
[DP the cause of the riot]i isn't [Sc [DP girls] tj
18 For the following contrast between: a * there is John in the garden b there is only John in the garden see Moro (1991), following the analysis of only suggested by James Higginbotham (p.c). 19 For the sake of clarity, recall that our approach differs with respect to Higginbotham's (1987) approach in that the postcopular DP of a there-sentence is analysed as a subject and not as a predicative nominal. 20 Interestingly, extraction from the class of DPs that can occur in f/tere-sentences is generally considered to be more acceptable than from those that give a Definiteness Effect: a
[DP which wall]; do you think John saw [Dp many/few/three [NP pictures of tj]?
b ?* [DP which wall]i do you think John saw [DP the/every/most [NP pictures of If these contrasts can be traced back to a failure of proper government (see Giorgi and Longobardi 1991 and Manzini 1992) then the Definiteness Effect could ultimately be regarded as a violation of a locality condition on movement at LF. However, since this approach needs an accurate empirical analysis that cannot be provided here, we will simply rely on the distinction given in the text. 21 The so-called 'list interpretation' associated with there-sentences involving definite DPs as in there are Tim, Bill and Tom in the room can perhaps be
regarded as a sort of last-resort interpretation for there-sentences containing definite DPs. In this case, these there-sentences are to be regarded as instances of locative inversion constructions (like in the room are Tim, Bill and Tom) involving movement of a phrase to the CP periphery. They are thus rather pseudo-there sentences', not involving the process of lexicalization of the copula that there-raising implies. See also chapter 5 (in particular, fn. 12). 22 Recall that there in fact has characteristics both of maximal projections and of heads, in that it occupies a spec-position but cannot contain either specifiers or complements. 23 As was mentioned in chapter 2, this analysis involving rightward movement seems to be unnecessary if the subject-in-spec-VP hypothesis can be maintained. For the sake of simplicity, this text will continue to assume the traditional analysis. 24 More accurately, one should say that a proper name does not include a D° with
280
Notes to pages 156-8
an adjectival character. In fact, proper names can, per se, be interpreted as predicates, as was suggested in chapter 1 (fn. 55). 25 The simple fact that ci is a clitic is not sufficient per se for the Definiteness Effect to be avoided. In French, the equivalent of there, namely y, is also a clitic but nevertheless the Definiteness Restriction is active: * [[II Vj a [un homme/*Jean tj]] dans le jardin] (// there have a man/Jean in the garden) This is due to the fact that French is a non pro-drop language and the operation of rightward movement is thus banned (cf. *// chante Jean (il sings Jean)). 26 Belletti (1988) independently developed an argument intended to relate the lack of the Definiteness Effect in Italian to the pro-drop parameter; this argument crucially relies on Case theory. The central assumption is that Nominative case can only be assigned to spec-IP and expletives do not transmit Case to postverbal position. Since the only case available for the latter position would then be 'Partitive', this would exclude definite DPs, which are not compatible with it: a
[IP therenom [VP arrive DPpart/*nom]]
If a DP following an unaccusative verb is Nominative, Belletti concludes, this must have passed through spec-IP and then been rightward-moved. We cannot immediately apply this theory to copular sentences, because data show that Nominative case is indeed compatible with a VP-internal position: b
[IP [DP l a causa della rivolta]i [Vp sono [ sc io tj]]] (the cause of the riot is I-nom)
The same conclusion should be extended to unaccusatives, if the analysis proposed in chapter 5 is adopted (see fn. 35 to chapter 5 in particular). 27 As a less common variant, one can alsofindqualcheduna delle ragazze (literally: some-of-one of-the girls), where una is not incorporated and is perhaps Casemarked by qualche - witness the morpheme cT (which appears to be the typical Case assigner). 28 The following case could be regarded as a potential counterexample: [delle ragazze] entrarono nella mia stanza (of the girls came into my room) However, in this case, in fact, the phrase delle ragazze has a very different role from that under discussion in the examples in the text. Delle ragazze is rather to be interpreted as 'some girls', i.e. as containing a quantifier. Thus, there is no reason to assume that an empty category is contained in the DP. 29 For some speakers, a sentence like non ci erano molte ragazze can also have a locative meaning, i.e. 'many girls are not there'. This means that for those speakers the strategy of rightward movement has been applied. This will not affect the argument here, since what we are interested in is the fact that it is impossible for molte in non ci sono molte delle ragazze to have narrow scope. 30 Judgments are quite delicate: although the majority of my informants detected a violation of the Definiteness Restriction, there are also speakers who find this sentence fully acceptable. Apparently, they rescue the sentence as a case of iist
Notes to pages 159-62 281 interpretation' (see fn. 21): for example, they would not accept third person singular in contrast with cases not involving of: there are/*'s many of the girls in the garden there's/are girls everywhere 31 Recall that the property in (iii) might ultimately turn out to be redundant if the class of D° with an adjectival character is proved to coincide with the class of D° allowing extraction; see also fn. 20 in this chapter. 32 For those who are interested in etymology, it might be interesting to notice that location is also contained in this lexical element (exist) since it comes from the prefix ek- (out of) and the root sist- (stay), thus exist would literally mean something like 'occupy a place'. 33 The lack of agreement in existential sentences will be discussed a few paragraphs below. 34 The pro-drop parameter would of course become fully relevant once the Definiteness Effect was taken into consideration: only pro-drop languages freely allow what are apparently violations of the Definiteness Effect, for reasons that I have tried to explain in this chapter. 35 In Jespersen (1924: 156), it is reported that in some Western German dialects gibt agrees with a postverbal noun phrase (as in es geben viele Apfel: it give many apples; 'there are many apples'). 36 I am very grateful to Giuseppina Turano for the Albanian data. 37 The list of languages explored, for example, by Freeze (1992) is quite impressive: it includes Black English, Catalan, Chamorro, Chichewa, Finnish, French, German, Hindi, Hungarian, Icelandic, Italian, Japanese, K'ekchi', Kiliwa, Mandarin Chinese, Mayo, Mixtec, Navajo, Paluan, Palestinian Arabic, Persian, Pokomchi, Portuguese, Quechua, Quiche, Russian, Scots Gaelic, Shanghainese, Spanish, Swedish, Tagalog, Tongan, Trukese, Turkish, Yucatec. Although Freeze (1992) offers a unique amount of data, it seems to me that his analysis cannot be straightforwardly maintained. For example, I do not see how an intermediate projection like P' can be raised to spec-IP, a singularity that can easily be avoided once a unified theory of copular sentences such as the one proposed in Moro (1990) is adopted. Freeze (1992) seems to be unaware of such a theory, although he cites Hoekstra and Mulder (1990) who adopt the analysis of there as a raised predicate from an earlier paper which circulated as Moro (1989). See also chapter 5 for further comments on Freeze's analysis. See Zwart (1991a), (1991b) for a critical extension of the theory of there as a raised predicate in Dutch. 38 Jespersen's intuition has apparently remained forgotten since it was formulated. For a historical survey of Jespersen's theory of clause structure see the appendix to the present volume. 39 The idea that 'the subject is hardly treated grammatically like a real subject' also correlates with a specific assumption that is central in Jespersen's (1924) system. In fact, in this framework the essential property of a subject is that it triggers verbal agreement (see also m.i in the appendix). Recall that in many languages, such as English and Italian, for example, existential sentences are anomalous with respect to agreement (examples are in word-for-word correspondence):
282
Notes to pages 163-72
a there's books everywhere b c'e libri ovunque In both (a) and (b) the subject is plural but the verb is singular. From our point of view, this fact is not problematic. One of the defining properties of existential sentences is that the subject cannot move as a whole, otherwise splitting would not be necessary and the quantificational reading would be readily available. The mismatch in (i)a and b can then be regarded as further evidence for the structural property in question: lack of agreement signals the fact that the subject is disconnected from the spec-IP position. This also explains why agreement cannot fail in Italian if the DP is definite (as noticed in Burzio 1986: 77, 132-3): oi [cij sono/*e] [Sc U tj] loroj] (ovunque) (there are/*is them everywhere) In this case, there can be no splitting: the subject has first been raised to spec-IP and agreement is thus activated to establish an extended chain and thus satisfy the ECP. 40 Recall that there is independent evidence for assuming that pro can play the role of a propredicate. See also the last chapter of this book for a case of locative use of propredicative pro in Italian. I will not approach the problem concerning the 'lexicalization' of the copula in languages with non-overt propredicative elements. 41 A note on terminology: if the analysis presented here proves tenable, the terminology based on the concept of 'definiteness' may turn out to be rather misleading. This terminology relates this phenomenon to 'reference' for it treats DPs as if they were inherently referring to things (see Reuland and ter Meulen 1987 and references cited there). What is supposed to be 'definite' vs 'indefinite' is the set of individuals in the real world to which the DPs refer. Thus, the girls is banned because it refers to a definite set of girls as opposed to, say, three girls. We have seen that this phenomenon has nothing to do with reference. Indeed, it has turned out to be the consequence of purely syntactic facts. This should be kept in mind when using the term 'Definiteness Effect' and this term should be treated as having no extralinguistic connotations. 4 The 'quasi-copula': the role offiniteclauses in sum-sentences 1 In general, an even stronger claim is made: that the finite clause cannot receive Case. This is often referred to as the 'Case Resistance Principle'. For a detailed discussion see Koster (1978). 2 Of course, the problem cannot be simply dismissed by assuming that seem is biargumental, otherwise sentences like the following should be acceptable: * Mary seems that John left 3 An appropriate context to distinguish between the two is not difficult to imagine. The inverse sentence can easily be imagined as occurring in a dialogue such as the following: 'Are you sure that the culprit is John?' 'Yes, I'm quite sure it's John.' As for the canonical interpretation, it would be enough to suppose that somebody in a museum points to a statue of John and says: 'It's John.'
Notes to pages 173-5
283
4 There are certain peculiar constructions that seem to be quite acceptable. Take, for example, the commercial Coke is it. Although it seems to me that it is hard to decide whether we are facing a genuine canonical copular sentence, I have been told by many speakers that these sentences can generally be encountered in jargon (advertising jargon, in the case of our example). (I am indebted to an anonymous reviewer and to Jim Higginbotham for a discussion on this topic.) 5 There is a difference here with respect to seem but it can be attributed to independent factors: a b
[Dp the worst hypothesis^ seems *(tj to be) [Sc [CP that John left] tj] [Dp that John left]j seems *?(tj to be) [ sc t{ [Cp the worst hypothesis]]
This restriction seems to be highly idiosyncratic and is related to semantic properties; witness the acceptability of sentences like this seems the best thing you can do; see also examples in (111) in this chapter and the related discussion. 6 For an alternative approach to this and related issues, see also Lasnik (1989). Incidentally, in certain Northern Italian dialects (for example, in Pavese) extraposition can be signalled by a noun phrase occurring between the verb and the embedded sentence, i.e. insi (so): a
la Maria l'ha dit insi cha la gnara no (the Maria subj.cl. has said so that subj.cl. comes-future not) 'Mary said she will not come'
Crucially, insi cannot occur between the verb and the embedded sentence if the verb is the copula or the quasi-copula: b c
gh'e (*insi) cha la Maria la gnara no (there is so that the Mary subj.cl comes-future not) it's that Mary will not come' a m'par (*insi) cha la Maria la gnara no (subj.cl. to-me seems so that the Mary subj.cl comes-future not) 'it seems to me that Mary will not come'
7 In general, it is assumed that relativization can turn a finite CP into a predicate by X-abstraction (the assumption comes from logic: see, for example, Quine 1950: 132-6 and Chierchia and McConell Ginet 1990): a [DP the man [CP Opj that I never met tj]] will leave tomorrow b * [DP the man [CP that I never met John]] will leave tomorrow This process appears to be unavailable if the predicative link is mediated by the copula. In fact, CPs can never occur as predicates in the typical context of copular sentences: c
* [IP the manj is [ sc tj [Cp Opj that I never met tj]]]
8 Although in Standard Italian both variants given in (23a-b) are attested (i.e. ci is not obligatory), it is interesting to notice that ci becomes obligatory if the sentence is interrogative:
284 a b
Notes to pages 180-8 (c') e che Gianni e partito (ci is that Gianni left) cosa *(c')e? (what ci is)
These inverse sentences involving embedded clause structures have other interesting properties (for example, they cannot be negated or put in the past tense). Exploring them in detail would take us too far from our main topic. 9 This, of course, does not prevent seem from occurring with a subject it. Consider for example: [IP [IP iti seems [ sc tj obvious]] [Cp that John left]] Here, it is raised from the subject position. The predicative position within the small clause is occupied by a full AP (obvious). Incidentally, notice that a sentence like it seems so is to be considered as an inverse structure, where so is a place holder for a finite clause (cf. IfsolCP, then. . . and John said sol CP. . .). 10 Behnis (1986) convincingly argues that when het is present, the CP is extraposed. Thus, the analysis of blijken given by Bennis is very close to that of affirm and obvious which we independently motivated here: the pronominal forms are generated in postverbal argumental positions. 11 The difference compared to Bennis (1986) here is clear: he assumes that 'het should be analysed as a referential expression in all its occurrences' (1986: 163) and that this analysis 'carries over for English if (1986: 313 - see also Cardinaletti 1990, who extended this analysis to German, Vikner 1995 and Tomaselli 1990). As Bennis noticed, this derives the correct result for blijken but cannot be straightforwardly adopted for schijnen. The extra assumption is required that schijnen selects only IPs and DPs. This assumption is not necessary in the analysis being presented here. 12 Guglielmo Cinque (p. c.) has pointed out that the case involving obvious could also be explained as a violation of a factive island. He noticed, however, that this violation persists in the presence of a modal such as would: * howj would it be obvious [that John left tj] This will exclude the explanation based on factivity in favour of our argument. 13 We have not analysed here sentences involving likely, such as: a
it is likely that John left
Nevertheless, it seems that likely behaves like obvious and the passive of affirm with respect to adjunct extraction from the embedded clause, showing that extraposition has taken place. Thus, we will not regard the it in (a) as a propredicate but rather treat it on a par with it in it is obvious that John left. On the other hand, it is a well-known fact that likely is similar to seem and to the passive of affirm and contrasts with obvious in that it allows raising from the infinitival construction: b
John is likely/afflrmed/*obvious to leave
Since we are limiting our analysis to the occurrence of embedded sentences with finite verbs, we will not explore this further distinction.
Notes to pages 190-2
285
14 Lasnik and Uriagereka (1988: 151) also noticed that as far as Binding Theory is concerned, the following sentence should be grammatical, contrary to facts: * Johnj seems that hej likes tj In fact, the anaphor tt is bound by he coindexed with John in its minimal domain. 15 Notice that anaphors and variables appear to satisfy locality conditions in a different way, since VP-adjunction is available to wh-movement as an A'position: a * Mary thinks that Johnj seems that we love tj b WhOj does Mary think that it seems we [Vp tj [VP love tj]] This might not be contradictory per se, but it should be noticed that it is hardly compatible with a derivational approach. 16 Along with Super Raising, there is also Super Passive (see Rizzi 1990: 84ff.): a it seems that Johnj was told tj that it was raining b * Johnj seems that it was told tj that it was raining Super Passive is theoretically important for it shows that the lack of 6-government (i.e. government by a head assigning a 6-role) is not responsible for the ungrammaticality of Super Raising. 17 To avoid confusion, notice that movement of an argument from a finite CP to the subject position of the matrix clause (as in * Johnx was affirmed t\C° t left) is excluded for a different reason. As has been shown, the finite clause occurring with passives cannot stay in the object position. Rather, it is extraposed, it being generated in the object position. Thus, movement of the subject of the embedded clause to the subject of the matrix clause position would correspond to lowering. 18 In * John is a picture of t the cause of the riot there is an overriding Subjacency violation, due to the fact that the embedded subject is not L-marked. We can disregard this difference because the same strong violation is yielded when the small clause is the complement of a lexical verb like believe: * Johns was believed [ sc [DP a picture of tj] [DP the cause of the riot]] Nevertheless, notice that * the riott is a picture of John the cause of tj does not involve a Subjacency violation (as was extensively discussed in chapter 1) but the sentence is still ungrammatical. 19 Notice that wh-movement from the embedded CP is possible when the proper configuration involving /^-raising is applied: a
whOj does itj seem [ sc [CP tj left] tj]?
In such a case, the same process of lexicalization we observed for there-raising arguably takes place. In fact, f/zere-sentences would be fully parallel to these cases: b * Johnj was [Sc [DP a picture of tj] there] c whOj do you think therej was [Sc [DP a picture of tj] tj]? d * Johnj seems [ sc [CP tj appears tj to win] it] e whOj do you think that itj seems [ sc [CP tj appears tj to win] tj]?
286
Notes to pages 195-200
20 The observation might arise that these two sentences are trivially semantically equivalent in the same sense as are all sentences 4p' with respect to sentences of the form 'p is true' (i.e. in Tarski's sense). The point here is simply that it is the syntactic structure that gives to the first sentence the same compositional semantics as the second. This follows from the role of it as a propredicate and, obviously, it is limited to the structure containing this element. Special thanks to Gennaro Chierchia for many helpful suggestions on the matter. 21 We can now assign to the two sentences discussed by Ruwet the proper representation: a b
22
23
24
25
26
[IP [IP itj seems [ sc tj true]] [Cp that John left]] [IP it; seems [ sc [CP that John left] tj]]
In the sentence involving true, the CP is extraposed and spec-IP is occupied by the subject of the small clause, i.e. it. To avoid confusion, it should be noted that /br-clauses can occur in subject position of predicates like unexpected or illegal, yielding for example: for him to win would be unexpected. The interpretation of these predicates is different from that of predicates such as true (we might refer to the former as nonpropositional). In fact, such cases should be contrasted with that he wins would be unexpected, which appears to be barely acceptable. Again, to avoid confusion it should be noted that a CPj+wh] might be the subject of true as in what John said is true but in this case the predicate is not applied to the propositional content of the clause. The shift from S to IP had an undesirable consequence: in the previous framework, government of the lower trace by seem was immediately available since only maximal projections can block government and the surviving clausal constituent was S, thus not a maximal projection. In the more recent framework, the clausal boundary is itself a maximal projection, namely IP. Several different solutions have been proposed to get round this problem: one can simply assume that IP is 'defective', following Chomsky (1986b), or adopt Cinque's (1990a) generalization to the effect that an XP is a barrier if it is not selected by a head not distinct from [+V]. See also the discussion in chapter 1 of this book and Moro (1993a). As Noam Chomsky has pointed out (p.c), the minimalist framework would provide an alternative account based on Case theory that also avoids recourse to S'-deletion: in fact, in such a framework, there would no longer be a need to distinguish CP from IP complements. The difference (and perhaps the advantages) with respect to the proposal I will suggest here will be clear. See also Pesetsky (1995) for alternatives to S'-deletion. Extraction in cases involving to must be kept distinct from ellipsis, as in: a
John wanted to [VP see Mary] but I didn't want to [Vp e]
In fact, licensing conditions for ellipsis do not necessarily involve proper government, at least not in the same way as extraction does. This can be proved by considering the following contrast:
Notes to pages 200-5
287
b John saw [DP Mary's [NP picture]] but not [DP Tom's [NP e]] c * WhOi did John see [DP Tom's [NP picture of tj]? Thus, (a) does not support the assumption that to can properly govern. 27 The possibility of negation following to is due to the fact that negation can also play a different role (as a constituent negation); witness the possibility of its cooccurring with pre-to negation: I believe John (not) to (not) see Mary See Zanuttini (1992) and Laka (1992) for a detailed analysis of the syntax of negation. 28 If one adopts the Split-Infl hypothesis (proposed by Pollock 1989 following Emonds 1985, and independently by Moro 1988), the question becomes even more complicated: if Agr and Tense head their own projections, what does to stand for? 29 In Moro (1988: 103), the hypothesis was put forward to the effect that small clauses can be regarded as 'predicative phrases'. See Bowers (1993) for a correlation between this idea and the role of as. 30 If we adopt the idea that adjuncts should be represented as small clauses (see Chomsky 1981), then the proper representation would rather be: [IP [IP Johnj telephoned] [ sc PROj as the doctor]] Since I do not see any empirical advantage here, I will just use the simpler representation. 31 I will leave it open as to whether as should be regarded as heading a small clause or, alternatively, as heading a maximal projection, where the latter is entirely contained in the small clause: a b
[ sc DP [asP [as, as DP]]] [ asP DP[ as ,[asDP]]]
Notice that in the first case as is entirely parallel to of in ^/-insertion. 32 This analysis implies that C° can take a small clause as a complement: a b
[cp for [ sc John to leave]] is unusual I know [CP whati C° [s c PRO to do tj]
Although C° is usually assumed to take only IPs as complements, I do not see any empirical reason to reject this implication out of hand (see also Kayne 1984: 114). 33 For our purposes, we can disregard the Case-marked DP which can optionally follow the quasi-copula (but not the copula) in cases like: it seems [DP to John] that Mary left It is reasonable to assume that this DP is outside the complement of seem, i.e. outside the small clause. Arguably, it is in a higher position, say V , within the VP (see Chomsky 1981, 1986a for the role of to as a Case marker). 34 Leaving aside diachronic variations, stylistic factors may also complicate the situation. Consider for example: There seemed a certainty in degradation' (T. E. Lawrence in Seven Pillars of Wisdom, cm, 1927, cited in Borges' Ficciones).
288
Notes to pages 207-17
35 In this case, di cannot occur before the infinitive: * di volare e vero. Cinque (1990b) observed that the distribution of infinitival di depends on the position of the infinitival clause, in that it is obligatory only if the infinitival clause is generated in object position: a
e vietato (*di) fumare (is prohibited di smoke) ha vietato *(di) fumare (has prohibited di smoke)
Even though I do not intend to develop a full analysis here (on clausal complementation in Italian see Acquaviva 1989), it would not seem at all implausible to trace the difference between the copula and the quasi-copula back to the presence of an abstract head in the small clause following seem. However, Kayne (1983: 119, fn.7) noticed that in French de is normally omitted with sembler (seem). This suggests that the role of de and di might be relevant only at a 'surface' level. 36 Notice that di cannot occur before the infinitival clause with the copula: partire e (*di) morire (leave is di die) 'to leave is to die' See the previous footnote for some observations on this matter. 5 A view beyond: unaccusativity as an epiphenomenon 1 The original we-extraction test has been refined over the years. Lonzi (1986) pointed out that the contrast is always present when auxiliaries are inserted but is not always clear otherwise. For example, ne starnutirono molti (of-them sneezed many) seems to be nearly acceptable, as opposed to * ne hanno starnutiti molti. 2 For example, the so-called pseudo-passive construction in English was observed to conform to such a division among intransitives: a b
the bed was slept in by the shah * the bed was jumped in by the children
For other tests supporting this dichotomy see again Perlmutter and Postal (1984) and Rosen (1981) among others. 3 Although a taxonomy based on two negative terms ('un-accusatives' and 'unergatives') sounds to me very unnatural, I will not follow here Burzio's (1986: 30) proposal to call unaccusatives 'ergatives' since I think it wise to avoid possible confusion with more traditional uses of this term. Notice also that, strictly speaking, 'unaccusative' refers to a type of construction like 'active' or 'passive', rather than to a type of verb. Unless otherwise required for the sake of clarity, the general practice will be followed and the term 'unaccusative verb' will be used as an abbreviation of the more correct formulation 'verb yielding an unaccusative construction'. 4 This is of course a very simplified representation which reproduces the state of
Notes to pages 217-24
289
the theory during the 1970s. Many subsequent refinements should be considered. Following the so-called 'subject-in-VP-hypothesis' (Koopman and Sportiche 1991) it would be possible to take the base position for subjects to be spec-VP. For the sake of simplicity, I will not indicate this position as the source of movement. Moreover, Rizzi (1990) no longer accepts the idea that the subject of telefonare is adjoined to VP. Since these refinements do not affect our argument, we will not pursue them here. 5 'A binding relation! is a binding relation other than a relation between elements of independent 0-roles' (Burzio 1986: 63). This is intended to exclude essere assignment in cases like the following: * Giannij e accusato se stessoj (Gianni is accused himself) 6 'A nominal contiguous to the verb is a nominal which is either part of the verb morphology, i.e. a clitic, or a direct object' (Burzio 1986: 56). 7 See Graffi (1988) for a discussion of related topics. Graffi proposed that we should distinguish the two different notions of subject as 'structural subject' (related to spec-IP position) and 'thematic subject' (related to a VP-internal position). 8 We adopted the simplified representation for inverse copular sentences in Italian not involving propredicative/?ro. Adopting this less precise representation will not affect our argument here. 9 We saw in the second chapter that only with indefinite DPs does the subject of a d-sentence stay in the small clause, while in the case of definite DPs it is rightward-moved. The cases involving definite DPs can be ignored here. 10 Notice that this argument is independent of whether ci is a subject expletive or a raised predicate. In fact, even in the traditional analysis the subject would not be adjacent to V°. 11 Recall that this contrast strongly undermines the theory of ci as an expletive (see chapter 2). 12 This is essentially the proposal of Moro (1990). Hoekstra and Mulder (1990) arrive at a similar conclusion on the basis of an independent set of facts. Their idea is that unaccusatives can be analysed as 'locative inversion' constructions where spec-IP is occupied by a PP ((ia)), on a par with copular constructions like those in (ib) (see Hoekstra and Mulder 1990: 28ff.): [IP [pp in the room]i entered [ sc a man tj] [IP [pp in the roomji was [ sc a man tj] Although I am very sympathetic to this analysis, which in fact explicitly subsumes an earlier version of my proposal to analyse there as a raised predicate (circulating as Moro 1989; cf. Hoekstra and Mulder 1990: 33ff.), I cannot immediately adopt it. The main empirical reason is the fact that 'locative inversion' as in (b) is impossible in Italian, apart from the focus/topic interpretation involving extraIP projections (see Rizzi 1995): c
* [IP [PP nella stanza^ era [ sc un uomo tj]]
This is not surprising within our approach, for we assume that 1° must be in a
290
Notes to pages 224-32
spec-head relation with a phrase containing features and that Case is assigned in spec-IP. Notice that this does not depend on the pro-drop parameter, since the presence of a PP in spec-IP is also banned in French (see also 1.3.3.2): d * [ip [pp dans la rue] est un homme] (in the street is a man) e * [IP [PP dans la rue] est arrive un homme] (in the street is arrived a man) Thus, there would appear to be no reason not to maintain that the major empirical fact supporting the small clause approach to unaccusativity is the possibility for predicative DPs, rather than PPs, to raise to spec-IP, something that is an invariant property across languages as shown in the following simplified paradigm: f g h
[IP [DP t n e cause of the riot]j is [Sc a man tj]] [IP [DP l a causa della rivolta]j e [ sc un uomo tj]] [IP [DP l a cause de la revoke] c'j est [ sc un homme tj]]
Interestingly, it should be noticed that agreement in sentences of the kind PP copula - DP is sensitive to the features of the postcopular DP, as opposed to inverse sentences, suggesting that these could rather be analysed as cases of further movement to an extra-IP projection (see Rizzi 1994, Quirk and Greenbaum uni 1973: iy ID. 414, t i t , and ctnu Rochemont rvAJCiiciiiuiu and cinu Culicover ^uiicuvci 1990 177 among others): i j k 1
[DP the
cause of the riot]j is [ sc a man tj] [DP the cause of the riot]j is/*are [ sc John and Mary t [Pp in the room]j is [ sc a man tj] [Pp in the room]j are/*is [ sc John and Mary tj]
cooular sentences (and, (and. crucial! All in all, inverse copular crucially, d-sentences) will still be considered here as the major empirical motivation for analysing unaccusatives as selecting a small clause. 13 A rather murky phenomenon related to wh-movement can be observed in English: a [DP which girlsjj do you think that there are [Dp pictures of tj]? b ?? [DP which girlsjj do you think that there arrived [Dp pictures of tj]? At the moment, I will not attempt to explain this contrast. 14 I follow Kayne (1985) in assuming that an AgrP is involved. 15 Marantz (1984: chapter 6) suggests that the mapping between 6-roles and structural positions varies parametrically. 16 The major advantages of this new approach are clearly illustrated in Hale and Keyser (1991). In particular, this would provide an elegant answer to two fundamental empirical questions: 'Why are there so few thematic roles?' and 'Why the UTAH?' (Hale and Keyser 1991: 38). 17 Many terminological and conceptual differences distinguish the two approaches. For example, we assume that predication always involves small clauses, never VPs. 18 Notice that the subject can be raised to spec-IP:
Notes to pages 232-3
291
[IP molte ragazzej [loCj-arrivano] [ sc tj tj]] (many girls arrive) This supports the hypothesis of incorporation of the lower predicate. Consider the following cases: b * [IP ioj [sono] [ sc tj tj]] (lam) c [,P iOj [cij sono] [ sc tj tj]] (I there-am) This contrast shows that only when the predicate is incorporated can the subject be raised. Otherwise, the process of raising would obliterate the chain headed by the predicate itself. 19 Pursuing this line of research, one might argue that the morphological processes of incorporation are not completely opaque. A first cursory survey indicates that all intransitives that incorporate a locative particle are unaccusatives like discendere (descend), per-venire (reach), ac-correre (run); similarly, all verbs that derive from adjectives: annerire (black-en), sbiancare (whit-en), ingigantire (magnify), etc. Thus, it would be reasonable to assign the following representation to the two classes: a b
[vPVVIUcDPIpptJ] [yoAVV^UcDPUpti]]
Interestingly, if this proves tenable the distribution of the 'ergative si' in Italian (see Burzio 1986) could be descriptively captured as occurring only with the [VoA°-Vo]-type verbs. 20 Interestingly, Keyser and Roeper (1992) following Kayne (1985) argue that the suffix re- (as in arrange and rearrange) originates as a predicate in a small clause. This provides independent support for the analysis given here of verbs resulting from lexical incorporation. Similarly, the affix -en would be the head over which the predicate of the lower small clause {black) incorporates to yield a verb {black-en). 21 There is a contrast here between English and Italian. When the copula is employed, the two languages are parallel in that there/ci are obligatory: a * (therej) are [ sc many girls tj] b pro [*(cij) sono] [Sc molte ragazze tj] On the other hand, when unaccusatives are used, ci is at best only optional (as we observed in the text): c * (therej) arrive [Sc many girls tj] d pro [(cij) arrivano] [ sc molte ragazze tj] The Italian case has been explained here by assuming that the locative predicate is in fact incorporated into the verb. There is no need to exclude lexical incorporation for English. The presence of there is simply explained by the fact that in this language pro is not licensed. Even if we assume lexical incorporation, there can continue to play the role of an expletive without this undermining the
292
22
23
24
25
Notes to pages 233-7
essential proposal we are making about it. Crucially, we can maintain that there is not the expletive of the subject of the predication (as in the standard theory) but rather the expletive of the predicate itself. Indeed, the very fact that there cannot occur with all unaccusatives (cf. there arrived/*blacken many girls) is hardly compatible with the assumption that there is the expletive of the subject of predication. If this assumption holds, the incompatibility of there with many unaccusatives must be traced back to an anomalous semantic restriction holding between two maximal projections (the DP subject and the VP predicate; see Burzio 1986: 159 and references cited there). If on the other hand there is analysed as a locative propredicate contained in the small clause, this restriction can rather be regarded as a standard lexical restriction holding between the predicate of a small clause and the head selecting the small clause itself. Marantz (1984) proposes a slightly different approach based on the grammatical function of subject of the predication rather than 0-roles. Thus, the paradigm under discussion would be explained by assuming that the lexical entry for affondare contains the specification [± logical subject], Interestingly, a correlation between the change of valence by a verb and incorporation is explicitly noted in traditional works which we could easily rephrase in terms of a small clause analysis: 'many intransitive verbs are used transitively when compounded with a preposition [in ancient Greek]' (Smyth 1920: 355). For example: djro-xcoQ8iv (from-leave), dia-paiveiv (over-pass), xaxa-JtoXefxeiv (sub-due), eju-axQaxeueiv (march against). No attempt will be made here to explain linguistic variation. For a detailed discussion concerning unaccusatives (and related issues) in Italian dialects see La Fauci (1984) and references cited there. In synthesis, for every argument a, there would be only one predicative structure:
If X = A, D, V, P we have the traditional small clause structure; if X = 1°, we have those structures generated by adjunction, such as: b
[ip [DP many girlsjj [IP John didn't meet tj]
Interestingly, the usual paraphrase for (b), i.e. 'many girls are such that John didn't meet them', reveals the implicit predicational nature of Q-raising. Qraising can be regarded as a homomorphism from sentences to copular constructions of the kind . . . be such that. . . whose subject is the Q-raised DP. 26 To put it in Chierchia's (1989) terms, it would appear that the unaccusative esserci is not unstable between transitive and unaccusative valency, as opposed to all other unaccusatives. 27 I will henceforth deal with avere as a main verb. Whether or not this analysis helps in understanding the selection of auxiliary essere and avere is an issue that I cannot pursue here. For a theory of auxiliary selection see Kayne (1993); for further remarks see also fn. 34 in this chapter. 28 The example reported in (50c) is taken from the dialect spoken in Pavia, northwestern Italy, and includes a subject clitic a. For an updated and detailed
Notes to pages 238-42 293 structural analysis of Northern Italian dialects see Poletto (1993). The fact that avere can occur expressing possession without ci might be regarded as a variant involving abstract clitic incorporation (as in arrivare). 29 Independently from Moro (1990), (1991), Freeze (1992) pursued a similar analysis. He suggested that ci occurring with avere should be analysed as a locative because it can 'be directly preceded by the negative' (Freeze 1992: 568). In Italian, however, there are indeed clitics that hold the subject position (such as si (a-b); see Burzio 1986, Cinque 1988 and references cited there), which can nevertheless be preceded by negation (c): a b
c
30
31
32 33
34
Gianni telefona bene dalla stazione (Gianni telephones well from the station) (* Gianni) si telefona bene dalla stazione ((*Gianni) si telephones well from the station) 'one telephones well from the station' non si / *si non telefona bene dalla stazione (not one / *one not telephones well from the station)
Thus, the order of negation with respect to the clitic is not sufficient to support the hypothesis. It is worth noticing that this analysis brings esserci into line with Chierchia's (1989) generalization that unaccusatives tend to be unstable (see the previous section). The crucial point here is that what appears to be unstable is not esserci but rather the structures associated with it: when the external argument is present esserci is turned into avere. Recall that Freeze's (1992) proposal considered ci as a sign of P° incorporation in Italian. Since the same element can be incorporated both with essere and avere, it is not clear to me how this could be used to support the idea that avere is a copula with an incorporated preposition. This possibility was suggested to me by Anthony Kroch. He also remarked that the occurrence of get with have tends to support the hypothesis proposed here. For example, it would be interesting to see whether the account given here of the Definiteness Effect can be applied to those languages like Hungarian (see Szabolcsi 1983) where possessive constructions are sensitive to the presence of a definite DP. Many interesting questions that cannot be pursued here are raised by this theory. I will simply point out three of them. First, these constructions involving averci can be followed by adjunct elements, paralleling the case of esserci: a
b
Gianni c'ha Maria [Cp che esce dal cinema] (Gianni there-has Maria that comes out from the movie) 'Gianni is urged by Maria's coming out of the movie' Gianni c'ha Maria [AP arrabbiata con lui] (Gianni there-has Maria angry with him) 'Gianni is touched by the fact that Maria is angry with him'
The obvious fact here is that the material following avere in the first case is essentially the same as the complement of perception verbs (see Guasti 1993 and references cited there). Since the predicative relation between the embedded CP
294
Notes to pages 243-4
and the AP with Maria is mediated by the incorporation of a lower head d, a very intriguing analysis would be to consider perception verbs as incorporating a lower head as well. Second, interesting alternations between the presence/absence of an external argument with deontic avere were reported in Moro (1991), where we find the following pair: c
d
Gianni c'ha da spostare una macchina (Gianni there has a car da move) 'Gianni has to move a car' c'e da spostare una macchina (there is a car da move) 'there is a car to be moved'
The theory presented here can easily be extended to cover these cases. Third, if we extend this analysis to the domain of auxiliary avere we should be led to adopt the following representation (leaving open the option between extraction and control): e
Giannij [cij ha] [ sc [ sc PRO/tj telefonato] tj] (Gianni ci has PRO telephoned) 'Gianni has telephoned'
In such a case, telefonato would play the role of the predicate of the small clause [sc PRO telefonato] which is the subject of the small clause complement of avere. Now, in Standard Italian ci cannot occur with auxiliary avere but this structure is in fact overtly realized in many Italian dialects (see Poletto 1993 and references cited there). Whether or not we should assume that an abstract clitic occurs in Standard Italian is a matter that cannot be pursued here. 35 As we have already noticed (chapter 3, fn. 26), Belletti (1988) suggested that Nominative cannot be assigned to the object position either. The Nominative we see in postverbal position would not be assigned inside VP but rather to spec-IP, the DP then being rightward-moved. The following case shows that this theory cannot be adopted for copular sentences: [IP proj sono [ sc io t;]] (pro am I +nom ) As a matter of fact, the DP contained in this example is indeed Nominative but we know independently that it cannot have undergone (rightward) movement, which is excluded for inverse copular sentences. A critical approach to Case assignment in unaccusatives has also been pursued by Delfitto (1992). 36 I am excluding here the hypothesis that a process of expletive replacement in the sense of Chomsky (1986a) takes place (along the lines I have indicated in chapter 2). I am grateful to Luigi Burzio for extensive and generous discussions on this topic. 37 The occurrence of there in infinitival contexts with arrive is a rather murky matter: a *? I believed [therej to arrive [Sc [many girls] tj]] b I believed [there} to be [ sc [many girls] tj]]
Notes to pages 248-9
295
This parallels the problem of the distribution of propredicative it with infinitival seem (see chapter 4): c *? I believed [itj to seem [ sc [that John left] ti]] For our purpose here, we will just extend the same descriptive generalization adopted to account for the cases involving seem, i.e. that propredicative DPs are compatible only with functional heads. This approach receives further confirmation if we contrast (ia) with (iii): d ? I believe [there, to have t{ arrived [Sc many girls tj]] According to my informants, the sentence in (d) is improved with respect to (a); this can be explained as a consequence of the fact that have is 'functional' and not 'lexical'.
Appendix: a brief history of the copula 1 Needless to say, this section does not aim to be a full scale philological treatise: the questions that I briefly report on here have been discussed at length by many authors (see among others Robins 1967, Lepschy 1990 and references cited there). In this section I will rely heavily on Kneale and Kneale (1962), Ackrill (1963) and Graff! (1991). Most citations will be found in the notes and, in the case of this appendix, should be read as if they were integral parts of the main text. A special thank you is due to Giorgio Graffi and Giulio Lepschy for many helpful comments on this appendix. 2 There is a very important caveat: the theory of the copula and the corresponding terminology are intimately related to the Indo-European languages, in particular to Greek and Latin. Whether or not a certain element in a given language is 'a copula' clearly cannot be determined without providing a definition of what the term is taken to mean. In order to have an idea of the problems inherent in identifying copulas in various languages, see for example the treatment of the copula in Celtic (Rouveret 1996 and Carnie and Harley 1994) or the debate about the 'so-called pronoun of the third singular that serves as the "copula" in Aramaic' (Benveniste 1966: 165). 3 I will not discuss here whether Aristotle's theory was ultimately related to words or things: in fact, via Boethius' translation of the Organon, it entered into the linguistic field influencing all successive theories (see During 1966 and Belardi 1985 for a detailed discussion on this matter). 4 I will quote Ackrill's (1963) translation in the text and the original ancient Greek version (taken from the Loeb edition as transcribed from the Bekker edition) in the footnotes. 5 'djtocjxxvtixog Se oi) Jtdg, dXX' evcp to dA.'nOeveivf] ipeuSeaGca VJI&QXEI, oim ev djcaai de vnaQ%ei oiovf] £v%i\ Xoyog \iev, 6Xk' oike dXr]0f|c; oirce ipevdr|c;'. 6 For a discussion of constituents before Bloomfield see Graffi (1990). 7 'jieot ydiQ OVVQEOIV xai 6iaiQeaiv eon TO ipevdog xai TO dXrjOec;'. 8 "AvdyxT] 6e Jidvra Xoyov djto(|>avTixdv ex Qr^aToc; eivai f\ jtTcoaecoc; Qf^aTog'. 9 'eoriv del xwv xa9' ETEQOV Xeyonevcov anjxeiov'.
296
Notes to pages 249-53
10 'Qfjjia de ton TO JtQaanfxaivov XQOVOV [. . .] vyieia \iev ovoua, TO de ir/iaivei 11 'ov6ev yaQ diac^eoei 81J18LV avOoamov pa6i£eiv f] &V6QO)JIOV (3a6i^ovxa eivai'. As Giorgio Graffi pointed out to me, the infinitival form (eivai) that is used in this example seems hardly consistent with the idea that the copula provides a tense specification. Although one could perhaps think of this infinitive as inflected, unlike other non-finite forms, an alternative possibility which is in fact adopted by Ackrill (1963) is to assume that 'Aristotle uses the infinitives "to be" and "not to be" [. . .] as stand-ins for indicative forms, "is", "is not", "was" etc' (Ackrill 1963: 122). 12 'E(()' oacov de TO eon \ii\ aonorrei, OIOV em TOO vyiaivei xai Padi^ei, em TOVTCOV TO auxd Jioiei oika) TiOejAevov cog dv ei TO eon jiQoarjjrceTo'. 13 Xeyo) 6e oiov eon 6ixaiog avOQcoJiog. TO eon TQITOV <$>v)\d axryxeiaOai'. 14 'MeTaTiOep,eva 6e Td ovojiaTa xai TCI Qr\\iaTa TOUTOV ar](iaiv8i, oiov IOTL XevKog dvGQomoc;, 8GTLV dv0Qcojrog Xeuxog'. 15 'dveu 6e QTj^iaTog ov6e|xia Kaxatyaoic, ovbe outoc^aaig'. 16 See for example J. L. Ackrill's comment on De Interpretatione (Ackrill 1963). Aristotle always avoids employment of a special name for the copula: rather he refers to actual instances of eivai (be). Not all translators seem to be aware of this fact. In the De Interpretatione {De Int. 3,16b, 25), for example, Aristotle says that eivai 'jtooaanumvei 6e oi>v0eaivTiva\ Ackrill's translation says 'but it additionally signifies some combination' (Ackrill 1963: 45). Cooke, instead, renders it as 'but imply a copulation [sic] or synthesis' (Cooke 1938: 121). Apart from the question related to the invention of the term 'copula', the term ovvOeotg (synthesis) employed here seems to imply that Aristotle already conceived eivai as a copula in the Abelardian sense. This possibility is excluded here, following Ackrill's hypothesis that in this passage Aristotle is not talking about 'the copulative but about the existential "is"' (Ackrill 1963: 123). 17 For a discussion on this matter see Kneale and Kneale (1962) and Pinborg (1972). 18 Correspondingly, the common use of to be (or being) as a shortened form of to be (or being) existent is to be treated in the same way. The following passage pointed out to me by Giulio Lepschy illustrates the point: Dixit Deus ad Moysen: ego sum qui sum. Ait: sic dicesfiliisIsrael: qui est, misit me ad vos (Exodus, in, 14). From a linguistic point of view, however, this should not be surprising. A similar case is that of to do (doing) as a shortened form of to do (or doing) something. Interestingly, the term 'existence' is not etymologically related to the copula esserel'eivai but to a locative predicate sisto and a locative particle ek-. 19 The third activity, reasoning, is in a sense derivative from the first two: in fact, it is defined as the production of a new judgment by means of two other independent judgments. For references and discussion, see Simone (1990: 333). 20 The passages from Aristotle which are relevant for this conception are Poetica 20, 1457a, 14-18 and De Int. 12, 21b, 9, cited in an earlier paragraph. 21 See Graffi (1991: 119-25) for a detailed discussion. I can only add here that since Leibniz and Wolf worked together at the University of Halle, it might be worth investigating whether Leibniz played any role in this tradition. Of course, this is pure conjecture and the issue will not be taken up here.
Notes to page 254
297
22 Although we cannot develop the matter here it is interesting to notice the interest in predication on the part of mathematicians (see for example Peano 1930). In fact, copular sentences have always attracted a great deal of interest outside linguistics. Russell attributed to Leibniz such an influence. According to him, for Leibniz, 'every proposition is ultimately reducible to one which attributes a predicate to a subject' (Russell 1900: 9). For a detailed discussion of Leibniz's theory of logic and language see Ishiguro (1990). Interestingly, she explicitly states that 'Leibniz includes among his subject-predicate propositions those of the form Cicero is Tullf (Ishiguro 1991: 101-2). 23 It is interesting to notice that Russell's example does not involve two definite noun phrases (or 'definite descriptions', as Russell would say). To express identity, one could have used a classical example involving two proper names like Cicero is Tully or the Fregean classic the evening star is the morning star. This reveals that Russell's idea of identity is really deeply rooted in his logical system and substantially passes over linguistic data. 24 Daniel Osherson pointed out to me that Sommers (1982) offers an alternative approach to this issue that appears to fit in with the theory presented in this book. Although a critical comparison obviously calls for a much more thorough investigation, the following quotation should suffice to illustrate Sommers' basic thesis (as limited to the theory of the copula): 'an identity proposition will now be defined as a monadic proposition that has proper names both in subject and predicate position' (Sommers 1982:122). Moreover: 'Tully is Cicero, Istanbul is Constantinopolis and the morning star is the evening star are considered as examples of predicating proper names' (Sommers 1982: 126). 25 See, for example: Montague (1973) (and a detailed discussion in Dowty, Wall and Peters 1981: 229, and Partee 1986; see also Moro 1987). In Montague's framework, simple sentences like John is Bill and John is a man would be ambiguous between identity and predication respectively. In fact, the special translation of be (i.e. APXxP{AAy [x=y]}) will yield two distinct outputs when combined with John (assigned: XP[P{b}]) and a man (assigned: XQ3x[man'(x) A Q {x}]). The two outputs are the following transparent formulae: j = b and man'(J) respectively. 26 Benveniste (1966) illustrates how confusing (and confused) the terminology concerning the notion of identity in fact is: 'il faut poser deux termes distincts que Ton confond quand on parle de "etre": l'un est la "copule", marque grammaticale d'identite; l'autre, un verbe de plein exercice' (1966: 187). As a verb, Benveniste says, it means 'avoir existence, se trouver en realite' (1966: 187). Curiously, Benveniste appeals to the notion of identity to illustrate the copulative use of be: 'Completement differente est la situation de la "copule", dans un enonce posant l'identite entre deux termes nominaux. Ici l'expression la plus generate ne comporte aucun verbe. C'est la "phrase nominale", telle qu'elle est representee aujourd'hui, par exemple, en russe ou en hongrois, ou un morpheme zero, la pause assure la jonction entre les deux termes et en asserte l'identite [sic] - quelle que soit, au point de vue logique, la modalite de cette identite: equation formelle ("Rome est la capitale de l'ltalie"), inclusion de classe ("le chien est un mammifere"), participation a un ensemble ("Pierre est Francois"), e t c ' (1966: 187-8).
298
Notes to pages 254-6
27 'There are in fact three lexical verbs be, one class 0, one 1 and one class 2. Broadly, the class 0 means "can be characterized as, has the attribute of being"; class 1 be means "exists, happens, is found or located"; class 2 be means "identifies or is identifiable as, can be equated with'" (Halliday 1967: 66). 'The equative clause . . . has the form "x equals y", as the leader is John; the equals relation is an asymmetric one meaning "is to be identified as" . . . the clause what John saw was the play is an identifying clause' (1967: 223). 28 '[B]e is commonly used to introduce a characterization or attribute of the subject, as in the example just given, but with complement noun phrases it also commonly introduces an identification of the subject' (Quirk and Greenbaum 1973: 353). 29 It would be sufficient to consider the contrast between sono io (am I; 'it's me') and *io sono (I am) discussed in the first chapter to understand how relevant syntax is. 30 Graffi (1991) reports that a catalogue of various definitions of the concept of 'sentence' includes more than 230 different entries (this total is based on Ries' (1931) and Seidel's (1935) lists). No examination of such a catalogue could possibly be attempted here. 31 'In grammar we should not of course forget logic, but steer clear of everything that may hamper our comprehension of language as it is actually used' (Jespersen 1937: 120). 32 Although this is true for the majority of European languages, this statement is too strong to be universal. By the time Jespersen wrote, this was well established. La Grasserie (1900) had already explored and published his analysis on the socalled 'conjugation objective'. Nevertheless, even in languages with objectagreement, the subject would still stand out thanks to other properties related to morphology such as the order of morphemes. It remains, however, the case that there is simply no way of applying this empirical test to languages like Chinese, where there is no verbal agreement at all. 33 In Moro (1988) I gave an empirical argument based on Binding theory to support Jespersen's position against the analysis of the copula as a predicate of identity. I will summarize it as follows. Suppose the copula could cooccur with two DPs both of which are referential (in the sense of Geach 1962 and Higgins 1973). A classic example would be: [DP the morning star] is [DP the evening star] Binding theory shows that a pronoun contained in a predicative nominal cannot be bound by the clausal subject, because the latter would be included in its local domain (defined as a Complete Functional Complex; see section 1.4.2 and fn. 86 in chapter 1). Suppose now we construct the following sentence, intending it to be an equative statement, that is, a sentence where neither DP is predicative of the other: b
* [DP the morning star]j is [DP itSj source of light]
Clearly, there is no way for the pronoun to be bound by the subject. On the other hand, if a predicate indicating identity is put in the sentence: c
[DP the morning star]; is equal to [DP itSj source of light]
Notes to pages 257-8
299
coreference is possible as indeed in pure transitive constructions, which of course do not involve predicative nominals, like: d
[DP the morning star]j lost [DP itSj source of light]
In (iii), indeed, we have a genuine identity statement as opposed to the case of the bare copula: neither DP is the predicate of the other, witness the fact that the local domain for binding is limited to the DP itself in each case. Eventually, notice also that if sentence (a) is analysed as an identity statement, the same should be said for the embedded sentence S in a case like the following which does not contain any copula: e
34 35
36 37
38
39 40
I consider [s [the morning star] [the evening star]]
I would like to emphasize that what I am saying here is that identity is not predicated by the copula or equivalently that one of the two noun phrases involved in a copular sentence always plays the role of a predicate. Whether or not the notion of identity can be employed to understand the relation between the two noun phrases is a different matter that I will not pursue here. See also the discussion about the only man who knew the secret was Tom which is considered as ambiguous (Jespersen 1937: 137). It might be worth noting that a further typical Aristotelian assumption is not maintained by Jespersen: 'Logicians are fond of analysing all sentences into three elements, subject, copula and the predicate; the man walks is taken to contain the subject the man, the copula is, and the predicate walking. A linguist must find this analysis unsatisfactory, not only from the point of view of English grammar, where is walking means something different from walks, but also from a general point of view' (Jespersen 1924: 131). Clearly, the distance from a Russellian approach is enormous here: in no sense are logical categories to be imposed on linguistic categories. We will not consider 'small clauses' here (see chapter 1 and references cited there). In the very first model (Chomsky 1957), the inflectional morphemes were generated within the verb phrase and incorporated in the verb by a process called 'affix-hopping'. The shift to the representation given by Chomsky (1981) and Chomsky (1986a) is not totally new in linguistics. As Graffi (1985) pointed out, the idea that inflectional morphemes do not basically belong to the verb phrase but to the sentence itself had already been proposed by Hjelmslev (1943) - if not by others - in the context of a critique of the definition of 'verb' proposed by Meillet (1921). See also Lyons (1977). Notice that, as with all hypotheses concerning the grammar of natural languages, Chomsky (1957) claimed that this idea does not follow from logical necessity; indeed, it can only be assumed to be true to the extent that it allows the right predictions to be made, as is generally the case in the natural sciences. See Graffi (1980) for a first proposal to implement Greenberg's (1963) theory concerning the order of constituents in a phrase in the X'-framework. See also Kayne (1994) for a new theory that avoids linear order parameters. The so-called 'split-INFL hypothesis', which gives the syncretic components of INFL their own projections (Agr°, T°), and which was proposed by Pollock
300
41
42 43
44
45
Notes to pages 259-60
(1989) (as a way of implementing a suggestion originally made by Emonds 1985) and independently also by Moro (1988), is totally irrelevant here. In fact, from a predicational point of view this hypothesis is completely equivalent to the syncretic model. The definition of 'clause' given by Bloomfield is exceptionally concise and thus worth quoting: 'when a linguistic form occurs as a part of a larger form, it is said to be in included position; otherwise it is said to be in absolute position and to constitute a sentence' (Bloomfield 1933: 150; see also Bloomfield 1933: 173). For empirically based reasoning supporting subject-object asymmetries see section 1.1 in this book. In the text, I have indicated some major works in the field. For a brief and by no means exhaustive survey of copular sentences in generative grammar see also: Higgins (1973), Stowell (1978), Ruwet (1982), Pollock (1984), Longobardi (1985), Burzio (1986), Rapoport (1987), Shlonsky (1987), Heycock (1991) and Gueron (1992). 4 [I]t is in fact true that in the simplest phrase structure grammar of English there is never any reason for incorporating "be" into the class of verbs, i.e. it will not follow from this grammar that be is a V. Just as one of the forms of the verb phrase is V + NP, one of the forms is be + Predicate' (Chomsky 1957: 67). For a theory that analyses the copula as a direct spell out of an inflectional morpheme, specifically T°, see Moro (1988). The taxonomy proposed in the pioneering work of Higgins (1973) is much more complicated. He concluded that there are four different types of copular sentences of the kind NP - copula - NP: 'predicational', 'specificational', 'identificational' and 'identity'. Longobardi (1985) introduced the traditional idea that the copula never expresses identity into the generative framework, thus reducing Higgins' taxonomy to two basic types which cannot be traced back to a unified syntactic structure: 'predicative' and 'identificazionali' (sic) (see Moro 1988 for a detailed discussion). Eventually, a unified theory of copular sentences, relying on Stowell's (1978) and Burzio's (1986) analysis of the copula as a raising verb, was proposed in Moro (1988) (see the first chapter of the present volume for a substantially revised version of this earlier proposal).
References
Abelard (1956) Dialectica, L. M. de Rijk, Assen. Abney, S. (1987) The English NP in Its Sentential Aspect', Ph.D. dissertation, MIT, Cambridge, Mass. Ackrill, J. (1963) Aristotle's Categories and De Interpretation, Clarendon Aristotle Series, Oxford University Press. Acquaviva, P. (1989) 'Aspetti della complementazione frasale', Tesi di laurea, University of Pisa. Anderson, S. and Kiparsky, P. (eds.) (1973) A Festschrift for Morris Halle, Holt, Reinhart & Winston, New York. Aristotle (1938) De Interpretation, ed. H. P. Cooke and H. Tredennick, Loeb Classical Library, 325, Cambridge, Mass, and London. Baker, M. (1988) Incorporation, University of Chicago Press. Barwise, J. and Cooper, R. (1981) 'Generalized Quantifiers and Natural Language', Linguistics and Philosophy, 4, pp. 159-219. Belardi, W. (1985) Filosofia, grammatica e retorica nel pensiero antico, Lessico Intellettuale Europeo, 37, Edizioni delFAteneo, Rome. Belletti, A. (1988) 'The Case of Unaccusatives', Linguistic Inquiry, 19, pp. 1-34. (1989) Generalized Verb Movement, Rosenberg & Sellier, Torino. (1994) 'Case Checking and Clitic Placement', ms., University of Geneva. Belletti, A. and Rizzi, L. (1981) 'The Syntax of ne: Some Theoretical Implications', Linguistic Review, 1, pp. 117-54. (1988) 'Psych-verbs and 0-theory', Natural Language and Linguistic Theory, 6, pp. 291-352. Beninca, P. (1980) 'Nomi senza articolo', Rivista di Grammatica Generativa, pp. 51-63. (1989) Dialect Variation and the Theory of Grammar, Foris, Dordrecht. Bennis, H. (1986) Gaps and Dummies, Foris, Dordrecht. Benveniste, E. (1966) 'Eire et avoir dans leurs fonctions linguistiques', in Benveniste, Problemes de linguistique generale, pp. 187-207, Gallimard, Paris. Berwick, R. (1982) 'Locality Principles and the Acquisition of Syntactic Knowledge', Ph.D. dissertation, MIT, Cambridge, Mass. Bloomfield, L. (1933) Language, University of Chicago Press. Borer, H. (1986) 'I-subject', Linguistic Inquiry, 17, pp. 375-416. Borsley, R. and Roberts, I. (eds.) (1996) Celtic and Beyond, Cambridge University Press. Bowers, J. (1993) 'The Syntax of Predication', Linguistic Inquiry, pp. 591-656. 301
302
References
Brekle, H. E. (ed.) (1966) Grammaire generate et raisonnee ou la Grammaire de Port-Royal, facsimile of the 1676 edition, Friedrich Frommann Verlag, Stuttgart. Bresnan, J. (1970) 'On Complementizers: Toward a Syntactic Theory of Complement Types', Foundations of Language, 6, pp. 297-321. Brody, M. (1994) 'Dependency, Projection and Phrase Structure', Working Papers in Linguistics, 6, University College, London. Burzio, L. (1986) Italian Syntax, Reidel, Dordrecht. Cardinaletti, A. (1990) Impersonal Constructions and Sentential Arguments in German, Rivista di Grammatica Generativa Monograph Series, 2, Unipress, Padova. Cardinaletti, A. and Guasti, T. (1995) Small Clauses, Academic Press, Chicago. Carnie, A. and Harley, H. (1994) 'Nominally Complex Copular Constructions', talk given at conference: Grammaire et Langues 1, Paris 8. Chao, W. and Horrocks, J. (eds.) (in press) Levels of Representations, Mouton de Gruyter, The Hague. Chierchia, G. (1989) 'The Semantics of Unaccusatives', ms., Cornell University. (1993) 'Reference to Kinds Twenty Years Later', Conference on Events in Linguistics, Bar Ilan University, Israel, pp. 301-18. (1995) 'Syntactic Bootstrapping and The Acquisition of Nouns' Meaning: The Mass-Count Issue', in Lust etal. (1995). Chierchia, G. and McConell Ginet, S. (1990) Meaning and Grammar, MIT Press, Cambridge, Mass. Chomsky, N. (1957) Syntactic Structures, Mouton, The Hague and Paris. (1973) 'Conditions on Transformations', in Anderson and Kiparsky (1973); republished in Chomsky (1977a), pp. 232-86. (1977a) Essays on Form and Interpretation, North Holland, Amsterdam. (1977b) 'On wh-movement', in Culicover and Akmajian (1977), pp. 71-132. (1981) Lectures on Government and Binding, Foris, Dordrecht. (1982) Some Concepts and Consequences of The Theory of Government and Binding, Linguistic Inquiry Monograph Series, 6, MIT Press, Cambridge, Mass. (1986a) Knowledge of Language: Its Nature, Origin and Use, Praeger, New York. (1986b) Barriers, Linguistic Inquiry Monograph Series, 13, MIT Press, Cambridge, Mass. (1988a) Some Notes on the Economy of Derivation and Representation, MIT Occasional Working Papers in Linguistics, 10, Cambridge, Mass. (1988b) Language and Problems of Knowledge: The Managua Lectures, MIT Press, Cambridge, Mass. (1993) 'A Minimalist Program for Linguistic Theory', in Hale and Keyser (1993), pp. 1-52. (1994) Bare Phrase Structure, MIT Occasional Working Papers in Linguistics, 5, MIT, Cambridge, Mass. Chomsky, N. and Lasnik, H. (1993) 'The Theory of Principles and Parameters', in Jacobs, von Stechow, Sternefeld and Vennemann (1993), pp. 506-69. Cinque, G. (1978) 'Towards a Unified Treatment of Island Constraints', in Dressier and Meid (1978). (1980) 'Extraction from NP in Italian', Journal of Italian Linguistics, 5, pp. 47-99.
References
303
(1988) 'On si Constructions and the Theory of Arb.\ Linguistic Inquiry, 19, pp. 521-81. (1990a) Types of A-bar Dependencies, Linguistic Inquiry Monograph Series, 17, MIT Press, Cambridge, Mass. (1990b) 'Ergative Adjectives and the Lexicalist Hypothesis', Natural Language and Linguistic Theory, 8, pp. 1-39. (1991) 'Lo statuto categoriale del ne partitivo', in Saggi di linguistica e di letteratura in memoria di Paolo Zolli, Biblioteca Veneta, 2, Antenore, Padua. (1992) 'Functional Projections and N-movement within the DP', talk given at the 15th GLOW conference, University of Lisbon. (1994) 'Romance Past Participle Movement and Clause Structure', talk given at conference: XX Incontro di Grammatica Generativa, University of Padua. Cocchi, G. (1994) 'An Explanation of the Split in the Choice of Perfect Aux.', Probus, 6, 3, pp. 87-102. Cooke, H. P. (trans.) (1938) Aristotle, De Interpretatione in Categories, On Interpretation, Prior Analytics, Loeb Classical Library, Harvard University Press, Cambridge, Mass., pp. 112-79. Culicover, P. and Akmajian, A. (eds.) (1977) Formal Syntax, Academic Press, New York. De Mauro, T. and Thornton, A. M. (1985) 'La predicazione: teoria e applicazione all'italiano', in Sintassi e morfologia della lingua Italiana d'uso: teorie ed applicazioni descrittive, proceedings of the 17th Congresso Internazionale di Studi della SLI, Urbino, Sept. 1983, Bulzoni, Roma, pp. 487-519. Delfitto, D. (1992) 'Assegnazione di Nominativo, inversione e rianalisi', talk given at conference: XVIII Incontro di Grammatica Generativa, University of Ferrara. Den Dikken, M. (1995) 'Copulas', talk given at the 18th GLOW conference, Troms0, Norway. Diesing, M. (1990) The Syntactic Root of Semantic Partition, Ph.D. dissertation, University of Massachusetts at Amherst. Dowty, D. R., Wall, R. E. and Peters, S. (1981) Introduction to Montague Semantics, Reidel, Dordrecht. Dressier W. U. and Meid, W. (eds.) (1978) Proceedings of the Twelfth International Congress of Linguistics, Innsbrucker Beitrage zur Sprachwissenschaft, Innsbruck, pp. 344-8. During, I. (1966) Aristoteles - Darstellung und Interpretation seines Denkens, Universitatsverlag, Heidelberg. Emonds, J. (1985) A Unified Theory of Syntactic Categories, Foris, Dordrecht. Farkas, D. et al. (eds.) (1978) Papers from The Fourteenth Regional Meeting, Chicago Linguistic Society. Freeze, R. (1992) 'Existentials and Other Locatives', Language, pp. 593-5. Friedemann, M. and A. Siloni, T. (1994) 'Les phenomenes d'accord dans les temps composes', talk given at conference: Grammaire et Langues 1, Paris 8. Geach, P. (1962) Reference and Generality, Cornell University Press, Ithaca. Giorgi, A. and Longobardi, G. (1991) The Syntax of Noun Phrases, Cambridge University Press.
304
References
Giusti, G. (1992) La sintassi dei sintagmi nominali quantificati, Ph.D. dissertation. University of Venice. Graffi, G. (1980) 'Universali di Greenberg e grammatica generativa', Lingua e Stile, 15, pp. 371-87. (1985) 'Hjelmslev e le frasi nominali', // Protagora, 25, 7/8, pp. 65-89. (1986) 'Una nota sui concetti di Qf\\ia e Xoyogin Aristotele', Athenaeum, LXXIV, pp. 91-101. (1988) 'Structural Subject and Thematic Subject', Linguisticae Investigationes, 12, 2, pp. 397-414. (1990) 'L'analisi in costituenti immediati prima di Bloomfield', Lingua e Stile, 25, pp. 457-569. (1991) La sintassi tra Ottocento e Novecento, II Mulino, Bologna. (1994) Le strutture del linguaggio: sintassi, II Mulino, Bologna. Greenberg, J. H. (ed.) (1963) Universals of Language, MIT Press, Cambridge, Mass. Grimshaw, J. (1990) Argument Structure, Linguistic Inquiry Monograph Series, 18, MIT Press, Cambridge, Mass. Guasti, M. T. (1993) Causative and Perception Verbs: A Comparative Study, Rosenberg & Sellier, Torino. Gueron, J. (1992) Types syntaxiques et types semantiques: la phrase copulative comme "palimpseste"', Revue quebecoise de linguistique, 22, 1, 1992. Haegeman, L. (1990) 'Non-overt Subjects in Diary Contexts', in Mascaro and Nespor (1990), pp. 167-79. Hale, K. and Keyser, J. (1991) On The Syntax of Argument Structure, The Lexicon Project Monograph Series, Center for Cognitive Science, MIT, Cambridge, Mass, (eds.) (1993) The View from Building 20: Essays in Honor ofSylvain Bromberger, MIT Press, Cambridge, Mass. Halliday, M. A. K. (1967-8) 'Notes on Transitivity and Theme in English', Journal of Linguistics, Part I: 2, pp. 37-215; Part II: 3, pp. 111-21 A; Part III: 4, 153-308. Heim, I. (1987) 'Where does the Definiteness Restriction Apply? Evidence from the Definiteness of Variables', in Reuland and ter Meulen (1987), pp. 21^2. Heycock, C. (1991) 'Layers of Predication: The Non-lexical Syntax of Clauses', Ph.D. dissertation, University of Pennsylvania, Philadelphia. (1994) 'Rethinking Inverse Copular Construction', talk given at conference: Grammaire et Langues 1, Paris 8. Higginbotham, J. (1987) Tndefiniteness and Predication', in Reuland and ter Meulen (1987), pp. 43-70. (1990) 'Philosophy of Language' in Osherson and Lasnik (1990), pp. 243-57. Higgins, F. R. (1973) 'The Pseudo Cleft Construction in English', Ph.D. dissertation, MIT, Cambridge, Mass.; published by Garland, New York, 1979. Hintikka, J. and Moravcik, J. (eds.) (1973) Approach to Natural Language: Proceedings of the 1970 Stanford Workshop on Grammar and Semantics, Reidel, Dordrecht. Hjelmslev, L. (1943) 'Omkring sprogteoriens grundlaeggelse', in proceedings of Universitet i Anledning af Universitets Aarfest, November 1943, pp. 3-113.
References
305
Hoekstra, T. and Mulder, R. (1990) 'Unergatives as Copular Verbs', The Linguistic Review, 7, pp. 1-79. Huang, J. (1987) 'Existential Sentences in Chinese and (In)definiteness', in Reuland and ter Meulen (1987), pp. 226-53. (1989) 'Pro-drop in Chinese: A Generalized Control Theory', in Jaeggli and Safir (1989), pp. 226-53. Hyams, N. (1986) Language Acquisition and the Theory of Parameters, Reidel, Dordrecht. Ishiguro, H. (1990) Leibniz's Philosophy of Logic and Language, second edition, Cambridge University Press. Jackendoff, R. (1990) Semantic Structures, MIT Press, Cambridge, Mass. Jacobs, J., von Stechow, A., Sternefeld, W. and Vennemann, T. (eds.) (1993) Syntax, Vol. I, de Gruyter, Berlin and New York. Jaeggli, O. and Safir, K. (eds.) (1989) The Null Subject Parameter, Kluwer Academic Publishers, Dordrecht. Jespersen, O. (1924) The Philosophy of Grammar, Allen & Unwin, London. (1937) Analytic Syntax, Munksgaard, Copenhagen; republished 1984, University of Chicago Press. Kahn, C. (1973) The Verb 'be' in Ancient Greek, Reidel, Dordrecht. Kayne, R. (1984) Connectedness and Binary Branching, Foris, Dordrecht. (1985) 'L'accord du participe passe en frangais et en italien', Modeles linguistiques, 7, 1, pp. 73-89. (1989a) 'Null Subjects and Clitic Climbing', in Ronat and Couquaux (1989). (1989b) 'Facets of Romance Past Participle Agreement' in Beninca (1989). (1993) 'Toward a Modular Theory of Auxiliary Selection', Studia Linguistica, 47. (1994) The Antisymmetry of Syntax, Linguistic Inquiry Monograph Series, 25, MIT Press, Cambridge, Mass. Keyser, J. (ed.) (1978) Recent Transformational Studies in European Languages, MIT Press, Cambridge, Mass. Keyser, J. and Roeper, T. (1992) 'Re: The Abstract Clitic Hypothesis', Linguistic Inquiry, 23, pp. 89-125. Kimball, J. (ed.) (1975) Syntax and Semantics, vol. IV, Academic Press, New York. Kneale, W. and Kneale, M. (1962) The Development of Logic, Clarendon Press, Oxford. Koopman, H. and Sportiche, D. (1991) 'The Position of Subjects', Lingua, 85, pp. 221-58. Koster, J. (1978) 'Why Subject Sentences Don't Exist', in Keyser (1978), pp. 53-64. La Fauci, N. (1984) 'Sulla natura assolutiva del controllore dell'accordo del participio passato in italiano', La Memoria, 3, Annali della Facolta di Lettere e Filosofia dell'Universitd di Palermo. La Grasserie, R. de (1900) Conjugaison pronominale, J. Maisonneuve Libraire Editeur. Laka, I. (1992) 'Negation in Syntax: On The Nature of Functional Categories and Projections', Ph.D. dissertation, MIT, Cambridge, Mass. Lasnik, H. (1989) 'Case and Expletives: Notes Toward a Parametric Account', paper presented at the Princeton Comparative Syntax Workshop.
306
References
Lasnik, H. and Uriagereka, J. (1988) A Course in GB Syntax, MIT Press, Cambridge, Mass. Lepschy, G. C. (ed.) (1990) Storia della linguistica, vol. I, II Mulino, Bologna. (1992) La linguistica del Novecento, II Mulino, Bologna. Lightfoot, D. (1991) How to Set Parameters: Arguments From Language Change, MIT Press, Cambridge, Mass. Longobardi, G. (1985) 'Su alcune proprieta della sintassi e della forma logica delle frasi copulari', in Sintassi e morfologia della lingua Italiana d'uso: teorie ed applicazioni descrittive, proceedings of the 17th Congresso Internazionale di Studi della SLI, Urbino, Sept. 1983, Bulzoni, Roma, pp. 211-13. (1988) Symmetry Principles in Syntax, Clesp, Padua. (1991a) N° movement in Syntax and LF, University of Venice Working Papers in Linguistics, 9. (1991b) 'In Defense of The Correspondence Hypothesis', in May and Huang (1991). Lonzi, L. (1986) 'Pertinenza della struttura Tema-Rema per l'analisi sintattica', in Stammer (1986). (1991) 'II sintagma avverbiale', in Renzi and Salvi (1991). Lust, B., Hermon, G. and Kornfielt, J. (eds.) (1995) Syntactic Theory and First Language Acquisition, vol. I, Lawrence Erlbaum Associates, New Jersey. Lyons, J. (1977) Semantics, Cambridge University Press. Mahajan, A. (1990) 'On the A/A'-Distinction and Movement Theory', Ph.D. dissertation, MIT, Cambridge, Mass. Manzini, R. (1983) 'Restructuring and Reanalysis', Ph.D. dissertation, MIT, Cambridge, Mass. (1992) Locality, Linguistic Inquiry Monograph Series, 19, MIT Press, Cambridge, Mass. (1993) 'Parasitic Gaps and Locality Theory: A Conclusion', in Working Papers in Linguistics, 5, University College, London, pp. 55-73. Manzini, R. and Wexler, K. (1987) 'Binding Theory, Parameters and Learnability', Linguistic Inquiry, 18, pp. 413^4. Marantz, A. (1984) On The Nature of Grammatical Relations, Linguistic Inquiry Monograph Series, 10, MIT Press, Cambridge, Mass. (1995) 'A Reader's Guide to A Minimalist Program For Linguistic Theory', in Webelhuth (1995), pp. 349-82. Mascaro, J. and Nespor, M. (eds.) (1990) Grammar In Progress, Foris, Dordrecht. May, R. (1986) Logical Form, Linguistic Inquiry Monograph Series 12, MIT Press, Cambridge, Mass. May, R. and Huang, J. (eds.) (1991) Logical Structure and Linguistic Structure, Reidel, Dordrecht. Meillet, A. (1921) Linguistique historique et linguistique generale, Champion et Klincksieck, Paris. (1934) Introduction a Vetude comparative des langues indo-europeennes, seventh edition, Hachette, Paris. Milsark, G. L. (1974) 'Existential Sentences in English', Ph.D. dissertation, MIT, Cambridge, Mass.
References
307
(1977) 'Toward an Explanation of Certain Peculiarities in the Existential Construction in English', Linguistic Analysis, 3, 1, pp. 1-30. Montague, R. (1973) 'The Proper Treatment of Quantification in English', in Hintikka and Moravcik (1973). Moro, A. (1987) 'Tempo e predicazione nella sintassi delle frasi copulari', Tesi di laurea, University of Pavia. (1988) 'Per una teoria unificata delle frasi copulari', Rivista di Grammatica Generativa, 13, pp. 81-110. (1989) "Therelci as Raised Predicates', ms., MIT, Cambridge, Mass. (1990) T/iere-raising: Principles Across Levels', talk given at the 13th GLOW conference, St. John's College, Cambridge. (1991) 'The Raising of Predicates: Copula, Expletives and Existence', MIT Working Papers in Linguistics, 15, MIT, Cambridge, Mass. (1993a) / predicati nominali e la struttura della frase, Rivista di Grammatica Generativa Monograph Series 7, Unipress, Padua. (1993b) 'Heads as Antecedents: A Brief History of the ECP', Lingua e Stile, 28, pp. 31-57. (1996) 'Virtual Conceptual Necessity: la semplificazione della grammatica generativa nei primi anni novanta', Lingua e Stile, 30, 4, pp. 637-74. (in press) 'There as a Raised Predicate', in Chao and Horrocks (in press). Osherson, D. and Lasnik H. (ed.) (1990) An Invitation to Cognitive Science, vol. I: Language, MIT Press, Cambridge, Mass. Partee, B. (1986) 'Ambiguous Pseudocleft with Unambiguous be\ ms., University of Massachusetts at Amherst. Peano, G. (1930) 'Algebra de Grammatica', Schola et Vita, 5, pp. 323-36. Perlmutter, D. (1978) 'Impersonal Passives and The Unaccusative Hypothesis', Berkeley Linguistics Society, 4, pp. 157-8. Perlmutter, D. and Postal, P. (1984) 'The 1-Advancement Exclusiveness Law', in Perlmutter and Rosen (1984), pp. 126-70. Perlmutter, D. and Rosen, C. (eds.) (1984) Studies in Relational Grammar 2, University of Chicago Press. Pesetsky, D. (1982) 'Paths and Categories', Ph.D. dissertation, MIT, Cambridge, Mass. (1995) Zero Syntax: Experiencers and Cascades, MIT Press, Cambridge, Mass. Pinborg, J. (1972) Logik und Semantik im Mittelalter, Fromman-Holzboog, Stuttgart. Platzack, C. (1987) 'The Scandinavian Languages and The Null-Subject Parameter', Natural Language and Linguistic Theory, 5, pp. 377-402. Poletto, C. (1993) 'La sintassi del soggetto nei dialetti italiani settentrionali', Quaderni Patavini di Linguistica, Monografie, 12, Unipress, Padova. Pollock, J.-Y. (1984) 'Sur quelques proprietes des phrases copulatives en francos', Langue Frangaise, 58, pp. 89-125. (1989) 'Verb Movement, UG and The Structure of IP', Linguistic Inquiry, 20, pp. 365^24. Quine, W. V. (1950) Methods of Logic, fourth edition, Harvard University Press, Cambridge, Mass. (1987) Quiddities, The Belknap Press, Cambridge, Mass.
308
References
Quirk, R. and Greenbaum, S. (1973) A University Grammar of English, Longman, London. Radford. A. (1990) Syntactic Theory and the Acquisition of English Syntax, Blackwell, Oxford. Rapoport, T. (1987) 'Copular, Nominal and Small Clauses: A Study of Israely Hebrew', Ph.D. dissertation, MIT, Cambridge, Mass. Renzi, L. (1990) Grande grammatica italiana di consultazione•, vol. I, II Mulino, Bologna. Renzi, L. and Salvi, G. (eds.) (1991) Grande grammatica italiana di consultazione, vol. II, II Mulino, Bologna. Reuland, E. and ter Meulen, T. (eds.) (1987) The Representation of (In)definiteness, MIT Press, Cambridge, Mass. Riemsdijk, H. van (1978a) A Case Study in Syntactic Markedness, Foris, Dordrecht. (1978b) 'On The Diagnosis of Wh-movement', in Keyser (1978), pp. 189-206. Ries, J. (1931) Was ist ein Satz, Taussig & Taussig, Prague. Rizzi, L. (1982) Issues in Italian Syntax, Foris, Dordrecht. (1985) 'Conditions de bonne formation sur les chaines', Modeles Linguistiques, 1, pp. 119-59. (1986) 'Null Objects in Italian and The Theory of pro', Linguistic Inquiry, 17, pp. 501-8. (1990) Relativized Minimality, Linguistic Inquiry Monograph Series, 16, MIT Press, Cambridge, Mass. (1991a) 'Proper Head Government and the Definition of A-position', talk given at the 14th GLOW conference, Leiden, Netherlands. (1991b) 'Residual Verb Second and the Wh-criterion', Technical Reports in Formal and Computational Linguistics, 2 (University of Geneva). (1995) 'The Fine Structure of Comp', ms., University of Geneva. Roberts, I. (1988a) 'Thematic Interveners', Rivista di Grammatica Generativa, 13, pp. 111-36. (1988b) 'From Rules To Constraints', Lingua e Stile, 23, 3, pp. 445-64. Robins, R. H. (1967) A Short History of Linguistics, Longmans, Green & Co., London. Rochemont, M. S. and Culicover, P. W. (1990) English Focus Constructions and The Theory of Grammar, Cambridge University Press. Ronat, M. and Couquaux (eds.) (1989) La grammaire modulaire, Editions de Minuit, Paris. Rosen, C. (1981) 'The Relational Structure of Reflexive Clauses: Evidence from Italian', Ph.D. dissertation, Harvard University. Rothstein, S. (1983) 'The Syntactic Form of Predication', Ph.D. dissertation, MIT, Cambridge, Mass. Rouveret, A. (1996) 'BOD in the Present Tense and in Other Tenses', in Borsley and Roberts (1996). Russell, B. (1900) A Critical Exposition of the Philosophy of Leibniz, London. (1919) The Philosophy of Mathematics, Allen & Unwin, London. Ruwet, N. (1975) 'Les phrases copulatives', in Recherches Linguistiques (University of Paris-Vincennes), 3, pp. 143-91; reprinted in Ruwet (1982). (1982) Grammaire des insultes et autres etudes. Editions du Seuil, Paris.
References
309
(1982) Grammaire des insultes et autres etudes, Editions du Seuil, Paris. Safir, K. (1985) Syntactic Chains, Cambridge University Press. Salvi, G. (1990) 'Le frasi copulative', in Renzi (1990). Seidel, E. (1935) Geschichte und Kritik der wichtigsten Satzdefinitionen, Biedermann, Jena. Shlonsky, U. (1987) 'Null and Displaced Subjects', Ph.D. dissertation, MIT, Cambridge, Mass. Simone, R. (1990) Seicento e Settecento, in Lepschy (1990). Smyth, H. (1920) Greek Grammar, Harvard University Press, Cambridge, Mass. Sommers, F. (1982) The Logic of Natural Language, Oxford University Press. Stammer, J. (ed.) (1986) Tema-Rema in italiano - Theme-Rheme in Italienischen, Symposium, Frankfurt am Main, 1985, Gunter Narr Verlag, Tubingen. Stowell, T. (1978) 'What Was There Before There Was There?', in Farkas et al. (1978). (1981) 'Origins of Phrase Structure', Ph.D. dissertation, MIT. Sweet, H. (1892) A New English Grammar, Clarendon Press, Oxford. Szabolcsi, A. (1983) The Possessor that Ran Away from Home', The Linguistic Review, 3, pp. 98-102. Tomaselli, A. (1990) La sintassi del verbo finito nelle lingue germaniche, Rivista di Grammatica Generativa Monograph Series 4, Unipress, Padua. Travis, L. (1984) 'Parameters and Effects of Word Order Variation', Ph.D. dissertation, MIT, Cambridge, Mass. Vendryes, J. (1921) Le Langage, introduction linguistique a I'histoire, La Renaissance du Livre, Paris. Vikner, S. (1995) Verb Movement and Expletive Subjects in The Germanic Languages, Oxford University Press. Webelhuth, G. (ed.) (1995) The Principles and Parameters Approach To Syntactic Theory: A Synopsis, Black well, Oxford. Williams, E. (1975) 'Small Clauses in English' in Kimball (1975), pp. 249-73. (1980) 'Predication', Linguistic Inquiry, 11, pp. 203-38. (1984) T/*ere-insertion\ Linguistic Inquiry, 15, pp. 131-53. (1994) Thematic Structure, Linguistic Inquiry Monograph Series, 23, MIT Press, Cambridge, Mass. Zamparelli, R. (1995) 'Layers in The Determiner Phrase', Ph.D. dissertation, University of Rochester, Pennsylvania. Zanuttini, R. (1992) 'Syntactic Properties of Sentential Negation. A Comparative Study of Romance Languages', Ph.D. dissertation, University of Pennsylvania, Philadelphia. Zwart, J. W. (1991a) 'Expletive Raising and Expletive Replacement in Dutch', in Proceedings of ESCOL 8, pp. 393-404. (1991b) 'Dutch Expletives and Small Clause Predicate Raising', in Proceedings of NELS22.
Index
A/A'-distinction 140,190-1 as an insufficient dichotomy, 263 n8 Abelard 248,251-3,254,255,296 nl6 Abney,S.24 Ackrill, J. 249,250,295 nl, n4,296 n i l , nl6 Acquaviva, P. 288 n35 acquisition of language and parameter setting, 67 and/?ro-drop parameter, 68-70 and the Definiteness Effect, 131-3,150, 155-6 and the subset principle, 132 and unaccusative constructions, 227-9, 232-3 see also semantic parameter address 266 n35 adjectival character of quantifiers definition of, 143,147-8 technical sense of, 278 nlO adjective phrases and agreement in small clauses, 54 [±N, ± V] features and adjective phrases, 77 adjuncts and small clauses, 52,276 n i l as base generated, 109 as opposed to predicates in existential sentences, 107-8,110 extraction of the adjoinee, 57 extraction of the adjunct, 57 adverbials as small clauses, 113 extraction from, 113-14,186-8 adverbs and the split-Infl hypothesis, 178 as diagnostics for functional projections, 55 in small clauses, 54-6 in verb-second constructions, 185 agreement and parameter setting, 67-70 and predication, 54 and the alternation between avere and essere, 238-41
310
and the internal structure of noun phrases, 268 n47 anomalous agreement in existential sentences, 281 n39 as a diagnostic for the subject, 256,281 n39 as a spec-head relation, 61,67 as relevant for locality theory, 40-1,47-8 in unaccusative constructions, 245 lack of agreement heads in copular sentences, 53-4 object agreement and past participle agreement, 271 n65 representation of, 22,40 rightward agreement in copular sentences and the theory of pro, 60-7 Akmajian, A. 302 Anderson 302 anaphors anomalous behaviour in predicative noun phrases, 86-9 anomaly of copular sentences cross-linguistic anomalies, 25-7 language-specific anomalies, 27-8 see also asymmetries, subject-object; copular sentences antecedent-government andECP,34 by heads in extended chains, 34 antisymmetry 276 n i l argument structure in Hale and Keyser's project, 229-31 inNPs,80 in VPs, 79-80,233-5 Aristotle 248,249-51,252,253,255,295 n3, 296 n i l as and small clauses, 203,287 n29,287 n30, 287n31,296nl6 as a predicative marker, 203 as different from the copula, 203 as equivalent to to, 203-4 see also come asymmetries, subject-object in copular sentences, 25-9 in transitive sentences, 17-19
Index asymmetries, subject-object {continued) see also anomaly of copular sentences; clause structure Aux-to-comp constructions 67-8 auxiliary selection 234,238 see also averci', esserci; possessive constructions averci as similar to esserci, 236-41 see also avere; possessive constructions avere as a main verb in possessive constructions, 237 as an auxiliary, 234,293 n34 as an enriched copula, 241 as selecting a small clause, 237,293 n34 see also averci backward reflexivization 274 n87 Baker, M. 22,108,226,227,264 n20,276 nl4 barriers 50-1 Barwise,J.277nl Belardi,W.295n3 Belletti, A. 22,83,108,215,271 n63,280 n26, 294 n35 Beninca,P.24,116 Bennis, H. 185,186,205,284 n i l Benveniste, E. 238,251,254,255,295 n2,297 n26 Berwick, R. 132 Binding theory and anaphors in predicative noun phrase, 86-9 and Condition C violation, 100,114 Bloomfield, L. 295 n6,300 n41 Borer, H. 100 Borsley, R. 301 Bounding theory as distinct from ECP, 49-52,262 n7,276 nl9 as part of locality theory, 262 n6 Bowers, J. 31,52,267 n42,287 n29 Brekle,H.E.252 Bresnan,J.263n3 Brody,M.56,109,263n4 Bromberger,S.269 Burzio, L. 63,102,216-18,224,226,233, 242-5,264 nl8,275 n4,282 n39,288 n3, 289 n5,291 nl9,292 n21,293 n29,294 n36,300n45 Burzio's generalization 242-3 c-command 263 n 12 c-domain 26 canonical copular sentences and extraction from postverbal predicative noun phrase, 52 definition of, 35
311
see also copular sentences; inverse copular sentences Cardinaletti, A. 31,52,266n30,284nil Carnie,A.295n2 Case and clitic movement, 245 and predicative noun phrases, 41-4 and Super Raising, 192 and the visibility condition, 41 and there, 120 and unaccusatives, 242-5 assignment of, 42-3 by default, 42 case agreement, 41 Case theory, 242-5 types of Cases and the distribution of Agr°, 43 Case agreement 41 chains and the expletive replacement hypothesis, 99-101 interference of, 269 n50 uniformity of, 190-1 Chao,W.302 Chierchia, G. 132,234,265 n27,278 n8,283 n7,286 n20,292 n26,293 n30 Chomsky, N. 16,18-22,34,38,39,40,41,42, 47,50-2,56,69,77,86,96-102,109,113, 115,122,126,128,130,131-3,134,141, 164,167,168,177,178,186,190,194, 198,199,200,202,209,210,234,238, 245,251,258-61,262 n7, n8,263 n3, n4, n8,264 nl6, n20,265 n21, n22, n23, n30, 266 n31, n35,267 n39,268 n46,269 n51, 270 n53, n58,273 n80, n83,275 n3, n7, 277 n21, n3, n4, n6,286 n24, n25,287 n30, n33,294 n36,299 n37, n38,300 n44 ci a comparison with English there, 118ff. alternative analysis as a raised predicate, 107ff. and the lexicalization of the copula, 115-18 distribution of, 103-7,275 nlO, 293 n34 standard analysis as an expletive, 102-3 with avere, 237 with clausal subjects, 175 with unaccusatives, 222-3 Cinque, G. 16,51,54,73,74,82,88,115,141, 143,186,263 n2,264 n20,265 n25,268 n44, n47,269 n50,272 n70,273 n78,277 n4,284 nl2,286 n24,288 n35,293 n29. Cinque's generalization 82 see also pseudo-extraction clausal arguments with AP predicates, 182-3 with passives, 168-9 with seem, 169-70
312
Index
clause structure and the Comp periphery, 258-9,263 n3 and the INFL node, 20-1 on the definition of clause, 298 n30,300 n41 richer representation of, 22 simple representation, 263 n5 see also flexible clause structure; rigid clause structure clitic doubling 154 Clitic Left Dislocation 74 see also lo cliticization and agreement, 71-2,240-1,245 and Case, 245 and predicative noun phrases, 71-5 and unaccusatives, 245 Cocchi, G. 271 n65 coda 113 coindexing and Binding theory Condition C violation, 97,100,114 and Burzio's generalization, 242-3 and predication, 97 come 207-8 Complete Functional Complex and predicative noun phrases, 86-8 updated version of, 87 configuration and grammatical functions, 19,258-9 and 6-roles, 226-32 control and predicative noun phrases, 87-9 andseem-sentences, 194-5,200,208-9 Cooke,H.P.250,296nl6 Cooper, R. 277 nl copula as a sign of affirmation, 251-3 as a sign of identity, 253-5 as a sign of tense, 249-51 as a spell out of inflection, 264 nl7 different meanings of, 248ff. invention of the term, 251-2 names of the copula in Aristotle's works, 296 nl6 see also anomaly of copular sentences; copular sentences copular sentences as expanded small clauses, 31-2 types of, 23 see also anomaly of copular sentences; canonical copular sentences; flexible clause structure; inverse copular sentences correspondence hypothesis 263 nlO Couquaux 305 Culicover,P.W.289nl2 DeMauro,T.250
defining properties and existential sentences, 159 and passive sentences, 135 notion of, 133-4 Definiteness Effect 282 n41 and Italian, 131,150ff. and language acquisition, 131-2,155-6 distribution of, 131 see also semantic parameter Delfitto,D.294n35 Den Dikken, M. 265 n25,275 n87 determiner phrases and inverse copular sentences, 23-4 and the relation with noun phrases, 24 and wh-phrases, 58 see also noun phrases Diesing,M. 116,147 Dowry, D.R. 297 n25 DP splitting analogy with wh-movement, 145-6,279 n20 and existential meaning, 142ff. see also Definiteness Effect; existential sentences Dressier, W.U. 303 During, 1.295 n3 Dutch 185,205 ECP as a 'long distance action' 265 n29 as distinct from the Subjacency Condition, 40,49-52,262 n7,276 nl9 different formulations of, 34,40,266 n35 Emonds, J. 178,287 n28,300 n40 empty categories and agreement, 40-1 and Universal Grammar in ECP, 40-1 asD°, 116-18,147-8 see also pro; PRO; traces, Empty Category Principle, see ECP English comparisons between Italian ci and English there, 118ff. comparisons between Italian sembrare and English seem, 205ff. esserci and extraction, 103-4 as an unaccusative verb, 218ff. see also copula; essere essere and avere, 241-2 and the unaccusative hypothesis, 220-1 existential meaning and DP splitting, 142ff. and the verb exist, 136 definition of, 136 existential sentences and levels of representation, 138-9
Index existential sentences (continued) and Universal Grammar, 133 as copular sentences, 136,160-3 types of, 160-3 see also DP splitting; existential meaning; Jespersen's generalization Expletive Replacement Hypothesis 99ff. and affixation, 101 and locality, 100 and the principle of Full Interpretation, 99 unsolved problems for the, 102,127 expletives as raised predicates, 107-8 ci, 107ff. het, 185-6 pro,223-4 see also it, there extended chains and agreement, 34 and locality, 38-9 Extended Projection Principle 96,260-1 extraction from adjuncts, 57,113-14,186-8 from predicative noun phrases, 52ff. from subject of existential sentences, 103ff.,279n20 from subject of inverse sentences, 49ff. of adverbial elements, 186-8 of ne, 28,60,269 n50,272 n67 extraposition 175,183-4,283 n6 Farkas,D.3O3 features and Case theory, 42-3 and ECP, 40-1 matching and predicative linking, 54 pro as a collection of, 66 [±N, ± V] features and predicative noun phrases, 77 flexible clause structure and copular sentences, 10-11,34-5,90 and existential sentences, 107-8,166 andsmn-sentences, 173-4,212 and unaccusative constructions, 223-4 see also clause structure; rigid clause structure floating quantifiers 267 n43 Freeze, R. 241,281 n37,293 n29, n31 Friedemann, M. 271 n65 Full Interpretation Principle and auxiliaries, 178 and expletives, 178-9 and predication, 99 functional heads and clause structure, 20-2,267 n42 and locality theory, 40-1,265 n29 and noun phrases, 24,73,268 n47
313
see also agreement Geach, P. 59,265 n26,298 n33 Giorgi, A. 77,82,87-8,272 n70, n74,274 n86, 279 n20 Giusti, G. 143,267 n43,269 n50 government 264 n20 see also antecedent-government; ECP; head-government Graffi, G. 250,253,255,259,263 n4,265 n28, 267n40,289n7,295nl,n6,296nll,n21, 298n30,299n37,n39 grammatical functions configurational definition of, 19,259 relevance of Chomsky's proposal, 19,259 types of, 19 Grasserie, R. de la 298 n32 Greenbaum, S. 133,254,289 nl2,298 n28 Greenberg, J. H. 263 n4,299 n39 Grimshaw, J. 226,272 n71, n75 Guasti, M. T. 31,52,266 n30,293 n34 Gueron,J.300n43 Haegeman, L. 69,273 n79 Hale, K. 16,221,224,226-32,242,247,271 n66,290nl6 Hale and Keyser's project 226ff. see also 0-theory Halliday,M.A.K.254 Harley,H.295n2 head-government 34 see also antecedent-government; ECP; government, heads andd, 108-9,116 and movement, 34,40,265 n29 and there, 125-6 functional heads, 20-2,24,287 n29, n31 lexical heads, 115-16,291 nl9 see also lexicalization of the copula Heim, 1.127 Hermon, G. 306 /tef 185-6 Heycock,C.93,300n43 Higginbotham, J. 53,97,143-4,147-8,149, 159,164,279nl8,nl9,283n4 Higgins, F. R. 59,92-3,259,265 n25,298 n33, 300n43,n45 Hintikka,J.3O4 Hjelmslev,L.299n37 Hoekstra, T. 281 n37,289 nl2 Horrocks,J.302 how many and extraction from existential sentences, 102,127 see also what, which Huang, J. 69,161
314
Index
identity evidence against the copula as an equative predicate, 298 n33 in Jespersen's thought, 256 in Russell's thought, 254,297 n23 in the modern tradition, 254 indexing as label for features, 265 n25 see also coindexing infinitival clauses and seem, 194ff. and sembrare, 205ff. as small clauses, 201-4 INFL, see inflection inflection as a syncretic node, 20-1 as an autonomous node, 20-2 see also split-Infl hypothesis !>ui283n6 inverse copular sentences and existential sentences, 108,121,138, 162-3 and extraction from within postverbal noun phrase, 103-4, 124 and extraction of postverbal noun phrases, 121ff. and seem, 173-4,212 and unaccusative sentences, 223-4 definition of, 35 peculiar properties of, 45ff. with clausal subjects, 174-6 see also flexible clause structure inversion, see rightward movement Ishiguro,H.297n22 it and extraposition, 183-4 in constructions with be, 172-6 in constructions with seem, 170, 173ff. in obvious-type constructions, 182-3 in passive sentences, 181-4 synopsis of different its, 184 Italian 19-20,27-9,102-3,205-10
Keyser, J. 16,221,224,226-32,242,247,273 n66,290nl6,291n20 Kimball,J.305 Kiparsky302 Kneale, M. 251,295 nl, 296 nl7 Kneale, W. 251,295 nl, 296 nl7 Koopman, H. 79,224,288 n4 Kornfielt,J.306 Koster,J.282nl
L-marking and ci, 115-16 and there, 125 see also heads; lexicalization of the copula X-operator as analogous to Principle JI, 277 n8 La Fauci, N. 264 nl4,292 n24 Laka,I.287n27 Lasnik, H. 77,168,181,283 n6,285 nl4 Latin and Case agreement in copular sentences, 41-2 left-branch position 36,262 n2 Lepschy,G.C. 19,295nl,296nl8 lexical 77,115ff. lexical categories 50-1,77 lexicalization of the copula byd,115ff. by there, 124ft see also heads; L-marking, licensing of empty categories and Universal Grammar, 40-1 of pro, 66-7,68-70 of traces, 40-1 Lightfoot,D.277n2 list interpretation and agreement, 279 n21 see also locative interpretation; locative inversion lo andECP,71ff. as a propredicative element, 71 as an N° rather than D°, 72-3 Jackendoff, R. 226,272 n71 as an uninflected clitic, 72 Jaeggli,O.69 as opposed to inflected clitics, 29,71ff. Jacobs, J. 305 in CLLD constructions, 74 Jespersen, 0.54,66,91-2,159-63,251,254, locality 255-8,259,261,267 n41,281 n35, n38, as equivalent to ECP, 262 n6, n7,266 n35, n39,298 n31, n32, n33,299 n34, n35 276 nl9 Jespersen's generalization 161 in the broad sense, 16-17,262 n6, n7 see also existential meaning; existential in the narrow sense, 40-1,266 n35 sentences locative interpretation and d,138,153-4 Kahn,C254 and DP splitting, 153-5 Kayne,R. 16,18,34,42,56,71,108,109,141, and existential meaning, 137-8 206,213,241,275 n87,276 nil, 287 n32, and there, 155 288 n35,290 nl4,291 n20,292 n27,299 n39 see also existential sentences
Index locative inversion 299 nl2 Longobardi, G. 24,28,53,77,82,87-8,116, 117,143,147-8,263 nlO, 269 n49,270 n53,272 n70, n74, n83,278 n i l , nl7,279 n20,300n43,n45 Lonzi,L.54,288nl Lust, B. 306 Lyons, J. 299 n37 Mahajan,A.263n8 Manzini, R. 31,52,69,109,132,141,186,266 n34,n35,279n20 Marantz, A. 132,226,227,262 n7,272 n71, 290nl5,292n22 Mascaro,J. 306 May, R. 26,103 McConell Ginet, S. 277 n8,283 n7 Meid,W.3O2 Meillet, A. 251,253,256,299 n37 Meulen,T.97 Milsark, G. L. 259,95-6,277 nl, 278 nl6 minimalism 262 n7,265 n30 Montague, R. 254,273 n80,297 n25 Moravcik, J.304 Moro, A. 22,31,40,52-3,127,157,190,191, 209,237,262 n7,263 n8,264 nl7,265 n22, n24, n25, n26, n29,266 n35,267 n42,269 n50, n51,270 n58, n59,273 n80, n81,274n86,276nll,nl5,279nl8,281 n37,286 n24,287 n28, n29,289, nl2,293 n29, n34,297 n25,298 n33,299 n40,300 n44,n45 movement, see quantifier raising; whmovement Mulder, R. 281 n37,289 nl2 rce-extraction 28,60,269 n50,272 n67 negation and interaction with quantifiers, 26-7 sentential negation, 27,265 nl3 Nespor, M. 306 nominal sentence 251,253 non-lexical heads, see functional heads, heads non-pronominal empty categories, see traces noun phrases raising of predicates within noun phrase structures, 274 n87 0-role assignment in noun phrases, 37-8, 80ff. [±N, ± V] features and noun phrases, 77 null determiners and existential meaning, 147-8 and government by a lexical head, 116-18 null predicates as proper governors, 64-5 in inverse copular sentences, 64-5,67-8 in unaccusative constructions, 223-4
315
see also null subjects;pro null subjects terminological difficulties, 65 see also null predicates; pro opacity and anaphors in predicative noun phrases, 86-8 see also Binding theory; predicative noun phrases; PRO operators and predication, 139-40 definition of, 140 see also Principle n Osherson,D.297n24 paradox, apparent in Aristotle's definition of 'verb'250 parameter and language acquisition, 68-70 and semantics, 132-3,150,155-6 and the subset principle, 131-2 see also Definiteness Effect; parameter setting parameter setting and inverse copular sentences, 69-70 and the pro-drop parameter, 68-9 see also rightward agreement Partee,B.297n25 partitive phrases 156-9 passives and active sentences, 133-5,167-9 and predicative noun phrases, 78ff. past participle agreement and cliticization, 71,73,271 n65 and raising, 225 and unaccusatives, 214,217,225 as distinct from agreement with the object, 271 n65 Peano,G.297n22 Perlmutter, D. 214-16,288 n2 Pesetsky, D. 263 n2,273 n85,286 n25 Peters, S. 296 n25 Pinborg,J.296nl7 Platzack,C69 Poletto, C. 293 n28,294 n34 Pollock, J.-Y. 22,178,259,267 n39,269 n51, 270 n58,287 n28,299 n40,300 n43 Port Royal grammar 252-3 possessive constructions 236-42 see also averci; avere Postal, P. 216,288 n2 PP preposing 279 n21,289 nl2 see also list interpretation; locative interpretation; locative inversion predication as mediated by a syntactic operation, 139-40
316
Index
structural conditions of, 85,235,290 nl7, 292 n25 see also Principle jt predicative nominals, see predicative noun phrases predicative noun phrases, 11,13,14,35ff., 77ff.,298n33 and Case assignment, 41-4 and cliticization, 71ff. and 9-role assignment, 77ff. features, 52-4,65,77 flexible clause structure, 11 opacity, 86-9 pronouns, 65,172 proper names, 172-3 see also pseudo-extraction prepositional phrases and the coda in existential sentences, 107-8 [±N, ± V] features and prepositional phrases, 77 Principle x as an empirical principle, 139-40 see also operators; predication; variables pro as a null predicate, 65 as obligatory in Italian, 68-9 condition of licensing, 66-7,68-70 see also parameter setting PRO and opacity, 87-9 and predicative noun phrases, 87-9 projection of d, 108 of there, 125-6 Projection Principle and there, 96,260-1 see also Extended Projection Principle pronouns aspredicates,65,172 proper government, see ECP, government proper names as predicates, 172-3 see also pronouns propredicates ci, 107ff. egli, 175 it, 170,172-6,183ff. lo,71 pro, 65 there, 121 pseudo-extraction and Cinque's generalization, 82 and movement, 82-3 see also psych-nouns psych-nouns 83 see also pseudo-extraction psych-verbs 83
quantifier raising (Q-raising) and the anomaly of copular sentences, 25, 27 see also c-domain; scope marker quasi-copula seem as a quasi-copula, 170ff. see also seem Quine,W.V.133,254 Quirk, R. 133,289 nl2,298 n28 Radford, A. 69 raising31,35,107-8,173-4,223-4,225^6 see also Super Raising Rapoport,T.300n43 reciprocals and opacity 273 n85 referentiality 265 n25,298 n33 relative clauses 283 n7 Renzi,L.254 Reuland, E. 97 Riemsdijk, H. van 268 n46,277 n4 Ries,J.298n30 rightward agreement cross-linguistic contrast between English and Italian, 28,60-1 the role of the propredicativepra, 60ff. rightward movement and the Definiteness Effect in Italian, 1534 and the that-trace effect in Italian, 19-20 elimination of, 276 nl8 rigid clause structure as opposed to flexible clause structure, 10-11,34-5 see also clause structure; flexible clause structure Rizzi, L. 16,19,34,39-40,63,65,69,83,108, 123,152,186,188,190,215,259,262 n8, 263 n3, n8, n9,265 n22, n25,266 n35,268 n46,269 n50,270 n53, n56,273 n79,276 nl8,285nl6,288n4,289nl2 Roberts, 1.200,262 n7,267 n36 Robins, R.H. 295 nl Rochemont, M. S. 289 nl2 Roeper,T.291n20 Romance 27-9 Ronat, M. 305 Rosen, C. 288 n2 Rothstein, S. 57,168,260,273 n80 Rouveret, A. 295 n2 Russell, B. 249,253-5,259,261,297 n22, n23, 299 n35 Russian 54,253 Ruwet, N. 28,194,195,274 n87,286 n21,300 n43 S'-deletion 198-9 alternatives to, 201,205
Index S'-deletion {continued) and seem, 197ff. as a problem, 199-200,286 n24 Safir, K. 69,102,120 Salvi, G. 254 saturation and predicative link, 260-1 and predicative nominals, 81,84-6,273 n80 scope marker inadequacy of the theory of there as a, 126 there as a, 101 seem differences with respect to Italian sembrare, 205-10 with finite clauses, 169ff. with infinitival (for-) clauses, 194ff. see also sembrare Seidel,E.298n30 semantic parameter and language acquisition, 132-3,150,1556 and the Definiteness Effect, 131-3,150, 155-6 see also acquisition of language; semantic restriction semantic restriction and the Definiteness Effect, 131-3,155-6 see also acquisition of language; semantic parameter sembrare as analogous to seem, 205 as different from seem, 205ff. Shlonsky,U.300n43 Siloni,T.271n65 Simone,R.296nl9 small clause and as, 203,287 n29, n30, n31 and the order of DPs within it, 36-8 and to,203 and X'-theory, 52-7 as adjunct structure, 52 as projected by a head, 52-3,267 n42,287 n31 as projected by Agr°, 53 as projection of predicate, 52,56,287 n29 infinitival clauses as small clauses, 201^1 see also adjuncts Smyth, H. 292 n23 Sommers,F.297n24 spec-head agreement and Case theory, 42-3 and locality, 40-1 and verb inflection, 39,61-2 spec-head relation, see spec-head agreement specifier 258 split-Infl hypothesis and copular sentences, 22
317
and the theory of clause structure, 22-3 different empirical sources, 270 n58 Sportiche, D. 79,224,288 n4 Stammer, J. 309 Sternefeld,W.3O2 Stowell, T. 31,34,52,96-7,300 n43 Subjacency and the anomaly of copular sentences, 49ff. as distinct from the ECP, 49-52,262 n7,276 nl9 Cinque's one-barrier version, 51 intensional version, 50 original version, 50,262 n7 see also ECP; locality; Bounding theory subject-in-VP hypothesis 288 n4 subject-object asymmetries, see asymmetries, subject-object; subject-inVP hypothesis subset principle 131-2 see also acquisition of language; parameter setting Super Passive 285 nl6 Super Raising an alternative analysis of, 191-3 traditional account of, 188-91 see also Super Passive Sweet, H. 66 syllogism 251 Szabolcsi,A.241,293n33 0-roles hierarchy of, 226-8,229-31,232-3,290 nl6 in unaccusatives, 228-9,232-3 types of, 226-7,290 nl6 0-theory and Case theory in Burzio's generalization, 242-5 and language acquisition, 227,232-3 Hale and Keyser's project, 229-31 traditional theory, 226-9 that-trace effect 20,278 n9 there and types of unaccusatives, 291 n21 as a propredicate, 145,278 nl4 as a raised predicate, 118ff. traditional analysis, 95-103 Thornton, A. M. 250 to and negation, 204 and the split-Infl hypothesis, 287 n28 and VP preposing, 203-4 as equivalent to as, 203-4 Tomaselli, A. 267 n40,284 n i l traces and agreement, 34,40-1 and ECP, 34 and Universal Grammar, 40-1 licensing of, 40-1
318
Index
traces (continued) see also licensing Travis, L. 108 unaccusative constructions alternative analysis, 222ff. and esserci, 218-22 Case, 242-5 diagnostics, 214^16 representation, 216-17,228-9 see also 6-theory; unaccusative hypothesis unaccusative hypothesis as related to a unique factor, 217 as related to distinct factors, 221 problems for, 220-1 unergative288n3 Universal Grammar 40-1,133,159,226 unstable valency and the alternation between essere and avere, 292 n26 and unaccusatives, 234 as a consequence of the alternative analysis, 234-5 Uriagereka, J. 168,181,285 nl4 UTAH 226-8 see also 0-roles; 0-theory variables 139-40 variation, cross-linguistic, in semantics, 131-3 limits of variation in existential sentences, 133,159-63 Vendryes,J.251,253,256 Vennemann,T.3O2 verb phrases 0-role assignment in, 226ff. [±N, ± V] features and verb phrases, 77
see also subject-in-VP hypothesis verb-second languages 185-6 Vikner,S.284nll visibility condition 41 see also Case vonStechow,A. 305 Wall, R.E. 297 n25 Webelhuth,G.3O9 Wexler, K. 69,132 wh-elements and the complex structure of noun phrases, 58-9,127 see also how many; what; which wh-movement and the anomaly of copular sentences, 25, 45,49 what and DP splitting, 145-6,279 n20 and the Definiteness Effect, 58-9,145-6 as lower than D°, 58-9 see also how maw)'; wh-elements, which which and the Definiteness Effect, 58-9,145-6 as associated to D° 58-9 see also how many; wh-elements; what Williams, E. 31,97,101,126,225,226,235, 260,273n80,275n8,n9 X-bar theory 56,258 see also adjuncts; antisymmetry; small clause Zamparelli, R. 275 n87 Zanuttini, R. 55,287n27 Zwart,J.W.281n37
