Issues in Italian Syntax [Reprint 2012 ed.] 9783110883718, 9783110139143

Citation preview

Issues in Italian Syntax

Studies in Generative Grammar The goal of this series is to publish those texts that are representative of recent advances in the theory of formal grammar. Too many studies do not reach the public they deserve because of the depth and detail that make them unsuitable for publication in article form. We hope that the present series will make these studies available to a wider audience than has hitherto been possible. Jan Köster Henk van Riemsdijk editors

Luigi Rizzi

Issues in Italian Syntax

1982 FORIS PUBLICATIONS Dordrecht - Holland/Cinnaminson - U.S.A.

Published by: Foris Publications Holland P.O. Box 509 3300 AM Dordrecht, The Netherlands Sole distributor for the U.S.A. and Canada: Foris Publications U.S.A. P.O. Box C-50 Cinnaminson N.J. 08077 U.S.A.

ISBN 90 70176 33 5 (Bound) ISBN 90 70176 22 4 (Paper) © 1982 Foris Publications - Dordrecht No part of this book may be translated or reproduced in any form, by print, photoprint, or any other means, without written permission from the publisher. Printed in the Netherlands by Intercontinental Graphics, H.I. Ambacht.

Contents

Introduction I. A restructuring rule II. Violations of the Wh island constraint and the subjacency condition

vii 1

49

III. Lexical subjects in infinitives: government, case and binding

77

IV. Negation, WA-movement and the null subject parameter

117

Bibliography

185

Introduction

The following essays are focused on various descriptive and theoretical issues which arise in the study of Italian syntax. The theoretical framework which forms the background of this study and to which appeals are constantly made is the so-called "Extended Standard Theory" (EST), the theory of grammar developed by Noam Chomsky and his school in the last decade.1 This theory has evolved rapidly in recent years, achieving a significant level of explanatory power in extensive empirical domains. Two very general aspects of this recent development have been particularly influential in guiding the preparation of the following essays. The first aspect stems from what in retrospective appears to have been the real starting point of this development, Chomsky's article "Conditions on Transformations" (cf. Chomsky (1973)), which defined a program and a style of research. "Conditions" had the major merit of showing, for the first time in a concrete and detailed manner, how the problem of principled explanation in syntax could be successfully addressed over large empirical domains. The success of this program opened a new, exciting perspective for syntactic theory: the gap between presenting the intricacy of descriptive problems and the aim of arriving at an explanation could be bridged without sacrificing either of these two complementary aspects of syntactic research. The common goal of the four chapters of this book is to show that the "Conditions" program, at various stages of its elaboration, can be fruitfully extended to the study of Italian syntax: the system of theoretical principles has become so pervasive in its empirical consequences that syntactic processes of a new and very different type from those which originally motivated the system can also be shown to fall within the scope of its explanatory power. Moreover, the attempted extension not only sheds light on important aspects of the structure of Italian, but also has consequences for the general theory, in that the broadening of its empirical domain can contribute to clarifying significant theoretical issues. Such a "feed-back" effect leads us to the second general factor which has influenced the following studies. For a lucky coincidence, the deepen-

viii

Introduction

ing of the grammatical model in the late 1970s,has proceeded in conjunction with a renewed interest in comparative syntax. Careful formal analyses of various syntactic domains of different languages, specially in the Romance and Germanic areas, have provided new types of empirical evidence for testing the general theoretical framework and have suggested significant refinements of the theory and sometimes also radical modifications of specific hypotheses. Moreover, for the first time in the development of the generative framework, this work has provided a sufficient empirical basis for seriously addressing the theoretical problem of syntactic variation across languages. Descriptive problems traditionally belonging to the domain of comparative syntax and linguistic typology have been revisited from a sophisticated theoreticalperspective, whichhaspermittedimportant progress in the explanation of known patterns of variation, as well as the discovery of new ones. The second goal of the following essays, explicated in full in the fourth chapter of this book, but present throughout, is to make a contribution to this program, through the investigation of various syntactic properties which appear to cluster coherently across languages. Even if the discussion is generally focused on Italian data, the comparative and typological counterpart of the problems is always present in the background, and often explicitly addressed. A few remarks on the general model, its recent development, and its present form are now in order. It is assumed that knowledge of language as it is manifested in ordinary linguistic behavior is to be characterized by a generative grammar, a formal system organized in various subcomponents, each of which contains a finite number of rules of a specific type. The grammar generates the sentences of the language and specifies, for each sentence, a set of structural representations which explicitly characterize the properties of the sentence on different linguistic levels. There are four fundamental structural representations: the Phonetic Representation (PR) and the Logical Form (LF) specify properties of the sound and of the meaning of the sentence, respectively; this (partial) characterization of the sound-meaning relation is mediated by two fundamental syntactic representations: the Deep Structure (DS), an abstract characterization of the "argument structure" of the sentence, satisfying the selectional requirements of the lexical items involved, and the S(urface) Structure (SS), a more "concrete" (i.e., closer to PR and LF) representation of the syntactic structure. 2 D Structures are characterized by two components: a categorial component containing specific instantiations of general rule schemata for structural configurations provided by the X-bar system,3 which generates abstract syntactic structures; and a lexicon, a list of lexical items specified w.r.t. phonetic syntactic and semantic properties; DS's are generated by inserting lexical items into the syntactic structures characterized by the categorial component. The four fundamental structural representations are mediated by different rule systems, as indicated in the following diagram:

ix

Introduction

DS

O)

Transformations

I

SS Phonology PR

Syntax of LF LF

Transformational rules map DS's onto SS's applying in accordance with the so-called "Trace Theory" of movement rules, which is discussed in more detail below. The "Phonological" component is in fact a cover term for various rule systems: apart from phonology proper, it includes at least deletion rules, and possibly "stylistic" rules, and other subcomponents (see Chomsky & Lasnik (1977) for a discussion of this organisation). The Syntax of Logical Form includes rules explicitly characterizing quantifier scope (May 1977), rules assigning antecedents to anaphoric elements, and possibly other processes, including "reconstruction" rules.4 Particular grammars are rule systems which conform to this general schema. The range of variation of particular grammars is defined by a higher level system, Universal Grammar (UG), which specifies a restrictive formal characterisation of the rule types available for particular grammars and their possible modes of interaction, thus providing an explicit definition of the notion "possible (grammar of a) natural language". UG also specifies a set of general constraints on rule application and general well-formedness conditions on the different structural representations characterized by particular grammars. Many of the proposed principles of UG on the form and functioning of rules and grammars and on the well-formedness of grammatical representations are discussed in detail in the following essays, and it is not necessary to introduce them here. The only general property of the theory which seems to me to deserve some preliminary consideration is the trace theory of movement rules, a cornerstone of the recent framework. 5 The basic idea of trace theory is that when a phrase is moved, the node which dominated it does not disappear but is left in place and coindexed with the moved phrase: it is the "trace" of the moved phrase. Consider for instance the following S Structures.6 (2)

(a) Abbiamo sentito [ s [ N P Mario] parlare di sé] "We heard Mario speak of himself' (b) Abbiamo sentito [ s [ N P lui] parlare di sé] "We heard him speak of himself' (c) Lo¡ abbiamo sentito [ s [ Ν Ρ ( · e] parlare de sé] "We him ; heard e¡ speak of himself'

χ

Introduction

The subject of the embedded infinitival clause can be realized as a lexical NP, as in (2)a, or as a pronoun; in the latter case, the pronoun can be left in place, as in (2)b (which corresponds to an acceptable sentence if/«/is contrastively stressed), or moved into the clitic position of the main clause, as in (2)c: the vacated NP position in this structure is the trace of the moved element. Traces do not receive phonetic realisation (the embedded subject NP is not "pronounced" in (2)c), but enter into a variety of grammatical processes at different levels, on a par with phonetically realized elements: for instance, the trace in (2)c is the clause-internal antecedent of the reflexive pronoun sé, on a par with the phonetically realized NP's in (2)a and b. Hence traces are "visible" in the LF component, where antecedentanaphor relations are established (or checked). Moreover, even if traces are not phonetically realized, they may also have an active role in the phonological component of (1), in that certain processes belonging to this component can be blocked by an intervening trace.7 Consider for instance the rule eliding the last vowel of the Italian indefinite quantifier una (one+ FEM) in contiguity with a vowel: (3)

, , una macchina ,omprero , . ... (a) Comprerò r I un automobile Ί will+buy one car"

The head N' of the object NP (macchina, automobile) can be pronominalized by the clitic ne, which is extracted and moved to the verb: (4)

Ne, comprerò [ N P una [ N , e]] "I of+it; will+buy one e "

Consider now a case in which the N' is followed by a postnominal modifier, e.g., an adjective, whose initial segment is a vowel: (5)

Comprerò [ N P un' [ N , automobile] azzurra] "I will+buy one car blue"

If ne is extracted, the elision rule cannot apply, in spite of the fact that una is apparently contiguous to a vowel: (6)

azzurra

Why is it so? Trace theory provides an immediate answer: the structure corresponding to (6) is (7)

Ne,· comprerò [ N P una [Ν-e] azzurra]

In (7) the contiguity between una and azzurra is interrupted by the inter-

Introduction

xi

vening trace, hence the context required by the elision rule is not fulfilled. In general, it can be said that, even if traces do not have phonetic content, they are not simply "invisible" for the rules of the phonological component, in that a trace may have the active role of blocking a process otherwise expected to apply. The presence of a trace can thus be detected also in this component of the grammar. Trace theory implies the consequence that S Structures are more abstract syntactic representations than the output of the transformational component in other models not involving traces (e.g., surface structures in the standard sense) in that they may contain positions which happen not to have a phonetic content. The adopted framework admits another type of phonetically null position, i.e., the subject position of so-called control structures: (8)

(a) Gianni ha promesso a Mario di NP partire " Gianni promised Mario NP to leave" (b) Gianni ha permesso a Mario di NP partire "Gianni permitted Mario NP to leave"

The phonetically null subject of the infinitival clause is interpreted in these structures as a pronominal element obligatorily bound by an antecedent (the "controller", underscored in (8)) in the next higher clause. This phonetically null NP position is generally referred to as PRO. PRO and trace differ at least in their derivational histories: the latter is produced in the course of the transformational derivation by the application of a movement rule, while the former is the consequence of a decision already taken at the level of lexical insertion. The "antecedent" of a trace, the moved phrase, is automatically provided by the transformational derivation; in contrast, the transformational derivation does not automatically provide the antecedent of PRO, which must be determined by an independent subtheory (the theory of control) on the basis of structural properties as well as lexical requirements of the matrix verb. Whether or not PRO and trace also differ in intrinsic constitution is a much-debated question which need not concern us for the time being (see chapter IV for a scketchy presentation of the issue, and Chomsky (1981a) for a detailed discussion). This fundamental theoretical background is shared by the studies which follow, but the four chapters, written at different times, differ slightly in certain technical details, reflecting in part the evolution of the framework over the past five years. Instead of undertaking the purely mechanical task of readjusting the text to achieve full homogeneity, we have preferred to keep the original formulations unchanged, indicating in the footnotes, in the very few cases of significant discrepancies, how things could be reformulated in the most recent version of the theory. The first chapter was written in 1976/77 during a research sojourn at the Université de Paris Vlll-Vincennes, and was published as an indepen-

xii

Introduction

dent essay in Keyser (1978). Here it is reproduced with minor modifications. It is shown that a number of verbs taking infinitival complements have apparently exceptional properties with respect to several syntactic processes. The main peculiarities are that these verbs allow a clitic pronoun originating in the infinitival complement to "climb" into the main clause; and that, in certain cases, these verbs can "inherit" the aspectual auxiliary selected by the embedded verb. These and various other peculiarities are traced to a unitary explanation by assuming that this class of verbs is characterized by a single abstract property: they optionally trigger a restructuring process which reanalyzes the main verb and the embedded infinitival verb as a single verbal complex, thus transforming a biclausal structure into a simple clause with a complex "verb". All the apparently exceptional behaviors are a direct consequence of the application of this abstract rule, given very natural supplementary assumptions which are discussed in detail.8 The second chapter was written in the fall of 1977 during a stay at MIT, and its background and content are strongly influenced by the theoretical debate which was dominant at the Department of Linguistics and Philosophy at that time.9 The basic empirical observation is that Italian allows a highly selective class of violations of the Wh Island Constraint of Ross (1967): a relative pronoun can be extracted from an indirect question, provided that the source position is not too "remote" from the final position. This state of affairs is shown to support Chomsky's (1973) Subjacency Condition, which ensures exactly the kind of locality requirement needed, given the appropriate choice of "bounding nodes"; moreover, the analysis developed involves the non-trivial theoretical consequence that wA-traces are to be treated differently from NP- and clitic traces w.r.t. the theory of binding (the Opacity Principle of Chomsky (1980a)) in that the latter, but not the former, count as "anaphors" in the relevant sense.10 Chapters III and IV have been elaborated within the framework first outlined by Chomsky in the Pisa workshop of April 1979, and then fully developed in the Lectures on Government and Binding (cf. Chomsky (1981a)). They then reflect a more homogeneous and structured theoretical background, developing in part some of the issues discussed in the first two chapters. The third chapter deals with the peculiar properties of a class of nonfinite clauses in Italian which allow lexical subjects in post-auxiliary position. Such apparently exceptional properties are shown to be explicable via a very simple assumption on the internal structure of this class of clauses, in interaction with several subcomponents of the GovernmentBinding framework: in particular, the theory of abstract syntactic Case, which is suitably extended, the theory of movement rules with its fundamental locality principle, Subjacency, and the theory of binding, which is minimally modified in order to capture the particular binding requirements operative in the construction at issue. The fourth chapter addresses the theoretical problem of syntactic

Introduction

xiii

variation across languages, in connection with certain structural peculiarities which systematically differentiate English and Italian: Italian, unlike English, allows phonetically null subject positions in tensed clauses, as a consequence of either a choice at the level of lexical insertion, or of the application of a movement rule removing the content of the subject position. More generally, the "Null Subject" property appears to systematically correlate across languages to a number of other characteristic properties, thus defining a coherent pattern of cross-linguistic variation, which in the terms of the adopted framework is assigned the theoretical status of a parameter of Universal Grammar. The main goal of this chapter is to reinterpret recent accounts of the "Null Subject" Parameter, proposing a different analysis which assigns a more central role to the process of subject inversion, freely operative in languages of the Italian type. In an appendix, the proposed approach is compared in detail with the account of the parameter developed in Chomsky's Lectures. In closing these introductory remarks, I would like to express my gratitude to the institutions which, in different forms and at different times, have supported the research the results of which constitute the content of this book: the Scuola Normale Superiore and the University of Pisa, the Université de Paris Vlll-Vincennes, the Massachusetts Institute of Technology, the Faculté Scientifique de Luminy-Marseille, and the Università della Calabria. The following studies owe much to many colleagues and friends with whom I had the opportunity of fruitfully discussing specific points and general background issues. I am particularly indebted to Giorgio Graffi, Nicolas Ruwet and Alfredo Stussi, who decisively influenced my formation as a linguist; to Henk van Riemsdijk and Jean-Roger Vergnaud for their major contributions, directly as well as through the creation of the GLOW circle, to the growth of a living theoretical debate in Europe; to Adriana Belletti and Guglielmo Cinque, who havepermitted fundamental improvements and inspired new directions of research through constructive critiques on virtually all details of my work, as well as through their own contributions on Italian syntax; and to Noam Chomsky and Richard Kayne, with whom I have contracted the greatest intellectual debt: apart from the many hours spent with each of them in discussing syntactic theory and Italian syntax, I owe them the very theoretical ground and raison d'être of my work in linguistics. NOTES 1. See, in particular, Chomsky (1977b), (1981a)and, for a more general and technically accessible discussion, Chomsky (1975), (1980c). 2. S Structures differ from the surface structures of previous models (e.g., the Standard Theory of Chomsky (1965)) in that, while being closer to the phonetic representations than D Structures are, they still are relatively abstract syntactic representations, containing more information than the simple organization in phrases of "pronounced" elements. See below for a brief discussion of this point.

xiv

Introduction

3. I.e., the formal theoiy of syntactic categories, their intrinsic content and internal structure. See Chomsky (1972"), ch. I, Jackendoff (1977), van Riemsdijk (1978). 4. I.e., rules "readjusting" S Structures in various ways, in order to permit the construction of the appropriate logical forms. See also Williams and van Riemsdijk (1981) for an organization of the components of the grammar different from ( 1 ), which renders reconstruction rules unnecessary, at least for the fundamental cases discussed in the literature. A still different proposal concerning the organization of the grammar, but elaborated within the same methodological guidelines of the more usual version of the theory, is argued for in Köster (1978). 5. On trace theory see, in particular, Chomsky (1977b), Fiengo (1974), (1977). 6. According to the usual convention, " e " is the identity element w.r.t. concatenation, hence " [ N P e]" designates an empty NP position. The notation " e " will be freely used, from now on, as an abbreviation for the substructure " [ N P . e]". In the following examples I will assume without discussion that perception verbs in Romance languages take infinitival complements with lexical subjects, as is proposed in Rouveret & Vergnaud (1980). See Kayne (1975) for a different view. 7. The following sketchy discussion reproduces the lines of much-debated arguments on the role of traces in phonosyntactic processes. See, in particular, Selkirk (1972), Chomsky & Lasnik (1977). More accurate and detailed accounts of various processes of truncation, elision and cliticization of Italian NP specifiers in connection with trace theory can be found in Rizzi (1980), Vanelli (1980). 8. The 1978 article and an earlier version of the restructuring hypothesis have occasionally given rise to discussions, attested in part in the literature. References mainly focused on this issue are the following: Zubizarreta (1980a) proposes to reinterpret our restructuring rule as a process which does not affect the constituent structure, but only the "thematic" structure, in the sense of Rouveret & Vergnaud (1980); Fresina (1980) argues for direct base generation of the "restructured" structures. Radford (1977), Burzio (1979) argue that restructuring is to be identified with the rule involved in the derivation of the causative construction. Of special importance is the discussion of Burzio (1981), in particular for the detailed investigation of the process of auxiliary change. Longobardi (forthcoming) discusses a processing strategy which applies to "restructured" structures. Finally, Kayne (1980b) suggests an interesting diachronic argument in support of the restructuring hypothesis (see ch. IV, appendix III, 5.2 for brief discussion). 9. The major issues were discussed and developed a few months later in Chomsky's article On Binding (cf. Chomsky (1980a). 10. Discussions of the hypothesis presented in this chapter and of possible alternative analyses can be found in Reinhart (1981), George (1980). Sportiche (1979) proposes an extension of our approach to various extraction processes in French (on which see also Godard (1980)), and Taraldsen (1981) develops an analysis along similar lines of extractions from wh islands and complex NP's in Norwegian. The idea that wh traces are not "anaphors" receives further support in Freidin and Lasnik (1981) and is incorporated in the general restructuring of the theory of binding elaborated in Chomsky's Lectures.

I. A restructuring rule*

There are many syntactic processes in Italian syntax which create a bifurcation in the class of verbs taking infinitival complements. Consider the following examples: (A) With some main verbs, an unstressed pronoun originating in the infinitival complement can be cliticized either to the main or to the embedded verb; with other main verbs, only the second cliticization is allowed: (1)

(a) Piero verrà a parlarti di parapsicologia. "Piero will come to speak to you about parapsychology." (b) Piero ti verrà a parlare di parapsicologia. (c) Piero deciderà di parlarti di parapsicologia. "Piero will decide to speak to you about parapsychology." (d) * Piero ti deciderà di parlare di parapsicologia.

(B) In "impersonai si' sentences, with some main verbs the direct object of the embedded clause can become the main subject; with other main verbs, this promotion is impossible: (2)

(a) Finalmente si comincerà a costruire le nuove case popolari. "Finally si will begin to build the new council houses." (b) Finalmente le nuove case popolari si cominceranno a costruire. (c) Finalmente si otterrà di costruire le nuove case popolari. "Finally si will get permission to build thenewcouncilhouses." (d) * Finalmente le nuove case popolari si otterranno di costruire.

(C) In Italian, verbs can take avere "to have" or essere "to be" as aspectual auxiliary: (3)

(a) Mario

Í ha I „ < voluto un costoso regalo di Natale.

"Mario has wanted an expensive Christmas present." * Reprinted by permission of MIT Press, Cambridge, MA from S.J. Keyser (ed.), Recent Transformational Studies in European Languages; Rizzi/A Restructuring Rule in Italian Syntax. © 1978 The Massachusetts Institute of Technology.

2

Issues in italian syntax

1 6

*ha t o r n a t o a c a s a · "Mario "is" come back home." However, some main verbs generally taking avere can optionally take essere when the embedded verb requires essere; other main verbs do not allow this process: (4)

(a) Mario ha voluto tornare a casa. (b) Mario è voluto tornare a casa. "Mario has/"is" wanted to come back home." (c) Mario ha promesso di tornare a casa. (d) * Mario è promesso di tornare a casa. "Mario has promised to come back home."

While at first sight these phenomena seem to be unrelated, a more careful consideration strongly suggests that an interesting generalization is to be captured: the classes of predicates which allow the exceptional behaviors exemplified in (lb), (2b), and (4b) are (with some qualification to be specified later) identical. In this chapter I want to show that the phenomena just described, and other related facts, can receive a unitary account. I will argue for the existence of a restructuring rule in Italian syntax, that is, a rule which changes the structure of a phrase marker without affecting its terminal string. This rule, governed by a restrictive but significant class of main verbs, will be shown to optionally transform an underlying bisentential structure into a simple sentence, creating a unique verbal complex consisting of the main and the embedded verb. In sections 1,2, and 3, the asymmetries observed in (1), (2), and (4), respectively, will be traced back to the operation of this single abstract rule; certain coherent interactions of these and other phenomena, predicted by the unitary treatment, will be shown to be inexplicable in any analysis describing the same facts as unrelated. Sections 4 and 5 will be devoted to the interaction of the restructuring rule with, respectively, the rule preposing the aprepositional dative loro, and Tough Movement. In section 6,1 will discuss some alternative account for the same class of facts, trying to show that the hypothesis proposed in this chapter is superior. Finally, in section 7, I will further investigate some properties of the restructuring rule.

1. Clitic Placement 1.1. The problem Before trying to give an account of the phenomenon exemplified in (1), I will briefly sketch the classical analysis proposed by Kayne (1969; 1975) of

A restructuring rule

3

the syntax of French clitic pronouns (which I will extend to Italian). According to this analysis, unstressed pronominal complements are basically introduced in postverbal position, as well as lexically specified complements, and are "cliticized" by a movement transformation, Clitic Placement (CP), which left-Chomsky-adjoins them to the verb:1 Clitic Placement ( CP) vbl - V - vbl - PRO - vbl 1 2 3 4 5 => 1 4 + 2 3 φ 5 For instance, sentence (5a) would be derived via CP from a structure similar to (5b), exactly parallel to sentence (5c): (5)

(a) Gianni gli presenterà Maria. (b) Gianni presenterà Maria a lui. (c) Gianni presenterà Maria a Francesco. "Gianni will introduce Maria to him/Francesco."

I will not try to justify this analysis, nor will I examine it in detail. The only thing I want to emphasize here is that, in general, CP seems to be subject to the condition that terms 2 and 4 of the structural description must be clause mates: (6)

(7)

(8)

(9)

(a) Credo che Gianni la presenterà a Francesco. "I believe that Gianni will introduce her to Francesco." (b) *La credo che Gianni presenterà a Francesco. (a) Sentivo Mario parlarle di parapsicologia. "I heard Mario speak to her about parapsychology." (b) * Le sentivo Mario parlare di parapsicologia. (a) Piero affermava di conoscerla molto bene. "Piero stated he knew her very well." (b) * Piero la affermava di conoscere molto bene. (a) Angela pareva averlo riaccompagnato a casa. "Angela seemed to have taken him home." (b) * ? Angela lo pareva avere riaccompagnato a casa.

In the following discussion I will adopt the theory developed in Chomsky (1973; 1975; 1977b). In this framework, conditions on specific rules such as the clause-mate condition just mentioned are entirely dispensed with, and overgeneration by grammatical transformations is avoided by general conditions on rule application, and well-formedness conditions on representations. Within this theory, derivation of (6b) is blocked by both the Tensed S Condition (TSC) and the Specified Subject Condition (SSC).2 Derivation of (7b), (8b), and (9b) is blocked by the SSC alone; for example,

4

Issues in italian syntax

consider the following plausible input structures to CP: (10)

(a) (io) sentivo [ s Mario parlare a lei di parapsicologia] (b) Piero, affermava [ s di P R O , conoscere la molto bene] (c) Angela,· pareva [ s e¡ avere riaccompagnato lo a casa]

Extraction of the clitic pronoun from the embedded clause is forbidden by the lexically specified subject Mario in (10a), 3 by the phonetically null P R O controlled by the main subject in (10b), and by the phonetically null trace bound by the surface main subject in (10c). 4 These paradigms interestingly support the already mentioned general conditions on rules, which have been introduced and first motivated on completely independent grounds. But the problem that we are going to face now seems to question their generality: with three classes of main verbs (some of which will be analyzed as Subject Raising verbs, and the others as Control verbs; see section 6.2 for relevant discussion), syntactically and semantically rather homogeneous, a clitic pronoun originating in the embedded sentence can b e extracted from it and cliticized to the main verb. The following sentences are all equally acceptable: Modals5 (11)

(a) Mario I v u o ^ e risolverlo da solo (questo problema). ι sa j " M a r i o wants to/can solve it by himself (this problem)." (b) Mario lo ( v u o ' e risolvere da solo. v ' I sa )

(12)

(a) Gianni J

j parlargli personalmente.

"Gianni has had/has been able to speak with him personally." (b) Gianni gli J

J parlare personalmente.

Aspectuals (13)

(a) Mario j

j batterla a macchina domani (la tesi).

a

" M a r i o will start/finish typing it tomorrow (his thesis)." (b) Mario la (14)

comincia a battere a macchina domani. I finisce di J

(a) Gianni J

a

j raccontargli stupide storie.

"Gianni is continuing/going to tell him stupid stories." (b) Gianni gli w "

°ntinua sta per

c

a

raccontare stupide storie. r

A restructuring rule

5

Motion Verbs venne ]

I I

andò J a chiamarli alla stazione, tornò j "Piero came/went/came back to call them at the station." venne 1

andò ! a chiamare alla stazione, tornò ) While the behavior of clitic pronouns in (6)-(9) is straightforwardly predicted by Kayne's formulation of CP plus general conditions on transformational rules, paradigms such as (11)-(15) require some further hypotheses. 1.2. Restructuring The hypothesis which will be argued for is the following: there exists a restructuring rule in Italian syntax, governed by modals, aspectuals, and motion verbs (with the variations mentioned in footnote 6), which optionally reanalyzes a terminal substring 6 V x (Ρ) V as a single verbal complex, hence automatically transforming the underlying bisentential structure into a simple sentence. For the time being, I will not try to give the rule a formal characterization, but only an intuitive one. Consider an intermediate structure such as (16), where the embedded subject NP is trace or PRO; 7 Restructuring can reanalyze the terminal substring included in the braces as a single verbal complex (provisionally labelled V; see section 7.1 for discussion of this problem): S

(16) NP

VP V

S NP

Gianni

{ V deve

VP V

NP

PP

presentare}

la

a Francesco

If Restructuring has applied to (16), yielding the simple structure (17a), nothing can now prevent the clitic pronoun from moving to the "main

6

Issues in Italian syntax

verb" dovere (in fact, the first lexical verb of the verbal complex), and further application of CP will yield (17b):8 (17)

(a) Gianni [ v deve presentare] la a Francesco, (b) Gianni la deve presentare a Francesco.

If Restructuring has not applied to (16), the structure remains bisentential (cf. (18a)), and the only sentence that can be derived, via CP, is (18b), the "long step" of the clitic being forbidden by the SSC. (18)

(a) Gianni deve [ s presentare la a Francesco] (b) Gianni deve presentarla a Francesco.

With the main verbs of (6)-(9), which do not allow Restructuring, the "long step" will be blocked in any case by the SSC. In arguing for the Restructuring hypothesis, I will follow two distinct strategies: (A) The Restructuring hypothesis states that sentence pairs such as (17b)-(18b) not only differ in the relative ordering of the formative la with respect to other terminal elements, but also differ radically in structure, along the lines indicated by (17a), (18a). It should be possible to verify whether or not this prediction is correct by applying appropriate constituency tests. In principle, if the hypothesis is correct, we should be able to show (a) that presentarla a Francesco is a constituent in (18b), but that presentare a Francesco is not a constituent in (17b) ; (b) that la deve presentare is a constituent in (17b), but that deve presentarla is not a constituent in (18b). Leaving aside point (b) (which will be discussed in section 7.1 ) for the time being, in the remaining paragraphs of this section I will investigate point (a). (B) Broadening the analysis somewhat, in the following sections I will show that the Restructuring hypothesis permits an immediate account for a set of apparently exceptional behaviors of the above-mentioned class of main verbs. 1.3. Wh movement In Italian, the phenomenon generally referred to as "pied-piping" (see Ross (1967)) can involve, along with many other constituent types, even a whole infinitival complement: from the underlying structure (19a), both (19b) and (19c) can be derived via Wh Movement: (19)

(a) questi argomenti [ s (io) verrò [ s a discutere dei quali] al più presto] mi sembrano molto interessanti. "These topics [ s I will come [ s to discuss (on) which] as soon as possible] seem to me very interesting."

A restructuring rule

7

(b) Questi argomenti, dei quali verrò a discutere al più presto, mi sembrano molto interessanti. (c) Questi argomenti, a discutere dei quali verrò al più presto, mi sembrano molto interessanti. Now, an unstressed pronoun originating in the embedded clause cannot cliticize to the main verb when the whole embedded clause is w/i-preposed: (20)

(21)

(22)

(a) Questi argomenti, a parlarti dei quali verrò al più presto,... (b) * Questi argomenti, a parlare dei quali ti verrò al più presto,... "These topics, to talk with you about which I will come as soon as possible, . . . " (a) Francesco, a parlarne col quale comincerò solo la settimana prossima, . . . (b) * Francesco, a parlare col quale ne comincerò solo la settimana prossima, . . . "Francesco, to talk about it with whom I will begin only next week, . . . " (a) Il Direttore, presentarti al quale, per il momento, proprio non posso, è comunque una gran brava persona, (b) * Il Direttore, presentare al quale, per il momento, proprio non ti posso, è comunque una gran brava persona. "The Director, to introduce you to whom, for the time being, I really cannot, is anyhow a very nice person."

Of course, when only the PP immediately containing the wA-word is preposed, sentences corresponding to (20b), (21b), (22b) are acceptable: (23)

(a) Questi argomenti, dei quali ti verrò a parlare al più presto,... (b) Francesco, col quale ne comincerò a parlare solo la settimana prossima,... (c) Il Direttore, al quale, per il momento, proprio non ti posso presentare,...

Within my hypothesis, the explanation of this fact is straightforward. Consider for instance the following input structures to Wh Movement: (24)

(a) Questi argomenti [ s verrò a parlarti dei quali] al più presto... (b) Questi argomenti [ s ti verrò a parlare dei quali] al più presto...

In (24a) the string a parlarti dei quali is a constituent, and as such can be whpreposed ; but in (24b), Restructuring having applied, the string a parlare dei quali is not a constituent, and the ungrammatical (20b) is automatically ruled out. An obvious objection to this argument is the following: it is not necessary to postulate a structural difference between (24a, b) to rule out (20b), etc.,

8

Issues in italian syntax

given that the simple extrinsic ordering Wh Movement-CP would suffice: application of Wh Movement to (25a) would yield (25b): (25)

(a) Questi argomenti [ s verrò [ s a parlare a te dei quali] . . . ] (b) Questi argomenti [ s [ s a parlare a te dei quali] verrò . . . ]

Further application of CP could only yield (20a); the unacceptable (20b) would be automatically excluded without the need for postulating a difference in constituent structure. But it can be shown that this ordering solution is not available: in fact, there are good reasons to believe that Wh Movement is ordered after CP. The following argument will be exactly parallel to the one given by Kayne (1975, section 4.3) for French. The clitic ne pronominalizes (among other things) partitives. When the partitive is in object position, appearance of ne is obligatory: (26)

(a) Ho visto molti dei corridori. "I have seen many of the racers." (b) (Dei corridori,) ne ho visti molti. (c) * (Dei corridori,) ho visto molti.

This result follows from the obligatory9 application of CP to the underlying structure ho visto [NP molti - ne]. But when the partitive is in subject position (i.e. basically preverbal), ne cannot appear: (27)

(a) Molti dei corridori hanno forato. "Many of the racers have got flat tires." (b) *(Dei corridori,) molti ne hanno forato. (c) (Dei corridori,) molti hanno forato.

Following Kayne's analysis, we can account for this paradigm by the fact that CP can move pronouns from right to left only, and by a late rule deleting occurrences of ne not previously cliticized. Let us now turn to the paradigm relevant for the relative ordering of CP and Wh Movement. In sentences where the partitive NP is w/j-moved, the possibilities of the appearance of ne remain exactly as in (26) and (27): (28)

(29)

(a) (Dei corridori,) quanti ne hai visti? (b) *(Dei corridori,) quanti hai visto? "Of the racers, how many have you seen?" (a) * (Dei corridori,) quanti ne hanno forato? (b) (Dei corridori,) quanti hanno forato? "Of the racers, how many have got flat tires?"

This behavior shows that CP is ordered before Wh Movement: if it were ordered after (or unordered with respect to) Wh Movement, it would be

A restructuring rule

9

impossible to account for the symmetry of (26), (28) and (27), (29): the structures underlying (28), (29) would (or could) not be distinguished at the relevant level of application of CP.10 We can conclude that CP is ordered before Wh Movement," so that (20b) and similar examples cannot be ruled out via rule ordering. Therefore, the argument for Restructuring is unaffected. 12 1.4. Cleft sentence formation I will adopt here the analysis of cleft constructions proposed by Ruwet (1974) (following a suggestion made by Chomsky (1972,34) for the pseudocleft construction). According to this analysis, given the underlying structure essere Δ che S, the rule Cleft S Formation extracts a constituent from S, placing it in the focus position Δ. For instance, from the underlying structure (30a), both (30b) and (30c) can be derived, while (30d) is correctly excluded because the string sua figlia al mare is not a constituent:13 (30)

(a) Essere Δ che (io) accompagnerò sua figlia al mare. "It be Δ that I will see his/her daughter to the sea." (b) E' sua figlia che accompagnerò al mare. (c) E' al mare che accompagnerò sua figlia. (d) *E' sua figlia al mare che accompagnerò.

Infinitival complements are among the constituents which can be clefted: (31)

(a) E' proprio a riportargli i soldi che sto andando, stai tranquillo ! "It is just to bring him back his money that I am going, don't worry!" (b) Quanto a questa storia, è discuterne con Mario che dovresti. "As for this story, it is to discuss about it with Mario that you should have." (c) La tua disavventura, è proprio a raccontarla a Francesco che ho cominciato. "Your mishap, it is just to tell it to Francesco that I have begun."

But when the clitic pronoun is cliticized to the main verb, the "infinitival complement" cannot be moved: (32)

(a) * E' proprio a riportare i soldi che gli sto andando, . . . (b) * Quanto a questa storia, è discutere con Mario che ne dovresti. (c) * La tua disavventura, è proprio a raccontare a Francesco che la ho cominciata.

Once more, this difference naturally follows from our hypothesis. For

10

Issues in italian syntax

instance, we can derive (31a) because a riportargli i soldi is a constituent and, as such, may be clefted. But consider the putative input structure for Cleft S Formation to derive (32a): (33)

essere A che (io) [ v gli sto andando a riportare] i soldi

The string a riportare i soldi is not a constituent in (33), and the unacceptable (32a) is automatically ruled out. As in the case of Wh Movement, it is now necessary to eliminate a possible objection concerning rule ordering: in fact, one could object that the unacceptable sentence (32) can be trivially ruled out by ordering Cleft S Formation before CP, without any need for postulating a difference in structure. But such an extrinsic ordering is wrong. From the underlying structure (34a), only sentence (34b), not (34c), should be derivable: (34)

(a) . . . essere Δ che (tu) devi leggere [ N P solo uno - ne] (b) Di tutti questi libri, in fondo, è solo uno che ne devi leggere. (c) * Di tutti questi libri, in fondo, ne è solo uno che devi leggere. "Of all these books, after all, it is only one that you have to read."

Judgments on (34b) and (34c) are correctly predicted by the ordering CP-Cleft S Formation, while the reverse ordering would incorrectly predict (34b) to be starred and (34c) to be acceptable. It follows that unacceptable sentences (32a-c) cannot be trivially excluded via rule ordering, and that the test is valid for constituency in the relevant cases.14 1.5. Right node raising Consider next the rule generally referred to as Right Node Raising (RNR). Following the account given of it by Postal (1974, section 4.8), this rule operates on coordinate sentences whose rightmost constituents are identical, in two steps: first, right-Chomsky-adjoining a copy of the identical constituents to the whole sentential structure, and then, deleting the identical constituents. For instance: (35)

(a) io lo ho solo invitato [ s a fare il suo dovere] —ma Gianni lo ha addirittura costretto [ s a fare il suo dovere] (b) Io lo ho solo invitato φ —ma Gianni lo ha addirittura costretto 0 —a fare il suo dovere. "I have only invited him—but Gianni has even forced him— to do his duty."

Only strings that are constituents can be right-node-raised;15 for instance, in comparison with (35b), consider the following unacceptable sentence,

A restructuring rule

11

where a nonconstituent has been raised: (36)

*Io ho solo invitato—ma Gianni ha addirittura costretto— Francesco a fare il suo dovere.

Keeping this in mind, let's apply this new test to the examples relevant for our discussion: (37)

(38)

(39)

(a) Mario sinceramente vorrebbe—ma a mio parere non potrà mai—pagargli interamente il suo debito, (b) * Mario sinceramente gli vorrebbe—ma a mio parere non gli potrà mai—pagare interamente il suo debito. "Mario sincerely would like—but in my opinion he will never be able—to pay him his debt." (a) Piero dovrebbe—ma francamente non credo che vorrà—parlarne con Gianni, (b) * Piero ne dovrebbe—ma francamente non credo che ne vorrà—parlare con Gianni. "Piero would have—but frankly I don't believe he will like— to talk about it with Gianni." (a) Francesco comincerà—e probabilmente per molto tempo continuerà—ad andarci di mala voglia, (b) * Francesco ci comincerà—e probabilmente per molto tempo ci continuerà—ad andare di mala voglia. "Francesco will begin—and probably for a long time will continue—to go there against his will."

Once more the relevant pairs of sentences seem to differ in structure along the lines predicted by the Restructuring hypothesis. But, as in the preceding cases, it is now necessary to show that the ungrammatical sentences of (37)—(39) cannot be ruled out by a simple ordering condition, i.e. with RNR being ordered before CP. Such an ordering is incorrect: it must be possible to derive the acceptable sentence (40b) from the underlying structure (40a): (40)

(a) Mario ha letto [ N P la maggior parte - ne]—e sicuramente ha anche ricopiato [ N P la maggior parte - ne] (b) Mario ne ha letto—e sicuramente ne ha anche ricopiato—la maggior parte. "Mario of it has read—and certainly of it has also copied— the greater part."

It is easy to verify that (40b) can be generated only if CP can apply before RNR.

12

Issues in italian syntax

1.6. Complex NP shift A different test is provided by the interaction of this rule (on which see Ross (1967), Postal (1974)) with CP. Complex NP Shift (CNPS) shifts a complex postverbal complement (where complex means long and/or containing a sentence) in sentence-final position. For instance, sentences (41b), (42b), (43b) derive via CNPS from structures similar to (41a), (42a), (43a), respectively: (41)

(42)

(43)

(a) Fra qualche giorno, verrò ad esporti la mia idea a Firenze, (b) Fra qualche giorno, verrò a Firenze ad esporti la mia idea. "In a few days, I will come to Florence to explain my idea to you." (a) Ho cominciato a discuterne con Mario da Gianni, (b) Ho cominciato da Gianni a discuterne con Mario. "I have begun at Gianni's house to discuss (of) it with Mario." (a) Volevo parlarti di questa vecchia storia da molto tempo, (b) Volevo da molto tempo parlarti di questa vecchia storia. "I wanted for a long time to talk to you about this old story."

But, if the clitic pronoun is extracted from the embedded clause, sentences corresponding to (41b), (42b), (43b) are unacceptable:16 (44) (45) (46)

(a) (b) (a) (b) (a) (b)

Fra qualche giorno, ti verrò ad esporre la mia idea a Firenze, * Fra qualche giorno, ti verrò a Firenze ad esporre la mia idea. Ne ho cominciato a discutere con Mario da Gianni, *Ne ho cominciato da Gianni a discutere con Mario. Ti volevo parlare di questa vecchia storia da molto tempo, ?Ti volevo da molto tempo parlare di questa vecchia storia.

The Restructuring analysis immediately accounts for this fact. Consider the structure underlying (41a,b) and (44a): (47)

. . . verrò [ s ad esporre la mia idea a te] a Firenze

If Restructuring does not apply, both (41a,b) can be derived via CP and, optionally, CNPS (regardless of the relative ordering of these two rules). But if Restructuring applies on (47), the string ad esporre la mia idea a te, having ceased to be a constituent, cannot be shifted into sentence-final position. Hence, the unacceptable sentences of (44)-(46) are correctly excluded. Notice that in this case there is no obvious alternative way of treating the same facts via rule ordering. One possibility would be to propose that shifted constituents become inaccessible to rules: if this were true, the ungrammatically of (44b), etc., would simply follow from the ordering CNPS-CP. But this alternative is untenable: as the following example shows, CP can extract a pronoun from a shifted complement (or, alterna-

A restructuring rule

13

tively, can apply before CNPS): (48)

(Della tua lettera,) ne volevo mostrare a Francesco soltanto la parte meno interessante. "(Of your letter,) I (of it) wanted to show to Francesco only the less interesting part."

Moreover, in examples exactly parallel to (44b), (45b), (46b), extraction of an element from the (previously or subsequently) shifted sentential complement by Wh Movement (a rule that we argued to follow CP) yields a perfectly acceptable result: (49)

(a) Francesco, a cui, fra qualche giorno, verrò a Firenze ad esporre la mia idea, . . . (b) Questa idea, di cui ho cominciato da Gianni a discutere con Mario, . . . (c) Gianni, a cui volevo da molto tempo parlare di questa vecchia storia, . . .

So, there seems to be no plausible alternative to the account given for paradigms (44)-(46) in terms of the Restructuring hypothesis. We can therefore take these facts as a new argument for our proposal.

2. Impersonal Si Sentences 2.1. A sketchy analysis A different kind of justification for the restructuring rule consists in showing that it allows a unitary account for some apparently very different phenomena. To follow this line of reasoning, it is now necessary to formulate some hypotheses on the syntax of the so-called impersonal si sentences.17 Consider the following paradigm: (50)

(a) Si dorme troppo poco. "Si ( = people) sleeps too little." (b) La gente dorme troppo poco. "People sleep too little." (c) * La gente si dorme troppo poco.

(50a,b) are roughly synonymous. In an intuitive sense, si is the subject of (50a), just as la gente is the subject of (50b). This intuition is confirmed by the fact that they cannot cooccur, as (50c) shows. But in many respects si behaves rather differently from a lexically specified subject: the latter pre-

14

Issues in italian syntax

cedes the negative element non and any kind of clitic pronouns, and it can be conjoined with another NP or separated from the verb by an adverb: (51)

(a) La gente non vi dorme volentieri. "People don't sleep willingly there." (b) In questo villaggio, il paesaggio e la gente sono egualmente piacevoli. "In this village, the landscape and the people are equally pleasant." (c) La gente raramente dorme volentieri. "People rarely sleep willingly."

The impersonal si obligatorily follows the negative element and at least some clitic pronouns, cannot be conjoined with a full NP, and cannot be separated from the verb: (52)

(a) Non vi si dorme volentieri. (b) *In questo villaggio, il paesaggio e si sono egualmente piacevoli. (c) *Si raramente dorme volentieri.

In short, it seems obvious that the impersonal si shares all the relevant properties of the clitic pronouns; assuming that, as Kayne(1975,section2.5) shows, the sequence clitic + verb is dominated by the category Vin surface structure, we can conclude that (a good approximation to) the surface structure of (50a) is the following: (53)

[ N P φ ] [ vp [ y si dorme] troppo poco]

There are at least two possible hypotheses concerning the derivation of structure (53): (A) The subject NP position, which does not dominate any lexical material in the surface, is basically empty, and the sñs basically introduced in clitic position. (B) The subject NP position is (basically, or at some successive stages of the derivation) filled by the impersonal si, which is subsequently cliticized to the verb by a movement transformation. In arguing for this second hypothesis, we will now consider the following sentences: (54)

(a) Non si è trattati con cordialità da quell'individuo. "Si is not treated nicely by that guy." (b) All'estero, non si è difficili da riconoscere. "Abroad si is not difficult to recognize." (c) In questo ministero, non si risulta mai essere completamente in regola. "In this ministry, si never turns out to be completely in order."

15

A restructuring rule

Under usual assumptions about passive, Tough Movement (or Deletion), and Subject Raising sentences, it is easy to show that none of (54a,b,c) could be derived within the base hypothesis. On the contrary, they can be perfectly well derived within the transformational hypothesis, on a par with the corresponding sentences with lexical subjects: (55)

(a) Gli ospiti non sono trattati con cordialità da quell'individuo. (b) All'estero, un italiano non è difficile da riconoscere. (c) In questo ministero, gli impiegati non risultano mai essere completamente in regola.

Consider, for instance, the following, rough derivation of (54a) (we will call Si Placement the rule extracting si from subject position, and leftChomsky-adjoining it to the verb; nontrivial details concerning verb agreement are omitted):18 Passive

(56)

(a) Quell'individuo non trattare si con cordialità Si Placement ^

(b) Si non essere trattati con cordialità da quell'individuo (c) é non si è trattati con cordialità da quell'individuo We can conclude that, to give a unitary account of (54) and (55), the impersonal si must be transformationally moved from subject position to clitic position, and the base solution is therefore inadequate. 2.2. Object preposing In all the examples given so far, the output structure of Si Placement is intransitive, i.e. without a direct object NP. We will now turn to consider the case of an intermediate structure (subsequent to the application of Si Placement): L[s[VpP···]]]

26. It is possible to keep Kayne's results if the reanalysis process is interpreted as suggested in the text: i.e., reanalysis does not destroy the original phrase marker, but simply superimposes another phrase marker to the same terminal string (hence, a different set of proper analyses). For instance, consider again (58)a. Reanalysis (which we may assume applies at any level of the derivation, whenever possible) superimposes on (i) the structure indicated in (ii): (i)

Wh {... [ s , [comp & e¡] [s e ¡ tvP · · · ] ] ]

(ii)

Wit γ .

S,[ S [

ypifl ...]]]

Now, both (i) and (ii) are ill-formed: (i) is ruled out by ECP, as suggested in the text* (ii) is ruled out since the wh operator does not have any variable to bind. In general, we assume that a reanalized structure gives rise to a well-formed output if at least one of the two simultaneous structures can be assigned a well-formed derivation, which is not the case for (58)a. 27. The Nominative Island Constraint is formulated as follows: (i)

NIC: "A nominative anaphor cannot be free in S'."

Whether the exceptional behavior of wh movement referred to in the text is a special case of the general immunity of wh traces from the binding principles for anaphors (cf. chapter II), or the apparent "immunity" is simply a consequence of the fact that in the structure (ii)

· ·-[s- e i C s s ei • - · ] ]

the nominative trace is bound in S' by the trace in COMP, is not relevant for the present discussion. In fact the issue simply disappears within the GB framework. 28. As for the trace in COMP, it is still a debated question whether traces in COMP are in the domain of ECP. If so, notice that our COMP accessibility hypothesis would suffice to make it properly governed by the higher verb. 29. The Tensed S Condition, originally proposed in Chomsky (1973) can be phrased in the following way: (i)

TSC: "An anaphor in the domain of TENSE cannot be free in S \ "

This condition has (essentially) the effect of capturing a number of "island" properties of tensed clauses. In order to eliminate the area of overlapping with the Specified Subject Condition (SSC), Chomsky (1980a) has proposed to restrict the domain of TSC to the subject position of tensed clauses (the VP internal positions of tensed and untensed clauses being already in the domain of SSC). This result was obtained by making reference to the notion "nominative" with the NIC. 30. We can provisionally adopt the following definitions of Governing Category (GC): (i)

GC: " a is the G C for β iff a is the minimal category containing β and a governor of β where a = NP or S."

116

Issues in italian syntax

31. But this assumption is not compelling if one adopts the alternative solution of fn. 22 of the reference quoted. 32. See also chapter IV, appendix III for a slightly different analysis of these facts, which does not presuppose that nominative Case assignment be different in nature from other Case assignment processes.

IV. Negation, FF/z-movement and the null subject parameter

0. In recent years the deepening of theoretical models for syntactic research has fruitfully coincided with a renewed interest in comparative syntax; in this perspective, the identification and analysis of systematic patterns of variation has become one of the fundamental goals of linguistic theory. Universal Grammar has been conceived of as a parametrized system, in which a specific core grammar can be derived by fixing a finite number of parameters: different values assigned to the parameters define different grammars and, in the optimal cases, complex patterns of variation are reduced to minimal differences in the parametric choices (see Chomsky 1981a for detailed discussion). The present study is devoted to the analysis of one of these cases. It is well known that some of the attested natural languages allow phonetically null subjects in tensed clauses, while others do not ; the two types are instantiated by Italian and English in (1 ). It has been shown in recent work that other properties systematically correlate with the null subject property: first of all, null subject languages (henceforth NSL's) generally have a free process of subject inversion, while non-NSL's generally do not (cf. (2)):1 secondly, non-NSL's generally show COMP-trace effects, while NSL's generally do not (cf. (3)):2 (1) (2) (3)

(a) (b) (a) (b) (a) (b)

everrà *e will come e verrà Gianni *e will come Gianni Chi¡ credi che e¡ verrà? *Who j do you think that ex will come?

In the perspective of the research program mentioned above, this systematic pattern of variation will be provisionally called the "Null Subject Parameter". In the last few years several attempts have been made to give (at least a partial) theoretical account of this parameter, first by means of specific stipulations, then in terms of general principles (Perlmutter 1971 ; Bresnan 1972; Chomsky & Lasnik 1977; Taraldsen 1978; Kayne 1980a; Pesetsky 1979). In this article I would like to propose a partial reinterpretation of the parameter based on Italian data. The two main points of the reinterpretation are the following:

118

Issues in italian syntax

(A) in spite of prima facie evidence, Italian has a COMP-t effect similar to the one found in non-NSL's, which does not show up overtly in cases of whextraction of the subject, but which is manifested in structures of a different type; (B) the contrast in (3) can be essentially reduced to the contrast in (2); that is to say, Italian can avoid the COMP-t effect by w/z-extracting the subject from postverbal position. In order to achieve these descriptive results, and to construct a system which incorporates them, it will be necessary to briefly review some recent proposals concerning the parameter and related phenomena (section 1), and then to discuss some aspects of the syntax and interpretation of negation in Italian (section 2). Section 3 will address the problem of characterizing the exact role of the verbal inflection in "null subject" sentences, and in sentences having undergone subject inversion. The last section will be devoted to showing how the proposed system can account for the apparent lack of COMP-t effects in NSL's. Appendixes I and II concern further consequences and possible improvements of the proposed analysis; appendix III discusses in detail some advantages and disadvantages of the proposed system in comparison with an alternative analysis developed in slightly different forms in Jaeggli (1980), Chomsky (1981a). 1. Some Recent Analyses 1.1 In the system of Chomsky (1980a) the general principle which is a natural candidate for a theoretical account of the Null Subject Parameter is the Nominative Island Constraint (NIC), and this possibility is developed in slightly different ways in Taraldsen (1978), Kayne (1980a), Pesetsky (1979). Chomsky (1981 a) has convincingly shown that the NIC, conceived of as a binding principle for anaphors, was a spurious generalisation, and that its effects should be factored out differently. In the system of Chomsky (1981a), the general principle which subsumes the empirical coverage of NIC as far as the "null subject" effects are concerned is the Empty Category Principle (ECP), a principle which constrains the possible occurrence of empty categories: (3)

ECP: "e must be properly governed." 3

The common core of the analyses proposed by Taraldsen, Kayne and Pesetsky can be easily rephrased within a theory which contains ECP instead of NIC. Such a transposition is given in Chomsky (1981a), and will be sketched out in what follows. Given very reasonable supplementary assumptions, the "English side" of the parameter follows directly from ECP. Structure (l)b is generated if the subject NP is not further rewritten by the base rules, which can be assumed to be optional. But the subject position of a tensed sentence in English is not a properly governed position (no lexical or coindexed category

Negation, Wh-movement and the null subject parameter

119

governs the subject), so that (l)b is ruled out by ECP. The same account holds when the subject position is transformationally vacated via rightward NP movement (structure (2)b); or via wA-movement of the subject across an overt complementizer: assuming for (3)b the following structure, derived Via successive cyclic application of wA-movement: (5)

W h o ¡ . . . [ s , [ C 0 M P e¡ that] [

will come]]

the trace in COMP does not govern the trace in subject position since the c-command requirement is not fulfilled, due to the presence of that.4 Why does Italian allow apparent ECP violations in (1), (2), (3)? A natural answer is the one suggested by Taraldsen (1978), which I will adopt in the revised form of Chomsky (1981a): languages may vary with respect to the governing properties of the verbal inflection. In languages like Italian the verbal inflection properly governs the subject NP, so that in the a sentences of (l)-(3) there is no ECP violation:5 (6)

[ s , . . . e¡... INFLj...]

In languages like English the verbal inflection is not a proper governer, so that the ECP violation in the b sentences of (1 )-(3) cannot be rescued in the same way. This hypothesis is appealing, among other ways, in that it gives precise theoretical status to the common observation that in NSL's minimal feature specifications of missing subjects are "recovered" via the verbal inflection. Whether or not there is a strict (or statistical) correlation between richness of inflectional systems and "null subject" properties can only be decided through an extensive cross-linguistic scrutiny, which is beyond the aims of the present work. Aspects of Italian syntax which bear on this question will be discussed in par. 3.6. 1.2. Kayne (1981a) makes the very important observation that the proposed account of (l)b-(3)b in terms of ECP violation carries over to structures created by rules of the syntax of Logical Form. 6 Kayne's argument goes as follows: In French, the negative quantifier-like NP personne ( = nobody), as well as other elements belonging to the same class, must be construed with the negative particle ne, in some sense to be made precise. If personne is in object position of an embedded clause, ne can be cliticized onto the embedded verb (as in (7) a) or, somewhat more marginally, onto the main verb (as in (7)b): (7)

(a) J'ai exigé qu'ils «'arrêtent personne. "I have required that they neg arrest nobody." (b) ?Je n'ai exigé qu'ils arrêtent personne. "I neg have required that they arrest nobody."

120

Issues in italian syntax

But if personne is the subject of the embedded clause, only the sentence with embedded ne is acceptable: (8)

(a) J'ai exigé que personne ne soit arrêté. "I have required that nobody neg be arrested." (b) * Je «'ai exigé que personne soit arrêté. "I neg have required that nobody be arrested."

In short, there is a curious subject-object asymmetry with respect to the possibility of construing personne with an occurrence of ne in a higher clause. Kayne notices that (7)a and (7)b clearly differ in interpretation: the quantifier-like negative element personne is assigned narrow scope in (7)a (i.e. its scope is the embedded clause only) and wide scope in (7)b: the respective (partial) logical forms, assuming May's (1977) format, 7 would be something like the following: (9)

(a) [s'[s j'ai exigé[ s , que personne^ ils arrêtent ς ] ] ] (b) [ s , personne^ j'ai exigé[ s , qui[ s ils arrêtent ej]]]]

The negative particle ne thus seems to play the role of an overt scope marker: the scope of a negative quantifier is the S which immediately contains an occurrence of «e construed with the quantifier. Given this analysis, the contrast (8)a-(8)b is automatically explained: the respective LF's would be: (10)

(a) [ s- [ s j'ai exigé [ s , que personne¡ [ s e¡ soit arrêté]]]] (b) [s, personne i [ s j'ai exigé [ s , que [ s ex soit arrêté]]]]

(10)b is ruled out by ECP, since the empty NP in embedded subject position is not properly governed. On the contrary, both(9)b, (10)a are well-formed with respect to ECP, since in both cases the empty NP is properly governed: in (9)b by the verb (a lexical category), in (10)a by the quantifier itself (a coindexed category). In conclusion, according to Kayne's analysis there is a true generalisation underlying ordinary COMP-t effects like the one shown in (3)b and such scope phenomena as the contrast just discussed. Although the relevant evidence is not uncontroversial, 8 if the proposed interpretation is correct, these facts constitute a very strong argument for a "syntactic" representation of quantifier scope, as in the theory of LF adopted in several recent works: May's (1977) Quantifier Rule is the LF analogue of w/z-movement, in that both rules move quantifier-like elements to presentential position, thus creating operator-variable structures. The well-formedness of these structures is then uniformly evaluated by such principles as ECP. Putting now together Taraldsen's account of the Null Subject Parameter and Kayne's analysis of quantifier scope, we would be led to the conclusion

Negation, Wh-movement and the null subject parameter

121

that, ceteris paribus, in a null subject language sentences corresponding to (8)b should be acceptable: in the logical form corresponding to (10)b the empty subject of the embedded sentence would be properly governed by the verbal inflection, and no ECP violation would be produced. In order to check the correctness of this prediction it is now necessary to make a survey of the basic properties of negation in Italian. 2. Aspects of Negation in Italian 2.1. Considering simple clauses first, we immediately notice that Italian has a peculiar process of "negative doubling" which complicates the transposition of Kayne's argument. A negative quantifier-like element (nessuno = nobody, niente = nothing, etc.) in postverbal position must co-occur with the ordinary marker of sentential negation (non) on the verb: (11)

(a) Mario non ha visto nessuno. "Mario neg has seen nobody." (b) Mario non ha fatto niente. "Mario neg has done nothing." (c) Mario non ha parlato con nessuno. "Mario neg has spoken with nobody."

But if the negative element is in preverbal subject position non does not appear: (12)

(a) Nessuno ha visto Mario. " Nobody has seen Mario." (b) Niente di buono potrà accadere. "Nothing good can happen."

It is clear that what counts for the appearance-disappearance of non is the position of the negative NP at S(urface) structure: if the direct or prepositional object of sentences like (11) is topicalized, non disappears: (13)

Con nessuno ho parlato ! "With nobody I have spoken!"

Reciprocally, if the subject of sentences like (12) is moved to postverbal position, non must appear: (14)

(a) Non ha telefonato nessuno. "Neg has telephoned nobody = Nobody telephoned." (b) Non potrà accadere niente di buono. " Neg will happen nothing good = nothing good will happen."

Before trying to give a formal account of this behavior, some remarks on

122

Issues in Italian

syntax

the interpretation of nessuno in these and other contexts are in order. First of all, in spite of the different number of negative elements occurring in the preceding sentences, the interpretation does not vary: i.e., the negative doubling found in (11)-(14) does not give rise to contradictory negation, and these sentences are interpreted as simple negations, on apar with (12), (13). Secondly, nessuno, niente, etc. are found not only in the negative contexts (and with the negative interpretation) considered so far, but also in interrogative contexts like the following:9 (15)

(a) Mi chiedo se nessuno abbia poi contattato Gianni. "I wonder whether anybody has eventually contacted G." (b) Mi chiedo se Gianni abbia poi contattato nessuno. "I wonder whether Gianni has eventually contacted anybody."

In such structures there is no negative interpretation, and no preverbalpostverbal asymmetry: in both cases nessuno is interpreted as an existential quantifier, as indicated by the glosses. Restricting our attention now to postverbal occurrences of nessuno {i.e., cases (11), (14), (15)b), we notice that it must be found either in a negative or in an interrogative context; if it is not, the structure is unacceptable: cf. (16)c: (16)

(a) Mi chiedo se Gianni abbia contattato nessuno. "I wonder whether G. has contacted anybody." (b) Gianni non ha contattato nessuno. "Gianni neg has contacted nobody." (c) * Gianni ha contattato nessuno. "Gianni has contacted nobody/anybody."

If we now compare (11), (14), (15)b and (16), a plausible analysis suggests itself to capture in a unitary way both uses of nessuno, niente, etc. An item belonging to this class is an existential quantifier with peculiar "polarity" requirements (but it cannot be fully assimilated to ordinary polarity items for reasons which will be briefly discussed in fn. 12). As a quantifier, it undergoes May's Quantifier Rule, applying in the syntax of LF, which moves it to a pre-clausal position, thus explicitly characterizing its scope. Its "polarity" property further requires it to end up at LF in the local context of a negative or interrogative marker (that is to say, the element which fulfills the "polarity" requirement of nessuno also acts as its "scope marker"). How the appropriate "locality" requirement is fulfilled in cases of interrogative polarity is rather straightforward: nessuno must be moved by QR to a clause initial position contiguous to (governed by) an interrogative complementizer. The case of negative polarity is more complicated, and its treatment requires a short digression. Consider such well-known contrasts as

Negation, Wh-movement and the null subject parameter (17)

123

(a) Many arrows didn't hit the target. (b) The target wasn't hit by many arrows.

According to the intuitions generally adopted in the discussions of the late 1960s, (17)b allows an interpretation in which not is construed with the quantifier many, thus producing the complex operator not + many = few (i.e., "few arrows are such that the target was hit by them"), while this interpretation is not available for (17)a (i.e., the latter cannot mean "few arrows are such that they hit the target"). In these cases, the linear ordering of the negation and the quantifier at S structure seems to be relevant. Now, an account of such intuitions within a QR approach would require some qualification: otherwise, QR itself, applying in the syntax of LF, would destroy the ordering (or c-command) information which is needed to distinguish the two cases. Suppose the appropriate device to be the following convention, whose structural description is checked, we may assume, whenever QR applies: (18)

"In the structure [ . . . neg . . . Q x . . . ] , where neg is a clausal negation c-commanding Q x , and Q x is a quantifier belonging to a certain class,10 Q x optionally incorporates neg.

This convention makes the appropriate distinction between (17)a and b: in the derivation of the LF of (17)a the structural description of (18) is never met, so that the not + many interpretation is unavailable; on the contrary, (18) can apply in the derivation of the LF of (17)b, as desired. Turning now to Italian, we may assume that the negative interpretation in such cases as (11), (14), (16)b is assigned to nessuno via convention (18): i.e., nessuno is a Q x . Putting together what has been said so far, we reach the conclusion that sentences (16)abc are assigned the following respective LF's: (19)

(a) Mi domando \_se [nessuno¡ [Mario ha contattato e¿]]] (b) [Non + nessuno¡ [Mario ha contattato e¡]] (c) *[Nessuno-l [Mario ha contattato e¡]]

(19)a and c are derived via QR from the corresponding S structures; (19)b is derived via QR and convention (18). (19)a-b are well-formed since the polarity requirement is fulfilled: nessuno is in the local context of a negative or interrogative marker; however (19)c (hence (16)c) is ruled out, since the polarity requirement is not fulfilled. In an exactly parallel way, this system associates the appropriate logical forms to (11), (14) and (15). 2.2: Let us now turn to (12)—(13). In these sentences there is no overt negative or interrogative marker, therefore they should be ruled out on a par with (16)c, in the way just discussed. Still, they are acceptable, and nessuno receives the interpretation of a negated existential, exactly parallel

124

Issues in italian syntax

to the interpretation of (11)—(14). Why is this so? Following a suggestion by R. Kayne, I will assume that what rescues (12)—(13) is a process similar to the rule "negative incorporation", proposed by E. Klima in his classical analysis of negation in English (cf. Klima 1964). As a first approximation, this process can be represented in the following way: (20)

nessuno

• [ + neg] when c-commanded by VP11

This rule applies at S structure associating the feature [ + neg] to a clauseinitial occurrence of nessuno, thus transforming the existential quantifier into a negated existential, and automatically fulfilling its requirement of a "negative or interrogative context". Rule (20) accounts for the wellformedness and interpretation of such examples as (12)—(13): for instance, (12)b would be assigned the following LF, via rule (20) and QR: (21)

[ Nessuno¡ [ e¡ ha visto Mario]] [ + neg] "There is no χ such that χ saw Mario"

It is important to notice that rule (20) must be optional: if it were obligatory, one could not account for the interpretation of such cases as (15)a indicated in the glosses, where no negation is involved and nessuno counts as an existential quantifier. As usual, if the optional rule (20) fails to apply in the derivation of (12)—(13), the resulting LF's are ruled out on apar with (19)c, since the polarity requirement is not fulfilled. 2.3. We can now address the central question formulated at the outset: is Italian different from French w.r.t. Kayne's examples? The answer is "no": exactly the same facts are found, modulo the independent differences between the systems of negation in the two languages. Take for instance the following examples, which correspond in structure to the French examples (7)M8)b: (22)

(a) Non pretendo che tu arresti nessuno. "(I) neg require that you arrest nobody." (b) Non pretendo che nessuno ti arresti. "(I) neg require that nobody arrest you."

(22)a is acceptable (marginally for some speakers) with nessuno construed with the higher occurrence of non, i.e., given our previous assumptions, with LF (23): (23)

[non + nessuno¡ [pretendo[che tu arresti e¡]] "There is no χ such that I require that you arrest x"

Consider now (22)b: it is a well-formed form (but see fn. 12), and in this

Negation, Wh -movement and the null subject parameter

125

respect Italian seems to differ crucially from French (cf. (8)b); but upon more careful consideration it turns out that (22)b is only acceptable in the totally irrelevant interpretation explicitly represented in LF (24): non and nessuno are construed independently of each other, non being the negation of the main clause, and nessuno receiving narrow scope interpretation as a negated existential via rule (20): (24)

[non pretendo [che nessuno¡ [e¡ ti arresti]]] [ + neg] "I do not require that there be no χ such that χ arrest you"

The relevant interpretation, with nessuno construed with the higher non, is impossible: (25)

* [ non + nessuno¡ [ pretendo[ che e- ti arresti] ] ] "There is no χ such that I require that χ arrest you"

Consider also the following examples with niente (= nothing):12 (26)

(a) Piero non crede che Gianni possa fare niente. "Piero neg thinks that Gianni can do nothing." (b) Piero non crede che niente possa spaventare Gianni. "Piero neg thinks that nothing can frighten Gianni."

(26)a is interpreted as indicated in (27)a, but (26)b cannot have the corresponding interpretation (27)b: it can only be interpreted with non and niente construed independently, as indicated in (27)c: (27)

(a) There is no χ such that Ρ thinks that G can do x. (b) T h e r e is no χ such that Ρ thinks that χ can frighten G. (c) Ρ doesn't think that there is no χ such that χ can frighten G.

The same subject-object asymmetry can be detected (even if the contrast is somewhat less sharp) when nessuno is an interrogative polarity existential quantifier. Discussing such cases as (15), we have seen that when the interrogative marker is in the same simple clause, no subject-object asymmetry is found. But if the interrogative marker is one clause up, the asymmetry shows up again: (28)

(a) Gianni mi ha chiesto se pensavo che tu avessi contattato nessuno. "Gianni asked me whether I thought that you had contacted anybody." (b) Gianni mi ha chiesto se pensavo che nessuno ti avesse contattato. "Gianni asked me whetherl thought that nobody had contacted you."

126

Issues in Italian syntax

As the glosses indicate, (28)a can be interpreted with nessuno as a wide scope existential, but this interpretation is not allowed for (28)b, where nessuno can only be interpreted as narrow scope negated existential (via rule (20)). In conclusion, putting aside the irrelevant interpretations (24), (27)c, etc., Italian turns out to be just like French in not allowing that a variable in preverbal subject position be bound by a "remote" quantifier of the nessuno class. It seems very plausible that this exactly parallel behaviour constitutes a true generalization. But if the solution in terms of ECPproposed for the French cases is to be extended to Italian, then the analysis of the null Subject Parameter presented at the outset must be revised, for the reasons already discussed.13 2.4. The next observation is that the impossibility of a "remote" binding solely concerns variables in preverbal subject position: if a subject containing elements like nessuno, niente, etc. is placed in postverbal position at S-structure, then the wide scope interpretation becomes available again; compare the following pairs: (29)

(30)

(a) Non pretendo che nessuno sia arrestato. "(I) do not require that nobody be arrested." (b) Non pretendo che sia arrestato nessuno. "(I) do not require that be arrested anybody." niente I (a) Piero non crede che mi possa spaventare. nessuno ) "Piero does not believe that (b) Piero non crede che

(32)

can

f r ight e n m e "

mi possa spaventare

"Piero does noi believe that (31)

not n ^S ) ( nobody \

mente

I nessuno \

can frighten me Ι » I anybody J

(a) Pensi che nessuno si presenterà? "(Do you) think that nobody will show up?" (b) Pensi che si presenterà nessuno? "(Do you) think that will show up anybody?" (a) Mi chiedo se Gianni pensi che nessuno lo stimi. "I wonder whether Gianni thinks that nobody esteems him." (b) Mi chiedo se Gianni pensi che lo stimi nessuno. "I wonder whether Gianni thinks that esteems him anybody."

Sentences a and ¿»only differ inform (essentially)w.r.t.thepre-vspost-verbal position of the subject, but they sharply differ in interpretation: as the English glosses indicate, the a sentences can only be interpreted with nessuno = narrow scope negated existential; on the contrary, the ¿sentences are interpreted with nessuno construed with the negative or interrogative marker in the higher clause. Notice that intuitions concerning the

Negation, Wh -movement and the null subject parameter

127

different interpretations of such pairs are sharper and easier to detect than ordinary scope intuitions, since they involve a different number of negations: for each pair, the interpretation of the a sentence involves one more negation than the interpretation of the b sentence (two vs one in (29), (30); one vj none in (31), (32)). 2.5 In conclusion, the result achieved so far can be summarized in the following descriptive statement: a variable in preverbal subject position cannot be bound by a "remote" operator of the nessuno type; a variable in postverbal (subject or object) position can: (33)

(a) * O p i . . . [ s , C O M P L Y . . . ] (b) Op¡ · . [ § ' COMP , . . V í ¡ . . . ]

The well-formedness of configuration (33)b is expected within an ECP approach: a base generated postverbal NP position (direct object, etc.) is properly governed by the verb; moreover, Kayne (1981b), Belletti & Rizzi (1981) provide evidence showing that a subject moved to the right ends up in a position governed by the verb (hence, properly governed). Therefore, further leftward movement of the postverbal subject in the syntax of LF does not yield ill-formedness, since ECP is not violated. The problematic case is configuration (33)a. Its ill-formedness would also follow, within an ECP approach, given the assumption that the preverbal position is ungovemed (or, at least, non-properly governed) in Italian too. But, again, this conclusion conflicts with the analysis of the Null Subject Parameter discussed before, which therefore requires some modification. 3. Specifying the Reference to INFL 3.1 The descriptive result of the preceding section seems to imply that INFL is not a proper govemer in Italian, at least in the relevant examples. The question which arises is then: which is the proper account of the Null Subject Parameter? This question will be addressed in the present section. Notice first of all that there seems to be a good argument, internal to Italian syntax, in favor of the relevance of INFL for the well-formedness of "null subject" sentences. We have seen in chapter III that in Italian a lexical subject can be found in certain uninflected clauses (gerunds and, at a marked stylistic level, infinitives) in which the auxiliary element is optionally moved to COMP, and assigns nominative case from this position to the subject. In such structures, a definite pronoun in subject position can never be phonetically null: (34)

(a) A proposito di Mario¡, ritengo [poter Γ^ΐ' disporre di fondi considerevoli] ' '' "As for MariOj, I believe [ to-be-allowed he¡/*0¡ to dispose of considerable funds] "

128

Issues in italian syntax (b) Riguardo a Francesca;, [avendo j ' j combinato questo pasticcio], sono nei guai fino al collo. 1 "Concerning Francesca;, [having she;/* no one); b. it is restricted to VP external position, while the rule for English is not: I saw no one. Notice that some locality requirement must be satisfied by rule (20), in order to block its application when the VP context is too "remote". A descriptively adequate locality specification would be "no S1 boundaries intervene, for i maximal." 12. It should be noticed that such sentences as (22)b, (26)b are only acceptable with a peculiar intonational pattern, i.e., with main stress on the main verb, and intonation break between the verb and the sentential complement: e.g., (ii)

non crédo // che niente possa spaventare Gianni

The account of this property would require a detailed discussion of the interaction between negation and intonation in Italian, which is beyond the aims of this work. A detailed analysis of negation in Italian will be proposed in an independent work, now in preparation.

Negation, Wh-movement and the null subject parameter

175

Concerning the examples at issue, it is worthwhile to mention the fact that in such contexts as (22)b, (26)b, ordinary negative polarity items behave differently from niente, nessuno, etc.: consider the following contrast (in the relevant interpretation and intonation) between niente and granché, a negative polarity item which can be roughly translated as anything significant: (ii)

non credo che *niente / ? granché sia stato fatto per aiutare i profughi "I neg believe that anything (significant) has been done to help refugies"

This suggests that the nessuno class and other quantifier-like expressions that are "negative polarity" items cannot be fully assimilated: the negative marker acts as a "scope marker" for the first class, but not for the second. 13. A hint that the Italian and French facts constitute a true generalisation is provided by the following observation. It seems to be a general property of the class of examples discussed in Kayne's article that whenever a Q-like element is found not only in preverbal position, but also in other argument positions of the same simple clause, the structure which forces wide-scope interpretation is improved. This effect can be detected with multiple wh questions, which generally show subject-object asymmetries of the type discussed in connection with negative quantifiers (Kayne 1981a): (i) (ii)

It isn't clear to me who knows if John saw who. *It isn't clear to me who knows if who saw John.

Now, if the most deeply embedded clause contains wh elements in both subject and object position, the result is significantly better: cf. the following contrast (examples from Kayne (1981a), based on an observation due to N. Chomsky, Pisa Workshop, spring 1979): (iii)

(a) *I know perfectly well which man said that which girl was in love with him/me (b) ?I know perfectly well which man said that which girl was in love with which boy

A similar (perhaps slighter) amelioration is found when the negative quantifier is involved: (iv)

(a) *Je n'exige que personne dise cela (b) ??Je n'exige que personne dise rien

Now, exactly the same effect is found in the corresponding Italian sentences: (v)

(a) Non pretendo che nessuno dica questo "I do not pretend that nobody say that" (b) Non pretendo che nessuno dica niente "I do not pretend that anybody say anything"

As the English glosses suggest, non and nessuno must be construed independently in (v)a, for the reason and with the modality discussed in the text; but the situation is changed in (v)b by the presence of another negative quantifier, and in this structure both niente and nessuno can be construed with the higher negation as negative polarity existential quantifiers. The parallelism with (iii) and (iv) is clear. Of course, this observation does not provide a compelling argument for the existence of a genuine generalisation in the cases at issue, due to the fact that the "amelioration effect" involved is poorly understood. But the parallelism at least suggests that the attempt to provide a uniform account for these cases is on the right track. The reason why an ECP violation causes ill-formedness in French (cf. (8)b), but not in Italian (where it simply eliminates one interpretation, cf. (22)b), etc.) is obviously due to independent differences in the two negative systems. The fundamental difference is that non in Italian is available both as clausal negation and as negative marker for such items as nessuno; on the contrary, ne is only available in French as "negative doubler" and scope

176

Issues in italian syntax

indicator of another negative element, the clausal negation being the complex form ne... pas. Therefore, there is no potential ambiguity for such forms as (8)b, since the clitic ne cannot function as negation of the main clause. The French equivalent of the acceptable reading of (22)a would be (vi)

Je n'exige pas que personne ne t'arrête

14.

The corresponding tensed clauses are, of course, acceptable:

(i)

(a) Gianni vorrebbe che avesse telefonato tuo fratello (b) Non è chiaro quando pioverà (c) Mario non sa perché è probabile che Gianni non ci aiuti (d) Sembra che siano successe cose terribili

While (40)d is unacceptable, the following is fine: (ii)

Sembrano essere successe cose terribili "Seem to have happened terrible things"

(ii) differs from (40)d in that the main verb agrees with the embedded postverbal subject. The derivation of this type of sentence, and the reason for the contrast will be discussed in par. 3.6. 15. This analysis has been proposed in Aoun (1979), following an original observation due to R. Kayne, and is developed in Jaeggli ( 1980a). On the general properties of Case theory see Chomsky (1980a, 1981a), Vergnaud (1979). 16. An alternative to the text theory could consist in identifying the primitive property of INFL in NSL's not in the optional specification [ + pronoun], but directly in the option of absorbing nominative Case. In this alternative, the feature [ + p r o n o u n ] would be taken as a purely descriptive device designating a Case-marked INFL, and everything else would be left essentially unchanged. The choice between the two alternatives depends in part on the still open question whether in general the feature [ + p r o n o u n ] is required for the identification of the class of elements with pronominal properties (e.g., with respect to the theory of binding). In fact, it might very well turn out that no primitive feature of this sort is needed, the "pronominal properties" of an item being a consequence of other aspects of its intrinsic constitution (e.g., such specifications as the features of person and number, and the like). For essentially expository reasons, in what follows we will continue to make reference to the feature [ +pronoun], leaving open the question whether it designates a primitive property, or is a convenient abbreviation for a Case-marked INFL. Some consequences of this footnote's alternative will be explored in appendix II. 17. Independent reasons supporting the idea that INFL absorbs nominative Case in "Null Subject" sentences are the following: a) Kayne (1980a) argues that the proper formulation of the Nominative Island Constraint must require that the "local antecedent" of a nominative empty category be Case marked. If this requirement is correct (see Kayne, op. cit., fn. 5, Pesetsky (1979), Chomsky (1981a) for discussion of potential problems), the natural transposition within an ECP framework would be that case II of the definition of "proper government" (cf. fn. 3) be reformulated with the additional requirement that a must be Case marked. Therefore, if INFL is a proper governer via case II of the definition, it must be assigned nominative Case (the only Case available in the local environment), as our analysis assumes. b) More direct evidence for the Case absorption property is provided in Z.de Fourier (1980). Our analysis predicts that, in "null subject" sentences, the empty subject position is un-Case marked, since INFL, being pronominal, must absorb Case. Fourier gives the following argument supporting this prediction: an extension of Longobardi's (1980a) filter, generally blocked by intervening Case-marked empty categories, is never blocked by

Negation, Wh-movement

and the null subject

parameter

177

an intervening null subject in tensed sentences (see also van Riemsdijk& Williams (1981) for relevant discussion of related issues). c) Another indication in favor of the text hypothesis is provided by the possibility of contracting certain wh elements in COMP with the auxiliary across a null subject in tensed sentences: (i)

(ii)

(a) Quando è arrivato? "When (did he) arrive?" (b) Quand'è arrivato? (a) Cosa hai fatto? "What (did you) do?" (b) Cos'hai fatto?

(in these cases the auxiliary cannot be moved to COMP by the rule discussed in ch. III because of the Doubly Filled COMP Effect, hence it is in its basic position, as the argument requires). On the assumption that, in general, contraction processes cannot take place across Case-marked empty categories (Jaeggli 1980b), this fact would again suggest that the intervening empty subject in (i)b, (ii)b is un-Case marked. 18. In order to simplify the discussion, we will consider only adjunction structures like (48), disregarding the possibility that the subject will be placed in the object position. On the latter type of structure, see Belletti & Rizzi (1981) and the thorough discussion in Burzio (1981). 19. An exactly parallel problem arises for the assignment of the appropriate thematic (0) role to the postverbal subject. Chomsky (1981a)has proposed that the proper assignment of θ roles is ensured by a well-formedness condition applying at LF, the θ Criterion, essentially requiring biuniqueness between θ roles and certain types of NP's. The θ Criterion can be phrased in the following way: (1)

(a) Each contentive element fulfills one and only one 0 role. (b) Each 0 role is fulfilled by one and only one contentive element. where 'contentive elements' are: lexically specified NP's, pronominals, lexical anaphors, variables, clauses.

Assuming, as Chomsky (1980b) does, that θ roles are assigned to NP's bearing appropriate grammatical functions, and that grammatical functions are configurationally defined in the language we are discussing, it would turn out that in (49), (50) the dummy elements (configurational subjects) are assigned the 0 role proper of the subject of to come. This would give rise to a double violation of the θ Criterion: a 0 role would be fulfilled by a dummy (non-contentive) element, in violation of clause a., and a contentive element (the postverbal subject) would be left without a θ role, in violation of clause b. The problem can be solved via a straightforward extention of convention (51): we can assume that, in the relevant configuration, the dummy element transmits the θ role assigned to its position to a coindexed NP in its domain. Notice that the theory of thematic relations, interacting with the process of "dummy" formation, produces a filtering effect which permits us keeping this process optional in the Italian case. Consider again (48): if [ + dummy] is not assigned to the pronominal INFL, the latter keeps the θ role assigned to the subject of telefonare, and no θ role can be assigned to the postverbal subject, in violation of clause a. of the 0 Criterion. Conversely, in a null subject structure like (i)

ei I N F L ha telefonato "e has telephoned"

if [ + dummy] is assigned to the pronominal INFL, the structure is ruled out again by the 0 Criterion, since a non-contentive element (the dummy INFL) keeps a 0 role, in violation of clause b. (i.e., (i) would be ruled out on a par with * there came). In conclusion, the θ

178

Issues in italian syntax

Criterion ensures that assignment of [ +dummy] apply exactly when it is needed in order to provide the correct interpretation. 20. Chao (1980) argues that Brazilian Portuguese seems to be exceptional with respect to the characteristic clustering of properties defining the "null subject" type in that it allows null subjects, in the relevant sense, but does not have a free process of subject inversion. A natural way of dealing with this case would consist in assuming that such a language does not have the process which transforms a definite pronoun into a dummy. In fact, this process clearly is an optional resources of UG: for instance, English too does not have it in its fully general form (it is raining, but* it came a man). If Chao is correct in arguing that subject inversion is highly restricted in at least one null subject language, this observation is relevant for a potential objection against our analysis. We have proposed an account which maximizes the similarities between subject inversion in Italian and the therein constructions in English and French. But there is a differential property which is not expected within this account: the English and French constructions obey a number of constraints on the nature of the verb and of the subject (e.g., the latter must be indefinite), while the Italian construction seems to be unconstrained. This difference is not predicted by our approach, within which we would expect to find only a loose (if any) cross-linguistic correlation between the null subject property and the character more or less constrained of subject inversion. In particular, this approach would be consistent with the existence of 1) non-NSL's with a subject inversion construction less constrained than in English and French, and 2) NSL's with a subject inversion construction more constrained than in Italian. Chao's observation, providing a candidate for case 2), seems to suggest that this picture is not implausible. 21. a binds β iff a c-commands and is coindexed with β; a is bound in β if there is a γ in β which binds a; a is free in β iff a is not bound in β; a is the governing category (GC) for β iff a is the minimal category contianing β and a governor of β, where a = NP or S. These structural relations are discussed in detail in Chomsky (1981)a; see also ch. III, section 5 and Belletti & Rizzi (1981) for some minor modifications. 22. This incompatibility is rather natural in terms of an intuitively plausible requirement of avoiding circularity in the binding relation. A somewhat related requirement is shown to be necessary for pronominal coreference in Higginbotham & May (1981). 23. Chomsky (Pisa lectures) has noticed that the binding relation, as formulated in fn. 21, is too general for the requirements of the binding principle, and must be restricted in some way in order to give empirically correct results (Chomsky refers to the restricted definition as "argument binding"). Our formulation (60) can be viewed as a partial characterization of the needed restriction. A full characterization would require some additional specification, to the effect that elements peripheral to S at LF (ordinary and wh quantifiers, NP's in TOP, etc.) do not count as "binders" in the relevant sense. 24. Why it is so is clear for the infinitival questions (71 )b-c: here, the COMP is already filled by the wh element, and further application of Aux to COMP is excluded by the wellknown Doubly Filled COMP Effect. As for(71)a-d, one possibility would consist in adopting the suggestion of Chomsky (class lectures, fall 1979), developed in chapter III, according to which ordinary control (and, we may add, raising) complements are base-generated, in the unmarked case, without a COMP position. In this case, the non-applicability of Aux to COMP would simply follow from the lack of a suitable "landing site". 25. On the incompability of "Arb" interpretation and "dummy" interpretation, see appendix, III. 26. That the postverbal subject is in the embedded clause can be shown in the following way. The clitic pronoun ne, corresponding to an N' modified by an indefinite quantifier, can be extracted from a postverbal subject in a number of cases, e.g., (i)

(a) Sono intervenuti alcuni rappresentanti delle Nazioni Unite "Have taken part some officials of United Nations" (b) Ne sono intervenuti alcuni "Of-them-have taken part some "

Negation, VJh-movement and the null subject parameter

179

the sentence corresponding to (i)a can be embedded as the complement of a raising verb (cf. (ii)a); in this case, ne can be extracted from the postverbal subject and cliticized onto the embedded verb (cf. (ii)b): (ii)

(a) Sembrano [essere intervenuti alcuni rappresentanti delle N U ] "Seem to have taken part some officials of UN" (b) Sembrano [esserne intervenuti alcuni ] "Seem to have-of-them taken part some "

The acceptability of (ii)b shows that the noun phrase [alcuni ] is in the embedded clause: if it were in the main clause, the trace of ne would be unbound, and the structure would be ill-formed w.r.t. the binding principle. In fact, (ii) minimally contrasts with the following: (iii)

(a) Hanno dichiarato di essere intervenuti alcuni rappresentanti delle NU "Have declared to have taken part some officials of UN" (b) *Hanno dichiarato di esserne intervenuti alcuni "Have declared to have-of-them taken part some "

In (iii)a the main verb is a control verb. Hence, because of its structural requirement, the postverbal subject cannot be in the embedded clause, and must be in the main clause. And, in fact, as (iii)b shows, cliticisation of ne onto the embedded verb produces an unacceptable result, as predicted by the binding principle. (Ne cannot be cliticized to the main verb either for independent reasons, discussed at length in Belletti & Rizzi (1981)). 27. The plural agreement on the main verb is ensured, when raising has applied, by an obvious consistency condition, the same which is necessary, in English, in order to account for such cases of "remote agreement" as There seems to be a riot, there seem to be riots. 28. Let us consider in detail all the binding relations (integers are assigned to the positions involved, in order to simplify the discussion): (i)

et IN FL; sembrano [e¡ [ +dummy] 1 2 3

NR] 4

Notice first that the embedded VP assigns a θ role to 3, which transmits it to 1 (in the way usually assumed for the raising construction); 1 transmits it to 2, as proposed in the preceding paragraph, and 2 sends it back to 4, via convention (51). Therefore, according to our definition, 3 is bound by 1 (actually, by 2 as well), and 1 is bound by 2, so that the requirements of clause (A) are fulfilled. 2 is free, since 1 c-commands it and is coindexed with it, but 2 is θ dependent from 1, as discussed, so that the binding relation does not hold; clause (B) is thus fulfilled. 4 is free: it is c-commanded and coindexed with 1, 2 and 3, but, because of transitivity, 4 is θ dependent from all of them, so that in no case will the binding relation hold. Clause (C) is thus fulfilled, and the structure is well-formed. 29. An alternative which might come to mind for the treatment of null dummies in Aux to COMP clauses could consist in eliminating reference to an abstract inflection for this case, and attributing the non-violation of ECP to the fact that the Aux in COMP properly governs the empty subject position: (i)

[ s , essendo [ s e [ v p piovuto]]]

This solution would adequately account for the well-formedness of this structure with respect to ECP, but it would not be sufficiently structured in other respects. First of all, it would not deal with the requirements of the theory of binding (the empty category, an anaphor, would be unbound); moreover, it would not account for its "dummy" interpretation. This solution is therefore less satisfactory than the one making reference to the abstract inflection, as

180

Issues in Italian syntax

proposed in the text. Moreover, this footnote's alternative would incorrectly predict that in any construction in which an auxiliary verb governs the subject a "null dummy" should be possible, even in a non-NSL. That this prediction is incorrect is shown by the unacceptability of "null dummies" in English structures in which subject-Aux inversion applies: (ii)

(a) Is *(it) raining? (b) Is *(it) clear that John will come?

On the contrary, the unacceptability of these sentences is correctly predicted by the text proposal, since the abstract inflection in English can never be pronominal. 30. Where F, F n are morphologically realized features of grammatical specification, including [ o r person]. Direct evidence for the fundamental role of the person specification is provided by modern Hebrew, according to the description of Borer (1979). In this language, the past tense inflection has grammatical specification of person, number and gender, and present tense inflection is only specified for number and gender. Now, a null dummy subject seems to be generally possible, e.g., (i)

Mistaber "Turns-out

se-'Avihu that-'Avihu

'ohev loves

bananot bananas

(Borer (1979), ex.(10))"

This shows, in our terms, that this language has the fundamental "null subject" property. But a null definite pronoun is only possible in past tense clauses: (ii) (iii)

('ata) 'axalta "(you) ate-2nd-m-sg (a) 'ata 'oxel "you eat-m-sg (b) *'oxel 'et "eat-m-sg acc

'et acc 'et acc

ha-banana the-banana" ha-banana the-banana" ha-banaba the-banana"

that is to say, the contrast between present and past tense clauses in Hebrew mirrors the contrast between tensed and Aux-to-COMP clauses in Italian. But the contrast in Hebrew is more minimal, since the presence vs absence of the person specification seems to be the only factor involved. 31. It is very plausible that the specification of the two parameters must be completed by various markedness assumptions. In particular, a reasonable refinement of (75)b would consist in eliminating from its formulation the direct reference to the grammatical specification F, F n ; this specification would then be made appeal to in ranking the possible values of the parameter along a dimension of markedness. The value "yes" would thus be characterized as unmarked (least marked) when the grammatical specification of the pronominal INFL is maximal, and its markedness would increase with the decreasing of the grammatical specification (again, the morphological realization of the feature [a person] presumably defining a threshold). (75)b so simplified would not necessarily be assigned a single value in a specific grammar, but could receive different values for different verbal paradigms within the same grammar (perhaps, even for different persons within the same paradigm, as seems to be required by certain Italian dialects). Thus, in Italian the value would be "yes" for the tensed paradigms, and " n o " for gerunds and infinitives; similarly, in Hebrew the value would be "yes" for the past tense paradigm, and " n o " for the present tense paradigm: in Dutch A and Icelandic the value would be " n o " for all paradigms, etc. 32. Adopting Freidin's (1978) hypothesis that strict cyclicity is not to be stipulated as an independent principle, (79) is uniquely excluded by the θ criterion in the way indicated. Otherwise, it would also be excluded as a violation of strict cyclicity (there insertion would apply in the embedded cycle after raising has applied in the main cycle). Notice that there is also another derivation of (79) (and (77)) which is excluded for independent reasons, i.e.,

Negation, Wh-movement and the null subject (i)

parameter

181

* Some friendSj seem [that there have arrived e¡]

in this case, some friends is moved to the right and there insertion applies in the embedded cycle; then some friends is raised from postverbal position at the main cycle. Here, the θ Criterion is not violated: there can tansmit the θ role assigned by the VP to the postverbal trace via convention (51); then the latter transmits it to the main subject in the usual way. Here, what rules out the structure is the binding principle, since the postverbal trace, an anaphor, is free in its minimal governing category. 33. If INFL is pronominal, (81) is also ruled out by the Case filter: INFL absorbs nominative Case, hence no Case can be assigned to the wh element in the higher COMP (but this conclusion does not hold if Kayne's (1981c) proposal of Case assignment in COMP by a higher verb is correct). Moreover, (81) would also be excluded by the θ criterion: either the pronominal INFL or the wh variable in (81) can receive the θ role proper of the subject of venire, but not both. Since both are contentive elements in the relevant sense, the θ criterion is systematically violated. 34. As for Case theory, nominative Case is transmitted from the dummy INFL to the postverbal subject via convention (51). The wh element chi in the higher COMP then "inherits" Case from its variable, as we may assume happens in general. 35. If Chao (1980) is correct in claming that some NSL's do not have a free process of subject inversion, then our approach would predict that all cases of apparent violation of the COMP-trace effect in these languages are not instances of wh movement, but rather are cases of the so-called "resumptive strategy", in which the operator-variable structure is not created by a movement rule, but is a base-generated structure in which an element in preclausal position is construed with a clause-internal resumptive pronoun. This is in fact the analysis argued for in the paper quoted. 36. It is often assumed in the literature that the que > qui rule of Kayne (1974) is obligatory. If this is correct, then (88) is also ruled out by the fact that an obligatory rule fails to apply in its derivation. 37. Instead, French has subject clitics. It is perfectly compatible with our approach to assume that subject clitics are generated under the INFL node, as originally suggested by Milner (Pisa workshop, spring 1979). See also Belletti (1980). 38. The notation . . . * ( . . . ) . . . is to be understood in the following way: if the content of the parenthesis is not realized, the structure is ungrammatical. Conversely, the notation . . . ( * . . . ) . . . indicates that the structure is ungrammatical if the content of the parenthesis is realized. 39. The stars in (97)-(98)b require some qualification. The relevant sentences are unacceptable when pronounced with continuous intonation, but are significantly improved by an intonation break after the wh quantifier: (i)

(a) (di queste pietre,) quante I ! sono cadute? "(of these stones,) how many // fell down?" (b) (di questi libri,) quanti // sono usciti? "(of these books,) how many // have come out?"

If the special intonation which rescues (97) (98)b does not have a structural correlate (i.e., if (97)-(98)b and (i)a-b have the same structure), then the problem discussed in the text simply does not arise: the relevant structures are grammatical, as predicted by the proposed analysis, which need not be revised. But it is not implausible that the special intonation which rescues these sentences does have a structural correlate. First of all, this is the characteristic intonation which separates certain items base-generated in TOP position from the clause; moreover, there are clear instances of interrogatives involving the wh quantifier quanto/i constructed via the resumptive strategy: quanto/i is base-generated in T O P position, and binds an S internal clitic resumptive pronoun:

182

Issues in italian syntax

(ii)

(di questi libri,) quanti // puoi dire di conoscer/i abbastanza bene? "(Of these books,) how many // you can say you know them well enough?"

A structure like (ii) requires the special intonation found in (i): if read with continuous intonation, (ii) becomes unacceptable. It is then plausible that the special intonation which rescues (97)-(98)b has a structural correlate, and characterizes an interrogative construction using the "resumptive" strategy (i.e., parasytic on the left dislocated construction; see Cinque (1978) for an extensive discussion of the corresponding case with the relative construction). If so, then the well-formedness of (i) is irrelevant, and our approach is to be modified in order to rule out (97)-(98)b, as is proposed in the text. But it should be considered that if this footnote's conclusion on (i) is incorrect, the discussion of this paragraph might become superfluous. 40. This is, informally speaking, the "unmarked" assumption. But alternatives have been proposed, e.g., in Kayne (1981) (see also Longobardi 1980a). If narrow scope assignment doesn't (necessarily) involve QR, as Kayne suggests, then the possibility of narrow scope for a negative quantifier in preverbal subject position is irrelevant for the point at issue. 41. The marginal status of this structure is plausibly to be attributed to the very peculiar intonational pattern which it requires, and the related very peculiar pragmatic conditions in which it can be appropriately uttered. We have already seen (fn. 12) that the interpretation with non-nessuno construed independently of each other requires a special intonational contour, with heavy stress on the main verb, followed by an intonational break. The preposing of a nessuno in the embedded clause requires, in turn, heavy stress on the preposed phrase. Therefore, (5) is only acceptable with the following intonation: (i)

non crédo II che a nessuno Gianni l'abbia detto

The sentence could be uttered as a denial of (4). 42. Notice that, if the proposed account is correct, in order to rule out such ungrammatical sentences as the starred versions of (94), (95)b, and (77), it is necessary to block only one derivational path, not two: the inflection would invariably be pronominal, hence the derived structure would be ruled out by the binding principle (and possibly by the θ Criterion and Case filter) for (94), (95)b, and by the θ Criterion for (77); the other a priori possible derivation discussed in the text, with INFL non-pronominal, simply would not arise. 43. Given the definition of Governing Category proposed in Belletti & Rizzi (1981), briefly discussed in chapter III, par. 5, the " P R O Theorem" would be slightly different: " P R O is ungoverned and is not in a context of Case assignment". This possible modification is not directly relevant here, and will not be discussed in what follows. 44. The superscript shared by the subject NP and INFL is the formal device characterizing agreement in this system. It is assumed that the co-superscripting procedure applies at D-Structure, and takes place whenever INFL is a governor (e.g., with tensed INFL containing overt agreement features, but not with infinitival INFL in English and French; on Italian non-finite inflections, see below). A noun phrase moved to the subject NP position in the course of the derivation inherits the superscript; moreover, if the subject NP is further moved, it carries the superscript along. 45. I.e., the functional principle requiring that the pronominal form used in a given position of an utterance be the "shortest possible", compatibly with the grammatical principles constraining the set of possible fillers of this position (see Chomsky (1981a) for discussion). 46. Notice that the PRO approach still has to block "local" applications of wA-movement, according to the discussion of par. 4.3. This result can be obtained in several different ways. One possibility would consist in complicating the process of Affix Movement, in order to mimic the device introduced in appendix II, or some similar mechanism. Alternatively, one could formulate the process of dummy P R O insertion as obligatorily applying to an empty superscripted N P in subject position, thus systematically "filling" this position when-

Negation, Wh -movement and the null subject parameter

183

ever it is empty at S-Structure, and rendering it unavailable for binding by a wh operator. Other possible approaches are discussed in Chomsky (1981a), eh. III, fn. 33. 47. See Chomsky (1980a) and, for a recent formulation of the theory of control, Manzini (1980b). 48. But this theory still has to account for the fact that the superscripted PRO appearing in (16)—(18) is restricted to the "dummy" interpretation, and does not have the full range of possible referents of a definite pronoun, thus contrasting with the superscripted PRO subject of a tensed clause. This problem is addressed below. 49. Notice that structures (e) and (f) are presumably also ruled out by the fact that, Aux-to-COMP having applied in the syntax, Affix Movement cannot properly apply in the phonology, hence the structure is barred by a plausible condition of morphological wellformedness which prohibits "dangling" affixes. 50. Notice that the lack of "referential independence" must be restricted to null pronominale, since an overt pronoun in subject position of the gerund freely allows an obviative interpretation: cf. (25) with (i)

Essendo lei così simpatica, Gianni è contento "Being her so nice + FEM, Gianni is happy + MASC"

It should be further noticed that this analysis implies that a sentence like (15)b is in fact ambiguous between a structure of control (case (a) (or (c)) of the matrix), and a "null subject" structure with a proximate superscripted PRO (case (d) of the matrix). 51. The relevance of the grammatical specification of INFL for the pronominal interpretation is further stressed by the comparison of Italian and Portuguese (A. Belletti, p.c.). Portuguese is like Italian in allowing "null subjects" in certain classes of infinitives, but differs from Italian in that such infinitives are overtly inflected with the grammatical specification of agreement (features of person and number); and, as is expected given (29), "null subjects" in Portuguese inflected infinitives are not restricted to the " d u m m y " interpretation, and have the full range of possible referents of definitive pronouns (and of "null subjects" of tensed sentences in both Portuguese and Italian). For discussion of the "null subject" properties of inflected infinitives in Portuguese see Quicoli (1972), Rouveret (1980), Zubizarreta (1980b). 52. This might lead one to try to sharpen the tentative suggestion that PRO' is in fact in the domain of the theory of control, the controller being INFL'. The role of INFL' in determining the interpretation of PRO' would then become naturally understandable as a case of the ordinary control relation (the referential properties of an occurrence of PRO obviously depend on the intrinsic properties of its controller). But this move would require dealing with a number of non-trivial technical problems which will not be discussed here. 53. "Revised" with respect to the original definition of the Pisa Lectures, given in fn. 21. For the present purposes, the accessibility relation used in (30) can be assimilated to c-command (but see fn. 12). 54. As for the fact that NP-(and CL-) movement is impossible from the embedded subject position (cf. chapter III, par. 5), this also follows, if appeal is made to the "agreement rule" (23) (strengthened to a biconditional): COMP is empty, hence INFL is [ + gov] (i.e., it counts as a SUBJECT); if Aux-to-COMP does not apply (cases (b)-(d) of the matrix), extraction of the subject via N P - ( C L - ) movement is ruled out by ECP. If Aux-to-COMP does apply, the only possible structure is (h) ((f) being ruled out by the condition on morphological well-formedness referred to in fn. 7). But if the subject is N P - ( C L - ) moved from this structure, the derived structure is ruled out by the binding principle (the trace would be free in its GC), on a par with (32). 55. Notice that the required definition of the notion SUBJECT, if combined with Chomsky's treatment of the parameter, would involve a non-trivial consequence: it would turn out that whenever Affix Movement applies in the syntax, the G C of the verbal complements is the VP ( = the minimal category in which they are governed and which contains an accessible SUBJECT). This would have consequences for cases like the following:

184

Issues in italioti

syntax

(i)

P R O ' [yp parla + INFL' di se stesso] " P R O speaks of himself"

In order to account for the well-formedness of (i), it would be necessary to assume that INFL' can be the antecedent of the reflexive pronoun; otherwise the latter would be free in its GC. Alternatively, if one wants to avoid this conclusion, it is possible to restrict the definition of SUBJECT in such a way that a governing INFL is the SUBJECT of the first phrase of type S' which contains it; hence in (i) the G C of the reflexive is S, where it is bound by the subject N P PRO'.

Bibliography

Abbott, Β. (1976) "Right Node Raising as a test for constituenthood", Linguistic Inquiry 7. Aissen, J. and D. Perlmutter (1976) "Clause Reduction in Spanish", Proceedings of the Second Annual Meeting of the Berkeley Linguistic Society. Anderson, S. and P. Kiparsky (1973) A Festschrift for Morris Halle, Holt, Rinehart and Winston, New York. Aoun, J. (1979) "On Government, Case-Marking, and Clitic Placement", mimeographed, MIT. Aoun, J., N. Hornstein and D. Sportiche (1981) "Some aspects of wide scope quantification", Journal of Linguistic Research 1. Baker, C. L. (1968) Indirect Questions in English, PhD dissertation, University of Dlinois. Barbaud, Ph. (1976) "Constructions supérlatives et structures apparentées", Linguistic Analysis 2. Belletti', A. (1980) "Morphological passive and Pro Drop: the impersonal construction in Italian", mimeographed, ΜΓΓ. Belletti, A. and L. Rizzi (1981) "The Syntax of ne: some theoretical implications, The Linguistic Review 1. Belletti, Α., L. Brandi and L. Rizzi eds. (1981) Theory of Markedness in Generative Grammar. Scuola Normale Superiore, Pisa. Benincà, P. (1980) "Il (soggetto) clitico a", unpublished manuscript. University of Padua. Berman, Α. (1974) Adjectives and Adjective Complement Constructions·, P h D dissertation, Harvard University. Besten, Η. den (1977) "On the interaction of root transformations and lexical deletive rules", mimeographed, University of Amsterdam. Bordelois, I. (1974) The Grammar of Spanish Causative Complements, PhD dissertation, MIT. Borer, H. (1979) "Empty subjects and constraints on thematic relations", Proceedings of NELS X, Ottawa. Bresnan, J. (1971) "Sentence stress and syntactic transformations", Language 47. Bresnan, J. (1972) Theory of Complementation in English Syntax, PhD dissertation, MIT. Bresnan, J. (1973) "A note on free relatives", mimeographed, ΜΓΓ. Bresnan, J. (1975) "Comparative deletion and constraints on transformations", Linguistic Analysis 1. Bresnan, J. (1976) "Evidence for a theory of unbounded transformations", Linguistic Analysis 2. Bresnan, J. (1978) "A realistic transformational grammar", in M. Halle, J. Bresnan, G. Miller eds, Linguistic Theory and Psychological Reality, The ΜΓΓ Press, Cambridge, Mass. Burzio, L. (1979) "Italian causative constructions", Journal of Italian Linguistics 3.2. Burzio, L. (1981) Intransitive Verbs and Italian Auxiliaries, P h D dissertation, ΜΓΓ. Chao, W. (1980) "Pro Drop languages and non-obligatory control", mimeographed, University of Massachusetts, Amherst. Chomsky, N. (1965) Aspects of the Theory of Syntax, The ΜΓΤ Press, Cambridge, Mass. Chomsky, N. (1972) Studies on Semantics in Generative Grammar, Mouton, The Hague. Chomsky, N. (1973) "Conditions on Transformations", in Anderson and Kiparsky (1973).

186

Issues in Italian

syntax

Chomsky, N. (1974) "The Amherst Lectures", Lectures given at the 1974 Linguistic Institute, University of Massachusetts, Amherst; Université de Paris VII. Chomsky, Ν. (1975) Reflections on Language, Pantheon Books, New York. Chomsky, N. (1976) "Conditions on rules of grammar", Linguistic Analysis 2. Chomsky, N. (1977a) "On wA-movement", in P. Culicover, T. Wasow, A. Akmajian eds. Formal Syntax, Academic Press, New York. Chomsky, N. (1977b) Essays on Form and Interpretation, North Holland, New York. Chomsky, N. (1980a) "On Binding", Linguistic Inquiry, 11.1. Chomsky, N. (1980b) "On the representation of form and function", The Linguistic Review 1. Chomsky, N. (1980c) Rules and Representations, Columbia University Press, New York. Chomsky, N. (1981a) Lectures on Government and Binding, Foris Publications, Dordrecht. Chomsky, N. (1981b) "A note on non-control PRO", mimeographed, ΜΓΓ. Chomsky, N. and H. Lasnik (1977) "Filters and Control", Linguistic Inquiry 8.3. Cinque, G. (1978) "La sintassi dei pronomi relative 'cui' e 'quale' nell'italiano moderno". Rivista di grammatica generativa 3. Cinque, G. (1980) "On the theory of relative clauses and markedness" The Linguistic Review. 1.3 Couquaux, D. (1978) "On trace interpretation", mimeographed, Université de AixMarseille II. Dresher, Β. E. and N. Nomstein (1979) "Trace theory and NP movement rules", Linguistic Inquiry 10. Emonds, J. (1976) A Transformational Approach to English Syntax, Academic Press, New York. Emonds, J. (1978) "The verbal complex [ v , V' - V] in French", Linguistic Inquiry 9. Erteschik, N. (1973) On the Nature of the Island Constraints, PhD dissertation, MIT. Fiengo, R. (1974) Semantic Conditions on Surface Structure, PhD dissertation, MIT. Fiengo, R. (1977) "On trace theory", Linguistic Inquiry 8.1. Fodor, J. (1978) "Parsing strategies and constraints on transformations", Linguistic Inquiry 9. Fourier, Z. de (1980) "Remarks on the 'Extended Linear Model'" mimeographed, ΜΓΓ. Freidin, R. (1978) "Cyclicity and the theory of grammar", Linguistic Inquiry 9.4. Freidin, R. and H. Lasnik (1981) "Disjoint reference and wh trace", Linguistic Inquiry 12.1. Fresina, C. (1980) Aspects de la grammaire transformationnelle de l'Italien, thèse de 3ème cycle, Université de Paris VIII. George, L. (1980) Analogical Generalizations of Natural Language Syntax, PhD dissertation, MIT. Godard, D. (1980) Les relatives complexes en Français, thèse de 3ème cycle, Université de Paris VII. Goldsmith (1978) "Complementizers and root sentences", unpublished paper, Indiana University. Guéron, J. (1975) "Une contrainte sur l'interprétation de phrases contenant NP et PP", Cahiers de Linguistique de l'Université du Québec. Guéron, J. (1980) "On the syntax and semantics of PP extraposition" Linguistic Inquiry 11.4. Higginbotham, J. and May, R. (1981) "Crossing, markedness, and pragmatics", in Belletti, Brandi, Rizzi (1981). Higgins, F. R. (1973) The Pseudo-Cleft Construction in English, PhD dissertation, MIT. Horn, G. (1974) The Noun Phrase Constraint, PhD dissertation, University of Massachusetts, Amherst. Hornstein, Ν. and D. Lightfoot eds. (forthcoming) Explanation in Linguistics, Longmans. Jackendoff, R. (1972) Semantic Interpretation in Generative Grammar, The MIT Press, Cambridge, Mass. Jackendoff, R. (1977) X-bar Syntax: a Study of Phrase Structure, Linguistic Inquiry monograph 2, The ΜΓΓ Press, Cambridge, Mass.. Jaeggli, O. (1980a) On Some Phonologically Null Elements in Syntax, PhD dissertation, MIT. Jaeggli, O. (1980b) "Remarks on to contraction", Linguistic Inquiry 11.1. Kayne, R. S. (1969) The Transformational Cycle in French Syntax, PhD dissertation, MIT.

Bibliography

187

Kayne, R. S. (1974) "French relative que". Recherches linguistiques 2-3. Kayne, R. S. (1975) French Syntax, The ΜΓΤ Press, Cambridge, Mass. Kayne, R. S. (1979) "Rightward NP movement in French and English" Linguistic Inquiry 10.4. Kayne, R. S. (1980a) "Extensions of binding and Case-marking", Linguistic Inquiry 11.1. Kayne, R. S. (1980b) "Vers une solution d'un problème grammatical: *je l'ai voulu lire, j'ai tout voulu lire", Langue française 46. Kayne, R. S. (1981a) "Two notes on the NIC", in Belletti, Brandi, Rizzi (1981). Kayne, R. S. (1981b) " E C P extensions", Linguistic Inquiry, 12.1. Kayne, R. S. (1981c) "On certain differences between English and French", Linguistic Inquiry 12.3. Kayne, R. S. and J. -Y. Pollock (1977) "Stylistic inversion, successive cyclicity, 'Move NP', and L F " , unpublished paper, Université de Paris VIII. Kayne, R. S. and J. -Y. Pollock (1978) "Stylistic Inversion, Successive cyclicity, and 'Move NP' in French", Linguistic Inquiry 9.4. Keyser, S. J. (1978) Recent Transformational Studies in European Languages; Linguistic Inquiry monograph 3, The ΜΓΓ Press, Cambridge, Mass. Kiparsky, P. (1973) "'Elsewhere' in phonology", in Anderson and Kiparsky (1973). Klima, E. (1964) "Negation in English", in J. Katz and J. Fodor eds. The Structure of Language, Prentice-Hall, Englewood Cliffs. Koster, J. (1976) "Remarks on wh movement and the locality principle", unpublished paper. University of Amsterdam. Koster, J. (1978) Locality Principles in Syntax, Foris Publications, Dordrecht. Lasnik, H. (1972) Analyses of Negation in English, PhD dissertation, ΜΓΓ. Lasnik (1980) "Restricting the theory of transformations: a case study", in Hornstein and Lightfoot (forthcoming). Lasnik, H. and R. Fiengo (1974) "Complement object deletion". Linguistic Inquiry 5. Lightfoot, D. (1980) "The history of NP Movement", in Hoekstra, van der Hulst and Moortgat eds., Lexical Grammar, Foris Publications, Dordrecht. Longobardi, G. (1980a) "Remarks on infinitives: a case for a filter" Journal of Italian Linguistics 5. Longobardi, G. (1980b) "Modularity: from a single rule to several conditions without loss of generalization: evidence from 'identificational predicates'", communication presented at the V GLOW Conference (Nijmegen). Longobardi (forthcoming) "Postille alla regola di ristrutturazione", to appear in Rivista di grammatica generativa. Maling, J., and A. Zaenen (1978) "The non-universality of a surface filter", Linguistic Inquiry 9.3. Manzini, M. R. ( 1980a) "Sulla struttura di un certo tipo di frasi infinitivali italiane", to appear in Studi mediolatini e volgari, Pisa. Manzini, M. R. (1980b) "On control", italian version to appear in Rivista di grammatica generativa. May, R. (1977) The Grammar of Quantification, P h D dissertation, ΜΓΓ. Napoli, D. J. (1973) The Two Si's of Italian, PhD dissertation, Harvard University, Perlmutter, D. (1970) "The two verbs Begin," in R. Jacobs and P. S. Rosenbaum eds., Readings in English Transformational Grammar, Ginn-Blaisdell, Waltham, Massachusetts. Perlmutter, D. (1971) Deep and Surface Structure Constraints in Syntax, Holt, Rinehart and Winston, New York. Pesetsky, D. (1979) "Complementizer-trace phenomena and the Nominative Island Condition," The Linguistic Review. 1.3 Pollock, J. -Y. (1976) "Théorie des traces et syntaxe du français, quelques problèmes," Recherches Linguistiques, 4: English Version in Keyser (1978). Postal, P. (1971) Cross-Over Phenomena, Holt, Rinehart and Winston, New York. Postal, P. (1974) On Raising, MIT Press, Cambridge, Massachusetts. Quicoli, C. (1972) Aspects of Portuguese Complementation, PhD dissertation, State University of New York.

188

Issues in italian syntax

Quicoli, C. (1976) "Conditions on Clitic Movement in Portuguese," Linguistic Analysis. 2. Radford, A. (1977) Italian Syntax: Transformational and Relational Grammar, Cambridge University Press, Cambridge. Reinhart, T. (1976) The Syntactic Domain of Anaphora, PhD dissertation, MIT. Reinhart, T. (1977) "A note on the two-COMP hypothesis," unpublished paper, ΜΓΤ. Reinhart, T. (1981) "A second COMP position," in Belletti, Brandi, Rizzi (1981). Riemsdijk, H. van (1976) "Extraction from PP and the head constraint," unpublished paper, University of Amsterdam. Riemsdijk, H. van (1978) A Case Study in Syntactic Markedness, Foris Publications, Dordrecht. Riemsdijk, H. van & E. Williams (1981) "NP-Structure," unpublished paper, Amsterdam University of Massachusetts. Rivas, A. (1974) "Impersonal sentences and their interaction with Clitic Movement in Spanish," unpublished paper, MIT. Rizzi, L. (1976a) "Ristrutturazione," Rivista di grammatica generativa, 1. Rizzi, L. (1976b) "La Montée du sujet, le si impersonnel et une règle de restructuration dans la syntaxe italienne," Recherches Linguistiques 4. Rizzi, L. (1978) "Nominative Marking in Italian Infinitives and the NIC," communication presented at the II GLOW Conference (Amsterdam), to appear in F. Heny ed. Binding and Filtering, Croom Helm, London. Rizzi, L. (1980) "Processi fonosintattici e teoria della traccia," forthcoming in Rivista di grammatica generativa. Rizzi, L. (1981) "On the doubly filled COMP effect," unpublished paper, Università della Calabria. Ross, J. R. (1967) Constraints on Variables in Syntax, PhD dissertation, ΜΓΓ. Rouveret, A. (1980) "Sur la notion de proposition finie," Recherches Linguistiques 9. Rouveret, Α. & J. -R. Vergnaud (1980) "Specifying reference to the subject," Linguistic Inquiry 11. Ruwet, N. (1972) Théorie Syntaxique et Syntaxe du Français, Seuil, Paris. Ruwet, N. (1974) "Phrases copulatives," Recherches Linguistiques 3. Ruwet, N. (1979) "On a verbless predicate construction in French," Papers in Japanese Linguistics vol. 6, Memorial Volume S.I. Harada. Selkirk, E. (1972) The Phrase Phonology of English and French, PhD dissertation, MIT. Sportiche, D. (1979) "On bounding nodes in French," unpublished paper, MIT. Stowell, T. (1979) "What was there before there was there," in D. Farkas, W. M. Jacobsen, and K. W. Todrys, eds. Papers from the Fourteenth Regional Meeting of the Chicago Linguistic Society, University of Chicago. Taraldsen, Κ. T. (1978) "On the NIC, vacuous application and the that-trace filter," unpublished paper, MIT. Taraldsen, Κ. T. (198Í) "The theoretical interpretation of a class of marked extractions," in Belletti, Brandi, Rizzi (1981). Vanelli, L. (1980) "A suppletive form of the Italian article and its phonosyntax," Journal of Linguistic Research, 1. Van Tiel-Di Maio, M. F. (1975) "Una proposta per la sintassi dell'italiano: V-raising" in Atti del IX Congresso annuale della Società di Linguistica Italiana - Roma 1975. Vergnaud, J. -R. (1979) Quelques éléments pour une théorie formelle des Cas, manuscript, University of Massachusetts, Amherst. Weinberg, A. & N. Hornstein (1981) "Case theory and Preposition stranding," Linguistic Inquiry 12. Zubizarreta, M. -L. (1980a) "Pourune restructuration thématique" Recherches Linguistiques1). Zubizarreta, M. -L. (1980b) "Remarks on Portugese infinitives", unpublished paper, MIT.